Proceedings of a Workshop
Predicting Human Health Effects from Environmental Exposures: Applying Translatable and Accessible Biomarkers of Effect
Proceedings of a Workshop—in Brief
Biomarkers of effect are measurable changes in an individual that indicate health impairment or disease. Although biomarkers have long been a crucial part of medical practice—blood pressure is a simple example—researchers have recently identified a variety of new biomarkers that signal the presence of conditions such as nervous system damage, autoimmune disorders, and cancer. Of particular interest is the potential of these new biomarkers to measure adverse health effects that may arise from exposure to environmental pollutants.
On August 12–13, 2020, the Standing Committee on the Use of Emerging Science for Environmental Health Decisions1 of the National Academies of Sciences, Engineering, and Medicine held a 2-day workshop to explore how new biomarker approaches can be applied to understanding the consequences of environmental exposures and improve environmental health decisions.2 The workshop brought together a multidisciplinary group, including experts in public health, environmental health, clinical medicine, and health disparities to discuss the state of the art in biomarkers and health. The workshop was sponsored by the National Institute of Environmental Health Sciences (NIEHS).
This Proceedings of a Workshop—in Brief summarizes the workshop presentations, both those that were recorded ahead of time and that were given during the workshop as well as the discussions that took place among the participants.
PRERECORDED PRESENTATIONS: NEW HORIZONS FOR BIOMARKERS OF EFFECT
Eight prerecorded presentations focused on selected types of biomarkers of effect that have potential or emerging use in environmental health research and decision making. The first of the eight presentations offered a broad overview on biomarkers of effect and important considerations for their development and use, while the other seven presentations each focused on new scientific research on biomarkers of effect on the following: the immune system, the nervous system, epigenetics, and genotoxicity and DNA damage.
Characteristics of a Robust Biomarker of Effect: Lessons Learned from the Development of the Kidney Injury Molecule 1 (KIM-1) Biomarker for Kidney Injury
A biomarker of effect is “any measurable biochemical, physiologic, or other alteration within an organism that, depending on magnitude, can be recognized as an established or potential health impairment or disease,” began John-Michael Sauer of the Critical Path Institute. Sauer emphasized that terminology must be clearly explained to ensure that everyone is aligned on what is being discussed. In pharmaceutical development, Sauer’s area of expertise, researchers typically use the term “safety biomarker” to refer to biomarkers measured before or after exposure to a medical product or an environmental agent to indicate the likelihood, presence, or extent of toxicity associated with an adverse effect. “Disease biomarkers” refers to another subcategory of biomarkers of effect that can be more diagnostic, monitoring, or prognostic in nature, he explained.
2Due to the COVID-19 pandemic, the workshop consisted of eight prerecorded presentations and a smaller number of live “virtual” presentations and discussion sessions.
Two other important terms are validation and qualification, which some people use interchangeably, but Sauer drew a distinction between them. Validation is “a process to establish that the performance of a test, tool, or instrument is acceptable for its intended purpose,” he said. Analytical validation is used when talking about the test, while clinical validation is used when talking about the tool. Qualification is a regulatory process utilized by the U.S. Food and Drug Administration (FDA) and other federal agencies “to demonstrate that biomarkers can be used” for a specified purpose. However, not all biomarkers are used for regulatory decision making. Hence, context of use, “the description of what a biomarker is to be used for and how it is to be applied,” is important, Sauer said. Although the qualification process does not specify the exact research needed, it does require the appropriate analytical and clinical validation for the given context of use. The clinical validation stage is key, he said, because it is how scientists prove “that the biomarker actually is performing as you think it should.”
Sauer next discussed the scientific considerations for establishing a valid biomarker. Using KIM-1 as an example, Sauer stated that his goal for the presentation was “to define the attributes of a scientifically valid biomarker to detect disease- and chemically-induced kidney injury in animals and humans.” KIM-1 is a safety biomarker developed to aid in the detection of kidney tubular injury in safety assessment species and healthy human volunteers. Sauer described how KIM-1 was developed, breaking down the development into three main stages: the discovery of the biomarker, the development and validation of the assay, and clinical validation.
KIM-1 is a transmembrane protein whose levels in the proximal tubule increase in response to an injury to the kidney; indeed, the levels of messenger RNA (mRNA) transcribed from the KIM-1 gene increase after an injury more than those of any other known gene. It was originally discovered in the late 1990s with a genomic approach that was looking for genes whose expression increased after kidney injury. A biomarker based on KIM-1 needed to give reliable and reproducible measures. Sauer offered a list of required criteria for an acceptable biomarker assay for clinical validation, including that it has acceptable accuracy, covers an appropriate measurement range, has acceptable selectivity and specificity, and that the sample be stable. The development and validation of a biomarker assay is an iterative process, he said, and there is a particularly intimate relationship between analytical and clinical validation.
The nonclinical and clinical validation process is a complex, multi-step process designed to ensure in the nonclinical part that the biomarker is responsive to injury to the target organ and not to off-target-organ injury and that the biomarker response is not related to just a single mechanism of tissue injury.3 In the clinical part of the process, performance of the novel biomarker should be superior to that of the current standard biomarker in detecting tissue injury. Sauer described the main aspects of clinical validation as including (1) correlation with the response one is trying to predict, (2) selectivity/specificity, or measuring what we think we are measuring, and (3) sensitivity, or whether we can detect what we are trying to detect. Sauer presented a summary of the attributes of a useful biomarker, which are shown in Box 1.
BOX 1 Attributes of a Useful Biomarker of Effect
1. The biomarker demonstrates a correlated response with the truth (hard endpoint).
2. The biomarker is selective or specific.
3. The biomarker is sensitive.
4. The biology and performance of the biomarker align with its use:
- Acceptable magnitude of changes in response to the insult.
- Acceptable intra- and inter-subject variability associated with the biomarker’s baseline and response.
5. The availability of reliable and reproducible measurement methodology.
6. Analytical and clinical validation have shown that the biomarker is fit-for-purpose for its proposed use.
Sauer said that his group had accumulated “convincing evidence” that the KIM-1 biomarker was sufficiently sensitive and specific to predict pathological changes in the kidney in test animals. In 2018 the FDA agreed, qualifying it for use by drug developers in submissions of investigational new drug applications, new drug applications, and biologics license applications.
Biomarkers of Immunotoxicity and Applicability to Per- and Polyfluoroalkyl Substances (PFAS) and Other Environmental Toxicants
Jamie DeWitt of East Carolina University spoke about biomarkers of immunotoxicity, with a specific focus on how such biomarkers can be used to identify the effects of various environmental toxicants, specifically PFAS.
DeWitt began by offering some background on the immune system, which protects a body from viruses, bacteria, and other pathogens but also plays roles in the maintenance of homeostasis in various body parts and in organ development,
protection against tumors, and the body’s response to various injuries. Exogenous agents, including pathogens and environmental toxicants, can trigger immune system actions, and immunotoxicity is defined as the ability of a substance to cause the immune system to behave inappropriately. This can include suppression of immune responses, which can weaken the body’s ability to fight infections or other diseases, or a heightening of immune response to the point that it is overly reactive. Another possible result of immunotoxicity is inflammation that goes on for too long, which can exacerbate an injury.
One important tool for evaluating the effects of environmental chemicals on the immune system is the host resistance test, in which an organism such as a laboratory test animal is challenged with an infectious agent and its immune response monitored. There are many diverse challenge agents, and they offer information about different types of immune system cells. However, these tests have various limitations—they are often costly and cumbersome to run, for instance—that do not make them ideal as biomarkers for the effects of environmental chemicals.
Recently a number of more effective biomarkers for immunotoxicity have been identified, based in large part on work by Mike Luster at the National Institute Occupational Safety and Health. These rely on looking at the effects of various toxicants on specific immune functions in test animals, such as an assay of plaque forming cells that looks at the amount of antibody produced in response to an antigen. “It’s really analogous to a vaccine response,” DeWitt explained. By testing a number of these assays, Luster found that some were much better at predicting immunotoxicity than others and that by combining two assays it was possible to greatly improve the ability to predict immunotoxicity.
“These individual assays are actually functional assays,” DeWitt said. “They are not some sort of a background measure of baseline immune function. They are tests of the ability of the immune system to respond to a particular challenge.”
As an example of how these biomarkers can be used, she offered the example of PFAS, a diverse group of chemicals used in a variety of applications and that tend to persist in the environment for many years. DeWitt focused on two in particular, perflurooctanoic acid (PFOA) and perflurooctanesulfonic acid (PFOS). Tests of biomarkers in rodents found that both PFOA and PFOS were immunotoxicants, and the results were compelling enough that in 2016, the National Toxicology Program concluded that they are presumed to be immune hazards in humans. Although there was other evidence, DeWitt said, it was mainly the data from the biomarker tests that led to those conclusions.
DeWitt ended with a statement that in testing chemicals for immunotoxicity, it is important to look specifically at functional outcomes, and that is precisely what these particular biomarkers do.
Biomarkers of Autoimmunity: Collaborations with Metals Mixture-Exposed Communities
Autoimmune diseases, a complex and diverse set of disorders in which the body’s immune system mistakenly attacks itself, have an equally complex set of causes, with genetic factors, diseases, and environmental exposures all playing roles in at least some autoimmune disorders. Esther Erdei of The University of New Mexico spoke about the search for biomarkers of autoimmunity, with a particular focus on autoimmunity related to exposures to various mining-related metals, such as uranium, arsenic, and mercury.
The issue is of particular concern for Native Americans living in the western United States, she said, where there are several hundred thousand abandoned sites that were formerly used for mining of metals or other hard minerals. More than 600,000 Native Americans live within 10 kilometers of at least one such abandoned mine, and the U.S. Environmental Protection Agency (EPA) has found that 40 percent of watershed headwaters in the western United States are contaminated by these mines. These facts combined with the reliance of many Native Americans on game animals, fish, and crops and livestock raised on these lands means that Native Americans in the West are much more likely than Americans in the rest of the country to be exposed to these metals and to perhaps develop autoimmune disorders related to those exposures.
Erdei said that research has shown that Native Americans, such as the Navajos who live in proximity to these mines and their waste, have increased risk for autoimmune disorders, hypertension, and various chronic diseases. Autoimmune diseases have particularly been linked to uranium in drinking water. To test for autoimmune disorders, Erdei’s group collected blood samples from Navajos exposed to uranium and ran two types of tests: antinuclear antibody (ANA) testing, which detects antinuclear antibodies that can attack the body’s own tissues; and testing with a special panel of autoantigens that could pinpoint the possible presence of specific autoimmune disorders.
A second study, the Navajo Birth Cohort study, enrolled 781 mothers, 764 babies, and 227 fathers, all of whom were born after the uranium mines were closed so that they had experienced no occupational exposure. Of a group of healthy males with no chronic diseases, 8 percent tested positive for ANA. The researchers also found that 28 percent of the adults had urine uranium levels greater than the 95th percentile found in adults in general in the United States. The levels of arsenic found in the Navajo adults were similar to those in the general population, but the form of arsenic was different—it was not the form found in seafood but rather a more toxic inorganic form. Levels of manganese, tin, and antimony, among other metals, were also higher in the Navajos.
In other Native American communities, a high prevalence of autoantibodies was linked with fish consumption and proximity to an arsenic-contaminated river, said Erdei. She added that more generally, there have been significant increases in ANA prevalence in younger adults across the United States over the past 25 years. These findings are significant, she said, because toxicity to the immune system can lead to various adverse health outcomes linked either to heightened immune activity of the sort that underlies autoimmune diseases and chronic inflammation or to immune suppression, which can leave a person vulnerable to disease and cancer.
Using a nationwide set of electronic medical records, Erdei and colleagues carried out a statistical analysis that found that Native Americans and other minorities have increased rates of various autoimmune diseases, including rheumatoid arthritis, lupus, and Crohn’s disease. In the future, she said, she hopes to expand on these analyses and look for other biomarkers of autoimmune diseases to gain a better understanding of their distribution around the country.4
Developing Fluidic Biomarkers of Central Nervous System Damage
In many cases of damage to the central nervous system (CNS), observing the damage and how far it has progressed requires the use of invasive procedures, said Syed Imam of the FDA. Thus, a collaboration was established to look for non-invasive fluid biomarkers of CNS damage. The collaboration involved the Health and Environmental Sciences Institute (HESI) as well as scientists from governmental bodies and regulatory agencies, academia, and the pharmaceutical industry.
To find biomarkers of neurotoxicity, the researchers used the neurotoxicant trimethyltin (TMT) to create damage to the brains of rats and examined levels of various substances in body fluids, such as blood plasma and serum, urine, and cerebrospinal fluid, and tissues, such as brain, liver, thymus, kidney, and spinal cord. They first verified that the TMT actually had caused brain damage in the rats by examining magnetic resonance imaging (MRI) scans of their brains as well as the brain tissue directly using stains that pinpointed the presence of inflammation and degenerating neurons. They next examined levels of glial fibrillary acidic protein (GFAP), which the CNS is known to produce in response to injury. They found that there were indeed increasing levels of GFAP in the brain after injury with TMT, but importantly, they also observed increasing levels of GFAP in blood serum, which implied that it was possible to detect brain injury non-invasively by observing levels of GFAP in serum.
Around the time that Imam’s group was finishing this work, two other studies were published about the clinical use of GFAP levels to predict traumatic brain injury in human patients.5 One study showed that doctors were able to determine what sort of treatment a patient with possible brain injury would need by examining serum GFAP, while the other showed that GFAP could be used to track the development of a brain injury over time. It was very encouraging, Imam said, that just as his group was demonstrating the viability of GFAP in a pre-clinical model, other work was showing its effectiveness in the clinic.
His group has since identified a number of other potential fluidic biomarkers for CNS damage. Levels of microRNAs (miRNAs) were found in the cerebrospinal fluid or serum that were either elevated or depressed after the induction of brain damage in the rats. However, one of the most promising candidates was tumor necrosis factor β2, which was detectable in the rats’ urine after they were given TMT. A biomarker for neurotoxicity or neurological damage that could be tested for in urine samples would be an important advance, Imam said.
Summing up, he said that a large number of potential fluidic biomarkers for CNS damage have now been identified. These have been found in serum, plasma, cerebrospinal fluid, and urine. The next steps will include qualifying these fluidic biomarkers of neurotoxicity, a second-phase study to validate a few promising candidates from the current study, and modeling the disease pathways for these biomarkers. The ultimate goal, he said, will be to qualify fluidic biomarkers of CNS toxicity for clinical trials.
Translocator Protein (TSPO): A Biomarker of Brain and Peripheral Inflammation
TSPO is a protein that is found in cells throughout the body. In the nervous system it is localized in glial cells, cells that help support the function of neurons, and specifically it is found mainly in microglia and astrocytes, two types of cells involved in the body’s response to injury and inflammation. As Tomás Guilarte of Florida International University explained in his presentation, because the expression of TSPO is increased in response to neuroinflammation and brain injury, the protein is used as a biomarker for damage to the brain and nervous system.
4 It should be noted that, as discussed later in the workshop, many people who do not have actual autoimmune diseases also have ANAs, so these tests are not a completely reliable biomarker.
5 Bazarian, J.J., et al. 2018. Serum GFAP and UCH-L1 for Prediction of Absence of Intracranial Injuries on Head CT (ALERT-TBI): A Multicentre Observational Study. Lancet Neurology 17(9):782–789; Yue, J.K., et al. 2019. Association Between Plasma GFAP Concentrations and MRI Abnormalities in Patients with CT-Negative Traumatic Brain Injury in the TRACK-TBI Cohort: A Prospective Multicentre Study. Lancet Neurology 19(10):953–961.
What makes TSPO particularly useful is that it can be bound to high-affinity ligands that can be labeled with radioisotopes, and this makes it possible to use such imaging techniques as positron emission tomography (PET) and single-photon-emission computed tomography (SPECT) to determine the distribution of TSPO and thus the location of brain injury and neuroinflammation. Because of this capacity, it has been used to map out brain damage and inflammation in a number of neurodegenerative and neurological diseases in humans.
To illustrate the power of TSPO as a biomarker of effect, Guilarte described a series of experiments in which TSPO was used to track brain injury in mice. His research group fed mice the chemical cuprizone, a copper chelating agent that targets mature oligodendrocytes and causes demyelination of white and gray matter. By feeding the mice cuprizone for varying times up to 4 weeks and then using radiography to examine slices of the mouse brains for TSPO, the group found that it could track the progress of the demyelination over time in great detail, seeing how much each area of the brain was affected, and could also see the slow remyelination of the oligodendrocytes after the cuprizone was removed from the animals’ diets. An important part of the study’s findings was that the TSPO was able to detect brain injury and neuroinflammation before any clinical or behavioral expression of the injury, which indicates that it is an early biomarker of effect.
In humans, TSPO has been used, among other things, to study brain injury in professional football players. Guilarte described one study his group did that found increased TSPO levels in retired football players who had widespread neuroinflammation in their brains and significant neuropsychological impairment and brain volume changes. A study in active and recently retired National Football League players found elevated TSPO levels in the absence of significant neuropsychological impairment and brain volume changes; this reflects the fact that neuroinflammation is an early event that occurs prior to neuropsychological deficits and brain volume loss.
In sum, Guilarte said, TSPO is a validated and widely used biomarker of brain injury and neuroinflammation that can be used to screen the neurotoxic potential of industrial chemicals and to assess neurodegenerative conditions in animal models and humans. In particular, it can be used to monitor the progression of brain disease and to assess the effectiveness of therapeutic strategies. But one disadvantage is that in vivo TSPO imaging by PET or SPECT is expensive and requires specialized facilities. To address this, Guilarte and coworkers are testing to see if TSPO levels in immune peripheral cells—which are much easier to assess—can serve as a surrogate biomarker for brain inflammation. That work is still ongoing, but if it proves possible, he said, the surrogate biomarkers will cost much less and have much wider applicability than the TSPO levels in the brain itself.
Identifying Epigenetic Biomarkers of Arsenic-Induced Fetal Birth Weight
Inorganic arsenic is known as the king of poisons, said Rebecca Fry, The University of North Carolina at Chapel Hill. It poses a major public health threat through groundwater contamination in a number of places around the globe, including Argentina, Bangladesh, Chile, China, India, and Mexico. In the United States, arsenic-contaminated water in private wells is also a problem in a number of states.
Exposure to arsenic has been associated with neurological disorders, cardiovascular disease, diabetes, and reproductive effects, Fry said. It is also a potent carcinogen, leading to cancers of the skin, bladder, lung, liver, prostate, and kidney, among others. Studies in mice have shown that in utero exposures to arsenic can produce various cancers later in life, including hepatocellular carcinomas and tumors in the adrenal glands, liver, lung, ovaries, and uterus. Furthermore, in some of these mouse tumors there was evidence of epigenetic changes through CpG methylation, meaning that regions of the DNA where a cytosine nucleotide is followed by a guanine nucleotide in the linear sequence of bases can be methylated to form 5-methylcytosines. This epigenetic change can modify expression of the affected gene. Other research has shown arsenic to be a developmental toxicant, with decreased birth weight observed in children born to mothers exposed to environmental arsenic, Fry’s major topic of discussion.
Arsenic-induced toxicity acts through a variety of mechanisms. But much remains to be learned about which of these mechanisms is relevant to the developmental toxicity of arsenic and the biological chain of events that links arsenic to developmental toxicity.
One approach to understanding these mechanisms is called the adverse outcome pathway (AOP), which is a conceptual construct that a researcher uses to trace the chain of causally linked events between a molecular initiating event—such as exposure to arsenic—and an adverse outcome. AOPs are useful in supporting chemical risk assessments, Fry said, because they provide mechanistic reasoning underlying the associations between exposures and outcomes. In her research program, Fry has been working with Steve Edwards at the EPA to develop an AOP linking arsenic exposure in utero with low birth weight.
In her study of 200 pregnant women in Mexico who had been exposed to arsenic in their drinking water, her group collected measurements of arsenic levels in drinking water and urine; birth weight and other birth outcomes in the women’s children; and DNA, RNA, and protein samples. Fry’s group examined arsenic-associated DNA methylation at CpG sites in the women’s children and found nearly 3,000 genes with altered methylation. Sixteen of these DNA methylated genes exhibited
a functional change in mRNA expression, 7 of which were associated with low birth weight in the infants. One of them was KCNQ1, a potassium channel–related gene, which has been associated with intrauterine growth depression as well as later-life metabolic health. Fry said “KCNQ1 is a really excellent example of a gene linked to both early fetal growth and later-life health outcomes that could provide a mechanistic basis for how arsenic is associated with its early and later-life health effects.”
Furthermore, by comparing the level of CpG methylation with the measured arsenic levels, Fry’s group created a dose–response curve that identified the arsenic dose at which the epigenetic modifications appeared. CpG methylation of the KCQN1 gene was associated with a benchmark arsenic dose of 86 parts per billion in drinking water—a level that many individuals in both Mexico and the United States are exposed to. The development of the dose–response curve is a step toward integrating this epigenetic biomarker into a risk assessment framework.
Development, Validation, and Application of an In Vitro Transcriptomic Biomarker for Genotoxicity Testing
A key issue in both environmental science and pharmaceuticals is whether a chemical is genotoxic, that is, it damages DNA, leading to mutations or cancer. Carole Yauk of the University of Ottawa described the development of an in vitro transcriptomic biomarker for genotoxicity that has been validated and applied and is now being reviewed by regulatory agencies in both the United States and Canada.
There are a variety of tests for DNA damage and mutation, such as the Ames assay, a bacterial reverse mutation test, but while these are sensitive, they lack specificity, leading to high rates of positive results that are not of interest—what Yauk called “irrelevant positives.” Furthermore, the tests are not particularly efficient and do not provide information on the mechanism causing the genotoxicity. Thus, she said, there is a need for high-throughput genotoxicity tests that provide mechanistic information.
To provide such a test, Yauk and her colleagues developed the TGx-DDI biomarker, an assay that can predict whether a chemical is DNA damage inducing (DDI). She and her team observed the response of human TK6 cells, which are human lymphoblastoid cells, in culture to 28 different DDI and non-DDI chemicals and identified 64 genes that were predictive of DNA damage. To perform a TGx-DDI assay, they expose TK6 cells to the target chemical and, with a transcriptomic analysis, observe how each of the 64 genes respond. To be conservative they take the data from the 64 genes and carry out three different analyses—a probability analysis, a principal component analysis, and a two-dimensional clustering analysis. If any of the three analyses classify the chemical as DDI, it is assumed to be DDI; only if all three analyses conclude that it is non-DDI is it classified as non-DDI.
TGx-DDI could be used either for determining the potential genotoxicity of a drug candidate or for assessing environmental chemicals, Yauk said, although her interests lie mainly with the latter.
Yauk next described how TGx-DDI was validated. It proved effective in identifying genotoxic agents and, importantly, screening out the irrelevant positives from the standard screening tests. It is sensitive to the level of the chemical being tested and identifies when the concentration is great enough to lead to genotoxicity. Her team also showed that TGx-DDI can be used on multiple platforms, including an Agilent microarray, an Affymetrix microarray, qPCR, and RNA-seq, and her team developed a software tool to make it easier to use.
The accuracy of the test is even greater when it is combined with the CometChip® assay, which is complementary to TGx-DDI. Together, the two provide high-throughput genotoxicity screening.
In terms of regulatory approval, Yauk said that TGx-DDI’s qualification plan is under review by the FDA, and her team has also submitted regulatory applications with Health Canada’s GeneTox21 program. Additionally, she noted, her team is not alone: several other groups have developed genotoxicity biomarkers, both in vitro and in vivo, and others look promising as well.
In conclusion, she said that transcriptomic biomarkers provide a rapid, efficient, and reproducible way to identify hazards or modes of action, and they are highly reproducible across technologies. The biomarkers have not yet been accepted or adopted for regulatory toxicity testing, and Yauk suggested that the reasoning for this includes the necessary expertise required for using them, concerns about the use of large data repositories, and worries about black box software tools. Thus, she said, it is unclear how to proceed in order to obtain regulatory acceptance of transcriptomic biomarkers. Further research is needed to determine how best to model dose–response relationships and to determine the thresholds of toxicity using these biomarkers, and it will be important to characterize the uncertainties in this approach.
New Biomarker for Genotoxicity Mode-of-Action Determination
Marc Audebert of the TOXALIM Research Centre in Food Toxicology in France discussed the γH2AX/pH3 biomarker for genotoxicity. He is particularly interested in the genotoxicity of food contaminants, but this biomarker could be used in many contexts.
As background, he noted that carcinogenicity assays have a variety of disadvantages. They are done in mice and rats, their cost is typically in the millions of dollars, and they may take a couple of years to complete. Genotoxicity assays, by contrast, are done in cell lines, cost thousands of dollars, and take only a couple of weeks to complete. Furthermore, the predictive capability of the genotoxicity assays is more consistent and reliable than that of the carcinogenicity assays. Thus, a number of new genotoxicity assays have been developed or are under development.
However, genotoxicity assays have a number of limitations. Because of inter- and intra-species differences, results do not always transfer well from lab animals to humans. The most useful assays are those that can be done with high throughput from small samples, and this is not always the case with genotoxicity assays. It is also important to be able to determine the genotoxic mode of action, such as whether the substance is an aneugen (i.e., causes daughter cells to have abnormal numbers of chromosomes) or a clastogen (causes chromosomes to break or otherwise be disrupted).
When a cell’s DNA is damaged, specifically in the case of a double-strand break caused by a clastogen, one of the cell’s responses is to phosphorylate the histone protein H2AX to create γH2AX. Histones are proteins found in eukaryotic cell nuclei that play a role in gene regulation. They are the main protein components of chromatin and act as spools around which the DNA winds. The γH2AX appears within minutes of damage to the DNA and can be detected with immunohistochemistry by using a targeted antibody. Audebert said that his lab has been using γH2AX in its genotoxicity studies for more than 10 years.
H3, another histone, is one of the major histones involved in the structure of chromatin. Aneugens affect the normal phosphorylation of H3 into pH3, either increasing it or decreasing it depending on the particular aneugen. The sensitivity of H2AX to clastogens versus the sensitivity of H3 to aneugens suggest a possible biomarker of genotoxicity to Audebert: by examining the levels of γH2AX and pH3, it should be possible to detect the presence of a genotoxin and determine if it is an aneugen or a clastogen.
To analyze the effectiveness of the γH2AX/pH3 biomarker, Audebert carried out a meta-analysis of 27 publications examining the change in levels of γH2AX and pH3 in response to a wide variety of chemicals—aneugens, clastogens, and non-genotoxics—in a number of different cell lines using a variety of techniques. As a predictor of genotoxicity, the biomarker had a sensitivity (i.e., avoidance of false positives) of 95 percent and a specificity (avoidance of false negatives) of 88 percent for an overall predictivity of 91 percent. Its performance was highly correlated with the standard micronucleus (MN) assay for genotoxicity.
Comparing the γH2AX/pH3 biomarker with the MN and other genotoxicity assays on a variety of characteristics, Audebert said that the new biomarker was generally equal to or better than the other assays on such things as what sorts of cells it can be run in, its predictivity, its reproducibility, and its throughput, and where it clearly surpassed the others was in its ability to discriminate among different modes of action—specifically in how it can tell the difference among aneugens, clastogens, and cytotoxic compounds.
In conclusion, he said that the γH2AX/pH3 biomarker is the most predictive in vitro genotoxicity test yet developed, it can be performed in every cell system, and it is the only assay that can easily and efficiently discriminate aneugen from clastogen.
LIVE WORKSHOP: AUGUST 12–13, 2020
The first day of the workshop included opening remarks from Rick Woychik the Director of NIEHS and two sessions—the first on the role that emerging technologies are playing in the discovery and translation of biomarkers and the second about data reliability and other considerations in using biomarkers in decision making. A third session on population variability and diversity was held on the second day of the workshop. The workshop concluded with an exploration of whether a specific emerging biomarker of effect for DNA damage is ready for use in environmental health decision making.
“NIEHS has a longstanding interest in the development of biomarkers of response to environmental stressors,” said Woychik. This interest is reflected in their extramural portfolio of research grants as well as in intramural research programs, two of which focus on biomarkers. Woychik noted that NIEHS-supported work began in 2006 when the trans-National Institutes of Health (NIH) Genes, Environmental, and Health Initiative funded a number of projects to develop protein, metabolite, and epigenetic markers reflecting responses to inhaled toxicants, pesticides, and endocrine-disrupting chemicals. Since then, NIEHS has funded numerous other projects aimed at developing additional biomarkers of environmental exposures, susceptibility, and adverse health effects such as biomarkers of mitochondrial dysfunction in animal models of Huntington’s disease and Alzheimer’s disease and DNA methylation markers that are altered in individuals with prenatal exposure to arsenic. Woychik also highlighted recent collaboration with the National Institute on Aging to determine whether telomeres can be used as “sentinels of not only environmental exposure but also of psychosocial stress and disease susceptibility.” Through their Superfund research program, NIEHS is supporting work to explore a variety of biological response indicators including DNA methylation markers that are altered in individuals with prenatal exposure to arsenic, biomarkers of altered glucose metabolism as a result of polychlorinated biphenyl exposure, and the changes in metabolic hormones among birth
cohorts after PFAS exposure. Woychik emphasized that new approaches “hold the promise to detect variation between individuals” that could be “enormously helpful” for future precision health efforts. However, Woychik acknowledged that much more still needs to done in biomarker discovery and validation. “Each of us responds to the environment in different ways and we have to become more sophisticated in how we develop and validate biomarkers that factor individual genetic, epigenetic, metabolic, and other biological variability into the equation,” stated Woychik.
This workshop, he predicted, should be an important part of the ongoing effort to use biomarkers to gain a better understanding of how the environment affects the health of individuals and the public as a whole.
The Importance and Promise of Biomarkers of Effect in Environmental Health
Nicole Deziel of Yale University and Patrick McMullen of ScitoVation, two members of the workshop organizing committee, gave an opening presentation to provide some big picture framing for the workshop and share thoughts and perspectives to stimulate ideas and discussion. Deziel began by acknowledging the tremendous technological advancements in evaluating and developing biomarkers of exposure that have been made with new tools and technologies. These include targeted and untargeted analysis, the exposome, and high-level screening and tracking of environmental exposures such as that performed by the National Health and Nutrition Examination Survey (NHANES). NHANES routinely measures exposures to more than 300 environmental chemicals in a nationally represented population. However, she added, linking biomarkers of exposure to health outcomes is still quite a challenge, and this is where biomarkers of effect have some important roles to play.
She highlighted two such roles. One is that biomarkers of effect can serve as surrogate endpoints for the health outcome of interest. This is interesting to her as an epidemiologist, she said, because it would speed research up considerably if she could reliably use a biomarker of effect as an indicator of an eventual health outcome rather than having to wait for years for the health outcome to appear. Such biomarkers also offer the possibility of intervening with, for example, exposure mitigation strategies in time to prevent undesirable health outcomes.
A second role of biomarkers of effect is that they help researchers trace the mechanisms underlying health effects caused by environmental exposures by highlighting some of the steps along the AOP. To illustrate the AOP, Deziel showed a figure representing the various steps from exposure to a toxicant to the resulting health outcomes in individuals and populations (see Figure 1). At each step along this path there are biomarkers of both exposure and effect that can offer insight into what is happening at that point.
Finally, Deziel suggested that workshop attendees might think about possible applications of biomarkers of effect in clinical medicine and public health. For example, DNA testing is now used to pinpoint genetic susceptibilities and allows suggestions to be made to individuals about actions they might take to avoid adverse health outcomes. She suggested that panels of biomarkers might be used in a similar way, pointing individuals or groups of people toward behaviors that could help them avoid negative health consequences.
McMullen next discussed the context in which biomarkers are used. He offered a list of categories into which biomarkers might be classified: diagnostic, monitoring, predictive, prognostic, pharmacodynamic/response, safety, and susceptibility/risk. This, he said, can be a useful framework when thinking about the various ways that biomarkers can be used. Referring to the list of attributes of a useful biomarker of effect presented by John-Michael Sauer (see Box 1), McMullen emphasized that some of attributes will be more important than others, and one should take that into account when choosing biomarkers.
The eight prerecorded talks identified a number of factors that can influence biomarkers of effect, he said, and these should be kept in mind when considering how to use and interpret them. Some biological effects are important because of inter-subject variability, such as diet, age, sex, body mass index, and gene-environment effects. Differences in sample preparation and experimental procedures can also introduce variability in results. Timing can also be crucial, with some biomarkers responding differently at different time points following exposure.
To close, McMullen offered a series of questions that he suggested the workshop participants should keep in mind as they discussed biomarkers of effect during the next 2 days:
1. How can biomonitoring of effects better include innovative tools and methods?
2. How can we better connect biomarkers of effect with measures of human exposures?
3. How do we incorporate rapidly developing technologies to increase the availability of biomonitoring information?
4. What can we do better to resolve challenges including reproducibility, translation, and the interpretation of risks?
5. How do we evaluate the utility of biomarkers across diverse populations?
Biomarkers of Effect Use in Environmental Health Research and Decision Making: MicroRNAs as an Example
Brian Chorley of the EPA discussed the uses of biomarkers in regulatory decision making, and touched briefly on miRNAs as a promising class of biomarkers of effect. He described three areas of “regulatory context of use”—biomonitoring to measure exposure; hazard prioritization, which assesses hazard identification and dose response; and hazard identification/weight-of-evidence, which combines multiple sources of information on exposure, hazard identification, and dose response to support regulatory decision making. This last area, he said, is where a suite of biomarkers is needed based on multiple models and a holistic approach.
The EPA has become interested in biomarkers of effect for several reasons, he said. They promise higher throughput, as well as lower cost human-relevant toxicity testing than the testing in use today. The hope is that they will enable predictive toxicology in risk assessment and identification of vulnerable populations. They should also make possible system-level integration in environmental toxicology, combining toxicity testing with other emerging big-data technologies.
Chorley listed six likely characteristics of an ideal biomarker. He said biomarkers should be specific, sensitive, predictive, robust, translatable, and non-invasive. He used an innovative radar plot visual to illustrate the use of biomarkers in decision making (see Figure 2). Each corner of the hexagon corresponds to an ideal attribute of a biomarker, and biomarkers for different purposes will ideally have different characteristics. For example, a biomarker used in biomonitoring should score high on the sensitive and non-invasive dimensions, while it would not be so important that it be translatable. Chorley continued that in the hazard prioritization context, a screening framework is important, so a biomarker should have high sensitivity to minimize false negatives and be robust. In a weight-of-evidence context, biomarkers should be translatable so that they indicate human adverse outcomes, both sensitive and specific to avoid both false positives and negatives, and robust to provide a clear assessment that can be used to define adverse dosages.
As newer technologies and new knowledge have appeared, a variety of novel biomarkers have come under discussion, Chorley said, and he pointed to miRNAs as a particularly promising example. These are markers of gene expression, not just of one gene but of multiple genes. They can be tissue-specific, and changes in miRNAs can be early and robust. They can be found safely in various biofluids, and there are many available methods to analyze them. They are promising enough that the HESI is focusing on developing a miRNA biomarker panel for kidney damage. It will likely be a long and difficult road to get such a panel qualified, he said, but there is enough promise that it is worth the effort.
Panel Discussion 1: The Role of Emerging Methodologies and Technologies in Biomarker Discovery and Translation
The first panel discussion was moderated by Norbert Kaminski of Michigan State University and a member of the workshop organizing committee. The panelists were Marc Audebert of the TOXALIM Research Centre in Food Toxicology, Rebecca Fry of The University of North Carolina at Chapel Hill, Tomás Guilarte from Florida International University, Syed Imam of the FDA, and Syril Pettit of the HESI.
Kaminski asked the panel members if they could point to new or emerging technologies that might accelerate the validation process and more quickly move putative biomarkers to acceptance and implementation. Fry suggested thinking about genome-wide technologies such as Illumina EPIC arrays, which can be used for CpG methylation assessment. In the case of transcriptomic biomarkers, the cost for RNA sequencing has come down, and quantitative real-time polymerase chain reaction could also be used to look at targeted gene expression. Pettit said that multiplex markers can be valuable, particularly to avoid use of a large number of individual markers that require a further evaluative process. Imam added that in looking at biomarkers for neurotoxicity or neurological disorders he has found MRI T2 relaxation studies to be highly useful; they represent an established technology that is noninvasive but offers a clear view of damage to the CNS. Guilarte agreed that neuroimaging technologies can be effective but warned that they are also expensive.
In response to an audience question about whether the AOP framework provides a hypothesis-driven approach to identifying and developing new biomarkers, Fry said that it does. “We can understand how various chemicals can have their effects on molecular pathways, cellular pathways, tissue, organ, and then population level,” she said.
Another question was related to the use of data sharing and collaboration to help drive the acceptance of biomarkers. Pettit said that such collaboration is vitally important in building confidence in biomarkers. “There has to be a community confidence in the utility of the marker to inform decision making,” she said, and that requires the developers of the biomarkers to obtain input from the community of practice. Guilarte added that having people from different parts of the world getting the same results is also crucial to acceptance.
Kaminski then asked whether large datasets might help move things forward or whether the field is still not sufficiently robust to help generate useful biomarkers. Fry said that she thought big data can certainly be helpful, but that it will be important to put the data in the context of a phenotype and a health outcome as the endpoint. Pettit said that she believes that today’s analytical tools could identify biomarkers and many researchers have the expertise to do that, but there is a learning curve for interpreting the output of some complex high-density datasets. Audebert added that a major value of big data is that the various omics make it possible to investigate multiple aspects of the AOP in detail.
Next Kaminski read a question from Alison Harrill of NIEHS on how to use multiple lines of biomarker information to create a coordinated picture of a patient’s disease. Imam said that in his work on fluid biomarkers of damage to the CNS, one goal is to isolate critical biomarkers of specific types of neurological damage rather than a broad biomarker of neurological damage in general. He also hopes for biomarkers that provide information at different stages of damage so that some might serve in the early detection of damage or disease while others could help identify how far along the disease has progressed.
Kaminski commented that this raised another question. “Often in the clinical diagnostic setting we are using multiple biomarkers to assess disease,” he said, “and yet in toxicology we are dependent on a single biomarker.” Are there any disadvantages to using multiple biomarkers? Audebert answered that because various biomarkers offer different types of information, having more biomarkers could be a great advantage in chemical risk assessments. Pettit commented that this may be a situation where there could be a great benefit from toxicologists communicating with the clinical community.
To end the discussion, Kaminski asked the panel if there are situations in which limited tissue availability makes it difficult to identify biomarkers. Are there specific technologies that may enhance the development and application of biomarkers in these difficult areas?
In neurology, Guilarte said, the “holy grail” of CNS toxicity is to find blood-based biomarkers that reflect what is happening in the brain, but researchers do not yet know whether such biomarkers exist. Fry suggested that one useful approach would be to look to exosomes—extracellular vesicles that are released from cells and carry information in various forms to other cells—for circulating biomarkers that can provide information about the cells that released the exosomes. For example, miRNAs that are only expressed in the placenta that can be found through an exosome analysis. Guilarte said it would be a good idea to look for brain-specific exosomes.
Evaluating and Improving the Biomarker Pipeline
In the second presentation session, Joshua Wallach of Yale University spoke about evaluating and improving the biomarker pipeline. He explained that he is not a biomarker expert but rather carries out research on research—a “meta-researcher,” as he termed it—who studies how research is conducted and how it can be improved. He offered some ideas on how to improve the biomarker pipeline from this perspective.
Wallach acknowledged the importance of biomarkers because they help in understanding the nature and extent of human exposure and risk from environmental toxicants and are often the only way to measure exposures and outcomes. Much of his focus is on clinical research, he said, and noted that he would often be using the term “surrogate marker” to describe a biomarker that has been validated and is used in practice. In clinical settings they are often easier and less expensive to measure than long-term outcomes, he said, and offered blood pressure as an example of a surrogate marker that serves as a predictor for cardiovascular outcomes.
Despite biomarkers’ numerous advantages, they have a number of important limitations, he said, pointing to such things as questions about their value in clinical practice, difficulty in finding effective biomarkers for diseases such as cancer, and concerns about bias and biomarker validity. Most of these limitations are related to transparency and reproducibility, he added, and improving biomarker data will involve working on reliability, transparency, and reproducibility.
Expanding on the idea of reproducibility, he said, it can be broken down into three different types: method reproducibility, relating to using the same experimental procedures on a new dataset; result reproducibility, or the ability to obtain corroborating results in a new study; and inferential reproducibility, or new investigators drawing different conclusions from those of the original authors following a reanalysis or replication. Various factors in the biomarker pipeline affect reproducibility. To talk about these issues, Wallach illustrated the pipeline as a literal pipe with different segments labeled discovery, validation, translation, evaluation, and implementation, and spoke about failures in that pipeline.
Each of the segments has various weaknesses, said Wallach. The discovery segment has limitations such as poor study design, small sample sizes, weak analyses, selective reporting, and “spinning” findings to suggest significance when it may not actually exist. A major issue in validation is the lack of reproduction and replication efforts by the same investigators or those using the same population. Weaknesses in translation include inadequate prioritization and a lack of transparency. In the evaluation stage, limitations include a lack of evidence linking surrogate markers with long-term outcomes of interest, a paucity of randomized trials, and a lack of gold standards against which to compare the biomarkers. Finally, the implementation segment is weakened by, among other things, a lack of guidelines and uncertainties about how to remove a biomarker from use if it has been found not to work.
To strengthen the biomarker pipeline, Wallach offered a number of suggestions. In the discovery phase, there should be more collaboration, standardization, and use of guidelines, and he called for preregistration of biomarker studies, as is often done in clinical trials. In validation, collaboration is also important, and it is critical to have validation with independent datasets and investigators and not calling something a biomarker until it has been independently evaluated. For the translation stage, overarching reviews of the entire biomarker research agenda in a field will be helpful, and researchers should systematically identify the strongest evidence with the lowest risk of bias. To improve evaluation, Wallach called for more and better trials and looking at the relationship between surrogate markers and long-term outcomes. Wallach concluded that in the implementation stage there needs to be transparency about de-implementation, or removing something from practice when it does not work.
Panel Discussion 2: Data Reliability and Transparency Considerations in Decision-Making Processes
The panel discussion that followed Wallach’s presentation was moderated by William Mattes of the FDA and a member of the workshop organizing committee. The panel members were Sarah Blossom of the Arkansas Children’s Research Institute, Jamie DeWitt of East Carolina University, Christina Jones of the National Institute of Standards and Technology (NIST), John-Michael Sauer of the Critical Path Institute, Wallach, and Carole Yauk of the University of Ottawa.
Mattes asked how one should identify, evaluate, and balance the risks and benefits of the use of a biomarker for environmental health decisions, and what evidence is needed for such an evaluation. DeWitt answered that the reliability of biomarkers used in environmental health decisions is not as crucially important as the reliability of biomarkers used for evaluating the safety of a drug under development because the associated risks are not the same. Yauk agreed that the bar is different for environmental chemicals because the risks are lower. Sauer expanded on those answers by saying that one should always decide what question is being asked with a biomarker. Everything is situational, and a biomarker should be fit-for-purpose. This means that the level of validation required will vary depending on the risks associated with decisions that are informed by biomarkers.
Wallach added that because one does not always require perfect confidence in a biomarker to begin using it, it is important to continue evaluating them after they have been introduced. The standard of evidence will differ for different
biomarkers, but there should be continuing evaluation for all. Sauer said that this approach is being encouraged in the pharmaceutical world—collecting data on safety biomarkers after they have been implemented.
DeWitt asked Blossom whether ANA tests to determine autoimmune disorders are a reliable biomarker. Blossom said no, as ANAs are used as an autoimmune disease biomarker, but many people that do not have autoimmune diseases also have ANAs, so these tests are really a poor biomarker. DeWitt responded that what is important is understanding the limitations of a biomarker, particularly if one desires to use them for both public health and disease prediction.
Yauk said that one issue with biomarker development in the environmental sector is that it tends to take place in a bubble and that the process would be improved if a larger community was involved. Yauk said that it is hard to know how to use an omics biomarker properly when it has just been discovered in a paper. Wallach agreed and said that a major step in that direction would be to make the sharing of data and protocols the norm in the field. Jones said that NIST encourages labs to work together to see how others carry out their work. Wallach also noted that there has been a push to allow surrogate biomarkers in drug tests to shorten the trials. This is not necessarily a problem, he said, unless there is no long-term follow up with the clinically relevant outcome to validate the work with the surrogate biomarkers after a drug is approved and on the market.
An attendee asked to what extent biomarker data fail to replicate because discovery was done in a limited population and the biomarker behaves differently in another population. Wallach said that the issue is tricky but that it is important not to rely on evidence from subsets of a population in a study and that one should be thinking about generalizability in study design. Sauer agreed and said that if a biomarker is to be used in a different population, it needs to be retested to confirm that it performs as expected.
Several panel members discussed issues arising from the fact that different labs often use different platforms for the same biomarker, which raises questions about the biomarker’s reproducibility across labs. Yauk said that one of her lab’s biggest challenges has been evolving technology. Because different labs use different technologies, it should be demonstrated that a biomarker will work well on all of the different platforms. Jones said that NIST is starting to have different labs use the same materials whatever the platforms are so that NIST can see the degree of agreement.
On a question about whether non-technical factors influenced acceptance and implementation of biomarkers, Sauer said that a huge culture change and education are needed in the clinical space to understand how biomarkers perform and how to use them well and correctly.
After the panel discussion, Gary Ginsburg of the New York State Department of Health and a member of the workshop organizing committee offered a brief review of the day’s presentations as well as of the prerecorded presentations. He noted that a lot of fascinating science was discussed on markers ranging from those for immune disorders to epigenetic markers, circulating exosomes, histone and transcriptomic markers, neuroimaging markers, and more. These markers can have application for systemic effects as well as for individual organs or organ systems. He noted that there are many opportunities for increasing contextual use of biomarkers of effect for hazard identification, screening, dose response assessment, and getting at understanding where the disease process and the chemical toxicity process intersect.
Inclusive Science: Developing Population Relevant Biomarkers of Effect
Lesa Aylward of Summit Toxicology, LLP, and a member of the organizing committee opened the next session on population variability and diversity, emphasizing that “inclusive science,” which the next speaker, Chandra Jackson of NIEHS, would address, is a topic that many people are thinking about more. Jackson went on to address the development of population-relevant biomarkers of effect. To set the stage, she began her presentation with the definition of “disparity” used by the National Institute on Minority Health and Health Disparities, which is “a health difference that adversely affects defined disadvantaged populations based on one or more health outcomes.” Health disparities are generally considered to be preventable, she said, and are therefore caused by an environmental determinant rather than an innate or genetic determinant. The NIH designates the defined disadvantaged populations to include African Americans, Hispanics, Latinos, Native Americans, Alaska natives, Asian Americans, native Hawaiians and other Pacific islanders, socioeconomically disadvantaged populations, underserved rural populations, and sexual and gender minorities.
An example of a health disparity would be the approximately 60 percent higher prevalence of type II diabetes among African Americans than among whites in the United States. One major factor in this disparity is the higher prevalence of obesity among African Americans, particularly women, as compared with white Americans. This difference in obesity rates in turn is due, at least in part, to various lifestyle and environmental factors, such as the consumption of less nutritious and more highly processed foods, less physical activity, poorer sleep, and greater exposure to environmental chemicals such as endocrine disruptors.
Jackson asked why lifestyle factors appear to differ by race. “We know the health behaviors described are influenced by one’s physical and social environment, and disadvantaged groups have less access to health-promoting goods and services and often greater exposure to health damaging—and, in this case, obesogenic—environments,” she explained. It is important
to keep this context in mind, Jackson said, because the internal biological signals that are the subject of biomarkers of effect are often influenced by external environmental factors. In short, she added, researchers interested in the effects of chemical or non-chemical stressors on health should move beyond an oversimplified reductionist approach and take into account the more complex systems and complete picture that includes multiple levels of influences that play a role in health outcomes.
Jackson pointed to endocrine-disrupting chemicals as an example of chemical stressors with a role in health disparities because these have been shown to disproportionately affect racial and ethnic minorities from early life onward due to differences in environmental exposures, including non-chemical stressors. Non-chemical stressors include various psychosocial stresses that are often experienced more strongly by disadvantaged populations. Such psychosocial stresses have been associated with cardiometabolic dysfunction, among other effects. Over time, the cumulative effects of these chemical and non-chemical stressors can result in a “weathering” of an individual’s body, making it more susceptible and less resilient to various diseases, such as diabetes and cardiovascular diseases. Jackson suggested that biomarkers of “allostatic load”—essentially the wear and tear on one’s body from repeated or chronic stress—could serve as an indication of the cumulative lifetime effects of all types of stressors experienced by an individual.
Current research approaches have not provided enough information to design interventions to address the fundamental causes of health disparities, Jackson said, so it is necessary to develop new approaches to understanding the complex nature of the upstream causes of these disparities. The exposome—the measure of all of the environmental exposures an individual encounters over his or her lifetime—represents an unprecedented opportunity to comprehensively investigate the fundamental causes of health disparities by better characterizing a range of environmental and social exposure pathways, she said. Such an approach would involve both biomarkers of exposure and effect. There are groups in the United States and Europe that are starting to take this approach.
Panel Discussion 3: Addressing Population Variability and Diversity
Moderated by Alison Harrill of NIEHS and a member of the workshop organizing committee, the third panel discussion focused on issues related to health disparities and diversity as well as biomarker considerations in a population space. The panel members were Chandra Jackson of NIEHS; Andres Cardenas, an assistant professor of environmental health sciences and computational biology at the University of California, Berkeley; Brian Chorley of the EPA; Esther Erdei of The University of New Mexico; Rebecca Fry of The University of North Carolina at Chapel Hill; and Lesliam Quirós-Alcalá of the Johns Hopkins Bloomberg School of Public Health and the Department of Environmental Health and Engineering at Johns Hopkins University.
Harrill opened with a broad question about what concrete steps biomarker researchers should take to better reflect population diversity and underserved populations. Jackson urged biomarker researchers to work with those who are formally trained in health disparities or inequities to make sure that their research takes into account the role of social, economic, and political factors in health outcomes and not to assume that disparities can be explained by genetic differences. Furthermore, she said, researchers who are interested in carrying out studies on vulnerable groups should approach such research intentionally and with an understanding of the mistrust of those groups for researchers, a mistrust grounded in a long history of mistreatment and neglect.
Erdei spoke of lessons learned working with Native American communities. It is key, she said, to be completely honest in describing the research you propose—its purpose, what it will involve, how samples will be taken and analyzed, and how the results of the research might be used in policy making. Furthermore, it is important not just to obtain the support of tribal leadership but to work directly with the broader group through community meetings. Finally, the research should offer benefits to the group and researchers should be honest in explaining what those benefits are expected to be. Quirós-Alcalá echoed what Jackson and Erdei said and added that when working with vulnerable populations, it is crucial to have staff that resemble the population. This helps in building trust, which is key to the success of the research.
Jackson added that researchers need to prove to members of vulnerable communities and their leaders that the research team will safeguard and protect their privacy. There is a great deal of concern about the misuse of genetic data, and if this is not addressed it can lead to strain in the relationship between researchers and communities.
Responding to a question from Harrill, Fry said that when one is working with disadvantaged populations in other countries it is necessary to have collaborators who are embedded in the culture and the community. She said that single studies are not able to cover all of the ground necessary to understand diversity in the vulnerability of various populations, so collaborations are essential to get a look at the range of vulnerabilities. She suggested that the NIH-funded Environmental Influences on Child Health Outcomes study could serve as a model for how multiple collaborations could take place in environmental health science.
Harrill asked what suggestions the panelists had for addressing health disparities as the field of toxicological assessment moves into an era in which it increasingly relies on in silico and in vitro models. Chorley said that it will be important to integrate susceptibility into these models and that, in turn, will require mechanistic knowledge regarding such susceptibility.
For instance, researchers need to consider how to incorporate factors such as socioeconomic stress and its biological effects into their models.
Fry said that her group is beginning work in which they take blood cells from genetically diverse populations and transform those into human pluripotent stem cells that can then be cultured into various body cell types. This makes it possible to look for differential susceptibility to environmental toxicants among genetically diverse populations.
Cardenas offered a word of warning, noting that data for high-throughput analyses are often drawn from populations that include few underrepresented communities. Biomarkers identified in such analyses may not work the same way in underrepresented communities. In response to a question on how to define a baseline for a biomarker’s value, Cardenas added that it is challenging. With the epigenome, for instance, one might think that a newborn would have a pristine epigenome, but even the newborn carries records of environmental exposures through both the paternal and maternal lines. In epidemiology, the baseline is often defined as a healthy population or a younger population that theoretically has not experienced many challenges.
Chorley agreed that establishing a baseline is an important step in developing biomarkers, and noted that researchers often select biomarkers with less variability or “noise” to make it easier to detect a deviation from baseline. On the other hand, if one is looking at biomarkers in a diverse population, a “noisy” or highly variable biomarker may be an indication of differences between subpopulations, and such a biomarker could be important in understanding those differences.
An audience member asked what characteristics would make a novel biomarker most desirable for studies of diverse populations. Jackson said that she would want a biomarker to be responsive to social conditions and sensitive to her outcomes of interest. She said she is interested in omics biomarkers—and metabolomics biomarkers in particular—because they are particularly useful in tracing out the pathway from an exposure to an outcome of interest. This can help policy makers decide how to respond to various environmental exposures.
Fry said that while most people think in terms of biomarkers of exposure or biomarkers of effect, there are some biomarkers that act as both. She offered as an example an imprinted gene that is a marker of prenatal arsenic exposure but also lies on the causal pathway to lower birth weight. Such a biomarker of exposure and effect is particularly important because it becomes the potential mechanistic link between the contaminant and disease.
Determining the Readiness of an Emerging Biomarker of Effect for use in Environmental Health Decisions: Concurrent Breakout Discussion and Debate
Patrick McMullen returned to the topic of Marc Audebert’s talk (prerecorded) to introduce the breakout session by briefly resummarizing the role of γH2AX as a global DNA damage biomarker. Briefly, the body’s cells produce γH2AX in response to DNA damage and circulating concentrations of γH2AX can be used as a biomarker of genotoxic damage. McMullen explained that the afternoon breakout session was designed to stimulate conversation about a couple of different contexts of use. To drive the conversation, breakout groups were instructed to use the characteristics of an ideal biomarker as described by Brian Chorley on Day 1 (see Figure 2). These were to use safety assessment and to select a method of population monitoring, and to address questions on how biomarker of effect technology can be applied in these different scenarios. For safety assessment, the chosen scenario involved drug development. The second, or “population,” scenario concerned a “fabricated” situation where the question being asked was whether γH2AX could be used to monitor health effects in workers in a facility that produces a carcinogen.
Participants were divided into four groups, each with an assignment related to the use of γH2AX as a biomarker of DNA damage. Each group was asked to respond either yes or no to a context of use prompt. Groups A and B were given the prompt “γH2AX is or is not ready for use in risk assessment.” Groups C and D addressed the prompt “γH2AX is or is not ready for use to screen/monitor at-risk populations.”
Is γH2AX measurement ready to be included in a panel of human cells to be added to the standard FDA genotoxicity battery?
Breakout Group A, as summarized by Lesa Aylward, made the case for γH2AX being included in an FDA genotoxicity battery by setting forth the positive attributes of the γH2AX biomarker: it is highly specific and sensitive; it can be used for dose-response measurements and the assessment of benchmark concentrations; it is easy to use; it can be implemented in peripheral blood cells; and so on. Breakout Group B, as reported by William Mattes, based its arguments against adding γH2AX to the battery of tests because it would add little value to the current battery but would increase cost; the micronucleus test that is already part of the battery actually measures double-strand DNA breaks, while γH2AX does not; and γH2AX measures responses that are transient and reversible and can produce false positives.
In the ensuing discussion, McMullen asked people about their experiences defining radar plots for γH2AX (see Figure 2). Rita Schoeny, an independent consultant, commented that it was important in the exercise to be highly specific about the
context of use for the biomarker, because the radar plot can change depending on the context. Reza Rasoulpour of Corteva Agriscience added that it is not just the context that must be taken into account, but the radar plot for γH2AX should be compared with radar plots for other choices of biomarkers.
Stephen Dertinger of Litron Laboratories, who was part of the group making the case for γH2AX, explained that the panel was not thinking about it in isolation but considering it in conjunction with another biomarker such as H3, which together could provide valuable information about the mode of action of a genotoxin. That information could be important in early drug testing.
Schoeny said that one of the considerations of the first panel was the cost of a false positive, which could be enormous if it caused the development of a potentially valuable drug to be halted. Les Recio of Integrated Laboratory Systems pointed out that γH2AX is a genotoxicity measure, not a mutagenicity measure, which makes timing highly important because DNA repair mechanisms will repair the DNA over a variable period of time depending on the type of damage and other factors. Dertinger agreed that there are issues with timing, but that is true in all of genotoxicity, and temporal information helps to guard against false negatives and provide additional information concerning mode of action.
Is the γH2AX biomarker sufficiently developed to be used for monitoring potential effects in workers?
Breakout Group C, as reported by Alison Harrill, offered a qualified endorsement of γH2AX being used to monitor for genotoxic damage in workers. The group members were enthusiastic about it because they understood how the biomarker relates to the mode of action, Harrill said, and they thought that using an occupationally-defined population was ideal because of the controlled exposure and the potential for well-controlled designs. Ultimately the group agreed that the biomarker was ready to be used in the occupational setting but that it made sense to do more research and better understand its behavior.
Nicole Deziel reported that Group D found many issues with the biomarker that led them to argue that it is not ready for use in biomonitoring of workers. There is not enough population data on the biomarker, leaving too many unknowns such as inter- and intra-individual variability. There is a lack of standardized protocols. The analytical method is not high-throughput. Finally, it is not clear what the relationship is between the DNA damage detected by the biomarker and actual cancer risk.
After those reports, McMullen opened the floor to discussion. He asked if it would be necessary with this biomarker to establish individual baselines. Kim Boekelheide of Brown University and a member of the organizing committee answered yes—this is a sensitive assay, and the background levels will be dependent on different exposures outside the workplace, such as smoking or living in a house that has high radon levels. Mattes commented that obtaining a true baseline would require more than a single measurement—perhaps two measurements taken 1 week apart to get a sense of any fluctuation in the individual. Solomon said that a study of γH2AX levels in roofers found a fair amount of inter-subject variability as well as some variability within individuals, which will make it challenging to establish baselines.
Ultimately, Solomon said, once the appropriate pieces are in place, the γH2AX biomarker has “incredible potential to inform appropriate workplace controls and also protect workers.” By providing an early indication of potentially hazardous exposures, it could allow companies to intervene and protect their workers before serious damage occurs.
In his closing comments, Boekelheide offered five main points or thoughts about major themes that emerged from the workshop. First, referring to John-Michael Sauer’s prerecorded presentation on the ideal characteristics of a biomarker, he pointed to histopathology as having provided a “biomarker” that has been around for more 150 years—it is “the truth.” The goal today is to develop biomarkers using the new science of molecular biology that will become the truth so that the field moves forward with a more sensitive, selective, and robust biomarker.
Turning to the radar plot described by Brian Chorley, he said that there are many other factors than just the six included in the plot that are important to biomarkers. Those need to be taken into account.
Third, when one is examining an action identified by a biomarker, is it an adaptive response or an adverse response? It may take multiple biomarkers to tell the difference.
Fourth, researchers often have inflated expectations for a biomarker when they get started, but it is a long process to develop and refine a biomarker, and limitations appear. With work, a biomarker and its applicability can be improved, and eventually one can achieve something valuable.
Finally, it will be important to develop biomarkers that respond to extrinsic factors, such as socioeconomic stresses. Such biomarkers will help in the conduct of intervention studies to see how changes in these extrinsic factors can result in different health outcomes.
DISCLAIMER: This Proceedings of a Workshop—in Brief was prepared by Robert Poole, Fran Sharples, and Keegan Sawyer as a factual summary of what occurred at the workshop. The planning committee’s role was limited to planning the workshop. The statements made are those of the rapporteurs or individual meeting participants and do not necessarily represent the views of all meeting participants, the planning committee, or the National Academies of Sciences, Engineering, and Medicine.
ORGANIZING COMMITTEE ON INTEGRATING THE SCIENCE OF AGING AND ENVIROMENTAL HEALTH RESEARCH
This workshop was organized by the following experts: Lesa Aylward, Summit Toxicology; Kim Boekelheide, Brown University; Nicole Deziel, Yale University; Gary Ginsberg, New York State Department of Health; Alison Harrill, National Institute of Environmental Health Sciences; William Mattes, U.S. Food and Drug Administration; and Patrick McMullen, ScitoVation.
Reviewers: The Proceedings of a Workshop—in Brief was reviewed in draft form by Jamie DeWitt, East Carolina University; Patrick McMullen, ScitoVation; and Julia Varshavsky, California Environmental Protection Agency, to ensure that it meets institutional standards for quality and objectivity. The review comments and draft manuscript remain confidential to protect the integrity of the process.
Sponsor: This workshop was supported by the National Institute of Environmental Health Sciences. For more information, contact the Board on Life Sciences at (202) 334-3947 or visit https://www.nationalacademies.org/bls/board-on-life-sciences.
About the Standing Committee on Emerging Science for Environmental Health Decisions
The Standing Committee on the Use of Emerging Science for Environmental Health Decisions is sponsored by the National Institute of Environmental Health Sciences to examine, explore, and consider issues on the use of emerging science for environmental health decisions. The Standing Committee’s workshops provide a public venue for communication among government, industry, environmental groups, and the academic community about scientific advances in methods and approaches that can be used in the identification, quantification, and control of environmental impacts on human health.
Presentations and proceedings such as this one are made broadly available, including at https://www.nationalacademies.org/our-work/standing-committee-on-the-use-of-emerging-science-for-environmental-health-decisions.
Suggested citation: National Academies of Sciences, Engineering, and Medicine. 2020. Predicting Human Health Effects from Environmental Exposures: Applying Translatable and Accessible Biomarkers of Effect: Proceedings of a Workshop—in Brief. Washington, DC: The National Academies Press. https://doi.org/10.17226/25962.
Division on Earth and Life Studies
Copyright 2020 by the National Academy of Sciences. All rights reserved.