Workshop on perinatal exposure to dioxin-like compounds. VI. Role of biomarkers.

Studies of perinatal exposures to dioxin-like compounds (DLCs), coplanar polycyclic halogenated aromatics whose prototype is 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD), have employed a variety of outcome measures to investigate effects on the reproductive/developmental, endocrine, immune, and neurobehavioral systems. The effects include infertility, growth retardation, fetal loss, changed sexual differentiation, reduced cognitive/motor function, dermatologic and other ectodermal effects, and decreased immune response. Significant biomarkers have included sperm count; CD4/CD8 ratio; and levels of testosterone, T4, and dopamine. Using specific dioxin or PCB congeners, these and other markers were used to investigate the mechanisms of the observed effects. The DLCs, which include some PCB congeners, are characterized by high-affinity binding to the Ah receptor; most biological effects are thought to be mediated by the ligand-Ah receptor complex. Other PCB congeners have low affinity for the Ah receptor, and operate by non-Ah receptor mechanisms. The biologic activity of a PCB mixture is the sum of the agonist and antagonist activities of the different constituents in the mixture. Animal studies with specific PCB congeners can help to clarify these activities. With similar approaches, biologic markers of effect can be developed and applied in epidemiologic studies to monitor for, and predict, adverse effects in humans.

Introduction A biologic marker, or biomarker, is generally defined as any functional or biochemical change in a cell or tissue that is indicative of exposure, response or potential susceptibility to a drug or an environmental agent ( Figure 1) (1). Generally, biomarkers are laboratory-based measures that are intended to increase the sensitivity, specificity, or power of a study to establish a causal association between an exposure and an outcome, and/or provide useful information about the mechanisms involved in exposures, responses, or susceptibilities.
This workshop focused on exposures to a subset of the polycyclic halogenated aromatic hydrocarbons, which include the polychlorinated biphenyls (PCBs), polychlorinated dibenzo-p-dioxins (PCDDs), and polychlorinated dibenzofurans (PCDFs) (Figure 2). The latter compounds are by-products of industrial syntheses and chlorine bleaching processes, and can be produced through burning of municipal or industrial waste. Many are persistent environmental contaminants, the most toxic and well studied of which is 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD). The workshop reviewed the evidence that perinatal exposures to these compounds may lead to adverse health outcomes.

Biomarkers of Exposure
Many of the dioxins, dibenzofurans and PCBs resist biodegradation and, because they are lipophilic, concentrate in adipose tissues. Tissue concentrations of these parent compounds serve as good measures of environmental exposures, and residue levels in body fat or in lipid fractions of breast milk or blood are used to quantify this exposure. A vast literature describes the levels of organochlorine residues found in body fluids or tissues of animals and humans, and documents the increased concentration of these contaminants in the fat of animals at successively higher trophic levels of the food chain.
The residue profile (nature and levels of residues) in tissues is unlikely to be identical to the composition of the original envi-  disorders), or have great medical or social costs (e.g., Alzheimer's disease). Thus, biomarkers of effect * signal that an adverse clinical effect has the potential to occur; * identify exposed or responsive individuals; * help to develop preventive measures by giving more detailed information (e.g., mechanism) on the health problem; * help to develop regulatory strategies for disease prevention (e.g., regulations based on biomarker end points).
Identifying Biomarkers of Effect and Susceptibility: Understanding Disease Mechanism(s) A key to identifying a suitable biomarker of effect or susceptibility is having a good understanding of the mechanisms) by which the adverse health outcome arises. In the medical model, the adverse outcome is disease, and is accompanied by the disease pathway paradigm: predisposing conditions, early warning symptoms, preclinical indicators and, finally, clinical symptoms. Some adverse outcomes (e.g., neurobehavioral) are not easily characterized as a series of stages on a pathway to an abnormal pathology or disease, and may not fit this model; instead they may represent an impaired capacity within what is considered the normal range (e.g., lowered IQ: see "Issues").
Optimally, a marker of exposure indicates that the agent has access to the causal mechanism; an effect marker is an element of this mechanism; and a marker of susceptibility bears some relationship to this mechanism. Rarely, however, is our knowledge of disease mechanisms sufficient for us to declare that a marker is strongly associated with either disease susceptibility or outcome. Instead, we place a marker on the continuum of biomarkers (exposure, early biological effect, or effect) depending upon the weight of evidence that connects the biologic measure with the disease outcome. The predictive value of any marker of effect depends upon the strength of evidence that links it with the adverse health outcome.
Most markers lack a strong enough association with disease outcome to be classified as markers of effect; consequently, if their response is found to vary proportionally with dose, they are classified as either markers of exposure or as markers of early biologic effect. Chromosomal aberrations, for example, are considered markers of early biologic effect rather than of cancer because our knowledge of their specific role in cancer causation is limited (2). Lowered sperm count, on the other hand, is a marker for infertility because it has a demonstrated relationship with this outcome (3)(4)(5).

Mechanism(s) of Action of Dioxin-like Compounds (DLCs): The Ah Receptor
Most cells contain a cytoplasmic protein, called the Ah receptor, which has a high affinity for TCDD (6). Other compounds that bind to the intracellular Ah receptor are termed dioxin-like compounds (DLCs). These include representatives of the coplanar, halogen-substituted multiring structures such as the PCDDs, PCDFs, and some of the coplanar PCB congeners. Most PCB congeners that are noncoplanar are reported to have low affinity for the Ah receptor. The TCDD-Ah receptor complex is believed to mediate many of TCDD's observed effects. The DLC-Ah receptor complexes produce a characteristic spectrum of responses in animal models. The potency of DLCs as effectors of teratogenicity, carcinogenicity, immunotoxicity, and biochemical changes (drug-metabolizing enzymes and growth factor pathways) strongly correlates with their binding affinity to the Ah receptor (7)(8)(9)(10).
The binding of a molecule of ligand to the cytoplasmic Ah receptor results in activation of the receptor and translocation to the nucleus. This is a complex event in which heat shock proteins dissociate from the receptor and another protein, Ah receptor nuclear translocator (ARNT) associates with the receptor. The ligand-AhR-ARNT complex acts as a transcriptional enhancer: it binds to regulatory regions on DNA, upstream of the CYPJAJ gene and other dioxin-responsive genes, and increases the transcription of these genes (11,12). The altered transcription of genes in the ligand-Ah-responsive gene battery is believed to be the mechanism by which DLCs produce their biological effects (13).
Non-DLCs operate by unknown mechanisms. Some non-DLCs are more active in producing certain biologic responses than related DLCs. For example, a noncoplanar PCB congener, PCB 28, is more active in altering dopamine levels in the forebrain of rats than dioxin-like coplanar PCB congeners (14). This suggests that the neurochemical response is mediated by a non-Ah receptor mechanism, which is consistent with the reported absence of the Ah receptor from neurons (15).
The discussion below summarizes studies that employ a variety of markers of effect to investigate disease outcomes. Only a few are directly linked to outcome (e.g., sperm count and infertility). Most markers described below signal changes in functions but are not yet strongly linked to specific adverse outcomes. For example, CD4/CD8 ratios, or levels of T4 or dopamine, are now considered markers of early biologic effects. Future studies may link them to adverse outcomes such as immune system depression or motor/cognitive deficits. CYPIA] activity is also considered a marker of early biologic effect (DLC bound to the Ah receptor), lacking a known association with disease.

Biomarkers and Perinatal Studies
Perinatal exposures encompass prenatal (transplacental) as well as postnatal (lactational) exposures. What is unusual about the application of biomarkers to the study of perinatal exposures?
Perinatal exposures are unique in that they impact upon organ systems that are in varying stages of development. The responses exhibited by these developing systems, for example, to perinatal exposures to DLCs are influenced by several factors including: the chemical nature of the agents; the timing and duration of the exposure during the developmental period; and the interrelationships between the developing systems. The broad range of responses that DLCs produce in multiple developing organ systems suggests that DLCs act via a central mechanism (e.g., hormonal) and that the developing systems interact.
In 1968 researchers discovered ectodermal abnormalities and cognitive/motor deficits among children who had accidental perinatal exposure to DLCs and non-DLCs in cooking oil (16)(17)(18). Since then, researchers have sought to do the following: * characterize the full range of clinical effects arising from perinatal exposures to these compounds; * identify the mechanisms) of action of these compounds in causing seemingly disparate effects; * understand the interrelationship of these effects in developing systems; * investigate the mechanisms) and the interrelationships between these complex clinical outcomes using biomarkers of exposure, susceptibility and effect.

Animal Studies
As reported in the workshop, perinatal exposures to DLCs in animal studies have provided information on effects in several developing systems, including the reproductive, endocrine, neurobehavioral, and immune systems. Adverse reproductive or developmental effects included reduced fertility in primates (19), endometriosis in primates (20), altered sexual differentiation in male and female rats (21)(22)(23), reduced sperm count in rats (24), and growth retardation in primates and rats (19,25). In the endocrine system, the effects included reduced T4 levels in the prefrontal cortex of rats (25), reduced testosterone levels in rats (22,23), and altered dopamine levels in the forebrain of primates and rats (14,26,27). Neurobehavioral effects included spatial learning/memory deficits in primates and rats (27)(28)(29)(30).
For several of the effects seen in animal studies, there appear to be corresponding outcomes in humans: reductions in infant height and primary immune response in Inuit infants whose mothers had breast milk with high PCB levels (41,42); and reductions in cognitive or motor functions in infants whose mothers ingested PCBcontaminated cooking oil or fish during pregnancy or lactation (16)(17)(18)43,44).

Biomarkers and Mechanisms
The wide range of effects seen in different organ systems of animals and humans underlies the current working hypotheses about the mechanisms of action of DLCs and non-DLCs (see Workshop Summary Report, this issue) including the following: * Biologic effects of DLCs are mediated by their binding to the Ah receptor, while the mechanisms of action of non-DLCs are not well understood; * DLCs act as "environmental hormones," disrupting normal endocrine functions. * Pluripotent effects on the reproductive/ developmental, endocrine, neurobehaviorial, and immune systems reflect the hormone-like activities of DLCs and non-DLCs as well as the highly integrative and interactive nature of these systems. * DLCs, as hormones, affect the differentiation of cells in various systems in a tissue-, cell-, and stage-specific manner; effects consequently are tissue-, cell-, and stagedependent; and timing of exposure can be critical to the expressed outcome. For example, a cell may express DLC-sensitive receptors for only a limited period in its life cycle. * The perinatal period is a sensitive period, a window through which one may examine the effects of DLCs on several developing organ systems. As seen in Table 1, a variety of biomarkers have been measured in each biological system (see Lindstrom et al., this supplement, for their use in specific studies). Results from these studies give insight into mechanisms of action of DLCs during the perinatal period.

Reproductive
Perinatal exposures to TCDD (and presumably DLCs) produce reproductive effects in male offspring that are different from those produced by PCB mixtures (presumably the non-DLC components): PCBs cause infertility in primates without decreasing the sperm count (19); TCDD lowers sperm count in rats without affecting fertility (22)(23)(24). The differences in response are attributable either to species differences or to differing mechanisms of actions of TCDD and PCBs. TCDD acts via the Ah receptor; some PCB congeners may act via non-Ah receptor mechanisms.

Endocrine
Prenatal TCDD in rats causes feminization and demasculinization of male offspring, as manifested by patterns of LH secretion and sexual behavior (21)(22)(23). Brain levels of T4 (thyroid hormone) and dopamine (neurotransmitter) were reduced in rats by perinatal exposures to specific noncoplanar (non-DLC) PCB congeners (14,(25)(26)(27). Dopaminergic effects were, in addition, age-dependent (26). Since T4 crosses the placenta and blood-brain barrier, changes in T4 hormone levels in the fetus may affect normal development in the brain (e.g., branching of apical dendrites). Studies in rats of the effect of specific PCB congeners on T4 and dopamine levels, apical dendrite development, and memory test scores could explore the relationship between these biomarkers and health outcomes. Immune Cells within the immune system are continually replenished and are a model for developmental effects and effects on differentiation. The DLCs can act as immuno-suppressors or enhancers in this system, depending upon the cellular target. The effects of DLCs on macrophages suggest an enhancement of inflammatory activity and expression of cytokines, particularly TNF and IL-i (35,36,45,46). The effects of DLCs on B-cells results in suppression of antibody production and a direct effect on phosphorylation in this cell type (34,47,48).
One of the most common observations is for a suppression of the anti-SRBC plaqueforming cell responses in TCDD exposed mice, which may be due to suppression of T-cell responses, although these effects are not observed in rats (31,49).
Changes in cell surface markers on Tcells are apparently a particularly sensitive response to the effects of TCDD in nonhuman primates, and have resulted in exploration of this effect as a biomarker in Volume 103, Supplement 2, March 1995 Thyroxine; unlike T3, passes blood-brain barrier. Free, as opposed to bound, T4; synthesis of T3 and T4 occurs within thyroglobulin from tyrosine residues. Total T4 (bound + free).
Thyroxine binding globulin; synthesized in liver; major thyroid hormone binding transport protein for humans; absent in rats. TSH (thyrotropin)-releasing hormone produced by hypothalamus (HT), which controls the release of TSH by the anterior pituitary; secreted by HT neuroendocrine cells into blood vessels of pituitary stalk, and travels to pituitary, where it stimulates cells to release TSH into main bloodstream. Thyroid-stimulating hormone; produced by anterior pituitary; stimulates cells in thyroid to synthesize and secrete thyroid hormone, which travels in blood to most cells of body; regulates thyroid function.
Measure of cognitive and motor ability (up to age 2). Measure of cognitive and motor ability (age 3-5 yrs). Spatial delayed alternation task; measures spatial learning/memory impairment in primates and rodents.
Cell surface markers on T-("thymus") lymphocytes; ratio of "helper" T/"suppressor" T indicates immune function (2 = "normal"; < 1 = compromised). CD29 are "memory" T-cells subset that have been previously activated by antigens; possible marker of sensitivity or effect to DLC exposures. Sheep red blood cell plaque-forming cell assay for ability of B-cells to produce antibody against antigen; most sensitive assay for effects of DLCs in mice. Family of proteins (includes TNF) synthesized in many cells in response to infections, inflammation, and stress; general effectors on metabolism and homeostasis. Markers for stem cells of lymphocytes in bone marrow; used to demonstrate that DLCs affect development of T-cells.
Tumor necrosis factor, produced by macrophages, is an immune system hormone (lymphokine) involved in inflammatory response; in animals, DLCs increase TNF and cause toxic effects (weight loss, death).
humans (39,50,51). DLC exposures change the ratios of lymphocyte subpopulations in rodents and primates (31)(32)(33)(37)(38)(39)(40), perhaps impairing the primary immune response and accounting for the decreased immunization "take" rates reported for Inuit children (41,42). Changes in the ratio of lymphocyte subpopulations (e.g., changed CD4/CD8 ratio) may mean that DLC and/or non-DLC components of PCB mixtures alter the differentiation pattern of cells of the immune system. A mechanism for TCDD-induced thymic atrophy is suggested by a study in mice in which the depletion of bone marrow prolymphocytes (stem cells) following TCDD exposure was measured using markers that are specific for prolymphocytes (terminal deoxynucleotidyl transferase, TdT, and recombinase activating gene, RAG) (32,33). In addition to reducing the number of lymphocytes in the thymus gland, TCDD appears to reduce the number of prothymocytes in the bone marrow available to repopulate the depleted thymus. In this way, TCDD's effects on bone marrow are believed to contribute to thymic atrophy.

Neurobehavioral
Perinatal studies with specific PCB congeners and biomarkers suggest a link between behavioral changes in rodents and primates with changes in levels of T4 or dopamine. For example, spatial learning/ memory deficits in monkeys and adult rats are produced by perinatal exposures to specific PCB congeners, especially PCB28 (Schantz), a non-DLC ortho-substituted congener. This same congener also alters dopamine levels in monkeys and rats in specific regions of the brain (26).

Issues
For perinatal exposures, are there markers of early events that predict later adverse human health outcomes? Can these markers be used to study delayed effects, i.e., predict adverse consequences in newborns and young children, in adults and possibly second generation children? Are these markers linked to exposure? Do they suggest possible public health interventions?
Two examples from the workshop illustrate the difficulties of establishing associations between biomarkers and disease outcomes. Human data suggest that exposure to DLCs changes the ratio of lymphocyte subpopulations (e.g., CD4/CD8 ratio). These changes fall within the range of variation seen in normal, "unexposed" human Environmental Health Perspectives populations. As a consequence, we do not know if they signal an adverse health outcome. If the changes fall outside the normal range, their health significance is still not clear. Are they associated with an adverse outcome? Should we consider a significantly altered lymphocyte subset ratio an adverse health outcome? How big must the change be?
The second example illustrates difficulties with outcome variables that integrate functions, such as IQ. A shift in an individual's IQ of a few points is not statistically significant and has little impact on the individual or on society. A decrease of two points in population IQ, however, can significantly impact public health. Do we have candidate markers that can predict this shift? Could changes in neurotransmitter levels (e.g., dopamine in forebrain or caudate nucleus) or other effects in the developing or adult brain be used to predict changes in IQ?
Workshop participants emphasized the difficulties in linking chemical exposures with complex responses that have multiple causes (e.g., neurobehavioral end points). For these responses, chemical exposures may not be the primary causes, and their effects may be obscured or confounded by stronger factors. (Alternatively, they may work in concert with other factors to produce an effect.) For example, neurobehavioral studies suggest there is an association between perinatal exposures to DLCs and changes in IQ in children. However, many socioenvironmental factors also affect IQ, and these may overwhelm and mask any contribution that is made by a chemical exposure. They act as confounders of chemical exposures. Before chemical linkages can be investigated, neurobehavioral measures are needed that can better control for social/environmental factors.
Finally, neurobehavioral researchers cite drawbacks in the application of biomarkers to their field. First, the medical model of disease, in which there is a pathway from early health impairment (preclinical signs) to frank pathology and disease, may not be applicable to neurobehaviorial end points. The various neurobehavioral "biomarkers," such as IQ, age at development of verbal skills, or short-term memory, are not considered precursors or early signs of disease. They are effects that are less favored. Decrements in these functions do not signal a transition from "normal" to "abnormal", but rather a shift to a less favorable position within the "normal" range.
The sensitivity and specificity of biomarkers in neurobehavioral studies was also questioned. The most useful cognitive measures are those for which population norms have been established. These measures generally sum the influences of a variety of factors that can affect development. They do not assay the functioning or non-functioning of specific neurobehavioral mechanisms. More specific tests are needed. Importantly, a more complete understanding of the mechanisms involved in the neurobehavioral changes themselves is needed before more specific tests can be developed.

Epidemiologic Studies in Progress
Several ongoing epidemiologic studies investigate the effects of perinatal exposures to DLCs and non-DLCs by employing a number of biologic markers. Koopman-Esseboom et al. (52) explore the effects of PCBs and dioxins on thyroid hormone metabolism (TT3, TT4, FT4, and TSH levels) and on neurologic and psychomotor development among children whose mothers had varying levels of coplanar PCBs and TCDD in their breast milk.
Buck et al. (this supplement) described an ongoing retrospective study of reproductive outcomes among 11,000 New York state anglers 18 to 34 years of age. End points examined for subsets of this cohort include fertility rates (NYS live birth and fetal death registries), spontaneous fetal mortality rates (NYS fetal death registries), pregnancy intervals, intrauterine growth retardation, secondary sex ratios, time to conception, intrauterine growth retardation, neurodevelopmental status, and disorders of ectodermal tissue. Rogan et al. (this supplement) examine the lactational effects of DDE (levels measured in breast milk) in a cohort of 230 mothers in Mexico. Preliminary results suggest that DDE acts as a prolactin antagonist. Rogan et al. continue to follow their cohort of 117 children in Taiwan with regular physical examinations to determine any peripubertal developmental effects from DLC exposures, including stages of puberty (as per Tanner scale). In a similar manner, they are also examining a cohort of 200 children born in Taiwan between 1985 and 1992 for peripubertal developmental effects that may result from DLC-exposures.
Clark et al. (this supplement) examine susceptibility markers in two cohorts, one in Seveso, Italy exposed to high levels of dioxins following a chemical plant explosion in 1976 and the second in Germany, with occupational exposures to dioxins in a chemical plant that produces 2,4,5trichlorophenol. Within each of these cohorts, certain individuals developed chloracne. Others, in the same cohort with similar exposures, did not. Genetic and biochemical differences between the responding and nonresponding individuals are sought to account for the differences in response. These include Ah receptor levels and the expression of the CYPlAl, TNFor IL-1 beta genes.

Conclusions
The potential adverse consequences on the reproductive, endocrine, immune, and neurobehaviorial systems of perinatal exposures to dioxins, dibenzofurans, and PCBs are being explored using a variety of biomarkers, including sperm count, CD4/ CD8 ratio, and levels of testosterone, T4, and dopamine. In animal studies, exposures to some of the listed compounds produce changes in the levels of these markers, suggesting that the chemicals affect fertility, impact the endocrine system, cause feminization and demasculinization of males, depress the immune system, and produce neuroendocrine and neurobehavioral effects. These and other markers will be employed in future experimental studies to explore the mechanisms of these effects. If validated, these markers can be used in epidemiologic studies to determine the effects in humans that arise from perinatal exposures to dioxins, dibenzofurans, and PCBs.