Degradation of mixture components
The degradation of the three biocides differed between the single exposure and the mixture exposure experiments. Whereas esfenvalerate degraded strongly in all concentrations in the single exposures, degradation of that compound was quite limited in the mixture exposure. The degradation of picoxystrobin did not differ between the single and the mixture exposure, yet a weak dose dependency occurred in both exposures with a slightly stronger degradation at lower concentrations. A dose-dependent degradation was also found for triclosan in both exposure scenarios. It is possible that the degradation of esfenvalerate in the mixture and of triclosan and picoxystrobin at the high mixture and single compound concentrations were hampered by the toxic effects of the three compounds to soil microbes. This has been observed, for example, for phenanthrene in soil 9. Degradation of triclosan at all concentrations was considerably stronger in the mixture exposure compared to the single exposure. It is difficult to explain this difference, because the precise removal processes of triclosan in soil are not known. Yet, the literature reports quite different degradation times for triclosan, indicating that slight differences in conditions affect degradation of triclosan considerably 3.
For picoxystrobin, the European Commission states an acute LC50 (14-d exposure) of 3.4 mg kg−1 (9.3 µmol kg−1) for E. fetida (http://sitem.herts.ac.uk/aeru/footprint/en/Reports/527.htm), which is very similar to the LC50 found in the present study (10.0 µmol kg−1 dry soil). Another study determined a much higher LC50 of 7.2 mg kg−1 soil (19.7 µmol kg−1) for E. fetida in artificial OECD-soil 29. The influence of soil organic matter on the availability of organic compounds to living matter may explain differences in the determined LC50s. Furthermore, lethal and sublethal effects occurred in a narrow concentration range, and concentration–response curves were consequently very steep. The determination of EC50s and LC50s, therefore, bears corresponding uncertainty. The mode of action of picoxystrobin on E. fetida is currently not known. The acute-to-chronic ratio (ACR) may be used to assess whether a chemical acts specifically or by narcosis on an organism. The ACR is commonly calculated as the ratio of the LC50 and a sublethal endpoint, such as the LOEC or EC50. In the present study, we used the LOEC for cocoon production, in order to compare ratios with ACRs calculated in a study by Roex and co-workers. 30. Using the threshold ACR for baseline toxicity proposed by Roex et al., to distinguish between narcotic (ACR < 2.58) and specific (ACR ≥ 2.58) mode of action 30, reveals that the ACR for picoxystrobin was at the threshold for narcosis (ACR = 2.6). In this case, the ACR does not clearly indicate whether or not picoxystrobin acts specifically on E. fetida.
The lethal toxicity of esfenvalerate was rather low, with a high LC50 of 85.8 µmol kg−1 dry soil. The European Health Commission states a LC50 of 10.6 mg kg−1 substrate (25 µmol kg−1) for E. fetida (http://sitem.herts.ac.uk/aeru/iupac/269.htm). It is difficult to explain this large difference, as exposure scenario, test duration, and soil type of the study are not known. In the present study, even the EC50 for reproduction (33.5 µmol kg−1 dry soil) was higher than the LC50 found in the literature. The ACR for esfenvalerate in the present study is 3.3 and may thus indicate a specific mode of action on E. fetida. Esfenvalerate is designed to target the sodium channels of the nervous system of arthropods. Arthropods and annelids share a synapomorphic ladder-like nervous system. The mode of action of esfenvalerate may therefore be the same on arthropods and annelids.
The LC50 for triclosan in the present study was 1,907 µmol kg−1 dry soil. Even higher was the LC50 for adult mortality of Eisenia andrei (866 mg kg−1 (2990 µmol kg−1) as determined by Amorim and coworkers 31. Noteworthy is that neither the study by Amorim et al. 31 nor in the present study did the exposure concentrations evoke 100% mortality. Hence, estimations of LC50 values are imprecise. The reported EC50 for reproduction in the closely related species E. andrei is 4 mg triclosan kg−1 (13.8 µmol kg−1) 31, which is more than four times greater than the EC50 determined for E. fetida in the present study (3.0 µmol kg−1). Again, differences in soil properties and among species may explain the differing EC50 values. Given the very high ACR of 602, it can be assumed that triclosan acts specifically on E. fetida. A recent study showed that triclosan alters enzyme activity and causes oxidative stress and DNA damages in E. fetida 32.
According to the results of the present study, the maximum spray doses of both Acanto (1000 ml per hectare) and Sumi-Alpha (DuPont, 500 ml with 50 g esfenvalerate L−1 Sumi-Alpha per hectare) are not expected to evoke any measurable effect on the reproduction or survival of one E. fetida generation (<EC1 estimated from the single exposure results in the present study) when applied singly. Triclosan concentrations in sewage sludge of up to 3.3 mg kg−1 (11.4 µmol kg−1) have been reported in sludge in Norway (http://www.klif.no/no/Publikasjoner/Publikasjoner/2010/April/Undersokelse-av-miljogifter-ved-fire-norske-renseanleggPFOA-Bisfenol-A-Triklosan-Siloksan-D5-Dodecylfenol-og246-Tri-tertbetylfenol/). Soil concentrations, however, are lower than sludge concentrations as the application of sludge to agricultural sites results in a dilution of triclosan in large soil volumes. Hence, triclosan concentrations in field soil will probably be below concentrations provoking effects on reproduction or survival.
The mixture used in the present study consisted of anthropogenic biocides with specifically designed modes of action toward three different organism groups. It was therefore expected that the concept of independent addition could describe mixture toxicity better than the concept of concentration addition. This assumption was supported by the ACRs calculated in the present study, which indicate that esfenvalerate and triclosan may act specifically and dissimilarly, whereas it remains unclear whether picoxystrobin acts specifically or by nacosis on E. fetida.
Commonly, when calculating mixture toxicity predictions using concentration addition and independent action, the fractions of the mixture components at the start of the exposure period are used. In the present study, this resulted in an overestimation of both the observed lethal and sublethal toxicity at low concentrations by both concepts, whereas concentration addition was a good predictor of lethal toxicity and independent action of sublethal toxicity at high mixture concentrations. Without taking the predictions based on the low-end fractions into account, one would conclude that dose–dependent deviations occur with antagonism at low-mixture concentrations. This is not uncommon, and examples can be found in the literature. For example, a study on the escape behavior of E. andrei exposed to a binary mixture comprising the pesticides dimethoate and spirodiclofen reported antagonism at low and synergism at high mixture concentrations 5. A study on the joint effects of glyphosate and spirodiclofen on the escape behavior of Porcellionides pruinosus also showed dose–dependent deviations from the mixture toxicity predicted by independent action, with synergistic effects at low and antagonistic effects at high concentrations 33. It has been stated that dose–dependent deviations from model predictions often occur when mixtures comprise only a few components and when higher organization level endpoints such as reproduction are used 6, 34. Deviations may indicate interactions, because both the concept of concentration addition and independent action assume non-interaction as a default 4. However, in the present study, an additional and probably the main factor driving deviations from model predictions is the strong and different degradation of the mixture components because the antagonism at low-mixture concentrations disappeared when recalculating concentration addition and independent action predictions using the low-end fractions. Yet, both concepts then strongly underestimated lethal toxicity at high-mixture concentrations. This is because the predictions assumed a mixture comprising 94.99% esfenvalerate, 5% pixocystrobin, and 0.01% triclosan at all concentrations. These fractions represent the mixture composition at the end of the experiment in the low-mixture concentrations. In the highest-mixture concentrations, however, the fractions of the components did not change considerably over the 28-d exposure. The single exposure experiments showed that triclosan provoked adult mortality only at very high concentrations, whereas picoxystrobin was very toxic to survival and probably the main driver of adult mortality: at the LC50 of the mixture, picoxystrobin is present at its EC34, whereas esfenvalerate and triclosan at concentrations provoking less than 2 and 1% mortality, respectively, when applied singly. While the mixture comprised 14 to 19% picoxystrobin in the highest mixture concentrations at the end of the experiment, the predictions with the low-end fractions assumed only 5% picoxystrobin at all mixture concentrations over the complete exposure period. Consequently, the predicted mortality at high concentrations reflects mainly the lethal toxicity of esfenvalerate. Hence, much lower mortality was predicted for earthworms exposed to high-mixture concentrations than observed. Thus, at high-mixture concentrations the predictions of lethal mixture toxicity based on the start fractions are more realistic.
The concept of independent action described sublethal toxicity at low-mixture concentrations much better when using the low-end fractions, where almost no triclosan was assumed to be present in the mixture. In contrast, at high-mixture concentrations the predictions based on start concentrations were slightly more accurate. This leads to the assumption that triclosan, which appears to be very toxic to the reproduction of E. fetida when applied singly, may cause sublethal toxicity first after some time of exposure. This assumption is supported by a study on the biochemical and genetic toxicity of triclosan on E. fetida conducted by Lin et al. 32. The study reported that the catalase activity of earthworms exposed to triclosan was higher than in the controls after a 2-d exposure, returned to control levels after a 7-d exposure, and finally decreased under control levels after a 14-d exposure. Catalase is an antioxidant enzyme that targets reactive oxygen species and thus indirectly takes part in the metabolism of contaminants. Lin et al. 32 suggest that E. fetida may tolerate oxidative stress at the first stage of exposure, while the antioxidant defense system is damaged after longer exposure. Hence, in low mixture concentrations, triclosan may be degraded before it could provoke effects, and the measured inhibition of reproduction is caused by the remaining two mixture components, mainly esfenvalerate. Comparing EC values of esfenvalerate when applied singly with the EC values of the mixture at concentrations up to the EC50, shows that mixture concentrations were only minimally higher than esfenvalerate concentrations.
At high-mixture concentrations, the concept of independent action fits the observed sublethal toxicity better when start fractions are used. This is because triclosan is less degraded in higher than in lower mixture concentrations, and the start fraction accords better with the measured fractions at these concentrations. Thus, the observed antagonism at lower concentrations may probably be tracked back to the concentration–dependent degradation of triclosan and picoxystrobin and to the time-dependent effects of triclosan.
When combining the predictions based on the start and the low-end fractions, concentration addition describes lethal mixture toxicity rather well, while independent action is the better predictor for the inhibition of reproduction. It has been suggested that the toxicity of mixtures comprising dissimilar compounds is higher (i.e., closer to concentration addition predictions) on mortality than on the inhibition of reproduction. The proposed explanation is that the inhibition of reproduction by a mixture is commonly based on more specific modes of action in comparison with mortality. Hence, concentration addition is more often observed when the endpoint mortality is studied, and independent action for the inhibition of reproduction 35. This is consistent with the results of the present study.
In the present study, a positive effect of the mixture on maturation time and growth were observed, while the reproduction of the F1 generation was more strongly inhibited than of the F0 generation. Brunninger et al. 12 reported positive effects of the herbicide terbuthylazine on the growth and cocoon production of E. andrei F1 individuals and a correlation of these two traits. They suggested that the increased growth was a result of increased metabolism due to chemical stress. It is likely that chemical stress caused an increase in metabolism also in the F1 individuals exposed to high-mixture concentrations in the present study, resulting in a faster growth and consequently earlier maturation than the F1 individuals in the controls. However, as the growth and maturation of F0 juveniles was not assessed in the present study, it cannot be concluded that this is a transgenerational effect but a general effect of the mixture on E. fetida. Yet, it can be assumed that the energy expense for growth and maturation led to a greater exhaustion of the F1 individuals, resulting in a lower cocoon production than in individuals of the F0 generation exposed to the same mixture concentrations. This assumption is supported by results of tests conducted by international laboratories for the purpose of inter-laboratory calibration (ring tests), which showed an inverse relationship between reproduction and body weight changes 36.
The mixture appeared to be more toxic to the reproduction of the F1 generation than the F0 generation. Mixture concentrations that did not provoke any effects on the F0 generation led to measurable toxic effects on the F1 generation, which is a result of the different slopes of the concentration–response curves for the two concentrations. This change in slope may indicate a change in mode of action. The ACR of the mixture increased from 1.45 for the F0 generation to 4.7 for the F1 generation in the mixture experiment. A study on the effects of phenanthrene to multiple generations of the springtail F. candida also showed a change in ACRs (LC50/EC50) from 1.4 in the F0 generation to 3 and 2.3 in the F1 and F2 generation, respectively, suggesting a shift from narcosis to specific effects 9. The mode of action of the mixture is a combination of the modes of action of the single substances and further investigation is needed to understand whether the increase in the mixture's ACR value from F0 to F1 indicates a shift toward more specific mechanisms. The increasing sublethal and unchanging lethal toxicity indicate however that stronger specific effects may be provoked in the F1 generation than in the F0 generation.
The F1 cocoons, and later the hatched juveniles, were exposed to the remaining concentrations of the three mixture components as the F1 generation grew up in spiked soil. Even though picoxystrobin and triclosan were degraded strongly in the lowest mixture concentrations during the maturation period, measurable concentrations of both components were found in higher mixture concentrations for up to 100 d, and esfenvalerate was not degraded completely in any of the mixtures over the complete maturation period. This means that the mixture components can affect juveniles at least for some time of the maturation period. As concluded from the results on the F0 generation in the present study and the findings of the study conducted by Lin et al. 32 triclosan is very toxic to sublethal endpoints when applied singly but may affect earthworms first after longer exposure time. Beside the biochemical effects of triclosan, Lin et al. 32 also studied the genotoxicity of triclosan to E. fetida by using comet assays. They found no severe DNA damages after 2-d exposures, while after 7-d and 14-d exposures a significant increase in DNA damages occurred in earthworms exposed to triclosan, including damages in the extreme damage class (>75% damage) 32. To our knowledge no studies have been published on potential genotoxic effects of picoxystrobin or esfenvalerate on earthworms or other soil-dwelling animals. But the evidence of triclosan inducing genotoxicity may lead to the assumption that certain DNA damages were evoked in the juvenile F1 earthworms, which then affected their reproduction when adult. This would explain deviations in the sublethal effects observed for the F0 and F1 generation.
Despite the rather fast degradation of the three biocides, the results of the present transgenerational study indicate that a repeated exposure to the investigated biocide mixture may lead to an increase in sensitivity in subsequent earthworm generations as a consequence of cumulative damages as reported in previous studies 9, 10. Acanto can be applied only once to a crop in the growing season, while up to three applications of Sumi-Alpha are permitted (1–2 leaf stage, 3–4 leaf stage, and when infected with fungi). A maximum application of 20 to 40 t of sludge per hectare and 10 years is permitted on grains in Norway (http://www.norvar.no/nv/Fag-prosjekter/Andre-fagressurser/Kunnskapsbase-slam/Spoersmaal-svar-om-slam/Bruk-av-avloepsslam-i-jordbruket-spoersmaal-og-svar). Consequently, the biocides may pose a risk to natural earthworm populations due to repeated exposures leading to damages being transferred to subsequent generations. Single generation exposure experiments are not capable of detecting such long-term risks, and their use in risk assessment practices may be questioned. Moreover, our results on the F0 generation show that predicting mixture toxicity using concentration addition and independent action is not always straightforward. Predicting the mixture toxicity to multiple generations is a more formidable challenge. The predictions made by concentration additions and independent actions are based on effects observed in the F0 generation and are thus only strictly valid for this generation. However, the purpose of using prediction models rather than assessing mixture toxicity in a chemical-by-chemical approach is to simulate realism, and predictions should therefore also be applicable to subsequent generations. Predicting safe exposure concentrations for a given population is a main goal of risk assessment practices. It has been suggested that concentration addition is the more conservative of the two models when the slopes of the individual toxicants' logistic concentration–response curves exceed 1.25 37, as it is the case for the sublethal toxicity in the present study. The concept of concentration addition may therefore be chosen to predict toxicity if consideration is given to the precautionary principle 19. The concept of concentration addition overestimates the EC50 towards the F0 generation regardless of which fractions are used in the model, and concentration addition can thus be ranged conservative for the F0 generation. For the F1 generation, however, the observed EC50 and the concentration addition predictions based on start fractions converge, while concentration addition predictions based on low-end fractions underestimate the EC50 for the F1 generation. Consequently, concentration addition is not strictly conservative for the F1 generation. This shows that even though concentration addition and independent action give more realistic predictions on mixture toxicity than chemical-by-chemical approaches, they are not adequate for reliably predicting mixture toxicity to natural populations that comprise multiple generations. The use of model predictions to assess long-term toxicity of the biocides can have severe consequences because the models do not account for transgenerational changes of effects.
Our results show that simulating more realism in ecotoxicological testing by considering multiple chemicals and generations adds valuable information for predicting the risk chemicals pose in natural systems. They also underline, however, the limitations of the commonly used prediction models for mixture toxicity. We therefore see the need of developing predictive tools that address these topics and thus allow more realistic predictions of mixture toxicity.