Hazard and risk assessment of chemical mixtures using the toxic equivalency factor approach.

There is considerable public, regulatory, and scientific concern regarding human exposure to endocrine-disrupting chemicals, which include compounds that directly modulate steroid hormone receptor pathways (estrogens, antiestrogens, androgens, antiandrogens) and aryl hydrocarbon receptor (AhR) agonists, including 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) and related compounds. Based on quantitative structure-activity relationships for both AhR and estrogen receptor (ER) agonists, the relative potency (RP) of individual compounds relative to a standard (e.g. TCDD and 17-beta-estradiol) have been determined for several receptor-mediated responses. Therefore, the TCDD or estrogenic equivalent (TEQ or EQ, respectively) of a mixture is defined as TEQ = sigma[T(i)]xRP(i)or EQ=sigma[E(i)]xRP(i), where T(i) and E(i) are concentrations of individual AhR or ER agonists in any mixture. This approach for risk assessment of endocrine-disrupting mixtures assumes that for each endocrine response pathway, the effects of individual compounds are essentially additive. This paper will critically examine the utility of the TEQ/EQ approach for risk assessment, the validity of the assumptions used for this approach, and the problems associated with comparing low dose exposures to xeno and natural (dietary) endocrine disruptors.


Introduction
The potential adverse impacts of chemicals are dependent on a number of factors, including levels and duration of exposure, chemical potency, timing of exposure, mechanism of action, and interactions between chemicals in a mixture. Hazard and risk assessment of chemicals carried out by regulatory agencies have focused primarily on the toxicities of individual compounds, whereas wildlife and humans are exposed to complex mixtures of man-made compounds that act through multiple pathways. Moreover, the human diet contains many natural products and cooking-derived compounds that exhibit many of the same toxic, mutagenic, and carcinogenic properties of industrial-derived contaminants (1)(2)(3). In most cases, humans are exposed to significantly higher levels of natural products than the man-made chemical toxicants that act through the same pathway. For example, early development of the Ames test for detecting bacterial mutagens generated considerable scientific, regulatory, and public concern over human exposure to the many different industrial chemicals that exhibited mutagenic activity in one or more of the highly sensitive bacterial tester strains (4,5). Subsequent studies demonstrated that some of the most mutagenic compounds in the human diet are not industrial-derived contaminants, but natural compounds that include a complex series of heterocyclic aromatic amines derived from cooking proteinaceous foods (e.g., fish, beef, poultry) (6)(7)(8). Thus, the public health concern regarding human exposure to mutagens must take into account intake and potency of both natural and man-made chemicals that act through various pathways.
Hazard and risk assessment of human exposures to chemicals must also take into account scenarios where chemical interactions may significantly influence toxic outcomes. For example, despite relatively high levels of human exposure to natural carcinogens in the diet, there are several other classes of natural products (e.g., flavones, antioxidants) that inhibit P450-mediated metabolic activation or induce detoxifying enzymes, and these compounds may provide protection against natural or manmade toxins (9)(10)(11). In contrast, workplace or environmental exposures to nontoxic levels of organochlorine solvents such as chloroform may lead to hepatotoxic effects if there is concurrent exposure to ketones because of their nonadditive (synergistic) interactions (12,13). Thus, chemical interactions are important determinants in evaluating the potential hazards and risks of exposure to chemical mixtures. This manuscript will outline the development, validation, and pitfalls associated with the toxic equivalency factor (TEF) approach for risk assessment of complex mixtures. The assumptions implicit in utilization of the TEF approach indude: the individual compounds all act through the same biologic or toxic pathway; the effects of individual chemicals in a mixture are essentially additive at submaximal levels of exposure; the dose-response curves for different congeners should be parallel; and the organotropic manifestations of all congeners must be identical over the relevant range of doses (14,15). TEFi values are either derived for a Environmental Health Perspectives * Vol 106, Supplement 4 * August 1998 m species-specific response or are a composite value obtained from TEFs for several responses, and individual TEFs are usually determined relative to the activity of a standard or reference compound. The TEF approach has been applied to different structural classes of compounds, including polynuclear aromatic hydrocarbons (PAHs), halogenated aromatic hydrocarbons (HAHs), and endocrine disruptors. The utility and problems associated with TEFs and TEQs will be discussed.

Toxic Equivalency Factor Approach for Polynuclear Aromatic Hydrocarbons
Individual PAHs such as benzo[a]pyrene (B[a]P) have been extensively investigated as carcinogens and as ligands for the aryl hydrocarbon receptor (AhR). The carcinogenic activity of PAHs is dependent on the oxidative metabolic activation of these compounds into genotoxic metabolites, which subsequently interact with DNA to initiate carcinogenesis. The carcinogenic potencies of individual PAHs have been determined in different bioassays and TEF values proposed for various PAHs are summarized in Table 1 (16)(17)(18)(19). The utility of this approach was demonstrated in studies using relatively simple reconstituted PAH mixtures in rodent carcinogenicity models (20,21). However, Warshawsky and coworkers (22) recently demonstrated that there are a number of important factors that can significantly modulate the genotoxicity of PAH mixtures, indcluding the presence or absence of B[a]P, the dose, and the solvents used in carcinogen administration. This variability of carcinogenic potency suggests that the TEF approach for PAHs may not be appropriate for some animal models. Studies in several laboratories have investigated the biochemical, toxic, and genotoxic activities of manufactured gas plant (MGP) residues, which contain complex mixtures of PAHs (23)(24)(25)(26)(27). Comparisons of the effects of MGP PAHs with B[a]P or a reconstituted mixture of PAH hydrocarbons suggested that the mixture induced synergistic responses or that other factors were important. Based on results of recent studies (24,25,28), the high activity/genotoxicity of the MGP-PAH mixture may be due to unidentified alkyl PAHs. For example, a reconstituted mixture of the 17 major PAHs in an MGP sample (24) did not induce liver tumor formation in the B6C3F1 male juvenile mouse model at a dose of 1071 mg/kg; in contrast, the field-derived sample induced a 45% incidence of liver tumors at the same dose (25,28). These results demonstrate that applications of TEFs for PAHs require a more detailed knowledge of the complete composition of these mixtures and the TEFs of all active components. Thus, the approach may be useful for defined PAH mixtures containing only parent hydrocarbons; however, for mixtures containing alkyl PAHs, the TEF approach is not valid because of the minimal data available on the identities and relative potencies of these compounds.

Toxic Equivalency Factor Approach for Halogenated Aromatic Hydrocarbons
Polychlorinated dibenzo-p-dioxins (PCDDs), dibenzofurans (PCDFs), polychlorinated naphthalenes (PCNs), and polychlorinated biphenyls (PCBs) are HAHs that are industrial compounds or industrial combustion by-products (Figure 1). These compounds are chemically and environmentally stable and have been identified in almost every component of global ecosystems, induding fish and wildlife and human serum, adipose tissue, and milk (29,30). HAHs are also routinely detected as residues in diverse food products, and the diet is the major source of human exposure to HAHs (30,32).
The mechanism of action of 2,3,7,8tetrachlorodibenzo-p-dioxin (TCDD) and related HAHs has been extensively investigated and the results support a pathway that involves initial ligand (HAH) binding to the intracellular AhR, which is widely expressed in mammalian tissues (33). The mechanism of AhR-mediated CYPlAI induction has been extensively investigated; the results show that the AhR is a ligand-induced nudear transcription factor in which transactivation is associated with interaction of the heterodimeric nuclear AhR complex with dioxin-responsive elements located in the 5'-promoter region of the Ah-responsive gene (34,35). unknown; however, it is assumed that many of the responses are because of altered gene expression.
Hazard and risk assessment of PCDDs and PCDFs initially focused on quantitation of TCDD in various environmental samples; however, with development of high-resolution analytical techniques coupled with studies on structure-toxicity relationships, it was apparent that the bulk of the toxicity induced by most PCDD/PCDF mixtures is not due to TCDD alone. Based on the well-characterized structure-activity relationships established for PCDDs and PCDFs (36)(37)(38)(39), a TEF approach has been developed for these compounds:  primarily been detected in environmental samples and are among the most potent congeners. The TEF for each 2,3,7,8-substituted congener compared to TCDD is variable among cell types, laboratory animal species, target organs, and responses. Research in our laboratory has extensively investigated the immunotoxicity-derived TEFs for several HAHs in mouse models (47)(48)(49)(50); TEFs for inhibition of the plaque-forming cell response to trinitrophenyl-lipopolysaccharide by 2,3,4,7,8pentachlorodibenzofuran in C57BL/6, DBA/2, and B6C3F1 mice varied by approximately 7-fold, and in some assays this congener was more potent than TCDD. Over a broader spectrum of responses, TEFs for individual PCDD/ PCDF congeners can vary by over 100-fold. The broad range ofTEF values for a specific congener compromises the use of a single TEF for this congener and may over-or underestimate the calculated TEQ for a mixture. Variable TEFs are due to many factors including differential pharmacokinetics and metabolism of HAHs in various in vivo and in vitro bioassays.
Validation of the TEF approach can be investigated by determining the in vitro or in vivo toxicities of reconstituted mixtures of PCDDs and PCDFs and comparing their observed versus calculated potencies. Eadon and co-workers (51) utilized a PCDF/PCDD mixture (primarily PCDFs) resulting from a PCB fire and compared the calculated versus observed effects for several end points in the guinea pig, including decreased body and thymus weights, increased serum triglycerides, decreased serum alanine aminotrasferase levels, and formation of hepatocellular cytoplasmic inclusion bodies. Their results showed that the experimentally observed TEQs per kilogram ranged from 2 to 21 ppm depending on end point; the calculated value using a set of provisional TEFs was 14.5 ppm. These results demonstrated a good correspondence between the observed and calculated values. Other reports using multiple end points in both in vivo and in vitro models demonstrated that for several PCDD/PCDF mixtures, there is a reasonable correspondence between calculated and experimentally determined TEQs (52)(53)(54)(55)(56)(57)(58)(59)(60). For more complex mixtures containing compounds that act through multiple pathways to give both similar and different toxic responses, the TEF/TEQ approach may not be appropriate. Moreover, it should also be noted that even for PCDDs/PCDFs, there is some  evidence that TEFs do not always predict relative congener potency in different rat strains (61). Several studies have also demonstrated that the coplanar PCBs (i.e., 3,3',4,4'tetra-, 3,3',4,4',5-penta, and (66) for coplanar PCBs and their mono-ortho-substituted analogs to show that these compounds contributed substantially to the TEQs of diverse industrial/environmental extracts. Provisional TEFs have been proposed for coplanar and mono-ortho coplanar PCBs and these values are used routinely for determining total TEQs (i.e., PCDDs, PCDFs, and PCBs) in various extracts (67 (68).

Toxic Equivalency Factor Approach for Endocrine Disruptors
It has been hypothesized that industrialderived estrogenic compounds (xenoestrogens) and possibly other naturally occurring estrogens may be responsible for a global decrease in male reproductive capacity (e.g., sperm counts) and increased incidence of breast cancer in women (85,86). The validity of these hypotheses has been questioned (1,87), and resolution of the role of hormonally active compounds in human disease requires further study.
Like AhR agonists, hormonelike compounds that act through specific cellular receptors should be good candidates for using a TEF approach. Verdeal and Ryan (88) previously compared human exposures to man-made and naturally occurring estrogenic compounds using a TEF/TEQ approach and diethylstilbestrol equivalents. The recent discovery of a second form of the estrogen receptor (ER), ERp (89), further complicates development of an estrogen equivalent (EQ) approach for estrogenic compounds. The specific responses that are mediated via ERa or ERN3 have not been delineated, and relative potency factors for these responses by different structural classes of natural and man-made estrogenic compounds have not been determined. Several groups have reported TEFs for both natural ligands for the ER (e.g., flavonoids, lignans) and for xenoestrogens, which are industrial-derived chemicals and their by-products (90)(91)(92)(93). Although individual TEFs have been assigned for each compound, most assay systems indicate that with few exceptions, both natural and xenoestrogens are > 1000 times less potent than 17p3-estradiol (E2). A close inspection of the data obtained for estrogenic compounds reveals that there are many problems in development of a TEF approach for these compounds and some of these problems are similar to those observed for HAHs and PAHs.

Pharmacokinetics, Metabolism, and Serum Protein Binding
The in vivo activity of natural and man-made endocrine active agents are significantly influenced by their uptake, distribution, and metabolism. For example, many of the organochlorine xenoestrogens exhibit low estrogenic potency based on results of in vitro bioassays; however, these compounds persist in the environment and bioaccumulate. In contrast, many naturally occurring estrogenic flavonoids in foods are rapidly metabolized. For example, studies in this laboratory (94) showed that naringenin, a flavonoid in grapefruit juice, exhibited estrogenic activity in in vitro bioassays, whereas at doses as high as 30 to 40 mg/animal naringenin, did not induce estrogensensitive responses in the rat uterus. In contrast, in female rats cotreated with E2 plus naringenin, there was significant inhibition of E2-induced uterine wet weight, progesterone receptor levels, peroxidase activity, and DNA synthesis. Bisphenol A and p-octylphenol are two estrogenic phenolic compounds that exhibit similar estrogenic potency in a number of in vitro assays (90). Vom Saal and co-workers (95) recently reported that prenatal to early postnatal exposure of mice to bisphenol A resulted in increased prostate weight in adult male offspring, whereas p-octylphenol was inactive in this model. The increased toxicity of bisphenol A compared to poctylphenol was associated with preferential binding of the latter compound to serum proteins and decreased uptake in target cells. In contrast, research in this laboratory in the immature female rat uterus indicated that nonylphenol was significantly more potent than bisphenol A, which exhibited weak ER agonist and partial antiestrogenic activity. Thus, the potencies of both compounds are highly variable and response/species specific, suggesting that a TEF approach would have to incorporate factors that address some measure of response specificity. Interacdons ofEndocrine-Active Compounds Application of a TEF approach assumes additive responses for compounds or mixtures that act through the same pathway at submaximal doses. Arnold and co-workers (96) initially reported that binary mixtures of weakly estrogenic organochlorine pesticides, including dieldrin, chlordane, toxaphene, and endosulfan, exhibited > 90fold and > 160to 1600-fold synergistic ER binding and induction of reporter gene activity in a yeast-based assay, respectively, compared to that observed for the compounds alone. This type of nonadditive interaction would negate a TEF/EQ approach for hazard assessment of these mixtures. However, studies from other laboratories (97,98) using the same compounds have not observed synergism. The paper on synergism was recently withdrawn (99); however, this does not preclude the possibility of other synergistic interactions.

Interacdons between Endocrine Response Pathways
The TEF/EQ approach is most applicable for hazard and risk assessment of a specific class of endocrine active compounds that act through a common receptor. However, as noted previously, there are a number of factors that complicate this approach and the problems are magnified with xenoendocrine active agents that act through steroid hormone receptor or thyroid hormone receptor-mediated pathways. For example, assessment of xenoestrogen exposure and potency (EQs) is complicated by tissue-specific agonist/antagonist activities, lack of data on intake and serum levels, and their relative contribution to total estrogen equivalents compared to much higher intakes of natural estrogenic chemicals in the diet (1). In addition, many compounds may interact with more than one hormone receptor and modulate multiple endocrine response pathways. For example, 2',3',4',5'-tetrachloro-2biphenylol binds to the ER and exhibits ER-agonist activities (100,101). The same compound also binds to the androgen receptor in a yeast-based assay but inhibits dihydrotestosterone-induced reporter gene activity in human Hep G2 liver cancer cells transiently transfected with the human androgen receptor and an androgenresponsive construct (102). Although 2',3',4',5'-tetrachloro-4-biphenylol did not bind the progesterone receptor, in a progesterone-responsive yeast assay this hydroxy-PCB inhibited progesterone receptor-mediated transactivation (103). These data illustrate how one endocrineactive compound can modulate multiple endocrine response pathways.
Another major problem for hazard assessment of xenoestrogens is associated with tissue-specific cross talk between different receptor-mediated pathways, which can lead to significant modulation of estrogen-induced responses. It has been pointed out that in human breast cancer cells, cross talk between the ER-and AhR-signaling pathway results in inhibition of estrogeninduced responses (1). Although this interaction is likely to be cell specific, it is possible that the estrogenic activity associated with xenoestrogens in the mammary gland will be inhibited by both xeno AhR and natural AhR agonists in the diet. Research in this laboratory has also focused on cross talk between the ER and other receptors that bind natural dietary constituents. Vitamin A, retinoids, phytol, and phytanic acid are vitamins or plant degradation products and are also important dietary constituents that act through the retinoic acid and retinoic acid X receptors, and there is cross talk between these receptors and the ER. For example, all transretinoic acid, 9-cis-retinoic acid, and phytol inhibit estrogen-induced responses in breast cancer cells, and any effects of xenoestrogens in these cells would be opposed by retinoic acid and retinoic acid X receptor ligands. The importance of this type of counteractive cross talk in other tissues and organs has not been determined. These are only some examples of cross talk between endocrine-signaling pathways that must be considered in an overall risk assessment of dietary exposure to xenoestrogens and other xenoendocrine active compounds, as well as natural compounds in food that act through the same signaling pathways.

Summary
Humans and wildlife are exposed in the diet to complex mixtures of natural and man-made chemicals. Hazard and risk assessment of these mixtures is a difficult process and the TEF approach has been utilized for several different classes of chemicals, including HAHs (AhR agonists), PAHs (carcinogens), and xenoestrogens (ER agonists). This review has pointed out both the utility and problems associated with the TEF approach for all three classes of chemicals. For example, although TEFs may be useful for regulating HAH emissions and cleanup levels, application of this concept for determining dietary TEQ intakes is complicated by the unknown contributions of naturally occurring AhR agonists, which exhibit both AhR agonist and antagonist activities (2). Moreover, the issue of cross talk between multiple endocrine pathways would further compromise the validity of the TEF approach when applied to dietary intakes of different classes of man-made and natural chemicals. Based on these uncertainties, the TEF approach should be used for limited applications and only after validation in animal models.