Lung tumorigenic interactions in strain A/J mice of five environmental polycyclic aromatic hydrocarbons.

The binary, ternary, quaternary, and quintary interactions of a five-component mixture of carcinogenic environmental polycyclic aromatic hydrocarbons (PAHs) using response surface analyses are described. Initially, lung tumor dose-response curves in strain A/J mice for each of the individual PAHs benzo[a]pyrene (B[a]P), benzo[b]fluoranthene (B[b]F), dibenz[a,h]anthracene (DBA), 5-methylchrysene (5MC), and cyclopenta[cd]pyrene (CPP) were obtained. From these data, doses were selected for the quintary mixture study based on toxicity, survival, range of response, and predicted tumor yields. The ratios of doses among PAHs were designed to simulate PAH ratios found in environmental air and combustion samples. Quintary mixtures of B[a]P, B[b]F, DBA, 5MC, and CPP were administered to male strain A/J mice in a 2(5) factorial 32-dose group dosing scheme (combinations of five PAHs each at either high or low doses) and lung adenomas were scored. Comparison of observed lung adenoma formation with that expected from additivity identified both greater than additive and less than additive interactions that were dose related i.e., greater than additive at lower doses and less than additive at higher doses. To identify specific interactions, a response surface analysis using response addition was applied to the tumor data. This response surface model contained five dose, ten binary, ten ternary, five quaternary, and one quintary parameter. This analysis produced statistically significant values of 16 parameters. The model and model parameters were evaluated by estimating the dose-response relationships for each of the five PAHs. The predicted dose-response curves for all five PAHs indicated a good estimation. The binary interaction functions were dominated for the most part by DBA and were inhibitory. The response surface model predicted, to a significant degree, the observed lung tumorigenic responses of the quintary mixtures. These data suggest that although interactions between PAHs do occur, they are limited in extent.

carcinogens exhibit all three of these effects depending on the chemicals, route of administration, sex, species, and target organ. The majority of these interaction studies have been performed using only two administered agents and a database of binary carcinogen interactions has been reported (4,5).
Polycyclic aromatic hydrocarbons (PAHs) are a pervasive class of environmental pollutant formed by the incomplete combustion of organic materials. Humans are exposed to PAHs from cigarette smoke, combustion products from gasoline, diesel fuel, coal, and oil, as well as from broiled and smoked foods (6). Many PAHs are carcinogenic in experimental animals and several PAH-containing mixtures (i.e., coke oven emissions, cigarette smoke, and coal tars) are human carcinogens (7)(8)(9). Although there have been a number of studies of interactions of PAHs within binary mixtures of PAHs, little work has focused on larger component mixtures and using lung tissues as tumor targets.
We selected five environmental PAHs to construct quintary (five-component) mixtures  [cd]pyrene (CPP) were selected for the extent and pervasiveness of their environmental occurrence, structural diversity, metabolic diversity, and range of tumorigenic potency. Using the strain A/lJ mouse lung as a target organ, we sought to answer the following questions: Are the mouse lung tumorigenic activities of these five PAHs additive? What is the extent of the deviation from additivity? Can specific interaction parameters be calculated? What is the effect of a nontumorigenic PAH on the tumorigenic activity of a quintary mixture of PAHs?
Methods for analysis of interactions have been reviewed including isobolographic analyses (10), interaction indexes (11), and response surface approaches (12). Response surface approaches have found use in identifying and quantitating chemical and drug interactions (13)(14)(15)(16). Response surface methods use regression techniques that are both descriptive and predictive and are not limited to number of independent variables or same mechanisms of action or dose-response slopes. We utilized a response surface methodology, using response addition, to both design and analyze a study that sought to identify and quantitate the interactions among five environmental PAH lung tumorigens. Using a 25 factorial experimental design (five PAHs at two doses each) strain A/J mice were treated with a series of quintary PAH mixtures and lung adenomas were enumerated after 8 months.
We report that PAH-induced lung adenoma formation in strain A/J mice can exhibit both greater than additive and less than additive interactions that are dose related. Interaction analyses by maximum likelihood methods using a response surface model identified statistically significant binary, ternary, and quaternary interactions.

Materials and Methods
Chemicals B[b]F (99%), DBA (97%), urethane (99%), and pyrene were purchased from Aldrich Chemical Co. (Milwaukee, Wisconsin) and B[a]P(298%) from Sigma Chemical Co. (St. Louis, Missouri). CPP (99%) was obtained from A. Gold (University of North Carolina, Chapel Hill, North Carolina) and 5MC (99%) from S. Amin (American Health Foundation, Valhalla, New York). Tricaprylin was purchased from Eastman Kodak (Rochester, New York). The crude pyrene was purified by column chromatography in hexane using silica gel, recrystallized from hexane, and sublimed in vacuo at 170°C to give a melting point of 149 to 151°C. The reported melting point was 149.6 to 1 50.3°C (17). Liquid chromatographymass spectrometry indicated this product was 99.7% pure.

Tumor Studies
Male strain A/J mice 6 to 8 weeks of age were obtained from Jackson Laboratories (Bar Harbor, Maine). Mice were housed in laminar flow rooms in groups of four in polycarbonate cages. Mice were maintained under standard conditions (20 ± 2°C; 50 ± 10% relative humidity; 12-hr light/dark cycle) and received food and water ad libitum. In the first study, individual PAHs were administered to male strain A/J mice at several doses. On the day of treatment, PAHs were sonicated in tricaprylin until complete solution was achieved, then mice were injected ip (0.2 ml/mouse). Urethaneand tricaprylin-treated mice served as positive and negative controls, respectively. After 8 months, all mice were sacrificed by cervical dislocation, the lungs removed, fixed in 10% neutral buffered formalin, and the surface tumors counted. No detailed histopathology was performed, as previous studies have identified these lesions as adenomas (18). In the second study (mixture study), a 25 factorial experimental design (five agents each at two doses) was used. Thirty-two groups of mice were randomized (20 per group) and dosing was performed in the order of the randomized group number. PAH mixtures were prepared by transferring individually weighed PAHs into each of 32 vials to construct the 32 dosing vials of quintary mixtures of PAHs. Quality assurance analysis by HPLC verified that the target doses were achieved (data not shown). Animals were treated and tumors were scored as described previously. In the second study smaller numbers of animals (10 mice per dose) were treated with the high dose of each PAH as indicator controls. Animal care and treatment were conducted in accordance with the guidelines established in the Guidefor the Care and Use ofLaboratory Animals (19). All animals were treated humanely with due consideration to the alleviation of distress and discomfort.

Dose Responses ofIndividual Polycycic Aromatic Hydrocarbons
Dose-response studies were performed with each of the five PAHs at dose ranges that resulted in a survival of 75 to 100% (Table 1). Statistical analysis indicated that 15 of the 22 groups were significantly different (p< 0.01) from the tricaprylin control group using a Bonferroni multiple comparison test on the square root transformed tumor response data. Both positive (urethane) and vehicle (tricaprylin) controls for lung adenoma response were in agreement with historical data (18).

Selecting Dose Levels for Quintary
Mixtures ofPolycyclic Aromatic Hydrocarbons Dose levels were selected that would satisfy the following: < 25% mortality; < 1 0% reduction in weight gain at the end of the study; a predicted range of tumor response between 2 and 100 lung adenomas per mouse; and the ability to observe an overall 2-fold greater than additive and a 4-fold less than additive tumor response. In addition, the PAH dose levels were prepared in ratios similar to those found in environmental air and combustion samples. In cigarette smoke (9), coal gasification emissions (21), ambient air (22), coke oven emissions (23), gasoline exhaust, and diesel exhaust (24) (22), gasoline exhaust, and diesel exhaust (24). 5MC has been detected in gasoline exhaust at approximately 6 to 10% that of B[b]F or B[a]P (24). Therefore, based on the doseresponse data in Table 1 (12,13). This resulted in 32 PAH mixture groups representing combinations of five PAHs at either high or low dose ( Table 2). This dose scheme would allow the calculation of five PAH dose parameters, ten binary interaction parameters, ten ternary interaction parameters, five quaternary interaction parameters, and one quintary parameter.

Analyses ofPolycyclic Aromatic Hydrocarbon Mixtue Tumor Data
Survival of mice in the 32 PAH mixture groups ranged from 70 to 100%, with a median of 85% and a mean (± SD) of 84.8 ± 9.1 ( Table 2). No dose dependency could be established between survival and doses of PAHs with the dose range tested. The mean body weights for the highest dosed group compared to tricaprylin control indicated a significant loss of weight at day 7 with a recovery to control values on day 14 and beyond (data not  (20). The relationship between the observed lung adenomas per mouse and the expected lung adenomas per mouse based on additive responses for each of the 32 PAH mixture groups indicated that many of the observed data points deviated from those expected based on additivity ( Figure 1). Two groups exhibited a statistically significant increase in expected tumor responses and 13 groups exhibited a significant decrease (p< 0.05) by the Students-Newman-Keuls multiple comparison test on the square root transformed data. Regressing the data to a linear function (R2 =0 .679) gave major deviations from the expected slope of unity (calculated slope = 1.70) and the expected y-intercept of zero (calculated y-intercept = -13.53  (26) was fit to the data parameterized according to Equation 1 (Appendix). The variance of the response was assumed to be of the form 0,u (i.e., a constant times the mean). The method of maximum likelihood was used to estimate the unknown model parameters using a ridge-stabilized Newton-Raphson algorithm (PROc GENMOD, SAS version 6.09). A power link function was used for g(,u) in Equation 1 with the best power estimate determined from a plot of the log likelihood versus the power parameter ( Figure 2). The peak of this plot was approximately 0.5. Therefore, the square root transformation of the response data (lung adenomas per mouse) was used in the subsequent analysis of the data. To determine the shapes of the dose-response curves using the square root transformation for each individual PAH, the tumor data in  (Table  3). All of the linear terms were highly significant (p< 0.00 1). Furthermore, 10 of the interaction terms were significantly different from zero at the 5% significance level. The five significant binary interac- Following the definition of additivity as given by Berenbaum (11) Table 1 Table 3 and compared to the observed data found in Table 1 (Figure 3). There was excellent agreement between the two data sets. The dose-response data for each of the individual PAHs (Table 1), although used in combination with the mixture tumor data to derive the model parameters, represent less than 15% of the total data set and therefore were not expected to dominate these predictions.

Prediction by the Response Surface
Model ofthe Tumorigenicity ofthe Quintary Mies Using Equation 1 and the parameters found in Table 3, the predicted response for each quintary mixture was estimated and compared to the observed responses ( Table 2) for each of the 32 quintary mixture groups (Figure 4). Even though only 10 Table 3 (i.e., Ao, A, /,P/,3, P4, f5).
The data points represent means ± SD of observed lung adenomas/mouse from Table 1. tumorigenic at doses between 10 and 200 mg/kg (Table 4). A similar model to that described in Equation 1 was used to fit a subset of the mixture data and the pyrene data. The subset consisted of the tricaprylin control group, the group exposed to the five PAHs, the group solely exposed to pyrene, and the group simultaneously exposed to the five PAHs and pyrene. Thus the data consisted of a 2 x 2 design and was analyzed accordingly. The effect of pyrene was not different from background (p= 0.614), the effect of the five-PAH mixture was to increase the numbers of lung adenomas from background (p< 0.00 1), and the interaction of the five-PAH mixture and pyrene was negative and significant (p= 0.007). Therefore, pyrene exerted a Observed mean lung adenomas/mouse  Table 3. Sixteen of the 32 parameters used in these calculations were significant at p < 0.05.
35% reduction in the lung tumorigenicity of the quintary mixture.

Discussion
Polycyclic aromatic hydrocarbons are ubiquitous environmental contaminants found in the air, soil, and water, and in hazardous waste sites. Since their discovery as carcinogens in 1915 (27), their toxicologic effects have been studied intensively. Many PAHs are carcinogenic in multiple species (28) and are suspected carcinogens in humans (29), which also makes them an important chemical class from a public health standpoint. The epidemiologic data on PAH-containing mixtures strongly suggests that they are human respiratory carcinogens. The experimental animal data also point to lung, subcutaneous tissue, mammary tissue, and liver (in newborn and juvenile rodents) as targets of PAHs by various routes of administration (ip, intramammary, intrapleural, oral, inhalation, dermal, and iv). Because humans are exposed to mixtures of PAHs, it is important to understand the interactions among PAHs in these mixtures to assess their risk to humans. Statistical methodology is available that allows the determination of specific interactions between groups of toxicologic and pharmacologic agents using response surface methods (12,30,31). To determine some of these potential PAH interactions, a study was constructed using five environmentally relevant PAHs administered as quintary mixtures to strain A/J mice with lung adenoma formation as the toxicologic outcome. The PAHs selected were B[a]P, B[b]F, DBA, 5MC, and CPP, based on their environmental occurrence, range of tumorigenic activities (32)(33)(34)(35)(36)(37), structural features (methylated vs nonmethylated, condensed vs linear, alternant vs nonalternant), and a diversity of routes of metabolic activation (38)(39)(40)(41)(42)(43)(44)(45)(46).
The strain A/J mouse system is a medium-term tumorigenesis bioassay where tumors can be detected and quantitated 8 months after treatment. This mouse carries a lung cancer susceptibility gene or genes that have not yet been identified (18). Studies have shown that lung adenomas in this mouse will progress to adenocarcinomas after 18 to 24 months, with some metastasis. In addition, alveologenic carcinomas in humans are similar in morphology to adenocarcinomas in mouse lung (18). Finally, studies have shown that lung tumors in strain A/J mice produced by PAHs exhibit high proportions of Ki-ras mutations with mutation spectra different from the spontaneous controls (47)(48)(49)(50). From these facts, we conclude that tumor formation in the strain A/J mouse has some relevance in the study of human lung cancer.
The results of these investigations indicate significant deviations from additivity that were both greater than additive and less than additive. The extent of these deviations ranged from +97.4 to -55% of that expected from additivity. Significant deviations were observed that were dose related i.e., lower doses, greater than additive; higher doses, less than additive. However, less than additive interactions dominated under most mixture conditions. Response surface modeling using multilinear regression techniques identified 6 statistically significant dose parameters and 10 significant interaction parameters of 32 possible parameters. The model and the estimated model parameters predicted the observed responses as well as the individual PAH dose-response curves. In additional studies we examined the need of all of the interaction parameters and found that significant fits to the data could not be obtained with fewer than the full complement of 26 interaction parameters (data not shown).
An analysis of the binary carcinogen interaction literature that encompasses multiple species, organs, and routes of administration has identified both greater than additive and less than additive effects for PAH-PAH interactions depending on target tissue species and route of administration (4). In single subcutaneous injection studies in female NMRI mice, Pfeiffer (37)  There are a number of seemingly conflicting reports on the interactive effects of PAH, either in binary mixtures or in combination with complex mixtures containing PAHs. For example, B[a]P and CPP exert a greater than additive effect toward the induction of mouse skin papillomas (4), whereas this mouse lung study identified a less than additive interaction.
Similarly, B[a]P and pyrene induce a greater than additive effect in papilloma formation in mouse skin tumor initiation studies (36), whereas pyrene, which induces neither mouse skin tumors (36) (36). Certainly, mixture composition, target tissues, species, strain, sex, and route of administration must play a role in the tumor outcome. Moreover, carcinogenesis is a multistage process that can involve absorption, distribution, metabolism, detoxification, elimination, macromolecular damage, mutation and/or chromosomal damage, DNA repair, altered gene expression, cytotoxicity, tissue injury, cell proliferation, and apoptosis. Many of these processes can be altered by enzyme induction and inhibition (3). Gibb and Chen (52) suggested that in the multistage model, a multiplicative effect of two or more carcinogens is consistent where each carcinogen acts on a different stage, whereas additivity occurs when each carcinogen acts on the same stage. Synergism has also been defined as occuring when the rate-limiting step in the generation of a single type of tumor differs for each of the two interacting carcinogens (53).
One approach to teasing out the dominant factors is to examine each separately. Future studies include examining quantitative DNA adduct formation, persistence, and repair over time and comparing the extent of DNA adducts formed by the quintary mixtures with the levels of adducts expected from additivity for each of the PAHs. Mixtures of PAHs enhance and inhibit covalent DNA binding (54). These studies are currently in progress.
In conclusion, a response surface model has identified a number of PAH-PAH interactions in a quintary mixture that accurately accounted for all of the observed responses. The observation of greater than additive responses from lower exposures is significant. However, because the magnitudes of all of the interactions were relatively small, these data suggest that although interactions of PAHs do occur, they are limited in extent.
DISCLAIMER: This manuscript has been reviewed by the National Health and Environmental Effects Research Laboratory, U.S. EPA, and approved for publication. Mention of trade names or commercial products should not be construed as endorsement or recommendation for use. Environmental Health Perspectives * Vol 106, Supplement 6 * December 1998 Appendix g(,u) = x'P [1] where: ,u= the number of lung adenomas for the nth mouse g(.) = a specified monotone function of the mean