An Approach for Assessing Estrogen Receptor-Mediated Interactions in Mixtures of Three Chemicals: A Pilot Study

Grantley D. Charles*,1, C. Gennings{dagger}, Timothy R. Zacharewski{ddagger}, B. Bhaskar Gollapudi* and Edward W. Carney*

* Toxicology and Environmental Research and Consulting, The Dow Chemical Company, 1803 Building, Midland, Michigan 48674; {dagger} Department of Biostatistics, Virginia Commonwealth University, Richmond, Virginia 23298; and {ddagger} Department of Biochemistry and Molecular Biology and National Food Safety and Toxicology Center, Michigan State University, East Lansing, Michigan 48824

Received September 26, 2001; accepted April 8, 2002


    ABSTRACT
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Most studies investigating interactions among endocrine-active chemicals have been limited to binary mixtures. This study reports on the preliminary evaluation an in vitro MCF-7 cell ER-{alpha} reporter gene system, coupled with a statistical methodology adapted for assessing interactions within ternary (3-chemical) mixtures. Two mixtures were initially chosen for assessment of the in vitro system’s ability to detect additivity (mixture A) as well as greater-than-additive (mixture B) responses. Mixture A was composed of 17ß-estradiol (E2), ethinyl estradiol, and diethylstilbestrol and served as a control for additivity, whereas mixture B (E2, epidermal growth factor, insulin-like growth factor-I) was selected to model greater-than-additive interactions based on previous in vitro studies. After generating complete dose–response curves for each chemical, ternary mixtures were then tested in a full factorial design (4 concentrations per chemical, 64 treatment groups). A response surface was estimated using a nonlinear mixed model, and the observed responses were statistically analyzed for departures from the responses expected under the assumption of additivity. Mixture A exhibited additivity in vitro when the chemicals were present at concentrations in the linear range of their individual dose-response curves. For mixture B, in vitro analysis resulted in the additivity hypothesis being rejected (p < 0.001) because of a greater-than-additive interaction, as expected. A limited in vivo evaluation of mixture A was performed in the immature mouse uterotrophic assay (27 treatment groups), which agreed with the in vitro assessment of no significant departure from additivity ( p = 0.903). These findings demonstrate the ability of this in vitro methodology to detect additive, greater-than-additive, and less-than-additive interactions within ternary mixtures, which now allows for the assessment of environmentally relevant mixtures.

Key Words: mixture; endocrine; estrogen receptor; additivity; diethylstilbestrol; 17ß-estradiol; response surface; synergy.


    INTRODUCTION
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Environmental and occupational human exposures to estrogenic chemicals through the diet, drinking water, and air are rarely limited to a single chemical. Although the "environmental estrogens" tend to exhibit weak activity when assessed individually, concerns have been raised that exposure to mixtures of such compounds can act additively or even synergistically to enhance toxicity (Payne et al., 2001Go). This increased interest in the toxicity of chemical mixtures has been reflected in recent amendments to the Food Quality Protection Act (FQPA, 1996Go), which required that the U.S. Environmental Protection Agency consider the "cumulative effects" of exposures to pesticides and other substances thought to possess a "common mechanism of toxicity." The endocrine disruption issue came to the forefront during the same period, resulting in the convening of the U.S. Environmental Protection Agency’s Endocrine Disruptor Screening and Testing Advisory Committee (EDSTAC, 1998Go), whose final report recommended the screening of chemical mixtures, including breast milk, gasoline, and ground water components.

One approach to the assessment of chemical mixtures is the toxic equivalency factor approach, which normalizes the dose of each component of the mixture against that of the most potent compound; the relative potencies are then summed to estimate the toxic potency of the mixture (Safe, 1998Go). Such an approach has also been suggested for use with endocrine active agents under an estrogen equivalency approach (Gaido et al., 1998Go; Soto et al., 1997Go).

Given the complexity of this issue, it is not surprising that research approaches in combination or mixtures toxicology have varied. These include the use of standard safety factor approaches and response surface analyses applied to multicomponent mixtures (Groten et al., 1997Go; Nesnow et al., 1998Go), analyses of binary mixtures using the concepts of response and concentration addition (Kortenkamp and Altenburger, 1998Go; Tully et al., 2000Go), as well as integrated quantitative structure–activity relationships and physiologically based pharmacokinetic-pharmacodynamic modeling for complex mixtures (Verhaar et al., 1997Go). Furthermore, because of the infinite number of potential chemical combinations, it would be impossible to test all of them using empirical approaches (Yang, 1992Go). The majority of previous mixtures studies and their effects on endocrine end points have been restricted to binary mixtures (Arcaro et al., 1998Go; Arnold et al., 1997Go; Ramamoorthy et al., 1997Go; Tully et al., 2000Go). This contrasts with "real-world" mixtures that are much more complex. Furthermore, there have been few attempts to apply rigorous statistical methods to the assessment of the interactions of mixtures of endocrine-active chemicals.

As a step toward developing methods to assess complex mixtures (greater than binary), this report describes the development of an in vitro approach for assessing interactions within ternary mixtures of chemicals that act through the estrogen receptor (ER). This methodology involves the use of a chimeric receptor-reporter gene transactivation system (Zacharewski, 1997Go) utilizing a 43 factorial dosing design (64 dosing groups) (Fig. 1Go). All possible combinations of four concentrations of three chemicals were modeled via a response surface methodology for the identification of interactions among the ternary chemical mixtures (Gennings et al., 2000Go, in pressGo) so as to identify additivity (no interaction) or departures from additivity (less than additivity, greater than additivity) (Fig. 2Go).



View larger version (32K):
[in this window]
[in a new window]
 
FIG. 1. Schematic of the factorial design for the in vitro mixtures studies with examples of chemical combinations in selected grids. X, Y, and Z represent the three different chemicals being evaluated. Concentration level 0 represents vehicle (no chemical), whereas 1, 2, and 3 represent 3 concentrations for each chemical based on individual concentration–response data generated in the MCF-7 reporter assay system.

 


View larger version (18K):
[in this window]
[in a new window]
 
FIG. 2. Possible combination effects of a mixture of 3 chemical agents. When there is no interaction among chemicals A, B, and C, an additivity surface is defined. If a less-than-additive interaction occurs in which more of a given chemical is required to produce the same response (antagonism), then the new surface will be above the expected no-interaction (additivity) surface. By corollary, if a greater-than-additive interaction occurs in which less of a given chemical is required to produce a similar response (synergy), then the new surface will be below the expected no-interaction (additivity) surface.

 
A series of 3 ER-{alpha} agonists was used in combination with the expectation of demonstrating additivity, whereas greater-than-additive interactions were expected in a mixture of 17ß-estradiol and growth factors based on previous reports (Aronica and Katzenellenbogen, 1993Go; Dupont et al., 2000Go; Ignar-Trowbridge et al., 1996Go). By demonstrating the ability to identify additive, less-than-additive, or greater-than-additive interactions in these "control" mixtures, we hope to apply such methods to environmentally relevant mixtures. Such approaches not only can provide information on the nature of chemical interactions but can also serve as an empirical test of whether chemicals share a common receptor-mediated mechanism (Dawson and Poch, 1997Go).


    MATERIALS AND METHODS
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Reagents.
The reagents 17ß-estradiol (E2), epidermal growth factor (EGF), insulin-like growth factor-I (IGF-I), diethylstilbestrol (DES), and ethinyl estradiol (EE) were all obtained from the Sigma Chemical Company (St. Louis, MO), and ICI 182,780 was purchased from Tocris Cookson (Ballwin, MO). All reagents used in the treatment of transfected cultures were dissolved in dimethyl sulfoxide (DMSO) or buffered saline in amber glass vials. Phenol red-free Dulbecco modified Eagle medium (DMEM), fetal bovine serum (FBS), and medium supplements were purchased from Life Technologies (Grand Island, NY). Dextran charcoal-stripped fetal bovine serum (FBS-DCC) was obtained from Hyclone (Logan, UT).

Cell culture.
MCF-7 cells (obtained from Dr. L. Murphy, University of Manitoba, Winnipeg, and originally acquired from Dr. C. M. McGrath, Meyer L. Prentis Cancer Center, Detroit, MI) were maintained in DMEM supplemented with 10% FBS, 2 mmol L-glutamine, 15 mmol HEPES augmented with 50 µg/ml gentamicin, penicillin/streptomycin (100 IU/ml/100 µg/ml), and amphotericin B (2.5 µg/ml). Cells were maintained at 3% CO2 and 95% humidity.

Transfection.
Cells were plated in triplicate in 96-well plates at a density of 6–8 x 103 cells/well in 5% FBS-DCC. After attachment and growth for 6 h, the cells were transfected using LipofectinTM (Life Technologies). To each well we added 50 ng of the ß-galactosidase (ß-gal) expression vector pCH110 (Pharmacia, Piscataway, NJ), 150 ng of 17m5-G-Luc, the Gal4-regulated luciferase reporter vector, and 5 ng of Gal4-HEG0, an ER expression vector (both provided by P. Chambon, INSERM, France). The plasmids were transfected in serum-free, antibiotic-free DMEM supplemented with 2 mmol L-glutamine. Cells were allowed to incubate overnight at 37°C in an humidified atmosphere of 3% CO2/air. Sixteen to 18 h after transfection, the medium was poured from the plate, the plate was inverted, and excess moisture was absorbed by placing the plate on sterile paper towels. The cells were then treated in triplicate with E2/DES/EE (mixture A) or E2/EGF/IGF-1 (mixture B) in 5% FBS-DCC. The pure antiestrogen ICI 182,780 (Wakeling et al., 1991Go) was used to verify that the reporter gene activity was strictly ER mediated because ICI 182,780 was able to completely inhibit responses to both individual and chemical combinations (data not shown). After treatment, wells were washed with phosphate-buffered saline, and 50 µl of lysis buffer (Promega, Madison, WI) was added to each well. Plates were then placed in a –70°C freezer for approximately 1 h. Thawing the plates after freezing at –70°C facilitated cell lysis. Aliquots from each well were divided into two 96-well plates for luciferase and ß-gal activity determination. The reference plasmid pCH110 was cotransfected as an internal control to correct for variations in transfection efficiency. The values presented are units of luciferase activity normalized to the ß-gal activity from individual wells and expressed as fold induction relative to control as the end point. Treatment regimens that resulted in observably reduced ß-gal activity relative to that of transfected cultures exposed to 10 nmol E2 under unaltered media conditions were considered cytotoxic and were not used for further analysis. In all experiments, the final concentration of the solvent (DMSO) did not exceed 0.3% in the culture medium; neither was the osmolality or pH of the dosing solutions changed by ± 6 mOsm/kg H2O or <= 0.06 units, respectively, relative to vehicle-treated controls.

Luciferase and ß-gal activity assays.
For transfected cells, 10 µl of lysate was combined with 100 µl of Luciferase Assay Reagent (Promega, Madison, WI), and luminescence was determined immediately using a Packard Topcount NXTTM luminescence counter (Packard Instrument, Meriden, CT). The ß-gal activity was measured using a chemiluminescent kit (Tropix, Bedford, MA). The ß-gal activity was initiated with 70 µl of galactosidase reaction buffer added to 10 µl of the cell lysate followed by a 30-min room temperature incubation. After the reaction was stopped by the addition of 100 µl of the Accelerator II stop buffer, the chemiluminescence was measured in the same manner as for luciferase.

Experimental Design
Reporter gene assay.
Test chemicals were evaluated in range-finding, reporter gene studies to establish individual chemical dose–response data. These data were used to facilitate the selection of chemical concentrations for use in the interaction studies. For mixtures A and B in the in vitro ER reporter assay, assessment of interactions between chemicals in the mixture was facilitated by the use of a factorial design. This entailed the selection of 4 concentrations of all the chemicals in a given mixture, including zero (vehicle control), in all possible combinations to give a total of 64 dosing groups (see Fig. 1Go). For mixture A (E2/DES/EE) the concentrations used were E2/DES (0, 10-11, 10-10, and 10-9 M) and EE (0, 10-12, 10-11, and 10-10 M). For mixture B (E2/EGF/IGF-I) the concentrations were E2 (0, 10-11, 3 x 10-11, and 10-10 M) and EGF/IGF-I (0, 10-10, 3 x 10-10, and 10-9 M). Each experiment was repeated at least 3 times with independent passages of MCF-7 cells.

Uterotrophic assay.
Litters of postnatal day 10–11 female CD-1 mice (obtained by cross-fostering and shipped with their foster dams) were purchased from Charles River Laboratories (Raleigh, NC). After arrival at the laboratory, the mice were acclimated for 1 week before testing. Litters with their foster dam were housed in separate polycarbonate cages with corncob bedding in rooms in which the relative humidity was maintained within a range of 40–70%. The room temperature was maintained at 22 ± 3°C. A 12-h light-dark photocycle was maintained for all animal rooms with lights on at 6:00 A.M. and off at 6:00 P.M. Room air was exchanged at 12–15 times/h. Mice were provided Purina Rodent Diet/Casein Base, 5K96, a low-phytoestrogen rodent diet in pellet form (a modification of the standard NIH-31 diet), with the soybean and alfalfa meal replaced with casein, and municipal drinking water provided ad libitum during the prestudy and study periods. Before dosing, the mice were weaned and randomized into dosing groups using a procedure designed to equalize groups based on body weights. Foster dams were euthanized by CO2 inhalation at the time of randomization.

An initial probe study was performed to gather individual chemical dose–response data so as to facilitate the setting of dose levels for the mixtures to be used in the main study. Dosing suspensions were prepared by initially dissolving the test material in ethanol and further diluting (1:10) with corn oil (10% ethanol–90% corn oil). Using E2, DES, and EE, 5 mice/group were dosed by oral gavage at the same time every morning for 3 consecutive days at 8 different doses: EE and DES at 0.03, 0.1, 0.3, 1, 3, 10, 30, and 100 µg/kg/day and E2 at 0.1, 0.3, 1, 3, 10, 30, 100, and 300 µg/kg/day. A single control group was treated with the vehicle. All mice were weighed on the day of randomization, on the 3 days of dosing, and on the day of necropsy (day 4). The mice were euthanized on day 4, and for each the uterus was excised and trimmed, the cervix removed, and wet weight determined. Each uterus was then nicked, blotted to remove luminal fluid, and reweighed as a measure of blotted weight.

For the evaluation of the mixture A, 27 groups of 6 mice were dosed in a full factorial study design in which the level of each test chemicals was 0, 0.25, or 2.5 µg/kg/day. These dose levels were chosen on the basis of the individual dose–response curves obtained in probe studies.

Statistical analysis using a nonlinear mixed model.
Further details of the statistical methodologies used in evaluation of the mixtures data from these experiments are described in Gennings et al. (in press). In vitro experiments were performed on 3 separate days with independent passages of MCF-7 cells to satisfy the assumption of independent experiments. Luciferase activity was normalized to ß-gal, that is, (Luc)/(ß-gal). Fold induction was calculated as Luc/ß-gal divided by the average Luc/ß-gal in the DMSO group. Uterotrophic assay data were expressed as fold increase in response relative to the mean of vehicle controls.

The definition of additivity or additive responses used in this analysis is that of Berenbaum (1985)Go and is based on the classical isobolograms for the combination of two chemicals (Loewe and Muischnek, 1926Go; Loewe, 1953Go). That is, in a combination of c chemicals, let Ei represent the concentration/dose of the ith component alone that yields a fixed response, y0, and let xi represent the concentration/dose of the ith component in combination with the c agents that yields the same response. According to this definition of additivity, if the substances combine with zero interaction, then

((1))

If the left-hand side of Equation 1Go is less than 1, then a synergism can be claimed at the combination of interest. If the left-hand side of Equation 1Go is greater than 1, then an antagonism can be claimed at the combination. Because Equation 1Go is the equation of a plane in c dimensions, this definition of additivity implies that under additivity contours of constant response are planar.

For the mixture studies, a 4 x 4 x 4 factorial design was used for three chemicals. Ideally, the location of the concentration levels in such a design should support the "active" part of the response surface. Because this region may not be known a priori, a design based on arithmetic spacing that at least roughly evenly spans the experimental region was preferred instead of log spacing.

A nonlinear model was chosen for the analysis of the mixture data because a sigmoid-shaped relationship was expected. The form of the model was similar to model 2 based on a Gompertz model parameterized as follows:

((2))
where yij is the average (across the triplicates) log fold induction from the jth experimental group and the ith experiment, j = 1,...,d; and i = 1,...,I; {gamma} is the unknown parameter associated with the maximum plateau effect in the experiment; {alpha} and the p-dimensional vector ß are unknown parameters associated with the intercept, slopes, and interaction terms on the complementary log-log scale; xj is a p-dimensional vector denoting the concentration of E2 in the jth group; and {varepsilon}i = [ei1,...,eid] is the unobserved random error vector from the ith experiment assumed to be independent and normally distributed with mean vector zero and variance {Sigma}.

The model given in Equation 2Go can be "linearized" by making the so-called complementary log-log transformation. For example, let x‘ß = ß1x1 + ß2x2. Notice that

((3))

Thus, {alpha} is an intercept and ß1 and ß2 are slopes for x1 and x2 on the complementary log-log scale. Following the logic of Carter et al.(1988), with some algebra, Equation 3Go can be put in the form of Equation 1Go, the definition of additivity. Namely,

where the concentration of each chemical alone that gives the specified mean response E(y)0 is

Now, if x‘ß = ß1x1 + ß2x2 + ß12x1x2, then

((4))

Considering responses above background such that –log(–log(E(y)0/{gamma})) > {alpha}, the denominator of the right-hand side is positive. Using concentration values (i.e., not log concentration) that are also nonnegative, the algebraic sign of the numerator, and hence the entire term, is determined by the sign of ß12. If ß12 is positive, then the right-hand side of Equation 4Go is less than 1, indicating a synergism at the combination of interest. If ß12 is negative, then the right-hand side of Equation 4Go is greater than 1, indicating an antagonism at the combination.

It is important to note that the algebraic relationship between the model and the definition of additivity in Equation 1Go does not hold if the concentration is modeled on the log scale. We follow the general rule that additivity is defined on the concentration scale unless there is reason to be interested in interactions on another scale. Clearly, if additivity holds on the arithmetic concentration scale, it generally will not hold on the log scale.

The model we used to fit the mixture data was given in Equation 2Go, in which

which is referred to as the "full model." Similar logic to that given in Equation 4Go can be used for this case. Here, the algebraic sign of the term on the right-hand side of Equation 4Go depends on 4 parameters: ß12, ß13, ß23,, and ß123. Thus, the model allows for dose-dependent areas of departure from additivity. Using this model, the hypothesis for additivity is that ß12 = ß13 = ß23 = ß123 = 0. A reduced (additivity) model is thus given in Equation 2Go in which x‘ß = ß1x1 + ß2x2 + ß3x3. Let L0 be the log-likelihood value from the additivity model; let L1 be the log-likelihood value from the full model; let p be the number of restricted parameters under additivity (here, p = 4). The likelihood ratio statistic for testing the hypothesis of additivity is given by LR = –2[L0 L1] ~ {chi}p2. Let {chi}p2({alpha}) represent the upper critical value of the chi-square distribution with p degrees of freedom. If LR exceeds {chi}p2({alpha}), then the hypothesis of additivity is rejected.

A nonlinear Gompertz model was chosen for these data because a sigmoid-shaped relationship was expected, and this model can take such a shape without imposing a symmetry function. To account for the intra-experimental relationship of cell growth and responsiveness, observations within an experiment were allowed to be correlated. Data from different experiments were assumed to be independent.


    RESULTS
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Individual dose–response analyses were performed in the ER-{alpha} reporter assay for all 5 chemicals that were consequently evaluated as components of both ternary mixtures. As illustrated in Figure 3AGo, the known estrogen agonists EE and DES generated expected responses in the assay system relative to vehicle and treatment, whereas EGF and IGF-I did not produce activity above background (see Fig. 3BGo). E2/DES at 1 x 10-11, 1 x 10-10, and 1 x 10-9 and EE at 1 x 10-12, 1 x 10-11, and 1 x 10-10 M were selected for use as concentrations that spanned the response range for these chemicals (see Fig. 3Go). For mixture B, E2 at 1 x 10-11, 3 x 10-11, and 1 x 10-10 M was used. While producing apparent proliferative but not increased luciferase responses, EGF and IGF-I were evaluated at 3 concentrations (1 x 10-10, 3 x 10-10, and 1 x 10-9 M) that approximated their previously characterized EC50 values for binding and activation through their respective receptor systems (Neelam et al., 1998Go; Rice and Garner, 1999Go). DMSO (vehicle) was used as a fourth "control" concentration.



View larger version (11K):
[in this window]
[in a new window]
 
FIG. 3. Dose-response curves for the activation of ER-{alpha} transcription by (A) E2, DES, and EE and (B) E2, EGF, and IGF-I. MCF-7 cell cultures were transiently transfected as described under Materials and Methods and incubated with the concentrations shown for 24 h. Results are from a representative experiment and expressed as percentage maximal response (10 nmol E2) ± SD, with each point being the mean of triplicate measurements. Experiments were repeated at least 3 times for all compounds.

 
The methodology used was ultimately designed to estimate interaction effects among mixtures of chemicals. An interaction can be thought of as the change in slope of the dose–response of a chemical in the presence of other chemicals. Of particular concern is that the slope estimates change somewhat from experiment to experiment for individual chemicals. In other words, it was important to determine the difference between a change of slope resulting from an interaction with another chemical from that caused by experiment-to-experiment variability.

The dosing scheme used (see Fig. 1Go) allowed for the calculation of concentration and chemical interaction parameters based on our response surface modeling methodology. The data were presented as the mean of triplicate responses for each treatment group (i.e., average fold induction) and fit into the model to generate the resulting parameter estimates (see Table 1Go) and response surface plots.


View this table:
[in this window]
[in a new window]
 
TABLE 1 Estimated Model Parameters for the Response Surfaces
 
In Figure 4A–DGo, a series of response surface plots for mixture A are shown in which the concentrations of E2 and EE are on the X and Z axes and the level of reporter gene response is shown in the Y axis. Points on the "floor" of the plot represent the individual chemical combinations, and the response from this specific combination can be thought of as supporting the response surface. Because it is impossible to draw in 4 dimensions, it was necessary to plot multiple surfaces showing E2 and EE at various concentrations of DES. An additive response of the chemicals in combination would be associated with a contour of constant response (additivity surface) that would be planar in 3 dimensions (see Fig. 2Go) and linear in 2. Departures from additivity would appear as bends in the additivity surface (3 dimensions; see Fig. 2Go) or line of additivity (2 dimensions; Fig. 5A and 5BGo). If the contour line or additivity surface bows downward toward the origin, then the chemicals would be considered to combine in a greater-than-additive (synergistic) manner and vice versa for a less-than-additive (antagonistic) relationship (see Fig. 2Go). Figure 5AGo is an example of an antagonistic interaction in which the contour lines bow upward away from the axis in contrast to the additive response exhibited by Figure 5BGo in which bending is minimal.



View larger version (52K):
[in this window]
[in a new window]
 
FIG. 4. (A–D) Fitted response surfaces at fixed levels of DES (0, 10-11, 10-10, and 10-9 M) over levels of E2 (0, 10-11, 10-10, 10-9, and M) and EE (0, 10-12, 10-11, and 10-10 M) using the nonlinear mixed model described under Materials and Methods. Each plot shows the interaction of two chemicals (E2 x EE). The base of the graph is a matrix on which all the test combinations of E2 and EE were plotted. The response for each combination is presented on the vertical axis. The effect of the third chemical DES is noted by visualizing the change in the response surfaces from A to D. Data were derived using four concentrations of each chemical in a factorial treatment regimen (64 groups). Analysis represents the average results from at least three replicate experiments.

 


View larger version (19K):
[in this window]
[in a new window]
 
FIG. 5. Two-dimensional contours of constant response plots. Each plot shows the interaction of two chemicals, E2 x EE, at a fixed concentration of the third chemical, DES. (A) The analysis from the full factorial study (64 treatment groups) over the full range of E2/EE concentrations at 10-11 M DES (corresponding to the response surface analysis, Fig.4BGo). (B) The analysis from the reduced factorial study (27 treatment groups) over the full range of E2/EE concentrations at 10-11 M DES (corresponding to the response surface analysis, Fig. 6BGo). Analysis represents the average results from at least three replicate experiments.

 
For mixture A, the response surfaces increased with increasing concentrations of E2 and EE but then bent downward at high levels of both chemicals. This downward response at high levels became more pronounced as the DES concentration increased (see Fig. 4A–DGo). The overall test of additivity based on the F test was rejected (p < 0.001), indicating a statistically significant departure from additivity among the three chemicals (see Table 1Go). Interaction parameters between chemicals (e.g., ß12 interaction between E2 and EE, ß23 interaction between EE and DES) are shown in Table 1Go. The fact that all of the 2-way parameter estimates were negative and significant (p <= 0.001) indicated that the chemicals antagonized each other, producing less-than-additive overall responses. The significant 3-way interaction term indicated that the interaction between any two chemicals changed over levels of the third chemical.

However, because it appeared that the overall departure from additivity was due to an antagonistic interaction among the chemicals at higher concentrations, it was of interest to determine whether these 3 ER-{alpha} agonists would be additive in the lower part of the concentration–response curve. To this end, these same data were reanalyzed in a 33 factorial analysis (27 treatment groups) in which the highest concentration of each chemical was excluded from the original data set. Figure 6A–CGo presents the new fitted response surfaces. The overall test for additivity among the three chemicals in combination was not rejected in this case (p = 0.759), confirming that the chemicals were additive in this region of the concentration–response curve. The linear nature of the contour plot (see Fig. 5BGo, representative of response surface in Fig. 6BGo) was more apparent compared with that of Figure 5AGo (representative of response surface in Fig. 4BGo). Also, the calculated 2-way interaction estimates shown in Table 1Go (e.g., ß12 E2/EE, p = 0.375; ß23 EE/DES, p = 0.813, ß13 E2/DES, p = 0.763) were no longer significant, confirming that none of the chemicals antagonized each other’s responses at these concentrations.



View larger version (27K):
[in this window]
[in a new window]
 
FIG. 6. (A–C) Fitted response surfaces at fixed levels of DES (0, 10-11, and 10-10 M) over levels of E2 (0, 10-11, and 10-10 M) and EE (0, 10-12, and 10-11 M) using the nonlinear mixed model described under Materials and Methods. Each plot shows the interaction of two chemicals (E2 x EE). The base of the graph is a matrix on which all the test combinations of E2 and EE were plotted. The response for each combination is presented on the vertical axis. The effect of the third chemical DES is noted by visualizing the change in the response surfaces from A to C. The analysis used the same data set as in Figure 4Go with the exclusion of the highest concentration from each chemical (27 treatment groups). Analysis represents the average results from at least 3 replicate experiments.

 
Further validation of the model system required evaluation of its ability to detect responses that were greater than additive. Figure 7A–DGo presents the estimated response surfaces of E2 and EGF interactions at various fixed concentrations of IGF-I. By examining the "edges" of the response surface (see Fig. 7AGo), one can see that E2, but not EGF, exhibited a positive assay response. Similarly, IGF-I alone had no activity in the receptor activation assay (see Table 1Go; ß1 [IGF], p = 0.906). However, the lines supporting the response surfaces increase sharply upward along the EGF axis as E2 and EGF in combination increase and then plateau at higher concentrations in contrast to Figure 6Go, in which this upward increase is far less pronounced along the equivalent EE axis. This indicates that the concentration of E2 required to yield a given response was decreased in the presence of EGF. The effect of IGF-I can be assessed by again noting the differences across the 4 plots. In this instance, there was only a slight change in the response surface resulting from the addition of IGF-I.



View larger version (82K):
[in this window]
[in a new window]
 
FIG. 7. (A-D) Fitted response surfaces at fixed levels of IGF-I (0, 10-10, 3 x 10-10, and 10-9 M) over levels of E2 (0, 10-11, 3 x 10-11, and 10-10 M) and EGF (0, 10-10, 3 x 10-10, and 10-9 M) using the nonlinear mixed model described under Materials and Methods. Each plot shows the interaction of two chemicals (E2 x EGF). The base of the graph is a matrix on which all the test combinations of E2 and EGF were plotted. The response for each combination is presented on the vertical axis. The effect of the third chemical IGF-I is noted by visualizing the change in the response surfaces from (A) to (D). Data were derived using four concentrations of each chemical in a factorial treatment regimen (64 groups). Analysis represents the average results from at least three replicate experiments.

 
The hypothesis of additivity for the responses of mixture B was rejected (p < 0.001) (see Table 1Go). Thus, a greater-than-additive interaction within the mixture was concluded from a response-surface analysis of the data. The 2-way interaction term between E2 and EGF (ß12) (see Table 1Go) was positive and significant (p < 0.001), further indicating that EGF increases the response to E2. The interaction of E2 with IGF-I (ß13) did not indicate significant departure from additivity (p = 0.813), implying that the majority of the observed interactions were a consequence of the E2 x EGF interaction.

It was also of interest to determine whether these chemical interactions that had been observed and studied in our in vitro system related to interactions in vivo. The rodent uterotrophic assay was chosen as the basic model for response assessment using mixture A. Individual dose–response data for each chemical were obtained via gavage studies in immature CD-1 mice, as illustrated in Figure 8Go. This figure presents the effect on both wet and blotted weight for the individual chemicals. Doses of 0, 0.25, and 2.5 µg/kg/day were chosen for evaluation in the ternary mixture using a 33 factorial design (27 dosing groups). These doses were based on the individual chemical dose–response data (see Fig. 8Go) with the goal of having individual responses in the low, linear region of the dose–response curves, which, based on the in vitro analyses, would allow for the most readily observable departures from strictly additive responses.



View larger version (13K):
[in this window]
[in a new window]
 
FIG. 8. Dose–response curves for the effect of the ER agonists E2, DES, and EE on uterine (A) wet and (B) blotted weights in immature female CD-1 mice. Mice (5/group) were dosed by oral gavage for 3 consecutive days, and uteri were excised for analysis on day 4. Data points represent the mean ± SD for treatment.

 
Fitted response surface plots of the absolute uterine wet weight are shown in Figure 9A–CGo. It is evident that increasing doses of E2 and EE had a positive effect on absolute uterine wet weight. As explained previously, the effect of DES can be visualized by making comparisons across the 3 graphs. None of the 2- or 3-way interaction parameters (see Table 1Go) were significant: ß12 E2/EE, p = 0.920; ß23 EE/DES, p = 0.547; ß13 E2/DES, p = 0.372; and ß123 E2/DES/EE, p = 0.634. Consequently, the overall test of additivity was not rejected (p = 0.903), indicating an additive relationship among the chemicals at the doses tested in the assay. This analysis was repeated using absolute blotted weight as well as wet and blotted uterine weights normalized to body weight data. In all cases (data not shown), the analysis of the results were virtually identical to those using absolute uterine wet weight.



View larger version (28K):
[in this window]
[in a new window]
 
FIG. 9. (A–C) Fitted response surfaces at fixed levels of DES (0, 0.25, or 2.5 µg/kg/day) over levels of E2 and EE (0, 0.25, or 2.5 µg/kg/day) using the nonlinear mixed model described under Materials and Methods. Each plot shows the interaction of two chemicals (E2 x EE). The base of the graph is a matrix on which all the test combinations of E2 and EE were plotted. The response for each combination is presented on the vertical axis. The effect of the third chemical DES is noted by visualizing the change in the response surfaces from A to D. The analysis used in vivo uterotrophic assay data derived from treatment of each chemical at 3 concentrations in a factorial design (27 treatment groups). Mice (6/group) were dosed by oral gavage for 3 consecutive days, and uteri were excised for analysis on day 4.

 

    DISCUSSION
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Preliminary evaluation of this approach to assess ternary mixtures required a demonstration of its ability to detect additivity and departures from additivity. Mixture A, containing three ER agonists (E2, DES, and EE), was chosen to test the model for its ability to detect additivity. Mixture B, containing E2, EGF, and IGF-I, was formulated to verify an expected greater-than-additive response (synergy). The effect of a well-characterized ER antagonist (ICI 182,780) on the response to E2 as part of a binary mixture was also investigated previously (unpublished data).

The results with mixture A indicated that the ER agonists interacted in an additive manner when the chemicals were present at levels ranging from just above their individual response thresholds through most of their linear response range. However, a less-than-additive or antagonistic interaction was observed at higher concentrations (generally in the upper half of the individual chemical dose–response range). It can be speculated that this was the consequence of receptor saturation at high agonist concentrations.

Because of the greater complexity of whole animal models, there is assumed to be a far greater range of possible chemical interactions. Hence, the mixture of E2/DES/EE was also evaluated in an in vivo uterotrophic assay system to assess the correlation with the in vitro data. The analysis of the 3 ER agonists in vivo resulted in an additive interaction between the mixture components similar to that found in vitro. Furthermore, this interaction held regardless of whether absolute or relative uterine wet or blotted weights were used in the analysis.

EGF and IGF-I have previously been shown to interact synergistically with E2 in binary mixtures (Aronica and Katzenellenbogen, 1993Go; Dupont et al., 2000Go; Ignar-Trowbridge et al., 1996Go). These interactions are thought to be due to cross talk mediated through integrated signaling pathways involving cyclic adenosine monophosphate, protein kinases, and receptor phosphorylation (Apostolakis et al., 2000Go; Kahlert et al., 2000Go; Nelson et al., 1991Go). The resulting analysis of the data from our model confirmed a greater-than-additive interaction in the overall mixture, which was mostly a consequence, in this case, of the interaction between E2 and EGF.

However, the assay system used in these studies consisted of a chimeric Gal4 receptor-reporter system lacking a functional AF-1 domain (Zacharewski et al., 1995Go). It has been demonstrated that the transcriptional activity of ER-{alpha} can be mediated by two activation functions, AF-1 and AF-2, located at the amino and carboxyl termini, respectively (Tzukerman et al., 1994Go), and furthermore, the ER-dependent transcription by IGF-I is predominantly mediated through AF-I (Ignar-Trowbridge et al., 1996Go). It should be noted that both EGF and IGF-I produced apparent proliferation in the assay as evidenced by 2- to 3-fold increases in basal ß-gal activity relative to control cultures (data not shown), indicating that they were bioactive in the assay system. The lack of interaction between IGF-I and E2 may have been due to the lower concentration of IGF-I relative to that used previously in the characterization of synergistic interactions (Aronica and Katzenellenbogen, 1993Go). Thus, the possibility cannot be excluded that a different promoter or receptor construct than the one used here would have been necessary to detect an interaction between IGF-I and E2. This chimeric system has been shown to exhibit differential responsiveness in cell lines other than MCF-7 (Connor et al., 1996Go). MCF-7 cells were used in this study because of their previously characterized responsiveness to this system (Balaguer et al., 1996Go), and the chimeric construct does not exhibit different binding affinities or ligand specificities relative to the native ER (Kumar et al., 1987Go) and proven to be less sensitive to serum-borne estrogen in media (Zacharewski et al., 1995Go).

It should be stressed that the test materials and test concentrations used in this study were chosen to demonstrate the ability of the experimental system and associated statistical methods to detect additivity and less-than additivity (antagonism) and greater-than-additive interactions. All three types of responses were, in fact, generated by the experimental model and were detected statistically.

The results of the current investigation also indicated that there can be multiple types of interaction among the components of a given mixture and that the nature of the interaction depends on the region of the response curve from which doses were derived. Hence, interactions should not be expected to be globally uniform across the dose–response spectrum. Therefore, studies addressing environmentally relevant mixtures need to be tested at environmentally relevant concentrations because high-dose interactions are likely to differ from interactions at low (particularly subthreshold) exposure levels. Therefore, proper design and interpretation of component mixture interactions requires knowledge of the individual chemical dose responses and appropriate selection of dose levels.

In conclusion, the data demonstrated the ability of an ER-{alpha} reporter gene system, coupled with response surface statistical methodology, to detect additivity and less-than-additive (antagonistic) and greater-than-additive interactions in ternary mixtures. Such methods are now available for the assessment of environmentally relevant mixtures.


    ACKNOWLEDGMENTS
 
Support for this research was provided jointly by the American Chemistry Council (Reference No. 9010) and The Dow Chemical Company, and their financial assistance is gratefully acknowledged. Tim Zacharewski is partially supported by the Michigan Agricultural Experimental Station. We thank Dr. Pierre Chambon (INSERM) for allowing the use of the ER and luciferase reporter constructs and Dr. L. Murphy for use of the MCF-7 cell line. We also thank Dr. Belen Tornesi and Ms. Ann Linscombe (Dow Chemical) for the excellent technical assistance provided during the course of these studies.


    NOTES
 
1 To whom correspondence should be addressed. Fax: (989) 638-9863. E-mail: gdcharles{at}dow.com. Back


    REFERENCES
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Apostolakis, E. M., Garai, J., Lohmann, J. E., Clark, J. H., and O’Malley, B.W. (2000). Epidermal growth factor activates reproductive behavior independent of ovarian steroids in female rodents. Mol. Endocrinol. 14, 1086–1098.[Abstract/Free Full Text]

Arcaro, K. F., Vakharia, D. D., Yang, Y., and Gierthy, J. F. (1998). Lack of synergy by mixtures of weakly estrogenic hydroxylated polychlorinated biphenyls and pesticides. Environ. Health Perspect. 106(Suppl. 4), 1041–1046.[ISI][Medline]

Arnold, S. F., Vonier, P. M., Collins, B. M., Klotz, D. M., Guillette L. J., Jr., and McLachlan, J. A. (1997). In vitro synergistic interaction of alligator and human estrogen receptors with combinations of environmental chemicals. Environ. Health Perspect. 105(Suppl. 3), 615–618.[ISI][Medline]

Aronica, S. M., and Katzenellenbogen, B. S. (1993). Stimulation of estrogen receptor-mediated transcription and alteration in the phosphorylation state of the rat uterine estrogen receptor by estrogen, cyclic adenosine monophosphate, and insulin-like growth factor-I. Mol. Endocrinol. 7, 743–752.[Abstract]

Balaguer, P., Joyeux, A., Denison, M. S., Vincent, R., Gillesby, B. E., and Zacharewski, T. (1996). Assessing the estrogenic and dioxin-like activities of chemicals and complex mixtures using in vitro recombinant receptor-reporter gene assays. Can. J. Physiol. Pharmacol. 74, 216–222.[ISI][Medline]

Berenbaum, M. C. (1985). The expected effect of a combination of agents: The general solution. J. Theor. Biol. 114, 413–431.[ISI][Medline]

Berenbaum, M. C. (1989). What is synergy? Pharmacol. Rev. 41, 93–141.[ISI][Medline]

Carter, W. H., and Carchman, R. A. (1988). Mathematical and biostatistical methods for designing and analyzing complex chemical interactions. Fundam. Appl. Toxicol. 10, 590–595.[ISI][Medline]

Connor, K., Howell, J., Chen, I., Liu, H., Berhane, K., Sciarretta, C., Safe, S., and Zacharewski, T. (1996). Failure of chloro-s-triazine-derived compounds to induce estrogen receptor-mediated responses in vivo and in vitro. Fundam. Appl. Toxicol. 30, 93–101.[ISI][Medline]

Dawson, D. A., and Poch, G. (1997). A method for evaluating mechanistic aspects of chemical mixture toxicity. Toxicol. Methods 7, 267–278.[ISI]

Dupont, J., Karas, M., and LeRoith, D. (2000). The potentiation of estrogen on insulin-like growth factor I action in MCF-7 human breast cancer cells includes cell cycle components. J. Biol. Chem. 275, 35893–35901.[Abstract/Free Full Text]

EDSTAC (1998). Endocrine Disruptor Screening and Testing Advisory Committee final report, August 1998. 63 Fed. Regist. 63(248), 71541–71568. Available at http://www.epa.gov/scipoly/oscpendo/history/finalrpt.htm.

FQPA (1996). Food Quality Protection Act, Public law 104–170, 104th Congress, August 3, 1996.

Gaido, K., Dohme, L., Wang, F., Chen, I., Blankvoort, B., Ramamoorthy, K., and Safe, S. (1998). Comparative estrogenic activity of wine extracts and organochlorine pesticide residues in food. Environ. Health Perspect. 106(Suppl. 6), 1347–1351.[ISI][Medline]

Gennings, C., Charles, G. D., Gollapudi, B. B., Zacharewski, T. R., and Carney, E. W. (2000). Analysis of a mixture of estrogen agonists in an ER-{alpha} reporter gene assay. Toxicologist 54(1), 224.

Gennings, C., Charles, G. D., Gollapudi, B. B., Zacharewski, T. R., and Carney, E. W. (in press). Analysis of resulting data from estrogen receptor reporter gene assays. J. Agric. Biol. Environ. Biostat.

Groten, J. P., Schoen, E. D., van Bladeren, P. J., Kuper, C. F., van Zorge, J. A., and Feron, V. J. (1997). Subacute toxicity of a mixture of nine chemicals in rats: Detecting interactive effects with a fractionated two-level factorial design. Fundam. Appl. Toxicol. 36, 15–29.[ISI][Medline]

Ignar-Trowbridge, D. M., Pimentel, M., Parker, M. G., McLachlan, J. A., and Korach, K. S. (1996). Peptide growth factor cross-talk with the estrogen receptor requires the A/B domain and occurs independently of protein kinase C or estradiol. Endocrinology 137, 1735–1744.[Abstract]

Kahlert, S., Nuedling, S., van Eickels, M., Vetter, H., Meyer, R., and Grohe, C. (2000). Estrogen receptor alpha rapidly activates the IGF-1 receptor pathway. J. Biol. Chem. 275, 18447–18453.[Abstract/Free Full Text]

Kortenkamp, A., and Altenburger, R. (1998). Synergisms with mixtures of xenoestrogens: A reevaluation using the method of isoboles. Sci. Total Environ. 221, 59–73.[ISI][Medline]

Kumar, V., Green, S., Stack, G., Berry, M., Jin, J. R., and Chambon, P. (1987). Functional domains of the human estrogen receptor. Cell 51, 941–51.[ISI][Medline]

Loewe, S. (1953). The problem of synergism and antagonism of combined drugs. Arzneimittleforshung 3, 285–290.

Loewe, S., and Muischnek, H. (1926). Uber kombinationswirkunger. I. Mitteilung: Hiltsmittel der fragstellung. Naunyn-Schmiedebergs. Arch. Pharmacol. 114, 313–326.

Neelam, B., Richter, A., Chamberlin, S. G., Puddicombe, S. M., Wood, L., Murray, M. B., Nandagopal, K., Niyogi, S. K., and Davies, D. E. (1998). Structure-function studies of ligand-induced epidermal growth factor receptor dimerization. Biochemistry 37, 4884–4891.

Nelson, K. G., Takahashi, T., Bossert, N. L., Walmer, D. K., and McLachlan, J. A. (1991). Epidermal growth factor replaces estrogen in the stimulation of female genital-tract growth and differentiation. Proc. Natl. Acad. Sci. U.S.A. 88, 21–25.

Nesnow, S., Mass, M. J., Ross, J. A., Galati, A. J., Lambert, G. R., Gennings, C., Carter, W. H., Jr., and Stoner, G. D. (1998). Lung tumorigenic interactions in strain A/J mice of five environmental polycyclic aromatic hydrocarbons. Environ. Health Perspect. 106(Suppl. 6), 1337–1346.

Payne, J., Scholze, M., and Kortenkamp, A. (2001). Mixtures of four organochlorines enhance human breast cancer cell proliferation. Environ. Health Perspect. 109, 391–397.

Ramamoorthy, K., Wang, F., Chen, I.-C., Norris, J. D., McDonnell, D. P., Leonard, L. S., Gaido, K. W., Bocchinfuso, W. P., Korach, K. S., and Safe, S. (1997). Estrogenic activity of a dieldrin/toxaphene mixture in the mouse uterus, MCF-7 human breast cancer cells, and yeast-based estrogen receptor assays: No apparent synergism. Endocrinology 138, 1520–1527.

Rice, K. M., and Garner, C. W. (1999). IGF-I regulates IRS-1 expression in 3T3–L1 adipocytes. Biochem. Biophys. Res. Commun. 255, 614–617.

Safe, S. H. (1998). Hazard and risk assessment of chemical mixtures using the toxic equivalency factor approach. Environ. Health Persp. 106(Suppl. 4), 1051–1058.

Soto, A. M., Fernandez, M. F., Luizzi, M. F., Oles Karasko, A. S., and Sonnenschein, C. (1997). Developing a marker of exposure to xenoestrogen mixtures in human serum. Environ Health Perspect. 105 (Suppl. 3), 647–654.

Tully, D. B., Cox, V. T., Mumtaz, M. M., Davis, V. L., and Chapin, R. E. (2000). Six high-priority organochlorine pesticides, either singly or in combination, are nonestrogenic in transfected HeLa cells. Reprod. Toxicol. 14, 95–102.

Tzukerman, M. T., Esty, A., Santiso-Mere, D., Danielian, P., Parker, M. G., Stein, R. B., Pike, J. W., and McDonnell, D. P. (1994). Human estrogen receptor transactivational capacity is determined by both cellular and promoter context and mediated by two functionally distinct intramolecular regions. Mol. Endocrinol. 8, 21–30.

Verhaar, H. J., Morroni, J. R., Reardon, K. F., Hays, S. M., Gaver, D. P., Jr., Carpenter, R. L., and Yang, R. S. (1997). A proposed approach to study the toxicology of complex mixtures of petroleum products: The integrated use of QSAR, lumping analysis and PBPK/PD modeling. Environ. Health Perspect. 105(Suppl. 1), 179–195.

Wakeling, A. E., Dukes, M., and Bowler, J. (1991). A potent specific pure antiestrogen with clinical potential. Cancer Res. 51, 3867–3873.

Yang, R. S. H. (1992). Strategy for studying health effects of pesticide/fertilizer mixtures in groundwater. Environ. Contam. Toxicol. 127, 1–22.

Zacharewski, T. (1997). In vitro bioassays for assessing estrogenic substances. Environ. Sci. Technol. 31, 613–623.

Zacharewski, T., Berhane, K., Gillesby, B. E., and Burnison, B. K. (1995). Detection of estrogen- and dioxin-like activity in pulp and paper mill black liquor and effluent using in vitro recombinant receptor/reporter gene assays. Environ. Sci. Technol. 29, 2140–2146.