Abstract
Following identification of potential for hazard, one of the key aspects in the assessment of health risk of any environmental agent of concern is an assessment of the dose response for effects observed with that agent. This refers not only to how the magnitude of the effect changes with dose, but also the shape of the dose response. This is especially important in extrapolating results from studies in experimental animals to exposures experienced by the human population.
2,3,7,8-Tetrachlorodibenzo-p-dioxin (TCDD) and dioxin-like compounds (DLCs) comprise a large class of structurally related compounds including polychlorinated dibenzo-dioxins, dibenzo-furans and some biphenyls, whose primary mechanism of action is via the aryl hydrocarbon receptor (AhR) (Grassman et al., 1998).
There is a considerable wealth of knowledge on the biological aspects of action of DLCs via the AhR and of the toxicologic consequences of exposure to DLCs in both in vitro cell systems and in vivo in experimental animals (Okey et al., 1994; Martinez et al., 2003). In addition, numerous studies of populations accidentally or occupationally exposed to DLCs have also been conducted and reviewed (IARC, 1997).
While a thorough discussion of these effects is beyond the scope of this paper, these studies show that DLCs are powerful growth dysregulators and depending upon the nature of timing, duration, magnitude of exposure, or experimental model used, lead to a variety of effects including carcinogenicity, immunotoxicity, and reproductive and developmental toxicity (Martinez et al., 2003).
Despite this wealth of data, there is still debate over the specifics of how this data can be used for the assessment of quantitative assessments of human health risk at ambient exposures, or in the establishment of levels of exposure that can be deemed to be without appreciable risk. This uncertainty is related to a lack of understanding of the specific mechanisms that are operational in the development of these toxicities, the concordance of these mechanism between humans and rodents, and also uncertainty on the nature of the shape of the dose response for these effects below the dose range over which empirical observations are made. Invariably the effects seen in experimental animals and most human epidemiologic studies are at exposure levels (both on external/administered dose and body burden basis) that exceed those experienced by the general population. As such, the nature of the dose response at ambient exposure levels has to be inferred from observed effects at high doses.
The nature of the dose response and in particular, the shape of the dose response, can be obtained by mathematically based dose response modeling of observed effects and subsequent interpretation of the model parameters about shape at “low doses”. One of models used most often in dose response modeling is the 4-parameter sigmoidal Hill model. The Hill model provides the foundation for the analysis of non cancer endpoints used in the US Environmental Protection Agency (USEPA) reassessment of the health risk posed by TCDD (McGrath et al., 1995; USEPA, 2001; DeVito et al., 2003).
The Hill model has the mathematical form
where E max is the maximum increase in response over background and ED50 is the dose where response = background + (E max/2). The Hill model is especially useful since the calculated shape parameter can be used to give an indication of the approximate shape of the dose response curve at low doses relative to the ED50.
For a shape parameter = 1 the model becomes essentially a Michaelis-Menton equation
At low doses relative to the ED50 the denominator approaches the ED50 and the model equation approaches a linear form (y=mx+c) where
For shape parameter values <1 the dose-response at low doses is considered supralinear. At shape parameters >1 the dose response at low doses is considered sublinear. Higher shape values exhibit increasingly severe sigmoidicity that can appear threshold-like in nature.
One issue to keep in mind when interpreting such models is that all dose response models, other than simple linear models of the form y=mx+c, are dose response curves (i.e., are non-linear) and when expressed in log space would appear sigmoidal in nature. It is knowledge of where one is on the dose response curve relative to the ED50 that allows interpretation of “low dose” shape.
Another key issue in interpretation of dose response is that it is an apparent dose response that is obtained from a model fit, not the true dose response. The true dose response is never known. Furthermore, the apparent dose response is dependent upon a number of inherent factors and assumptions that determine the model fit and the subsequent interpretation, including: the mathematical form of the model used, number of dose groups, inter dose spacing, location of doses on the dose response curve, variance in the data, and relationship between dose timing and maximum observed effect.
The most comprehensive evaluation on the shape of the dose response curve for TCDD effects was conducted as part of the 2001 USEPA heath assessment. In that analysis, almost 250 noncancer rodent endpoints were evaluated using both Hill and Weibull dose response models (USEPA, 2001; DeVito et al., 2003). The analysis was limited to data sets that had at least 3 dosed groups in addition to controls and represented repeat dose studies, adult single dose studies and single dose developmental studies. Endpoints were representative of known biochemical effects such as increased enzyme induction, pathological effects, immunological effects and toxicological effects. In all models the shape parameter could not be less than 1 (i.e., supralinearity could not be observed) and dose response models were considered to be “non-linear” at low doses if the shape parameter was >1.5. Approximately 40% of the endpoints had a shape parameter >1.5 and hence were considered to be low dose linear in behavior (i.e., 60% were considered to be low dose non-linear). In general this was the case for all types of endpoints under the different dosing scenarios indicating that one could not extrapolate these findings to predict the shape of the curve for a given endpoint. Low dose linearity or non-linearity could equally be expected.
Given that the apparent dose response cannot be predicted based on empirical experience, what can be expected based on the mechanism of action of TCDD and related compounds? TCDD and related compounds are ligands for the AhR, a ligand activated transcription factor whose key mechanism is to activate the transcription of specific responsive genes (Schmidt and Bradfield, 1996). It is believed that all responses associated with exposure to TCDD are primarily due to initial binding to and activation of the AhR, subsequent alterations in expression of TCDD-regulated genes, and altered signaling of biological pathways that interact with the AhR signal transduction mechanism (Poland and Knutson, 1982).
Alterations in expression of dioxin responsive genes occurs via a mechanism that involves a high-affinity interaction of the ligand with the AhR that functions as a ligand-activated transcription factor (Okey et al., 1994; Schmidt and Bradfield, 1996). Ligand binding initiates a signaling pathway in which the cytosolic AhR dissociates from heat shock proteins and translocates to the nucleus. At some point subsequent to ligand binding, the AhR associates with another protein, the aromatic hydrocarbon nuclear translocator protein (ARNT), to form the nuclear DNA-binding and transcriptionally active AhR complex. Both the AhR and ARNT proteins are members of the basic helix-loop-helix (bHLH) family of transcription factors (Hoffman et al., 1991; Burbach et al., 1992; Ema et al., 1992). The AhR-ARNT heterodimer binds with high affinity to a specific DNA sequence termed the dioxin response element (DRE). DREs have been identified in the enhancer regions of genes encoding several drug-metabolizing enzymes. The hallmark response to TCDD is the transcriptional induction of the cytochrome P450 1A1 gene (CYP1A1), which is mediated by binding of the AhR-ARNT complex to DREs present in the 5′ flanking region of the gene. While it has proposed that some effects of TCDD may be due to mechanisms that do not require direct DRE mediated effects, recent transgenic studies show the abrogation of TCDD induced toxic effects when using mice that express a form of the AhR receptor that cannot translocate to the nucleus and activate gene transcription (Bunger et al., 2003).
Given the central role of the AhR in mediating the biochemical and toxic effects of dioxins (Birnbaum, 1994), the dose responses for dioxins is in large part related to the degree of occupancy and activation of this receptor. While the Hill model can be interpreted as relative to receptor binding, simple “receptor occupancy” models for explaining the dose response for TCDD are likely too simplistic since, especially for in vivo studies, it ignores issues of pharmacokinetics and interrelated pharmacodynamics (Portier et al., 1993). One key example of this is the induction of CYP1A1 in the liver. Within the liver of rats exposed to TCDD, the expression of CYP1A1 protein, and also that of CYP1A2 and CYP1B1, exhibit a clear centrilobular pattern of induction with increasing response as a result of increasing dose due to an increase in the number of maximally induced cells in the liver, rather than a graded increase in response in all cells in the liver (Tritscher et al., 1992; Walker et al., 1998). Studies have shown that there is a gradient of TCDD within the liver (Santostefano et al., 1999), as a result of regiospecific induction of CYP1A2, a known TCDD inducible binding protein responsible for the hepatic sequestration of DLCs (Diliberto et al., 1997). In addition. it has been proposed that within each hepatocyte there is regiospecific variation in sensitivity and that there may be a molecular “switch” such that some cells respond near maximally at lower doses than other cells in other regions of the liver (Andersen et al., 1997a; Andersen et al., 1997b; Broccardo et al., 2004; Chubb et al., 2004).
These observations show that the nature of the dose response curve for TCDD and DLC induced endpoints is governed not only by receptor ligand mechanics, but also other physiological signals that impact upon downstream signal transduction and ultimate manifestation of the response. Therefore to fully understand the nature of the dose response for dioxins one has to understand the mechanisms responsible for development of observable responses and the key control systems at the cellular, tissue and organismal levels.
Through the series of papers in this issue we shall explore a variety of the issues pertinent to the interpretation of the dose response for dioxins and other persistent organohalogen compounds and mechanism of action for some toxicities that ultimately control the nature of the dose response.
Footnotes
ACKNOWLEDGMENTS
N. J. W. is supported by the Intramural Research program of the NIH and NIEHS.