Mixture Toxicity of Reactive Chemicals by Using Two Bacterial Growth

7 Oct 2005 - paradigms of mixture toxicity also hold true for reactive chemicals. Compounds with ... work in ecotoxicology (1, 2) took up the pharmaco...
0 downloads 0 Views 182KB Size
Download PDF
Environ. Sci. Technol. 2005, 39, 8753-8761

Mixture Toxicity of Reactive Chemicals by Using Two Bacterial Growth Assays as Indicators of Protein and DNA Damage MANUELA RICHTER AND BEATE I. ESCHER* Department of Environmental Toxicology, Swiss Federal Institute for Aquatic Sciences and Technology Eawag, CH-8600 Du ¨ bendorf, Switzerland

The mixture toxicity of reactive chemicals was investigated with a set of bioanalytical tests that quantify not only the toxic effects but also allow the identification of the preferred target of reactive chemicals in bacterial cells. Softer electrophiles such as acrylates react preferentially with thiol groups in proteins and peptides, and harder electrophiles such as epoxides preferentially attack DNA. In addition, some compounds, e.g., benzyl chloride, have no preference for a biological target and damage both DNA and proteins. A thiophenol was used as a model compound representing nucleophiles. We explored if the paradigms of mixture toxicity also hold true for reactive chemicals. Compounds with the same targets and the same modes of action should act concentration additive in mixtures, and compounds with different modes of action should act according to the concept of independent action. In addition, we investigated the potential for interaction of compounds of mixtures of electrophiles or electrophiles plus nucleophiles, which might lead to synergistic or antagonistic effects. The toxicity of mixtures of electrophiles with a single preferred target was consistent with the prediction for concentration addition. Unfortunately, the predictions for independent action did not differ much from those for concentration addition; therefore it was not possible to differentiate between these two models. Mixtures of two groups with different preferred target sites clearly showed concentration addition. In contrast, mixtures of compounds with multiple targets, i.e., compounds that show nonspecific reactivity toward any biological nucleophile, exhibited effects that lay distinctly between the predictions for concentration addition and independent action. We observed neither synergism (higher toxicity than predicted by concentration addition) nor antagonism (lower toxicity than predicted by independent action) for mixtures of electrophiles. Binary combinations of different electrophiles with the nucleophile 4-chlorothiophenol yielded smaller effects than those expected from the prediction for independent action. The degree of antagonism was correlated with the reaction rate constant of the electrophile with the thiol group of glutathione, which indicates that the interaction between the mixture components occur in the toxicokinetic phase and is purely a result of chemical reactivity between the mixture components. Overall, * Corresponding author phone: 0041-44-823 5068; fax: 0041-44823 5471; e-mail: [email protected]. 10.1021/es050758o CCC: $30.25 Published on Web 10/07/2005

 2005 American Chemical Society

we conclude that the concepts of mixture toxicity apply not only for baseline toxicity and receptor-mediated mechanisms, as has been shown in a large number of studies, but also for reactive mechanisms of toxicity, provided that one has checked beforehand that no chemical reactions occur between the mixture components.

Introduction During the last two decades, mixture toxicity concepts have been increasingly applied to ecotoxicological endpoints. Early work in ecotoxicology (1, 2) took up the pharmacological mixture concepts of concentration addition (CA) for mixtures of similarly acting compounds initially proposed by Loewe and Muischnek in 1926 (3) and independent action (IA) for mixtures of dissimilarly acting compounds proposed by Bliss in 1939 (4). In a series of studies, two research groups showed convincingly for multicomponent mixtures of defined and strictly similarly or dissimilarly acting compounds that these concepts hold also for integral endpoints of interest in ecotoxicology such as inhibition of luminescence in bacteria (5, 6), growth inhibition in algae (7-11), daphnia (11, 12), and water plants (13), and even receptor-mediated endpoints such as the yeast estrogen screen (14, 15) or the E-screen test for estrogenicity (16). More recently, the concepts of CA and IA have also been used to describe mixture toxicity of complex mixtures of similarly and dissimilarly acting compounds (17, 18). It has to be kept in mind that the mixture toxicity concepts were initially derived in pharmacology and strictly hold true for the toxicodynamics of effects only. Transfer of these concepts to integrative endpoints in ecotoxicology is only possible if alterations and interactions in the toxicokinetic phase are nonexistent or negligible. Toxicokinetic processes include uptake, distribution to target and nontarget sites, metabolism, and excretion. If any step in the toxicokinetic phase is different in the mixture than for single components, and particularly, if the components of the mixture interact with each other along their way to the target site, then deviations from the mixture toxicity concepts of CA and IA can be observed, resulting potentially in synergistic or antagonistic effects. For instance, the synergistic effects of atrazine in combination with organophosphates are presumably due to influences of atrazine on the activation step of organophosphates, which is a prerequisite for acetylcholine esterase inhibition (19, 20). Most mixture toxicity studies have been performed with baseline toxicants and chemicals that act via receptormediated mechanisms. Mixture studies with chemicals that exhibit reactive and multiple mechanisms are scarce and inconclusive. Chen and Yeh showed that reactive electrophilic chemicals acting via the same mode of toxic action are likely to show antagonism in a luminescence inhibition test with Vibrio fischeri (21). In addition some severely synergistic combinations of reactive chemicals with dissimilar modes of toxic action could be identified, e.g., malonitrile in combination with formaldehyde (21), but the underlying causes of the interactive effects have not been elucidated. With the same combinations of cyanogenic and electrophilic chemicals, binary mixture experiments were also performed in an algal toxicity test (22). Here, for the more reactive aldehydes, a tendency for synergistic effects was observed in combinations with malononitrile, and a tendency for antagonistic effects was observed in combinations with the more inert acetonitrile (22). With a similar test set of compounds, VOL. 39, NO. 22, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

8753

Lin et al. recently reported good correlations of toxic units of mixtures of reactive chemicals between inhibition of the enzyme luciferase and the in vivo luminescence in V. fischeri (23), which pointed out that the interactive effect occurred on the toxicodynamic level. Mixtures of electrophilic acrylates showed CA in a test based on glutathione depletion in rat hepatocytes (24). We have developed a set of bioassays for electrophilic chemicals that allowed the classification and quantification of two distinct reactive modes of toxic action (25). Softer electrophiles such as acrylates were classified as more reactive toward peptides and proteins, causing depletion of glutathione and damage to proteins, while harder electrophiles such as epoxides rather cause DNA damage. The electrophilicity scale is continuous (26), and accordingly many compounds were classified as nonspecific reactive; i.e., they act according to both reactive mechanisms (25, 27). Examples of this class of nonspecific reactivity include epichlorophydrin or benzyl chloride. The decisive indicator and discriminator of these modes of toxic action was a bioassay that made use of a set of Escherichia coli mutants (25). The comparison of growth inhibition by reactive chemicals in a strain deficient of glutathione (GSH) and its fully functional parent strain gives an indication of the relevance of GSH conjugation as a detoxification step and related toxic effects as well as direct reactivity with cysteine-containing proteins (25). Analogously, the stronger susceptibility of an E. coli strain that lacks important DNA repair systems as compared to its parent strain is a quantitative indicator of the mode of toxic action of DNA damage (25). We believe that these bioassays provide a perfect tool to investigate the possible interactive effects of reactive chemicals in mixtures because effects at specific target sites can be independently identified and quantified. With a set of 15 electrophiles of known reactive mode of action and one nucleophile, we explored the hypothesis of whether the toxicity of chemicals or groups of chemicals that react with the same target (GSH or DNA) and do not interact with each other can be described by CA. The alternative hypothesis was that electrophiles interact with each other, leading to other toxic products, possibly with other modes of toxic action. In this case, antagonistic effects should be observable. We did not rule out synergy altogether, but it is less likely that even more reactive molecules are formed in the course of the experiment. Even if two chemicals do not react with the same biological target, e.g., an electrophile plus a nucleophile, they might interact or even react with each other in the toxicokinetic phase, resulting also in nonclassical mixture effects. For groups of chemicals with distinctly different target sites and modes of action, we used the concept of IA as a null hypothesis. The study is further complicated by the fact that a large number of electrophiles investigated in the previous study showed multiple reactive mechanisms; i.e., they were active both in the assay for GSH-depletion-related toxicity and in the assay for DNA damage.

Experimental Section Chemicals. The set of electrophiles comprised acrolein (ACR, CAS Registry No. 107-02-8), acrylamide (ACA, CAS Registry No. 79-06-1), acrylonitrile (ACN, CAS Registry No. 107-13-1), benzyl chloride (BCl, CAS Registry No. 100-44-7), 4-chlorobenzenethiol (CTP, CAS Registry No. 106-54-7), 2,3-dichloro1-propene (DClP, CAS Registry No. 78-88-6), epichlorohydrin (EPI, CAS Registry No. 106-89-8), 1,2-epoxybutane (EOX, CAS Registry No. 106-88-7), (2,3-epoxypropyl) benzene (EPOX, CAS Registry No. 4436-24-2), ethyl acrylate (EA, CAS Registry No. 140-88-5), 2-hydroxyethyl acrylate (HEA, CAS Registry No. 818-61-1), isobutyl acrylate (IBA, CAS Registry No. 1068754

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 22, 2005

63-8), styrene oxide (SOX, CAS Registry No. 96-09-3), 3-methylbenzyl chloride (3MBCl, CAS Registry No. 620-19-9), 4-nitrobenzyl chloride (NBCl, CAS Registry No. 100-14-1), and trans-1,4-dichloro-2-butene (DClB, CAS Registry No. 11057-6). ACR, EA, HEA, ACA, BCl, CTP EOX, EPI, IBA, and NBCl and were purchased from Fluka Chemie AG, Buchs, Switzerland. DClP, 3MBCl, and SOX were obtained from SigmaAldrich Chemie AG, Steinheim, Germany. DClB and EPOX were bought from Aldrich Chemical Co., Inc., Milwaukee, WI. ACN was bought from Riedel de Hae¨n, Seelze, Germany. All chemicals were of the highest purity available (g95%) and used as received; ACA was used as a 40% aqueous solution. Because all compounds are prone to fast hydrolysis (28) and are relatively volatile, the chemicals were pipetted from a flask containing the pure liquid compound directly into the bioassay. Nominal concentrations were used because of the low hydrophobicity of the compounds and because transfer into high-performance liquid chromatography or gas chromatography vials resulted in a significant loss of compound (28). For the most volatile compounds, the nominal concentrations were corrected for the fraction in the gas. The air-water partition coefficients were calculated from the Henry constants reported in the Chemfate and Physprop databases (http://esc.syrres.com/efdb/Chemfate.htm and http://esc.syrres.com/interkow/physdemo.htm). The fractions in the gas phase were calculated with these partition coefficients and an aqueous volume of 5 mL, a gas volume of 7 mL, and a negligible volume of the organic phase (bacteria). Accordingly, corrections were made for EA (17% in the gas phase), IBA (32% in the gas phase), EOX (9% in the gas phase), BCl (19% in the gas phase), DClP (70% in the gas phase), and DClB (28% in the gas phase). Bacteria. A pair of E. coli strains, which differed only in their ability to synthesize glutathione (GSH), was used to indicate glutathione-depletion-related toxicity. The parent strain MJF276 (hereafter referred to as GSH+) has the capability to synthesize GSH, and the mutant strain MJF335 (hereafter referred to as GSH-) lacks both γ-glutamylcysteine synthase and GSH synthetase. Another pair of E. coli strains, which differed only in their ability to repair damaged DNA, was used to indicate the mode of toxic action of DNA damage. The parent strain MV1161 (hereafter referred to as DNA+) contains all DNA repair systems, while the mutant strain MV4108 (hereafter referred to as DNA-) lacks a number of DNA repair systems, among them systems of the adaptive response, the SOS response, and repair of 3-methyladenine. For details on the characterization of the bacteria and the derivation of the methods, see ref 25. Quantification of Growth Inhibition in E. coli Strains GSH+ and GSH-. All experiments were performed as described in ref 25. Briefly, E. coli were grown aerobically at 30 °C on a shaking incubator in minimal medium, which consisted of 33 mM KH2PO4, 60 mM K2HPO4, 7.6 mM (NH4)2SO4, 1.7 mM Na3-citrate, 1 mM MgSO4, 0.1 ‰ (w/v) vitamin B1, and 11 mM glucose with the pH adjusted to 7 and complemented with 25 mg/L tetracycline (MJF 276) or 25 mg/L kanamycin (MJF 335). Growth inhibition experiments were performed in 12 mL glass tubes with Teflon-coated screw caps and starting cell densities of approximately 108 cells/mL. Electrophiles (single compounds or mixtures) were directly pipetted into the cell suspensions in the glass tubes, and geometrical dilution series were quickly prepared. After an incubation of 6 h at 30 °C in closed tubes, growth was determined by light scattering at 600 nm (OD600) and growth inhibition was calculated with eq 1.

(

% growth inhibition ) 1 -

)

∆OD600,sample 100% (1) ∆OD600,control

∆OD600 refers to the difference in OD600 at the beginning and

at the end of the incubation period. The concentrations resulting in 50% inhibition of growth (EC50) were derived from a log-log fit of the concentration-effect curves (eq 2), using the software Prism 4.0b (GraphPad Software, San Diego, CA), which computed a fit of the experimental data of all parallel experiments by minimizing the sum of least squares under the prerequisites of fixed minimum at 0% and fixed maximum at 100% growth inhibition. Adjustable parameters were the slope m and the EC50.

% growth inhibition )

100% 1 + 10m(log EC50 - log concentration)

(2)

with the fraction pi of a mixture component i defined as follows

EC50i

pi )

∑ EC j)1

log log ECy ) σlogECy )

(100%y%- y%) + log EC

50

m

x(log(100%y%- y%)m ) σ -2 2

2 m

(3)

+ σlogEC502 (4)

The influence of GSH on growth inhibition of the (mixtures of) chemicals was characterized by the toxic ratio (TR) of EC50 values of MJF276 and MJF335, TRGSH (eq 5). For chemicals or mixtures with nearly identical concentration-effect curves (TRGSH < 2), GSH had no detoxifying effect. Chemicals and mixtures with TRGSH g 2 were assigned to the mode of toxic action “glutathione-depletion-related toxicity”. The threshold is slightly lower than the one proposed by ref 25 but reflects our recent results.

TRGSH )

EC50 (GSH+)

(5)

EC50 (GSH-)

Quantification of Growth Inhibition in E. coli Strains DNA+ and DNA-. Because the strain DNA- had the tendency to form filaments, colony-forming units (cfu’s) were used instead of optical density to assess growth inhibition by toxicants. All experiments were performed as described in ref 25. Growth inhibition was calculated with eq 6, and eq 2 was used to derive the EC50.

(

% growth inhibition ) 1 -

)

cfu (sample) 100% (6) cfu (control)

Because the reproducibility of this test was not very good due to the large number of mutations in the DNA- strain, a positive control of 2 µL ethyl methane sulfonate (EMS, CAS Registry No. 62-50-1, g98%, Fluka) per 5 mL of E. coli suspension (resulting in 3.8 × 10-3 M EMS) was performed in parallel to each assay. This concentration of EMS resulted in 100% inhibition of growth of DNA- and 0% inhibition of growth of DNA+. Only data from those experiments that yielded this result in the presence of EMS were further evaluated. Growth inhibition differences between DNA+ and DNA- were described by TRDNA (eq 7). TRDNA > 10 indicates the mode of toxic action “DNA damage”, while for compounds with lower TRDNA DNA damage is not the cause of the observed growth inhibition effect.

TRDNA )

EC50 (DNA+) EC50 (DNA-)

(7)

Mixture Experiments. Mixture experiments were performed with 3-9 compounds with a fixed ratio design,

50j

The total concentration in the mixture cmix is the sum of the concentration of the n components i, ci, which are present as fractions pi (eq 9). n

Effect concentrations ECy at effect levels y% other than 50% were derived by implementing y in eq 2 and rearrangement (eq 3), and associated errors were derived by error propagation (eq 4).

(8)

n

cmix )

∑

n

∑(p c

ci )

i mix)

i)1

(9)

i)1

The concentration-effect curves and EC50,mix values were derived with eq 2 with a total concentration cmix on the concentration axis. The experimental concentration-effect curves were compared with the predictions for concentration addition (CA) and independent action (IA). For CA, the sum of all toxic units, i.e., the ratios of ci to the effect concentration at any effect level y, ECyi (eq 3), must equal 1 (eq 10) (5) n

ci

∑EC i)1

)1

(10)

yi

Consequently, the effect concentration ECy,mix is calculated for each effect level y from eq 11 (5)

(∑ ) n

ECy,mix )

pi

-1

(11)

i)1 ECyi

By incremental application of eq 11 for effects y from 0% to 100%, a concentration-effect curve for concentration addition can be predicted. The standard deviation of the CA prediction of ECy,mix was approximated by a resampling method using a log-normal distribution of the errors of all input parameters and 1000 random resampling steps of eq 11 (29). The error σECyi was calculated with eq 4, and the error of pi, σpi, was estimated to be a 5% pipetting error. The resampling routine was written in Mathematica (version 5.0, Wolfram Research). The prediction of the alternative mixture concept of independent action can be calculated with eq 12 (6) n

E(cmix) ) 1 -

∏(1 - E(c ))

(12)

i

i)1

where E(cmix) corresponds to the predicted effect of the mixture and E(ci) to the effect of mixture component i. The confidence of the prediction of E(cmix) was estimated by the same resampling routine as described for CA above.

Results and Discussion Single Compounds. Concentration-effect curves were measured for 15 electrophiles with the different E. coli strains. 4-Chlorothiophenol (CTP) was used as a model compound for a nucleophilic molecule, and more information on CTP is given in the Supporting Information. Two representative examples, ACN in the GSH strains and EOX in the DNA strains, are depicted in Figure 1. Generally, our earlier results (25) were quite well reproducible and robust on the EC50 level. However, slight changes in the EC50 values and particularly the slope caused a considerable difference in the mixture predictions. To ensure the highest consistency within the mixture study, all concentration-effect curves for the single VOL. 39, NO. 22, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

8755

FIGURE 2. Binary mixture effects in the E. coli strains MV 1161 (DNA+) and MV 4108 (DNA-). The boxes extend from the 25th to the 75th percentile, and the average is marked by a line. The error bars extend to the 95% confidence interval. The left two boxes show the mixture of SOX and EOX, each in a concentration of half of their EC50; the right two boxes show the mixture of ACN and HEA, each in a concentration of half of their EC50. The line marks the 50% effect level, which corresponds to the prediction for concentration addition of the mixture.

FIGURE 1. Examples of concentration-effect curves of the single compounds. (A) ACN in ) GSH- (E. coli strain MJF 335) ( GSH+ (E. coli strain MJF 276). (B) EOX in O DNA- (E. coli strain MV 4108) b DNA+ (E. coli strain MV 1161). Error bars refer to standard deviations; the 95% confidence band is shown with broken lines. compounds were therefore newly determined in series with the mixture experiments, and the resulting EC50 values are reported in Table 1. The new TRGSH values were somewhat smaller than the values reported earlier, but the 95% confidence intervals were much smaller, and all 95% confidence limits reported in Table 1 lay well within the earlier reported 95% confidence intervals. Binary Mixtures in the Test for DNA Damage. The test for DNA damage turned out to be not robust enough for mixture studies. First, as is evident from Figure 1B, the

concentration-effect curves of the single compounds show a broad confidence band of the best fit. Second, it is not possible to perform enough replicates and obtain enough experimental data points because the procedure for plating bacteria on agar plates and counting cfu’s is very material consuming. Therefore we had to test only binary mixtures at the predicted 50% effect level instead of determining full concentration-effect curves. SOX and EOX were mixed in a concentration of half of their EC50 each, and in both the DNA+ and the DNA- strain, approximately 50% effect was measured as expected from CA (Figure 2). However, the confidence intervals are very large, and from the 19 replicates all could be used in the case of DNA+, but only 9 could be used in the case of DNA- (because the positive control did not respond adequately). Similarly, the mixture of ACN and HEA showed also the expected CA (Figure 2), but again confidence intervals were wide, and only 15 out of 19 replicates could be used in the evaluation of DNA- (19/19 in DNA+). These results were discouraging, and therefore no further mixture experiments were performed with the DNA( strain.

TABLE 1. EC50 Values of the E. coli Strains GSH+ (MJF 276), GSH- (MJF 335), DNA+ (MV 1161), and DNA- (MV 4108), Toxic Ratios TRGSH and TRDNA, and Classification of Mode of Actiona TRGSHb

log(EC50(GSH+)/mM)) log(EC50(GSH-)/mM))

log(EC50(DNA+)/mM)) log(EC50(DNA-)/mM))

TRDNAb

moa class GSH GSH GSH GSH GSH GSH

1.49 ( 0.02 0.27 ( 0.01 -2.02 ( 0.03 -0.14 ( 0.02 -0.18 ( 0.01 -0.09 ( 0.04

n.d. -0.05 ( 0.01 -2.56 ( 0.02 -0.43 ( 0.01 n.d. n.d.

n.d. 2.14 (1.97-2.31) 3.45 (2.89-4.11) 1.94 (1.73-2.18) n.d. n.d.

n.d. 0.72 ( 0.08 n.d. n.d. 0.89 ( 0.06 n.d.

n.d. 1.17 ( 0.15 n.d. n.d. 0.10 ( 0.19 n.d.

n.d. 2.8 (1.2-6.2) n.d. n.d. 6.2 (2.4-16.0) n.d.

EOX EPOX SOX

1.28 ( 0.03 0.00 ( 0.01 0.20 ( 0.02

1.18 ( 0.02 0.06 ( 0.04 0.14 ( 0.02

1.25 (1.06-1.48) 0.87 (0.69-1.09) 1.14 (1.00-1.30)

2.14 ( 0.08 n.d. -1.50 ( 0.14

0.38 ( 0.10 n.d. 0.81 ( 0.02

57 (31-105) DNA n.d. DNA 205 (104-403) DNA

BCl DClP DClB EPI MBCl NBCl

-0.39 ( 0.02 -1.24 ( 0.02 -1.05 ( 0.01 0.64 ( 0.01 -0.16 ( 0.02 -1.08 ( 0.02

n.d. n.d. n.d. n.d. n.d. n.d.

n.d. n.d. n.d. n.d. n.d. n.d.

n.d. n.d. n.d. n.d. n.d. n.d.

n.d. n.d. n.d. n.d. n.d. n.d.

n.d. n.d. n.d. n.d. n.d. n.d.

n.r. n.r. n.r. n.r. n.r. n.r.

CTP

-1.48 ( 0.02

n.d.

n.d.

n.d.

n.d.

n.d.

n.d.

ACA ACN ACR EA HEA IBA

a

The statistics are given in the Supporting Information. Mode of toxic action classification from ref 25, supported by the measured TR of this study: GSH, glutathione depletion related toxicity; DNA, DNA damage; u.r., unspecific reactivity; n.d., not determined. b 95% confidence interval.

8756

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 22, 2005

FIGURE 3. Experimental concentration-effect curves of the mixtures and predictions for CA and IA. (A and B) Three acrylates (0.4% ACR, 68% ACN, 31.6% EA). (C and D) Three acrylates and three epoxides (0.02% ACR, 5.2% ACN, 2.2% EA, 78.8% EOX, 5.3% EPOX, 8.5% SOX). (E and F) Three epoxides (91% EOX, 3% EPOX, 6% SOX). In all plots the experimental data are depicted as follows: ) GSH- (E. coli strain MJF 335) ( GSH+ (E. coli strain MJF 276). Error bars refer to standard deviations. The lines in parts A, C, and E are the predictions for CA; the lines in parts B, D, and F are the predictions for IA. In all plots, broken lines refer to predictions for GSH- (E. coli strain MJF 335), and full lines refer to predictions for GSH+ (E. coli strain MJF 276). In each set of lines, the left one is the lower 95% confidence limit of the prediction, the middle the prediction, and the right the upper 95% confidence limit of the prediction. The fitted curve of the experimental data is omitted for clarity of the presentation, but the data are reported in Table 2. We can conclude that the prediction for CA is consistent with the results obtained for binary mixtures of compounds with the same mode of action. There is no indication for severe synergistic or antagonistic effects, but of course anything may be hiding in the wide confidence intervals. This exercise also shows the importance of highly reproducible, fast, and robust tests for the evaluation of mixture toxicity. Fortunately, the GSH( strains fulfill the requirements and are suitable for mixture toxicity studies. Mixtures of Similarly Acting Compounds in the Test for Glutathione-Depletion-Related Toxicity. The three acrylates ACR, ACN, and EA, which all are classified to the mode of action “GSH-depletion-related toxicity” were mixed in the ratio of their EC50 values. The concentration response curves for GSH+ and GSH- are depicted in Figures 3A and 3B together with the predictions for CA (Figure 3A) and IA (Figure 3B). The corresponding EC50 values are listed in Table 2. The prediction for CA pointed to slightly higher toxicity than the prediction for IA, but the 95% confidence belts of the two predictions overlap partially. In addition, due to the low number of compounds in the mixture (10, 30) and the large scatter of the experimental data points, it is not possible to assign the experimental results clearly to any of the predictions, because the experimental data overlap partially with both predictions. It is clear, however, that neither synergy nor antagonism applies. The experimental data for GSH+ appear to point to a slightly higher toxicity than both predictions, and GSH- appears to fit the prediction for IA better, but statistically no conclusion can be drawn. A similar observation was made for a mixture of the three epoxides EOX, EPOX, and SOX (Figures 3E and 3F), which are all DNA-damaging electrophiles but showed no specificity in the test for glutathione-depletion-related toxicity; i.e., the

concentration-effect curves for GSH+ and GSH- should overlap, which was indeed experimentally observed. For both strains, the prediction of CA and IA are consistent with the experimental findings. As an earlier theoretical analysis of binary mixtures has shown, predictions for CA and IA tend to fall on each other for mixtures of compounds with flat concentration response curves and in some cases IA may even predict higher toxicity than CA (31). Such a case was encountered for the first time in practice in a mixture study with 12 phenylurea herbicides using the reproduction inhibition assay with the green algae Scenedesmus vacuolatus (32). As the authors speculated, the agreement between the predictions for CA and IA is a mathematical coincidence and does not bear any mechanistic implications (32). In addition to slope, three other parameters influence the relationship between the predictions for CA and IA (31, 32): the number of compounds in the mixture, the concentration ratio, and the effect level under consideration. The concentration ratio was fixed to the ratio of EC50 values in our experiments, and our predictions covered the entire range of effect levels from 1% to 99%. Therefore, the only variable was the number of mixture components, which is only limited by experimental restrictions. However, even when we increased the number of acrylates to six (ACA, ACN, ACR, EA, HEA, and IBA), a distinction of between the two predictions was not possible (data not shown). Mixtures of Similarly and Dissimilarly Acting Compounds in the Test for Glutathione-Depletion-Related Toxicity. In the next step, the aforementioned mixtures of three acrylates were combined with the three epoxides, mixed in the ratio of their EC50 values. This combination constitutes a mixture of two dissimilarly acting groups of similarly acting VOL. 39, NO. 22, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

8757

9

n.d. n.d.

The statistics of the experimental data are given in the Supporting Information; n.d., not determined. a

0.01 (-0.01 to 0.03) 0.15 (0.12 to 0.18)

0.31 (0.26 to 0.35)

n.d.

n.d.

n.d.

n.d.

n.d. n.d. n.d. n.d. n.d. n.d. 0.54 (0.51 to 0.57) 0.48 (0.45 to 0.50)

0.71 (0.67 to 0.74)

1.05 (0.92-1.21) 1.11 (0.98-1.27) 1.35 (0.93-1.96) 0.92 (0.85 to 0.97) 0.91 (0.87 to 0.95) 0.97 (0.94 to 0.99)

1.05 (1.00 to 1.10)

0.95 (0.89 to 1.00)

0.86 (0.82 to 0.90)

1.70 (1.43-2.01) 1.51 (1.38-1.64) 1.74 (1.21-2.44) 0.63 (0.57 to 0.68) 0.47 (0.45 to 0.49) 0.39 (0.35 to 0.43) 0.87 (0.83 to 0.92) 0.65 (0.63 to 0.68) 0.62 (0.56 to 0.68)

0.01 (-0.03 to 0.06) -0.09 (-0.13 to -0.06) 0.03 (-0.03 to 0.09) -0.59 (-0.65 to -0.53) -0.50 (-0.53 to -0.46) -0.39 (-0.43 to -0.35) 4.00 (3.40-4.72) 2.54 (2.25-2.86) 2.14 (1.47-3.11)

0.4% ACR, 68% ACN, 31.6% EA 0.02% ACR, 5.2% ACN, 2.2% EA, 78.8% EOX, 5.3% EPOX, 8.5% SOX 91% EOX, 3% EPOX, 6% SOX 0.01% ACR, 1.3% ACA, 23.1% ACN, 3.1% EA, 1.1% HEA, 0.7% IBA, 64.8% EOX, 4.6% EPOX, 1.3% SOX 80.3% EPI, 6.3% BCl, 1.0% DClP, 1.5% DClB, 1.4% NBCl, 9.6% MBCl

TRGSH

prediction for IA

TRGSH

prediction for CA

TRGSH

experimental prediction for IA prediction for CA

log(EC50(GSH-)/mM)

experimental prediction for IA prediction for CA

log(EC50(GSH+)/mM)

experimental mixture composition pi (mol %)

TABLE 2. Compilation of the Results of the Mixture Experiments Including Experimental EC50 Values of the E. coli strains GSH+ (MJF 276) and GSH- (MJF 335), Toxic Ratio TRGSH and Corresponding Predictions for CA and IA (95% Confidence Intervals Given in Parentheses)a 8758

FIGURE 4. Experimental concentration-effect curve of a mixture of organochlorine molecules (80.3% EPI, 6.3% BCl, 1.0% DClP, 1.5% DClB, 1.4% NBCl, 9.6% MBCl) and predictions for CA and IA in ( GSH+ (E. coli strain MJF 276). Broken lines refer to predictions for IA, and full lines refer to the predictions for CA. In each set of lines, the left one is the lower 95% confidence limit of the prediction, the middle the prediction, and the right the upper 95% confidence limit of the prediction.

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 22, 2005

compounds. As Figures 3C and 3D depict, in both the GSH+ and the GSH- strains, the experimental data are best described by CA. In this case, IA predicted a clearly lower toxicity and was distinguishable from the CA prediction (see also Table 2). This result is very interesting with respect to classification of mode of toxic action. Both groups of chemicals are electrophilic but have different preferential target sites. Therefore one would expect IA between the two groups. Such a scenario can be predicted by a two-stage prediction, which is discussed in more detail in the Supporting Information, but clearly the two-stage prediction did not describe the experimental data better than CA. Because a higher number of mixture components often allowed a better differentiation between CA and IA (5, 6), the number of acrylates was increased to six in the next experiment. There are no further compounds in our test set that show exclusively DNA damage apart from the three epoxides. Therefore the number of components in this group was kept constant, leading to unequal numbers of components in the two groups. The experiments with this ninecomponent mixture are technically challenging due to the high volatility and fast reactivity of the compounds, and it is not possible to prepare stock solutions of the compounds beforehand, but in each mixture experiment the components have to be mixed and weighed from the pure compound. The mixture predictions of the nine-component mixture in GSH+ (Table 2) confirm that a differentiation between CA and IA should be possible, but again the experimental data are clearly congruent with the prediction for CA. Nevertheless, the toxic ratios TRGSH of the mixture fully reflect the mode of toxic action of the mixture components. The mixture of acrylates has a TRGSH of 4.0, and the mixture of the epoxides a TRGSH of 1.1 (Table 2), consistent with the values of the single compounds (Table 1). The mixture of three acrylates and three epoxides had a TRGSH of 1.70, i.e., between those of the mixture components. To summarize all mixture experiments with similarly and dissimilarly acting compounds, due to the similarity of the predictions (including the two-stage prediction discussed in the Supporting Information) one cannot assign the appropriate model based on experimental evidence. Nevertheless, all findings are consistent with the theory that compounds with the same target sites and same modes of action act concentration additive and compounds with different target sites and different modes of action act independently. Concerning the target site, there is less clarity. Although the toxicity of compounds assigned to the mode of action “DNA damage” appears to be dominated by DNA damage,

FIGURE 5. Experimental concentration-effect curve for binary mixtures of CTP and (A) EPOX, (B) BCl, (C) HEA, and (D) ACR and predictions for IA in GSH+ (E. coli strain MJF 276). Error bars refer to standard deviations. Broken lines refer to prediction for IA and drawn lines refer to modeling of the experimental data with eq 2. In each set of lines, the left one is the lower 95% confidence limit of the prediction, the middle the prediction, and the right the upper 95% confidence limit of the prediction.

TABLE 3. EC50 Values in GSH+ (E. coli Strain MJF 276) for Binary Mixtures of the Nucleophile CTP and Various Electrophiles and Predictions for CA and IA (95% Confidence Limits Given in Parentheses) log(EC50(GSH+)/mM) mixture composition pi (mol %)

experimental

prediction for CA

prediction for IAa

98% EPOX, 2% CTP 95% BCl, 5% CTP 97% HEA, 3% CTP 30% ACR, 70% CTP

-0.05 (-0.14 to 0.05) -0.09 (-0.14 to -0.04) -0.17 (-0.20 to -0.15) -1.10 (-1.16 to -1.05)

-0.21 (-0.25 to -0.17) -0.62 (-0.57 to -0.67) -0.37 (-0.33 to -0.41) -1.76 (-1.81 to -1.71)

-0.09 (-0.12 to -0.06) -0.62 (-0.66 to -0.58) -0.34 (-0.37 to -0.31) -1.69 (-1.75 to -1.64)

these compounds can also conjugate with GSH. The chemical rate constants of the reaction with the GSH anion were in similar orders of magnitude for acrylates and epoxides. This might be a reason for the observation that the mixture of three acrylates and three epoxides did show CA and not IA. Mixtures of Compounds with Nonspecific Reactivity in the Test for Glutathione-Depletion-Related Toxicity. A mixture experiment with the six nonspecific reactive electrophiles EPI, BCl, DClP, DClB, NBCl, and MBCl, mixed in the ratio of their EC50 values and tested with the GSH+ strain (Table 2), yielded a concentration-effect curve without overlapping confidence intervals between the prediction of CA and IA (Figure 4). The test was performed with the GSH+ strain that is closest to the wild-type and has no deficiencies related to either GSH-depletion-related toxicity or DNA damage. Therefore, neither mechanism should have been preferentially detected. In this example, the activity of compounds with multiple modes of action cannot be described by the classical models of mixture toxicity, CA, and IA. This observation warrants further investigations to investigate if it is generalizable. Nevertheless, a mechanistic interpretation of the observation is possible. Electrophiles react according to different reaction mechanisms with biological nucleophiles, and the mechanisms may even differ between different nucleophiles (28). While most of the electrophiles that can be assigned clearly to one of the two specific mechanisms, GSH-depletion-related toxicity or DNA damage, act in a second-order nucleophilic substitution reaction with the nucleophiles GSH or 2′deoxyguanosine (which is a model for a DNA base), many of the organochlorine electrophiles, which are nonspecific reactive, react via a first-order nucleophilic substitution, at least with water. In this mechanism, the leaving group, typically the chloride ion, dissociates first, leaving a very reactive electrophilic carbenium ion behind that reacts with any nucleophile it encounters without any selectivity. Consequently, such compounds are reactive both toward cysteine-containing proteins and peptides, such as GSH, and toward DNA. Strictly speaking, such a scenario with the same underlying reaction mechanism but different target sites would not be classified as similar mode of action but also not as dissimilar mode of action.

Binary Mixtures of a Nucleophile and an Electrophile. As Hewlett and Plackett stated in their initial formulation of CA and IA, these two concepts imply no interaction between components in a mixture (33). There is no indication that mixtures of electrophiles show any interaction or compete for molecular targets. In contrast, for mixtures of electrophiles and nucleophiles one could assume that they might react with each other and this interaction or reaction might compete with their reactivity toward biological targets. This hypothesis was tested with the nucleophile 4-chlorothiophenol (CTP) in combination with the various electrophiles of this study. Binary mixtures of CTP and different electrophiles mixed in the ratio of their EC50 values showed experimental concentration-response curves that were congruent with IA or less effective than the prediction for IA. This is illustrated in Figure 5 for binary mixtures of CTP with the four electrophiles EPOX, BCl, HEA, and ACR. The EC50 of the experiments and predictions are compared in Table 3. CA showed again very similar predictions as IA, but IA was taken as a reference because of the dissimilar mode of action of the mixture components. For clarity, the predicted curve for CA was not included in Figure 5, but the predicted EC50 values for CA are reported in Table 3. The concentration-effect curve of the binary mixture of CTP with EPOX (Figure 5A), which is classified as a DNAdamaging compound, was consistent with IA. Evidently, the low reactivity toward GSH is paralleled by a low reactivity toward CTP. In contrast, BCl showed a clear antagonism (Figure 5B). This is consistent with expectation because BCl acts nonspecifically reactive because it reacts in a first-order reaction; i.e., it has the tendency to form a stabilized benzylium cation, which is highly electrophilic and not selective for any nucleophile. Both acrylates, HEA and ACR, showed an antagonistic effect (Figures 5C and 5D). ACR is more reactive toward GSH than HEA, which is also reflected in a stronger antagonistic effect. The degree of antagonism can be expressed by the ratio of the experimental EC50 to that of the prediction for IA (Table 3). These ratios are well correlated with the second-order rate constant of the electrophiles with GSH (28); i.e., the degree of antagonism seems to be directly related to the VOL. 39, NO. 22, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

8759

reactivity of the electrophile with a thiol nucleophile. This analysis supports our initial hypothesis that nucleophilic pollutants may compete with biological nucleophiles as targets for reactive electrophiles. These findings are also supported by the recent work of Chen et al. on the mixture effects of aldehydes with acetonitrile in an algal toxicity test that also showed antagonistic effects (22). In contrast to this, the more reactive malononitrile showed synergism in the latter study, but the cause of the differences was not elucidated (22). In our experience, true synergistic and antagonistic effects on a toxicodynamic level are very rare (34); the causative interactions occur mostly in the toxicokinetic phase (19, 35, 36). It has to be always considered additionally that apparent antagonism or synergism may be caused by the interaction of mixture components. This statement appears trivial for the mixtures of reactive chemicals investigated here, but similar cases might also be encountered in mixtures of heavy metals and complexing organic chemicals.

Acknowledgments We thank Angela Harder for helpful discussions, Nadine Bramaz for experimental assistance, and Nathalie Che`vre and Marion Junghans for reviewing the manuscript.

Supporting Information Available Additional information is provided on the working hypotheses, on the reference compound CTP, and on the two-stage predictions and toxicity values of Tables 1 and 2, complemented by their confidence intervals and characteristics of the concentration-effect curves. This material is available free of charge via the Internet at http://pubs.acs.org.

Literature Cited (1) Ko¨nemann, H. Fish toxicity tests with mixtures of more than two chemicals: A proposal for a quantitative approach and experimental results. Toxicology 1981, 19, 229-238. (2) Hermens, J. L. M.; Canton, H.; Steyger, N.; Wegman, R. Joint effects of a mixture of 14 chemicals on mortality and inhibition of reproduction of Daphnia magna. Aquat. Toxicol. 1984, 5, 315-322. (3) Loewe, S.; Muischnek, H. U ¨ ber Kombinationswirkungen I. Mitteilung: Hilfsmittel der Fragestellung. Naunyn-Schmiedebergs Arch. Exp. Pathol. Pharmakol. 1926, 114. (4) Bliss, C. I. The toxicity of poisons applied jointly. Ann. Appl. Biol. 1939, 26, 585-615. (5) Altenburger, R.; Backhaus, T.; Boedeker, W.; Faust, M.; Scholze, M.; Grimme, L. H. Predictability of the toxicity of multiple chemical mixtures to Vibrio Fischeri: Mixtures composed of similarly acting chemicals. Environ. Toxicol. Chem. 2000, 19, 2341-2347. (6) Backhaus, T.; Altenburger, R.; Boedeker, W.; Faust, M.; Scholze, M.; Grimme, L. H. Predictability of the toxicity of multiple mixtures of dissimilarly acting chemicals to Vibrio Fischeri. Environ. Toxicol. Chem. 2000, 19, 2348-2356. (7) Faust, M.; Altenburger, R.; Boedeker, W.; Grimme, L. H. Algal toxicity of binary combinations of pesticides. Bull. Environ. Contam. Toxicol. 1994, 53, 134-141. (8) Faust, M.; Altenburger, R.; Backhaus, T.; Blanck, H.; Boedeker, W.; Gramatica, P.; Hamer, V.; Scholze, M.; Vighi, M.; Grimme, L. H. Joint algal toxicity of 16 dissimilarly acting chemicals is predictable by the concept of independent action. Aquat. Toxicol. 2003, 63, 43-63. (9) Junghans, M.; Backhaus, T.; Faust, M.; Scholze, M.; Grimme, L. H. Predictability of combined effects of eight chloroacetanilide herbicides on algal reproduction. Pest Manage. Sci. 2003, 59, 1101-1110. (10) Junghans, M.; Backhaus, T.; Faust, M.; Scholze, M.; Grimme, L. H. Toxicity of sulfonylurea herbicides to the green alga Scenedesmus vacuolatus: Predictability of combination effects. Bull. Environ. Contam. Toxicol. 2003, 71, 585-593. 8760

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 22, 2005

(11) Cleuvers, M. Mixture toxicity of the anti-inflammatory drugs diclofenac, ibuprofen, naproxen, and acetylsalicylic acid. Environ. Toxicol. Pharmacol. 2004, 59, 309-315. (12) Cleuvers, M. Initial risk assessment for three beta-blockers found in the aquatic environment. Chemosphere 2005, 59, 199-205. (13) Drost, W.; Backhaus, T.; Vassilakaki, M.; Grimme, L. H. Mixture toxicity of s-triazines to Lemna minor under conditions of simultaneous and sequential exposure. Fresenius Environ. Bull. 2003, 12, 601-607. (14) Rajapakse, N.; Silva, E.; Kortenkamp, A. Combining xenoestrogens at levels below individual no-observed-effect concentrations dramatically enhances steroid hormone action. Environ. Health Perspect. 2002, 110, 917-921. (15) Silva, E.; Rajapakse, N.; Kortenkamp, A. Something from “nothing”sEight weak estrogenic chemicals combined at concentrations below NOECs produce significant mixture effects. Environ. Sci. Technol. 2002, 36, 1751-1756. (16) Payne, J.; Scholze, M.; Kortenkamp, A. Mixtures of four organochlorines enhance human breast cancer cell proliferation. Environ. Health Perspect. 2001, 109, 391-397. (17) Junghans, M. Studies on combination effects of environmentally relevant toxicants: Validation of prognostic concepts for assessing the algal toxicity of realistic aquatic pesticide mixtures. Ph.D. Thesis, University of Bremen, Germany, 2004. http:// elib.suub.uni-bremen.de. (18) Altenburger, R.; Walter, H.; Grote, M. What contributes to the combined effect of a complex mixture? Environ. Sci. Technol. 2004, 38, 6353-6362. (19) Pape-Lindstrom, P. A.; Lydy, M. J. Synergistic toxicity of atrazine and organophosphate insecticides contravenes the response addition mixture model. Environ. Toxicol. Chem. 1997, 16, 24152420. (20) Howe, G. E.; Gillis, R.; Mowbray, R. C. Effect of chemical synergy and larval stage on the toxicity of atrazine and alachlor to amphibian larvae. Environ. Toxicol. Chem. 1998, 17, 519-525. (21) Chen, C. Y.; Yeh, J. T. Toxicity of binary mixtures of reactive toxicants. Environ. Toxicol. Water Qual. 1996, 11, 83-90. (22) Chen, C. Y.; Chen, S.-L.; Christensen, E. R. Individual and combined toxicity of nitriles and aldehydes to Raphidocelis subcapitata. Environ. Toxicol. Chem. 2005, 24, 1067-1073. (23) Lin, Z. F.; Du, J. W.; Yin, K. D.; Wang, L. S.; Yu, H. X. Mechanism of concentration addition toxicity: They are different for nonpolar narcotic chemicals, polar narcotic chemicals and reactive chemicals. Chemosphere 2004, 54, 1691-1701. (24) Freidig, A. P.; Hofhuis, M.; Van Holstijn, I.; Hermens, J. L. M. Glutathione depletion in rat hepatocytes: A mixture toxicity study with R,β-unsaturated esters. Xenobiotica 2001, 31, 295307. (25) Harder, A.; Escher, B. I.; Landini, P.; Tobler, N. B.; Schwarzenbach, R. P. Evaluation of bioanalytical tools for toxicity assessment and mode of toxic action classification of reactive chemicals. Environ. Sci. Technol. 2003, 37, 4962-4970. (26) Schwarzenbach, R. P.; Gschwend, P. M.; Imboden, D. M. Environmental Organic Chemistry, 2nd ed.; Wiley: New York, 2003. (27) Harder, A.; Escher, B. I.; Schwarzenbach, R. P. Applicability and limitations of QSARs for the toxicity of electrophilic chemicals. Environ. Sci. Technol. 2003, 37, 4955-4961. (28) Harder, A. Assessment of the Risk Potential of Reactive Chemicals with Multiple Modes of Toxic Action. Ph.D. Thesis, Department of Environmental Sciences, ETH Zu ¨ rich, Switzerland, 2002. http://e-collection.ethbib.ethz.ch/cgi-bin/show.pl?type) diss&nr)14966. (29) Escher, B. I.; Bramaz, N.; Eggen, R. I. L.; Richter, M. In vitro assessment of modes of toxic action of pharmaceuticals in aquatic life. Environ. Sci. Technol. 2005, 39, 3090-3100. (30) Faust, M. Kombinationseffekte von Schadstoffen auf aquatische Organismen: Pru ¨ fung der Vorhersagbarkeit am Beispiel einzelliger Gru ¨ nalgen (Combination Effects of Pollutants on Aquatic Organisms: Assessment of Predictability on the Example of Unicellular Green Algae). Ph.D. Thesis, University of Bremen, Germany, 1999. (31) Drescher, K.; Boedeker, W. Assessment of the combined effects of substances: The relationship between concentration addition and independent action. Biometrics 1995, 51, 716. (32) Backhaus, T.; Faust, M.; Scholze, M.; Gramatica, P.; Vighi, M.; Grimme, L. H. Joint algal toxicity of phenylurea herbicides is equally predictable by concentration addition and independent action. Environ. Toxicol. Chem. 2004, 23, 258-264.

(33) Hewlett, P. S.; Plackett, R. L. A unified theory for quantal responses to mixture of drugs: Noninteractive action. Biometrics 1959, 15, 591-609. (34) Escher, B. I.; Hunziker, R. W.; Schwarzenbach, R. P. Interaction of phenolic uncouplers in binary mixtures: Concentrationadditive and synergistic effects. Environ. Sci. Technol. 2001, 35, 3905-3914. (35) Beerenbaum, M. What is synergy? Pharmacol. Rev. 1989, 41, 93-141.

(36) Tripathi, A. M.; Agarwal, R. A. Synergism in tertiary mixtures of pesticides. Chemosphere 1997, 35, 2365-2374.

Received for review April 19, 2005. Revised manuscript received September 6, 2005. Accepted September 6, 2005. ES050758O

VOL. 39, NO. 22, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

8761