What Contributes to the Combined Effect of a Complex Mixture?

Sep 4, 2004 - action seems plausible. Chlorophyll fluorescence quenching analysis supports to discriminate between prometryn,. N-phenyl-2-naphthylamin...
5 downloads 0 Views 423KB Size
Environ. Sci. Technol. 2004, 38, 6353-6362

What Contributes to the Combined Effect of a Complex Mixture? ROLF ALTENBURGER,* HELGE WALTER, AND MATTHIAS GROTE UFZ Centre for Environmental Research Leipzig Halle, Permoserstrasse 15, 04318 Leipzig, Germany

The effect of a mixture of 10 compounds, which have previously been identified in an effect-directed analysis as potentially relevant for a specific contaminated riverine sediment (Brack et al. Arch. Environ. Contam. Toxicol. 1999, 37, 164), were investigated for the underlying joint effect. Components identified in an organic sediment extract included several PAHs (benzo[ghi]fluoranthene, benz[a]anthracene, fluoranthene, pyrene, 2-phenylnaphthalene, anthracene, and phenanthrene) plus prometryn, N-phenyl2-naphthylamine, and parathion-methyl. Experiments were performed using a one-generation algal bioassay with the unicellular green algae Scenedesmus vacuolatus as well as chlorophyll fluorescence quenching analysis to describe the effects of the components and mixtures thereof. Analysis of the mixture effects based on concentration-response modeling of the effect data reveals that indeed effect contributions of several components can be expected although the mixture ratio is not equitoxic and the individual components vary greatly with respect to biological effect. Comparing predicted and observed mixture effects, the combined effect may not be attributed to a joint narcotic effect of the mixture components. Evidently, some of the components act specifically and dissimilar and may therefore be best described in their combined effect by response addition while for others a similar mode of action seems plausible. Chlorophyll fluorescence quenching analysis supports to discriminate between prometryn, N-phenyl-2-naphthylamine, and PAHs. A joint model for calculating the combined effect using concentration addition for the suspected unspecifically acting components in algae (PAHs and parathion-methyl) and subsequently response addition for this group and the other components clearly improves the description of the observed combined effect. Allocation of effect contributions to specific components using toxic units or effect contributions lead to different judgments. The observed combined effect of a 3-compound mixture of prometryn, N-phenyl-2-naphthylamine, and benzo[ghi]fluoranthene is indistinguishable from the effects of the original 10-compound mixture, demonstrating the need in site-specific assessment of complex contamination to account for the mode of action of contaminants. Implications for the confirmation step in effectdirected analysis of substances causing effects in complex contaminated samples are discussed.

Introduction Organisms are rarely exposed to individual chemicals, and effect assessment has therefore to account for the occurrence * Corresponding author phone: +49-341-235-2224; fax: +49-341235-2401; e-mail: [email protected]. 10.1021/es049528k CCC: $27.50 Published on Web 09/04/2004

 2004 American Chemical Society

of diverse contaminants with different toxic potentials. Occurrence of pollutants, however, may not signify contribution to toxic effects. Misfit between chemical contamination profiles and bioassay data is common (2). The identification of relevant contributors to observed mixture effects in site-specific assessment would offer scope for measures targeted at toxicity reduction, which is of particular interest for costly remediation efforts. The identification of toxicants in complex mixtures is attempted by linking exposure information to biological effects (3). Approaches that deal with complex contamination typically identify more than individual toxicants (i.e., mixtures). The toxicity of a mixture, however, depends not only on the exposure concentration of each mixture constituent and its ratio but also on the means of the toxicants to act jointly (4, 5). For the quantifcation of contributions from identified toxicants summation of toxic units (6, 7), sum of induction equivalent quantities (IEQs; 8) or total body residues (TBRs; 9) as derived from the individual components have been used. These procedures implicitly assume a concentration additive behavior of the components in the mixture and allow a comparison of expected and observed response for the selected effect level only. Current experimental evidence in conceptual mixture toxicity studies, however, have provided evidence in algae (10, 11) and bacteria (12) that the effect of chemicals with specific but dissimilar mode of action may be quantitatively better described using the reference model of response addition. This may even be the case in mixtures with components of unclear mode of action (13), which should be typical for complex mixtures occurring in the environment. Moreover, the notion that the quality of a mixture toxicity prediction might depend on the level of observed effect has been made (14, 15). Drescher and Bo¨deker (16) studied the quantitative relationships between the predictions from concentration and response addition for binary mixtures. They demonstrated for typical doseresponse functions of individual toxicants in aquatic toxicity tests that response addition tends to calculate a lower combined effect as compared to the expectation generated from concentration addition. From these theoretical considerations, it follows that an invalid assumption of a similar mode of action may produce false qualitative and quantitative assessments of toxicants, with the possible result that relevant contributors may be overlooked. The goal of this study was to apply the currently discussed conceptual approaches in mixture toxicity studies to a sitespecific complex mixture and derive a methodology suitable for toxicity identification efforts. The three main objectives were (i) to achieve detection of combined effects in mixtures with a ratio in line with the components’ prevalence in sediment extracts, (ii) to qualify and quantify the contributions of suspected toxicants to observed mixture effects based on the alternative concepts of concentration addition and response addition, and (iii) to use mode of action information to improve noninteraction modeling for identification of relevant toxicants.

Materials and Methods Test Chemicals. The substances used for single and mixture toxicity studies were identified as potential toxicants in sediments from the Spittelwasser, a freshwater site in the Bitterfeld area of Germany, applying an effect-directed fractionation and identification approach (1). Identity, source, lot, and purity information of the test chemicals are provided in Table 1. All other chemicals were purchased in the highest VOL. 38, NO. 23, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

6353

TABLE 1. Identity, Source, Lot, Purity, and Concentrations Found in Sediment Extracts of Mixture Components substance

CAS RN

source

lot

purity (%)

analyzed concn (mg/kg)a

anthracene benz[a]anthracene benzo[ghi]fluoranthene fluoranthene N-phenyl-2-naphthylamine parathion-methyl phenanthrene 2-phenylnaphthalene prometryn pyrene

120-12-7 56-55-3 203-12-3 206-44-0 135-88-6 298-00-0 85-01-8 612-94-2 7287-19-6 129-00-0

Aldrich Aldrich Promochem Aldrich Aldrich Riedel-de-Ha¨ en Aldrich AccuStandard Riedel-de-Ha¨ en Fluka

11102PI-011 01924PU TU Lot 3 12301-058 TI08223HU 9160x 11015AU PU 980609LB-AC 0195x 417950/140901

99 99 98 98 97 99.6 99 99 99.7 99

1.1 0.35 2.5 3.8 16 1.6 5.9 5.0 6.3 7.6

a The concentrations provided refer to freeze-dried, sieved (63 µm), pooled sediment samples from the Spittelwasser, which was Soxhlet extracted for 24 h with dichloromethane (1).

available purity from Aldrich (Steinheim, FRG), Riedel de Haen (Seelze, FRG), Merck (Duesseldorf, FRG), or Sigma (Deisenhofen, FRG). Test Organisms and Culture Conditions. Liquid cultures of the unicellular green alga S. vacuolatus Shih. et Krauss, strain 211-15, culture collection Pringsheim (SAG Go¨ttingen, Germany) were grown photoautotrophically at 28 ( 0.5 °C in an inorganic, sterilized medium adjusted to pH 6.4 under conditions specified earlier (17). Cells were synchronized by light:dark changes of 14:10 h and a periodic dilution to a standard cell density of 106 cells/mL before the onset of the light phase of the growth cycle (t0). Synchronization was checked by analysis of the cell size distribution at t0. Determination of Concentration-Response Relationships. Concentration-response relationships of the test compounds were experimentally determined using a 24 h test under synchronized conditions taking the inhibition of algal cell reproduction after one generation as effect parameter. The initial cell density was set to 7.5 × 104 cells/mL. Gastight test tubes (Pyrex 15, QVF, Wiesbaden, Germany) were used as test vessels. Culture volumes were 8 mL with a headspace of 3 mL. The test medium was the same as for cultivation but enriched with 1.9 mmol/L NaHCO3 providing a final pH of the medium of 6.9 ( 0.2. Illumination was ensured by a combination of two types of fluorescent light tube (L36W/41 Interna, L36W/11 daylight, Osram, Berlin, Germany) with an intensity of 13-18 W/m (22-33 kLux) providing a photosynthetic active radiation of 350 µeinstein s-1 m-2 at the surface of the test vessels. The test substances either dissolved in the test medium or in dimethyl sulfoxide (DMSO) were added to the cultures at t0. The concentration of DMSO in the test was 0.1%. The experimental design was characterized by 12 test concentrations as triplicates diluted after range finding tests to cover 5-80% effect plus 6 untreated cultures or 3 untreated and 3 solvent controls when used and a geometrical dilution series adjusted to the steepness of the concentration-response relationship. The data obtained in independently conducted experiments were pooled for data analysis. Aliquot samples of the cultures were taken in duplicates at t0 and at the end of the standard algal reproduction cycle (t24), and the mean cell number was analyzed twice using a CASY II-particle counter (Scha¨rfe Systems, Reutlingen, Germany). The inhibition of cell reproduction was calculated by expressing the difference in cell count increase to the control as percent of the cell increase in the control cultures. Analysis of Nominal Concentrations. The analytical investigations of test concentrations have been performed to gain experience with the water solubility of the mixture components in our test system; to check for a possible effect of the cosolvent DMSO on the concentration-response estimations; to achieve reproducible exposure concentrations in the biotests that may be used as a reliable basis for mixture 6354

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 23, 2004

prediction; to obtain information on the stability of exposure over the course of the experiment; and to check for any indication of physical interactions in the mixture. Nominal concentrations of stock solutions, concentration at the beginning, and concentration at the end of the algal biotest were assayed using reversed-phase high-performance liquid chromatography (rp-HPLC) equipped with a wavelength variable fluorescence detector. Acetonitrile-water mixtures of varying ratios (60:40 to 90:10) were used as eluents on a Merck LiChrospher 60, RP select B column (125-4 mm, 5 µm) or for the mixture analysis a LiChrospher PAH (250-3 mm, 5 µm) (Merck, Darmstadt, Germany) equipped with an equivalent precolumn, at a constant temperature of 25 °C and flow rates of 0.5 mL/min. Pumps, fluorescence detector, and peak analysis software were from Merck (LaChrom System, Darmstadt, Germany) too. Fluorescence detection was compound-specific optimized for maximal excitation in the UV (220-300 nm) and maximal emission at wavelengths between 300 and 500 nm. Chlorophyll Fluorescence Quenching Analysis. Analysis of variable fluorescence of algal suspensions to assess the similarity of interference of substances with photosynthetic primary reactions was conducted by performing pulse amplitude modulated fluorescence quenching analysis of chlorophyll a at 680 nm using a PAM fluorometer (Water PAM, Walz GmbH, Effeltrich, Germany). Algal suspensions at densities of 105 cells/mL supplied with 1.5 mmol/L NaHCO3 were exposed for 5-120 min to the test substances in the growth medium as described above. After a dark incubation of 5 min, a single multiple turnover saturation pulse was applied to determine the photochemical efficiency of photosystem II as quantum yield. After relaxation actinic light of an intensity of about 300 µmol einstein s-1 m-2 in the test tube, which was comparable to the growth conditions was applied and on top at 20 s intervals, 13 saturation pulses were applied. By this sequence, comparison of dark and light adapted fluorescence intensities were performed, and effect parameters of photochemical (qP) and nonphotochemical (qN, NPQ) quenching were derived (18). Data Analysis. The concentration-response relationships of the single substances were biometrically modeled by using a “best-fit” approach (19). For that purpose, 10 different 2or 3-parametric sigmoidal regression modelssincluding the commonly used Probit, Logit, and Weibull modelsswere fitted to the experimental data set. The parameters of the models were estimated using a generalized least-squares approach. For each individual set of data, the best-fitting model was chosen on the basis of statistical criteria. The statistical uncertainty of the EC50 was estimated using the bootstrap approach (19). Calculation of Expected Mixture Toxicity. The concentration of each component in the mixture can be expressed as fraction of the total mixture concentration. Consequently,

TABLE 2. Single Substance and Mixture Effects on Reproduction of Unicellular Green Algae Scenedesmus vacuolatus CAS RN

log Kowa

203-12-3 56-55-3 7287-19-6 135-88-6 206-44-0 129-00-0 612-94-2 120-12-7 85-01-8 298-00-0

5.41 5.54 3.51 4.38 5.16 4.88 5.20 4.45 4.47 2.86

substance benzo[ghi]fluoranthene benz[a]anthracene prometryn N-phenyl-2-naphthylamine fluoranthene pyrene 2-phenylnaphthalene anthracene phenanthrene parathion-methyl 10-component mixture

pib 0.046 0.006 0.111 0.304 0.079 0.103 0.159 0.026 0.140 0.026 1.000

EC50e Nc (Cd) (µmol/L) 25 15 36 10 10 55 8 33 31 36 48

parameterg modelf

θ1

θ2

θ3

(23) 0.045 [0.0474; 0.0429] Generalised Logit 998.7238 800.6945 0.0086 (11) 0.060 [0.0655; 0.0555] Generalised Logit 74.2437 65.8448 0.1069 (6) 0.069 [0.0719; 0.0670] Weibull 5.2045 4.807 (11) 0.16 [0.167; 0.153] Logit 15.0138 18.8574 (10) 0.17 [0.201; 0.153] Generalised Logit 269.0739 455.8077 0.0088 (44) 0.24 [0.284; 0.202] Weibull 2.787 5.143 (10) 0.36 [0.397; 0.317] Weibull 4.5441 10.9374 h (31) 2.8 [17.2; 1.35] Weibull -0.937 1.258 (12) 3.3 [3.58; 3.18] Generalised Logit -3.4876 5.1291 0.5935 (6) 31 [32.7; 30.5] Generalised Logit -38.4176 22.7277 0.1591 (12) 0.363 [0.342; 0.389] Weibull 1.646 4.574

a Ref 21, MedChem database; parathion-methyl: http://www.daylight.com/daycgi/clogp. b Fraction of total mixture concentration. c Number of data points. d Number of controls. e Mean effect concentration with 95% two-sided bootstrap confidence interval in [ ]. f Weibull: E ) 1 exp(-exp(θ1 + θ2(log10 (c)))). Generalised Logit: E ) 1/(1 + exp(-(θ1 + θ2(log10 (c)))))θ3. Logit: E ) 1/(1 + exp(-(θ1 + θ2(log10 (c))))). g θi denotes the parameter of the concentration-response function. h EC50 for anthracene is an extrapolation, as the maximum effects observed lie between 20 and 40%.

a total concentration of the mixture, at which a certain effect is generated, can be calculated using concentration addition according to

( ) n

ECxmix )

i)1

-1

pi

∑ECx

(1)

i

where ECxmix is the total concentration of the mixture provoking x% effect; ECxi is the concentration of component i provoking the x% effect, when applied singly; and pi denotes the fraction of component i in the mixture. The calculation of total mixture concentrations for various effect levels lead to a complete iteration of an expected concentration-effect relationship. For response addition, the following calculus applies: n

E(cmix) ) 1 -

∏(1 - E(c )) i

(2)

i)1

The effect at the total concentration of the mixture, E(cmix), is based on the effects of the components, which they generate at concentrations at which they are present in the mixture (E(ci)). If the latter is expressed as a fraction of the total mixture concentration, it holds: n

E(cmix) ) 1 -

∏(1 - E(p c

i mix))

(3)

i)1

This allows the calculation of an effect expected according to the concept of response addition for any concentration of the mixture. Joint modeling was used as an approach to calculate an expected toxicity for a mixture of components with heterogeneous modes of action using concentration addition and response addition for subgroups of components, according to a procedure described elsewhere (20).

Results Detection of Combined Effects. The biological efficacy of the 10 chemicals identified as potentially ecotoxic in a Spittelwasser sediment extract (1) against algae has been defined by complete concentration-response curves for the inhibition of algal reproduction. The derived parameters for the description of the concentration-response functions for the individual compounds and the mixture of them are provided in Table 2. The concentrations given are based on analytically checked nominal concentrations for the stock solutions and representative concentrations at the onset of the experiment. Nominal concentrations were corrected for

analytically determined concentrations when the deviation from the nominal concentration was larger than the variance of the analytical determination (typically about 20%). As deviations of nominal and analytically determined concentrations at the onset of the experiment (t0) were mainly related to sparse solubility of the components, for the mixture experimentation only stock solutions prepared in DMSO were used. Here no corrections of the nominal concentrations proved necessary. The stability of the components in the experiment for the 24 h of exposure was similarly determined as the analytically quantified difference between the application of the substance at t0 and after 24 h at the end of the biotest for concentrations in the effective but not completely inhibitory range. The loss varied for the individual compounds between 4% for parathion-methyl and 74% for benzo[ghi]fluroanthene. In particular, the PAH components benzo[ghi]fluroanthene, 2-phenylnaphthalene, anthracene, benz[a]anthracene, and pyrene showed losses above 20% over the test duration of 24 h. Repeated testing in 2 (2-phenylnaphthalene) to 12 (pyrene) independent experiments revealed that the individual concentration-response curves despite the lack of compound stability were still wellreproducible. This can also be deduced from the 95% twosided bootstrap confidence intervals for the median effect concentration (Table 2) that vary from 4.5 to 17% of the EC50 for benzo[ghi]fluroanthene, 2-phenylnaphthalene, benz[a]anthracene, and pyrene as compared to a variance between 3.5 and 17% for the other compounds. Only for anthracene is this picture different with an evidently unusually large confidence estimate. This may be explained by the fact that anthracene evokes a maximum effect in the algal reproduction test of 20-40% in the investigated concentration range of up to 2 mmol/L, so that the estimated EC50 values is already an extrapolation. Furthermore, in the concentrations from 0.2 to 2 mmol/L, higher response variability might be linked to variability in solubility. The concentration-response relationship in the lower part and the estimation of lower effect concentrations, however, do not indicate a variance that prohibits its use in the mixture experimentation. The 95% two-sided bootstrap confidence interval (0.1522; 0.0497) for the estimated EC10 of 0.0904 µmol/L of anthracene, for example, is comparable to the effect variance at the EC10 for the other components that range from 12% (for prometryn) to 43% (for fluoranthene). The modeling of the effect observations using the comparative approach of Scholze et al. (19) showed that out of the 10 regression models employed Weibull and Logit models provided the best fit of data, while Box-Cox transformation of the concentration scale did not improve VOL. 38, NO. 23, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

6355

FIGURE 1. Data for the inhibition of algal reproduction of the 10-component mixture (closed circles) and regression models of this mixture (dotted lines) and its components (solid lines). The modeled concentration-response relationships of the components are derived from the concentrations, at which the components are present in the mixture (mixture concentration × fraction for the component from Table 2) and the calculated effect for the single substances (using the function and parameters provided in Table 2). The numbers refer to the substances as follows: N-phenyl-2-naphthylamine (1), prometryn (2), benzo[ghi]fluoranthene (3), fluoranthene (4), 2-phenylnaphthalene (5), pyrene (6), benz[a]anthracene (7), phenanthrene (8), anthracene (9), and parathion-methyl (10). Open circles show the response of the controls. description of concentration-response relationships for our assay. The estimated mean effect concentrations (EC50 values in Table 2) for the inhibition of algal reproduction differ over a range of 3 orders of magnitude. The biologically most active substance benzo[ghi]fluoranthene generates an effect of 50% at 0.045 µmol/L. To yield the same effect, the concentration of the least toxic substance parathion-methyl had to be increased 700-fold (Table 2). The variance on the effect scale is usually highest for the untreated controls and amounted to a maximum of (10% of the mean value in the controls. The efficacy of the mixture of chemicals, composed with a ratio in line with the prevalence of the components found in sediment extracts (1; Table 2), was measured for a dilution series of 24 dilution steps in two independent experiments. The experimental data allow adequate description of the concentration-response function between 10 and 90% effect (Figure 1). The function describes a monotonic curve using the Weibull model. The variance with 6-7% at the EC50 is smaller than for most substances investigated individually, and neither the curve nor the variance in the investigated range gives any indication of unexpected disturbances such as physicochemical interactions. Analytical separation of the mixture using a gradient rp-HPLC and detection of the fluorescing components also did not provide any indication for alterations of the mixture components. To investigate whether the mixture tested provoked a detectable combined effect, the observed biological mixture activity is compared with the modeled effects of the components at the concentrations at which they are present in the mixture (Figure 1). From a mixture concentration of 0.1-0.5 µmol/L, the effects of the mixture are distinctly different from the effects of any component, which they generate at concentrations at which they are present in the mixture. In this concentration range, the observed mixture toxicity may therefore be attributed to a combined effect of 6356

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 23, 2004

more than one chemical. For mixture concentrations above 0.5 µmol/L, the effects observed for the mixture cannot be differentiated from the effects of N-phenyl-2-naphthylamine alone, without further experimental efforts. Below 0.1 µmol/L no distinction between individual and combined effects can be made. Comparison between Predicted and Observed Combined Effects: Narcotic Joint Action. Substances occurring at low concentrations have been suggested to contribute to a mixture effect via an unspecific, narcotic mode of action even if they exert a specific mode of action at higher concentration (22-24). Considering that for the combined effects of the mixture up to almost 60% effect the components apart from prometryn are present in concentrations not expected to evoke detectable effects, this hypothesis was to be tested. To generate an expectation for a narcotic or baseline toxicity of the mixture components, the concentrationresponse relationships of 12 non-electrophilic, non-reactive, non-ionized organic compounds were experimentally established for the same algal bioassay as used for the mixture study (Table 3). The resulting quantitative relationship between the estimated EC50 values for baseline toxicity and log Kow is given by

log EC50 (mol/L) ) -0.863 ((0.040) × log Kow - 0.897 ((0.092) (4) n ) 12, r 2 ) 0,978, r (adj) ) 0.977, S ) 0.219, F ) 470 This relationship compares well with other QSAR equations for baseline toxicant effects on algae reported in the literature (24). To analyze whether the observed mixture toxicity might be anticipated by baseline toxicity contributions of the

TABLE 3. Concentration-Effect Relationships for Inhibition of Reproduction of Unicellular Green Algae Scenedesmus vacuolatus for Non-electrolyte, Non-reactive, Non-ionized Organic Compounds parameterb

b

substance

CAS RN

log Kow

EC50 (µmol/L)

modela

θ1

θ2

θ3

2,4-dichlortoluene 1,4-dichlorbenzene monochlorobenzene benzene hexanol pentanol butanol propanol acetone 2-propanol ethanol methanol

95-73-8 106-46-7 108-90-7 71-43-2 111-27-3 71-41-0 71-36-3 71-23-8 67-64-1 67-63-0 64-17-5 67-56-1

4.24 3.38 2.92 2.13 2.03 1.56 0.88 -0.25 -0.24 0.05 -0.31 -0.77

0.000025 0.00013 0.00060 0.0011 0.0022 0.011 0.033 0.077 0.12 0.19 0.28 0.77

Hill Hill Hill Hill Hill Hill Logistic sigmoid Logistic sigmoid Hill Hill Logistic sigmoid Hill

0.024885 0.12558 0.55082 1.07554 2.133683 10.62842 27.46047 75.54954 116.2722 193.1716 287.4571 773.8221

6.3659 9.6376 15.9235 2.34987 1.655486 4.570527 0.114348 0.067738 3.89376 13.4331 0.013943 8.296587

0 0 0 0 -1.755307 3.581901 -46.59372 -3.651312 0 8.62415 7.188001 4.467508

a Functions and corresponding models. Hill: E ) (θ + (100 - θ ))/(1 + ((c/θ )-θ2)); Logistic sigmoid: E ) (θ + (100 - θ ))/(1 + exp(-θ (c - θ ))). 3 3 1 3 3 2 1 θi denotes the parameter of the concentration-response function.

FIGURE 2. Comparison between regression model of the observed mixture toxicity (black solid line) and expected toxicity of the mixture according to concentration addition (dotted line), response addition (interrupted line), and joint model (light solid line). The mixture toxicity assuming concentration addition for the baseline toxicity of the components is indicated (black dot) for the EC50 response. The expected baseline toxicity is calculated from eq 4. The inset depicts the expected combined effects for the different models at the observed effect of the mixture of 50%; CA ) concentration addition; RA ) response addition, JM ) joint model, Obs ) observation. components only, a mixture toxicity based on the assumption of narcotic activity of the components in algae was then calculated and was compared to the observed mixture toxicity (Figure 2). The resulting EC50 of 8.6 µmol/L for the expected combined effect based on concentration additive behavior for the baseline toxicity is 24 times above the value actually observed for the EC50 of the mixture (0.36 µmol/L). Basing the explanation of expectable combined effects on contributions from a similar but unspecific mode of action of the components only would therefore clearly underestimate the effect of the mixture at stake. Predictions Assuming Specific, Similar, and Dissimilar Mode of Action. The expected combined effects based on the assumption that all components act via a similar or dissimilar specific mode of action are depicted as concentration response curves in Figure 2. These predictions are

much closer to the observation than those for a purely joint narcotic action. However, calculation according to concentration addition, assuming all components to act similar, clearly overestimated the observed combined effect, while response addition assuming a dissimilar mode of action of the components underestimated the combined effect. The observed toxicity of prometryn and N-phenyl-2naphthylamine exceeded the toxicity expected on the basis of the QSAR derived for narcotic effects by more than 2 orders of magnitude, while for the other components this ratio was much smaller. For prometryn this is consistent with the known specifc binding to the D1 protein of the photosystem II reaction center and the subsequent inhibition of photosynthetic electron transport as the specific mode of action of triazine derivatives (25). For N-phenyl-2-naphthylamine, no information on a specific mechanism or mode of action VOL. 38, NO. 23, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

6357

FIGURE 3. Contributions of the mixture components (numbered solid lines) to the mixture toxicity (bold solid black line) at various effect levels expressed as toxic units (TUs). The sum of TUs of the five main contributing components is additionally depicted (dotted line). The numbers refer to the substances as follows: N-phenyl-2-naphthylamine (1), prometryn (2), benzo[ghi]fluoranthene (3), 2-phenylnaphthalene (4), fluoranthene (5), pyrene (6), benz[a]anthracene (7), and phenanthrene (8). TUs of parathion-methyl and anthracene are below 0.02 and not shown. TUs are calculated according to eq 5. is currently available. Our own investigations comparing the high algal toxicity described here with effect determined for luminescent bacteria, daphnia, and fish eggs using shortterm protocols according to ISO standards show that the more than narcotic activity of the compound is specific for the assay with S. vacuolatus. Furthermore, chlorophyll a fluorescence quenching analysis showed distinct differences between the effects of exposure to the different compounds. When exposing algal suspensions at the compound-specific EC50 concentrations for the inhibition of algal reproduction under actinic light, prometryn within 20 min caused a dramatic decrease in fluorescence yield of photosystem II, while N-phenyl-2-naphthylamine and benzo[ghi]fluoranthene treated suspensions remained unaffected under these conditions. For N-phenyl-2-naphthylamine short-term effects on nonphotochemical quenching were only evoked by concentrations significantly larger than the EC50 for the inhibition of algal reproduction. Furthermore, extended exposure times of several hours at EC50 concentrations led also to observable effects on chlorophyll a fluorescence which, however, might be interpreted as unspecific secondary effects. This pattern indicates a target site for N-phenyl-2-naphthylamine distinctly different from a photosystem II inhibitor like prometryn. On the basis of this, we classified prometryn and N-phenyl2-naphthylamine as dissimilar acting in comparison to the remaining seven PAHs and parathion-methyl for which again only an unspecific, narcotic mode of action in algae may be assumed. Using this grouping, a joint model for calculating the expected combined effect was employed (20). Comparing the calculated model-dependent mixture toxicities with the observed combination effect the joint model clearly gained an improved description of the observed mixture toxicity 6358

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 23, 2004

(Figure 2) than concentration addition for the complete concentration range. In comparison to the calculated mixture toxicity for response addition, the joint model again provided a better description of the observation for mixture up to a concentration of 0.4 µmol/L. At the observed EC50 of the mixture (0.36 µmol/L), the mixture toxicity expected according to the joint model deviated by 7% on the effect scale from the observed mixture toxicity. In comparison, at the same mixture concentration, the calculated mixture toxicity according to concentration addition or response addition assuming all 10 components to have either similar or dissimilar modes of action deviated by 50% and 18% on the effect scale, respectively, as can be derived from the inset in Figure 2. Quantifying Contributions of Mixture Components to the Combined Effect. From Figure 1 it may seem evident that the individual contributions from the components to the detected combined effects were heterogeneous and that not all components may have significantly contributed to the detected combined effect. Using the reference concept of concentration addition, contributions of the components to the observed effect of the mixture can be quantified as toxic units (TU) according to

TUi ) ECx(mix) × pi/ECxi

(5)

with ECx(mix) as a defined effect concentration of the mixture x; pi as the fraction of the component i in the mixture; and ECxi as the concentration of the component i at which it generates the same effect as the mixture, when tested singly. Figure 3 depicts the toxic units of the mixture components for the tested sediment extract concentration ratio as depending on the effect level. Depending on the considered

FIGURE 4. Effects of the mixture components (numbered solid lines) at various effect levels observed for the mixture. Expected effects of the 10-component mixture (bold solid line) and the mixture of the two main contributing components (dotted line) calculated according to response addition are additionally shown. Effect estimations are based on the fractional concentrations (mixture concentration × fraction pi of total mixture concentration, see Table 2) at which the components are present in the mixture and the determined concentrationeffect function (cf. Table 2). The numbers refer to the substances as follows: N-phenyl-2-naphthylamine (1), prometryn (2), benzo[ghi]fluoranthene (3), fluoranthene (4), anthracene (5), pyrene (6), and phenanthrene (7). Effect estimates of 2-phenylnaphthalene, benz[a]anthracene, and parathion-methyl are below 0.005% and not shown. Effects of the components (numbered solid lines) are calculated according to eq 6. effect level for the combined effect, the TU contributions differ for each component and vary between the components. For example, at the observed 20% effect level for the 10 component mixture, prometryn is estimated to have contributed with a TU value of 0.56 followed by N-phenyl-2naphthylamine having contributed with 0.46 TU. This ranking is effect level dependent. At the 80% effect level, for example, N-phenyl-2-naphthylamine provided a TU value of 0.9, more than prometryn with a TU value of 0.6. Looking at the sum of the toxic units 5 components (namely, N-phenyl-2naphthylamine, prometryn, benzo[ghi]fluoranthene, 2-phenylnaphthalene, and fluoranthene) accounted for more than 90% (Figure 3 dotted line) of the observed mixture toxicity (Figure 3 solid line). The values for the sum of toxic units range from 1.6 to 2.8 depending on the effect level considered. Values for the sum of the toxic units above 1 signal that the mixture effect concentrations is higher than calculated for concentration addition (i.e., the observed combined effects are lower than expected for concentration additive behavior). Using response addition, the effect that a component i generates (Ei) can be calculated as follows:

(6)

The modeled effect expected for a component when acting individually differs for each component and varies between the components depending on the observed effect level. For example, at the 50% effect observed for the mixture, prometryn is expected to have individually evoked a 20% effect, which is the highest contribution of all components, followed by N-phenyl-2-naphthylamine (assumed to have generated a 5% effect). At 80% effect of the mixture, N-phenyl2-naphthylamine primarily contributed with an expected 64% effect, whereas prometryn is expected to have contributed with an individual effect of 42%. The expected mixture effects of the two most contributing components N-phenyl-2naphthylamine and prometryn according to response addition (Figure 4 dotted line) do not deviate by more than 7% as compared to the effects expected for the 10-component mixture over the effect range regarded (Figure 4 solid line) (i.e., these two compounds would suffice to explain the expected response additive mixture effect). The difference between the observed mixture effect and the regressionmodeled inhibition of algal reproduction for the mixture indicates that the 10-component mixture was more effective than expected for response addition for effect levels below 70%.

with fi denoting the concentration-response function used to fit the observed concentration-effect data of compound i, as provided in Table 2, and (ECx(mix) × pi) as the fractional effect concentration of the mixture. Figure 4 depicts the effects of the individual components as calculated from the fractional effect concentrations according to eq 6 in relation to the observed mixture effect.

Mixture of Identified Effect Contributors. The highest contributions of all 10 components to the observed mixture toxicity were quantified for N-phenyl-2-naphthylamine and prometryn, irrespective of the approach applied (Figures 3 and 4). Using response addition for the two components, the expected mixture toxicity depending on the considered response level deviated by a factor of up to 2 (solid curve in Figure 5) from the observed and biometrically described 10-

Ei ) f(ECx(mix) × pi)

VOL. 38, NO. 23, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

6359

FIGURE 5. Comparison of the distances between the effect concentrations for the expected toxicity of the 2-component mixture of prometryn and N-phenyl-2-naphthylamine according to response addition (solid line), the observed responses of the 10-component mixture (open circles), and the 3-component mixture (closed circles) of prometryn, N-Phenyl-2-naphthylamine and benz[ghi]fluoranthene and the regression model of the observed mixture toxicty (y-axis). ECx values are calculated according to the following models and parameters: ECx(2-component mixture) (logistic sigmoid) ) 0.460117 - (1/11.73258 ln(1/E - 1)); ECx(10-component mixture - observation) cf. Table 1. component mixture. In particular for effects higher than 60%, no substantial difference from the complete mixture could be detected. At lower mixture concentrations, however, the calculated mixture toxicity using the joint model showed improved description of the observed mixture toxicity (Figure 2) while response addition of the two dominating substances prometryn and N-phenyl-2-naphthylamine underestimated the combined effect by a factor of up to 2. At least in this range of effect further contributors have to be considered. Using the TUs depicted in Figure 3, benzo[ghi]fluoranthene may be identified as the next most prominent potential contributing component. A mixture of this compound together with N-phenyl-2-naphthylamine and prometryn was experimentally tested, and the resulting effects compared to the efficacy observed for the 10-component mixture (Figure 5). The observed effects of the 3-component mixture cannot be differentiated from the toxicity observed for the 10component mixture without further experimental efforts. This result indicates that the combined effect of the 10-component mixture might result from the contributions of three components (N-phenyl-2-naphthylamin, prometryn, and benzo[ghi]fluoranthene).

Discussion Currently found discrepancies between the results from efforts of environmental chemistry and ecotoxicity testing to identify relevant toxicants in site-specific contamination assessments (2) pose the challenge to establish advanced techniques of linking exposure and effect information. In the discussion, we want to focus on the techniques for detection and quantification of combined effects and reflect the role of mode of action information for noninteraction modeling of combined effects for diagnostic use of sitespecific complex mixtures. The selection of compounds used in this study has been based on an effect-directed analysis of contaminated sedi6360

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 23, 2004

ment in the vicinity of former chemical production sites at Bitterfeld, Germany (1). Effect-directed analysis has been suggested as the strategy of choice for the identification of toxicants at complex contaminated sites (3) and indeed structurally heterogeneous compounds have been identified as being of potential relevance, some of which may be considered as ubiquitous such as the PAHs, while others are specific for this site (prometryn, parathion-methyl, and N-phenyl-2-naphthylamine) (1), thus representing a truly complex mixture at nonequitoxic ratios. The efficacy of the individual components as analyzed in the algal assay used in this study compare well for prometryn, parathion-methyl, and N-phenyl-2-naphthylamine to literature values (1). For the PAH compounds, the measured concentrations of some of the test chemicals at the end of the bioassay deviate substantially from the initial concentrations. Our evidence suggest that the observed losses in the concentrations of the parent compounds are mainly due to photo-transformation rather than to sorption or biotransformation. In terms of the estimated effect concentrations for these compounds, the reported values relate to the initial concentrations, which are certainly not a realistic value. The alternative of using, for example, time-weighted concentrations, however, not only requires knowledge of the kinetics of photo-transformation for the individual compounds but also implies that only the parent compound is bioactive. For anthracene, we could demonstrate that illumination of an aqueous suspension within a few hours led to the formation of various photomodification products of which several (namely, quinone and dione derivatives) could be identified as bioactive using effect-directed analysis (26). Thus, the answer to the question of whether photo-transformation of the PAHs compounds leads to bioactivation or deactivation is not straightforward but requires specific efforts to identify bioactive transformation products and determine their concentrations and stability over time. These observations

will become problematic for the analyses of the mixture toxicity, in particular if the decreases in concentrations of the parent compounds in the mixture studies are substantially different from those observed in the tests with the individual chemicals. Indication that this was not a problem in the investigated cases derives from three observations: First of all, repeated assaying of individual components showed that the variance of effect determination was comparable for all components irrespective of their stability over the exposure time of 24 h. This indicates that the elucidated effects should be reproducible for similar conditions. Second, the transmission spectra of the investigated mixtures did not indicate substantial differences, and the light absorption of the individual components in the mixture should therefore be comparable with the conditions when assayed alone. Third, the HPLC analysis of the mixtures did not indicate a change in the pattern of occurring photometabolites nor in the trend of loss of parent compounds. So in summary, we may conclude that the individual effect concentrations may comprise a systematic bias for the following considerations on the principles of allocation of effect contributions in a given mixture; however, it is irrelevant what the actual EC values are, provided they are reasonably reproducible. While effect-directed analysis evidently helps to elucidate structures of contaminants of potential concern, the confirmation of the suspected toxicants as contributing to the effect of the occurring mixture is not straightforward. The commonly used approach to sum up toxic units (1, 6, 27), toxic equivalents (28), or induction equivalents (8, 29) is based on several assumptions: (i) the concentration-response relationships are assumed to show monotonic functions; (ii) all components irrespective of the components concentration are expected to contribute to the overall effect (30); (iii) the comparison between expectation and observation for the mixture effect is based on a point estimate, suggesting that the combined effect assessment is valid for all response levels; (iv) the expected mixture effect is calculated using concentration addition (i.e., the components are anticipated to act similar) (5). The issue of monotonic concentration-effect relationships is no problem for bioassays using apical end points unless problems in the exposure regime occur (e.g., limited solubility of hydrophobic chemicals, or micelle formation of surfactants). It is common to achieve monotonic concentration-effect relationships (12, 31-33). For bell-shaped curves as may occur with enzyme assays, however, expectations as to the anticipated combined effects according to concentration addition are no longer unambiguous (34). The notion that all components of a mixture contribute to a combined effect irrespective of their individual concentration at which they occur is inherent to the formula for concentration addition (eq 1) (30) where fractions of the effect concentrations are used. Considerable experimental efforts have been spent to provide evidence for this assumption studying multiple mixtures of non-reactive organic chemicals (35) and mixtures of specifically acting s-triazine derivatives (33). However, although providing proof of principle, this kind of evidence is arbitrary with respect to the mixture ratios employed as they are optimized to provide an equitoxic ratio for the effect level of interest (i.e., the mixture concentrations are deliberately designed to provide equal contributions at a designed reference point, for example, EC50). As demonstrated for the contaminants found in the Bitterfeld Spittelwasser sediment in this study, site-specific mixtures in contrast will be found at non-equitoxic ratios. This statement is evident also for others mixtures identified in effect-directed identification efforts (6, 8). In these cases, only some of the mixture components may be expected to contribute to a combined effect, while for most other components the concentrations at which they occur in relation to their

individual concentration-effect function will make it difficult to detect their actual contributions experimentally. In this study, we could demonstrate that only three out of the identified 10 potential toxicants suffice to explain the observed combined effect of the mixture. Detectability might not be the same as irrelevance, but in assessment practices it provides at least clear scope of priorities. Criteria for differentiating between detectable and nondetectable contributors may be based on variance considerations for the individual as well as for the combined effects, on arbitrarily selected effect quantities, or on specified toxic unit fractions. While the former approach recognizes the experimental evidence, the advantage of the latter approaches would be to provide an easier to compare measure. Selecting a suitable procedure for a given assessment task should in any case take account of the robustness of the estimation of components contributions to mixture effects with regard to the individual concentration-response function and to the mixture ratio. Regarding the assumption of a response level independent combined effect implicit in current toxicant confirmation procedures, we could demonstrate that in fact the expected contributions to the overall effect for the assumption of no interaction changed considerably between the components over different response levels. This behavior results from the different slopes of the individual concentration-response functions (15, 16). The variation of steepness of concentration-response curves even for components of great structural similarity have been described using cellular algal and bacterial assays for triazines (33), chlorophenols (31), or nitrobenzenes (20) and may be seen as an indication for compound differences in pharamacokinetics and also for multiple modes of action. Escher and Hermens (24) argue that toxic equivalent concentrations may express the relative potency of a given component (i.e., the effect contribution) only if the concentration-effect curves of the compounds are parallel and have the same maximum. Different response maxima may be addressed by defining a fixed level approach (29, 36). The requirement of parallel curves, however, imposes a severe drawback that might be easiest to overcome by modeling concentration-effect curves as demonstrated here for the compounds as well as for the mixtures. Typically, the required information for this is available from the study of the components. Finally, currently practiced allocation of contributions of identified mixture components using the concept of concentration addition in toxicant identification (e.g., refs 7, 8, and 16) implicitly anticipate the components to act similar (5). It is far from clear what a similar or dissimilar action, respectively, may be in the context of calculating a reasonable expectation for a combined effect on the basis of the individual components activities. A common view regards chemicals as similarly acting if their modes of action can be described as unspecific (37) (i.e., the components exerting their effect via membrane perturbation as narcotic toxicants only (24) or if the concentrations of specifically acting compounds are so low that only these baseline toxicities contribute to an overall effect) (24). The results in this study indicate that the combined effects assessment based on the assumption of joint action from narcotic modes of action of the components would have clearly underestimated the mixture effect. However, the mixture ratio in this study provides relatively high fractions for prometryn the known specifically acting photosystem II inhibitor, so that no general conclusion may be drawn. Evidence found in the literature on this (22, 23) is also not conclusive, particularly as the alternative concept of response addition has been neglected. If the components of a mixture have different target sites and independent modes of action, their combined effect is expected to follow the concept of response addition or VOL. 38, NO. 23, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

6361

independent action (24, 38). Experimental evidence that response addition even at integral levels of response may produce good description of observed combined effects has been provided for specifically and strictly dissimilar acting pesticides (11) and antibiotics (12). Moreover, the finding reported here, that for mixtures comprising of heterogeneous structures and possibly components with different modes of action, the combined effect is lower than expected for concentration addition but higher than for response addition has also been found for a mixture of 11 structurally heterogeneous priority pollutants including atrazine, parathion, triorganotin compounds, and PAH compounds (13) as well as for a mixture of 14 nitrobenzenes with anticipated partial differences in their modes of action (20). For a predictive assessment one may therefore utilize both concepts to generate a “prediction window” for possible combined effects. Regarding its diagnostic use, however, reflection of the modes of action of the prominent components in a complex mixture seems necessary to avoid false positive or negative assessments about relevant contributors to a combined effect. The joint model applied in this study demonstrated that it is not necessary to exactly know a mode of action in order to improve the model based description of combined effect data but that a founded hypothesis on the similarity or dissimilarity of action may suffice.

Acknowledgments The best-fit modeling and uncertainty estimation was performed by Martin Scholze. This work was carried out under the EU Project EVK1-1999-00055 Bridging Effect Assessment of Mixtures to Ecosystem Situations and Regulation (BEAM). Thanks are due to the constructive critics of three anonymous reviewers.

Literature Cited (1) Brack, W.; Altenburger, R.; Ensenbach, U.; Mo¨der, M.; Segner, H.; Schu ¨u ¨ rmann, G. Arch. Environ. Contam. Toxicol. 1999, 37, 164. (2) O’Connor, T. P.; Paul, J. F. Mar. Pollut. Bull. 2000, 40, 59. (3) Brack, W. Anal. Bioanal. Chem. 2003, 377, 397. (4) Greco, W.; Bravo, G.; Parsons, J. C. Pharmacol. Rev. 1995, 47, 331. (5) Altenburger, R.; Nendza, M.; Schu ¨u ¨ rmann, G. Environ. Toxicol. Chem. 2003, 22, 1900. (6) Boxall, A. B. A.; Maltby, L. Arch. Environ. Contam. Toxicol. 1997, 33, 9. (7) Swartz, R. C.; Schults, D. W.; Ozretich, R. J.; Lamberson, J. O.; Cole, F. A., DeWitt, T. H.; Redmond, M. S.; Ferraro, S. O. Environ. Toxicol. Chem. 1995, 14, 1977. (8) Brack, W.; Schirmer, K. Environ. Sci. Technol. 2003, 37, 3062. (9) Van Loon, W. M. G. M.; Verwoerd, M. E.; Wijnker, F. G.; van Leeuwen, C. J.; van Duyn, P.; van deGuchte, C.; Hermens, J. L. M. Environ. Toxicol. Chem. 1997, 16, 1358. (10) Faust, M.; Altenburger, R.; Backhaus, T.; Boedeker, W.; Scholze, M.; Grimme, L. H. J. Environ. Qual. 2000, 29, 1063.

6362

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 23, 2004

(11) Faust, M.; Altenburger, R.; Backhaus, T.; Blanck, H.; Boedeker, W.; Gramatica, P.; Hamer, V.; Scholze, M.; Vighi, M.; Grimme, L. H. Aquat. Toxicol. 2003, 63, 43. (12) Backhaus, T.; Altenburger, R.; Boedeker, W.; Faust, M.; Scholze, M.; Grimme, L. H. Environ. Toxicol. Chem. 2000, 19, 2348. (13) Walter, H.; Consolaro, F.; Gramatica, P.; Scholze, M.; Altenburger, R. Ecotoxicology 2002, 11, 299. (14) Haas, C. N.; Cidambi, K.; Kersten, S.; Wright, K. Environ. Toxicol. Chem. 1996, 15, 1429. (15) Christensen, E. R.; Chen, D.; Nyholm, N.; Kusk, K. O. Environ. Toxicol. Chem. 2001, 20, 2361. (16) Drescher, K.; Boedeker, W. Biometrics 1995, 51, 716. (17) Altenburger, R.; Boedeker, W.; Faust, M.; Grimme, L. H. Ecotoxicol. Environ. Saf. 1990, 20, 98. (18) Kromkamp, J. C.; Forster, R. M. Eur. J. Phycol. 2003, 38, 103. (19) Scholze, M.; Boedeker, W.; Faust, M.; Backhaus, T.; Altenburger, R.; Grimme, L. H. Environ. Toxicol. Chem. 2001, 20, 448. (20) Altenburger, R.; Schmitt, H.; Schu ¨u ¨ rmann, G. Environ. Toxicol. Chem. (in press). (21) Leo D. A. MedChem Database, CLOGP, Version 4.61; Daylight Chemical Information Systems: Irvine, CA, 1999. (22) Hermens, J.; Leeuwangh, P. Ecotoxicol. Environ. Saf. 1982, 6, 302. (23) Dyer, S. D.; White-Hull, C. E.; Shephard, B. K. Environ. Sci. Technol. 2000, 34, 2518. (24) Escher, B. I.; Hermens, J. L. M. Environ. Sci. Technol. 2002, 36, 4201. (25) Wakabayashi, K.; Bo¨ger, P. Weed Biol. Manage. 2004, 4 , 8. (26) Brack, W.; Altenburger, R.; Ku ¨ ster, E.; Meissner, B.; Wenzel, K.D.; Schu ¨u ¨ rmann, G. Environ. Toxicol. Chem. 2003, 22, 2228. (27) Lee, J.-H.; Landrum, P. F.; Field, L. J.; Koh, C.-H. Environ. Toxicol. Chem. 2001, 20, 2102. (28) Safe, S. H. Environ. Health Perspect. 1998, 106, 1051. (29) Engwall, M.; Bromann, B.; Na¨f, C.; Zebu ¨ r, Y.; Brunstro¨m, B. Mar. Pollut. Bull. 1997, 34, 1032. (30) Ko¨nemann, W. H.; Pieters, M. N. Food Chem. Toxicol. 1996, 34, 1025. (31) Altenburger, R.; Backhaus, T.; Boedeker, W.; Faust, M.; Scholze, M.; Grimme, L. H. Environ. Toxicol. Chem. 2000, 19, 2341. (32) Berenbaum. M. C. Adv. Cancer. Res. 1981, 53, 269. (33) Faust, M.; Altenburger, R.; Backhaus, T.; Blanck, H.; Boedeker, W.; Gramatica, P.; Hamer, V.; Scholze, M.; Vighi, M.; Grimme, L. H. Aquat. Toxicol. 2001, 56, 13. (34) Altenburger, R.; Segner, H.; van der Oost, R. In PAHs: An Ecotoxicological Perspective; Douben, P. E. T., Ed.; John Wiley: Chichester, 2003; pp 297-328. (35) Deneer, J. W.; Sinnige, T. L.; Seinen, W.; Hermens, J. L. M. Aquat. Toxicol. 1988, 12, 33. (36) Brack, W.; Segner, H.; Mo¨der, M.; Schu ¨u ¨ rmann, G. Environ. Toxicol. Chem. 2000, 19, 2493. (37) Ko¨nemann, H. Toxicology 1981, 19, 229. (38) Po¨ch, G. Combined Effects of Drugs and Toxic Agents; SpringerVerlag: Wien, Austria, 1993; p 167.

Received for review March 26, 2004. Revised manuscript received June 3, 2004. Accepted June 11, 2004. ES049528K