Seven out of the nine tested chlorobenzenes had an effect on F. candida adult survival in both soils within the applied concentration range of 0 to160 (−1000) mg kg−1 dry soil. In OECD soil, eight out of nine compounds had an effect on reproduction, while for LUFA2.2 soil all nine compounds yielded an EC10 and EC50. The estimated threshold values (EC10, EC50, and LC50) expressed on a molar basis per kilogram of dry soil are given in Table 2. The EC10 and EC50 values for both soils increased with increasing logKow, going from 1,4-dichlorobenzene to pentachlorobenzene. For the LC50, this trend was not that prominent. In case of LUFA2.2 soil, the lethal effects occurred between 323 µmol kg−1 dry soil for 1,2,3,4-tetrachlorobenzene and 436 µmol kg−1 dry soil for 1,2,4-trichlorobenzene, but most compounds had a LC50 around 350 µmol kg−1 (Table 2). Comparing chemicals with the lowest and highest logKow, the LC50 differed only by 5 µmol kg−1 dry weight in LUFA2.2 soil and 34 µmol kg−1 dry weight in OECD soil. This finding confirms the older studies by van Gestel and Ma 14 with the earthworms Eisenia fetida and Lumbricus rubellus, in which the LC50 differences, especially between tetra- and pentachlorobenzene, were marginal. As shown in Figure 1, regressions of LC50 values against logKow for both soils are comparable when LC50 values are plotted as mmol L−1 interstitial water (estimated concentrations). In case of 1,4-dichlorobenzene, 1,2,3,4-tetrachlorobenzene, and pentachlorobenzene, LC50 values were actually overlapping or the same. Furthermore, a comparison with LC50 values derived from polycyclic aromatic hydrocarbons (PAHs) taken from Droge et al. 9 indicates that molecular size of organic pollutants might have different effects on the survival of F. candida (Fig. 1).
Table 2. Chlorobenzene concentrations causing 10% and 50% reduction of reproduction (EC10, EC50) and 50% mortality (LC50) of Folsomia candida exposed for 28 d to LUFA2.2 and OECD soil
|Mmol kg−1a||Mmol kg−1a||Mmol kg−1a||Mmol kg−1a||Mmol kg−1a||Mmol kg−1a|
| 1,4-dichlorobenzene||498 (456–539)||575 (514–635)||349 (320–377)||571 (524–617)||651 (598–703)||932 (856–1007)|
| 1,2,3-trichlorobenzene||233 (210–255)||378 (348–407)||341 (277–404)||211 (193–228)||432 (397–466)||701 (644–757)|
| 1,2,4-trichlorobenzene||209 (192–225)||443 (407–478)||436 (405–466)||254 (233–274)||487 (447–526)||451 (414–487)|
| 1,3,5-trichlorobenzene||225 (201–248)||261 (251–270)||346 (314–377)||230 (211–248)||309 (283–334)||679 (624–733)|
| 1,2,3,4-tetrachlorobenzene||123 (104–141)||183 (166–199)||323 (252–393)||98 (90–105)||281 (258–303)||828 (760–895)|
| 1,2,3,5-tetrachlorobenzene||60 (49–70)||289 (249–328)||394 (358–429)||88 (80–95)||269 (247–290)||625 (574–675)|
| 1,2,4,5-tetrachlorobenzene||170 (148–191)||183 (158–207)||NO||117 (107–126)||319 (293–344)||NO|
| Pentachlorobenzene||143 (130–155)||155 (146–163)||354 (313–394)||135 (124–145)||216 (198–233)||898 (825–970)|
| Hexachlorobenzene||ND||1320 (945–1694)||NO||NO||NO||NO|
Figure 1. Relationships between median lethal concentration (LC50) values for the effect on survival of Folsomia candida after 28 d of exposure to chlorobenzenes and logKow. Hexachlorobenzene and 1,2,4,5-tetrachlorobenzene are not included, as no lethal effect was observed up to a concentration of 1,000 mg kg−1 dry soil. White squares = Organisation for Economic Co-operation and Development soil; black squares = LUFA2.2 soil; grey triangles = polycyclic aromatic hydrocarbons (PAH; naphthalene, phenanthrene, and pyrene) data as reported by Droge et al. 9.
Download figure to PowerPoint
Two compounds showed toxicity that was dissimilar from the other chlorobenzenes. Hexachlorobenzene was not toxic to adult animals over the complete concentration range and only affected reproduction in LUFA2.2 soil. This is in agreement with earlier findings in soil toxicity tests with F. candida and other soil-dwelling organisms that indicated reduced or no toxic effects on exposure to compounds with a logKow > 5.5. Droge et al. 9, for instance, found no lethal or reproductive effects for PAHs with a logKow higher than 5.54 at a maximum concentration of 990 mg kg−1 dry soil. In toxicity tests with a related species, Folsomia fimeteria, Sverdrup et al. 8 found no toxic effect of benz[a]anthracene (logKow 5.9) and chrysene (logKow 5.8) at nominal concentrations of 980 mg kg−1 dry soil in a soil with 1.6% organic carbon. With a logKow of 5.73, hexachlorobenzene is positioned in the same hydrophobicity range and displays a comparable mode of action. The fact that effect concentration values could be calculated for LUFA2.2 soil might derive from indirect effects on F. candida or its eggs, or a slow physiological response. Both aforementioned studies with PAHs also indicated a logKow threshold of 5.2 with a toxicity of pyrene similar to pentachlorobenzene in our compound series (logKow 5.18). This “super-lipophilic” range beyond logKow > 5.2 needs careful attention in future studies. The occurrence of exceptional effects like the toxicity of hexachlorobenzene in LUFA2.2 or additive effects in mixtures cannot be excluded.
More exceptional is the effect of 1,2,4,5-tetrachlorobenzene, with a clear impact on reproduction, but not on the survival of F. candida. Possible explanations for this remarkable response that was not observed for the other two tetrachlorobenzenes might be (1) an efficient metabolism in the adult individuals, as also shown for polycyclic compounds 32, effectively detoxifying the compound, but affecting the reproduction; or (2) an alternative compound property such as the melting point, as presented by Mayer and Holmstrup 33. In their experiment, mortality of 41- to 44-day-old adult F. candida only occurred when exposed to compounds with a melting point below 110°C. Chemicals with higher melting points did not have an effect within 7 d of exposure. The results for 1,2,4,5-tetrachlorobenzene are consistent with this finding because its melting point is 140°C, whereas for the other isomers, it is 47 and 54°C, respectively. Additionally, Hurdzan et al. 22 compared the toxicity of tetrachlorobenzenes to the terrestrial oligochaete Eisenia andrei and the aquatic oligochaete Tubifex tubifex and found no effects on adult survival. However, in none of these studies were effects on reproduction determined, which makes it impossible to generalize the observed pattern for 1,2,4,5-tetrachlorobenzene: the induction of effects only on reproduction. Based on traditional approaches of logKoc and logKow, a narcotic effect would have been expected. Alternative compound properties, such as melting point, cannot solely explain the effect on the reproduction. In this respect, QSAR development requires taking into account not only the mode of action of a compound but also a time of action of the test system and the test organism's possible life stage of action. An underestimated issue in soil ecotoxicological tests is the differentiation in terms of developmental stages in the life cycle of the test organisms. It is possible that several compounds cause their toxic effect only on eggs and not on adults. Toxicity therefore might result from effects on adult, hatchling, or dormant stages, or be multilevel. Taking into consideration that eggs are not only morphologically and physiologically different and have a higher metabolic rate, but are also immobile, toxic compounds can influence them differently compared to mobile adults. It is therefore recommended to include egg tests in the future to discriminate sensitive life stages.
Quantitative structure-activity relationships
Figure 2. Relationships between the effective concentration for 10% (EC10) (a) and 50% (EC50) (b) values for the effect of chlorobenzenes on reproduction of Folsomia candida and the logKow in Organisation for Economic Co-operation and Development soil (white squares) and LUFA2.2 soil (black squares).
Download figure to PowerPoint
Tri- and tetrachlorobenzene isomers with respective similar logKow and logKoc values form distinct clusters that are more compact in OECD soil and for the EC10 (Fig. 2a) than in LUFA2.2 soil and for the EC50 (Fig. 2b). For the EC10 this strong grouping is interrupted just by 1,2,3,5-tetrachlorobenzene for LUFA2.2 soil, which has lower estimated EC10 values in interstitial water as compared with its isomers. The same compound, however, has a higher EC50. A steep concentration-response curve for 1,2,3,5-tetrachlorobenzene probably caused the small differences between EC10 and EC50 values in the LUFA2.2 soil. Chemical properties offer no further generalizable explanation here because similar results were not observed for OECD artificial soil. Likewise, 1,3,5-trichlorobenzene had lower effect concentrations than its isomers in both soils, but only at the EC50 level. Having a lower logKoc than its isomers, a higher interstitial water-based EC50 would be expected instead. Possibly, biodegradation with soil and species-specific thresholds for microbial activity occurred over the 28-d test period, influencing the concentration in soil. Unfortunately, data on biodegradation of 1,3,5-trichlorobenzene in soil or any natural environment are scarce. Volatilization cannot be excluded, but should have a similar effect on the EC10, which is not the case.
Another observation that can be made from Figure 2 is that the distance between the regression lines for the two test soils is different. The OECD has effect concentrations in interstitial water that are lower than those in LUFA2.2 for both effect concentration values. Estimated EC50 and EC10 values should, however, be more or less equal, because physiological responses to toxic compounds can be expected to be related to the concentration in interstitial water only. Therefore, the differences are most probably due to variations in the logKoc values, which were used to calculate the aqueous concentrations. Ter Laak et al. 34, for instance, reported logKoc to vary by a factor of 1.6 to 3.5 for soot- and coal-free low organic content soils. In the present study, the following soils with different origin of their organic matter were used: sphagnum peat used in OECD soil is derived from slowly decomposed freshwater biomass, whereas the LUFA2.2 soil contains natural organic matter derived from grassland soils. From sorption QSARs it can be derived that aquatic organic matter generally sorbs organic chemicals stronger than terrestrial organic material 12, an observation that might explain the above discrepancy between the two soils.
Solid-phase microextraction and model validation
Polyacrylate fiber-water partition coefficients (logKfiber) determined for the eight chlorobenzenes are shown in Table 3. Dichlorobenzene was excluded from the SPME analyses because its estimated half-life in the thin fiber coating was too short in comparison to the time needed for processing and extracting the fibers. Hexachlorobenzene is not included as it was already discarded from the QSAR due to its deviating behavior. The Kfiber values appear to be somewhat lower, yet proportional to octanol-water partition coefficients (logKow) (see Fig. 3). The values were used for the measurement of chlorobenzene concentrations in interstitial water of LUFA2.2 soil. The resulting EC10 and EC50 values are given in Table 3. In Figure 4, the tri-, tetra-, and pentachlorobenzene data are compared to the logKow of the compounds. Although correlations similar to those in Figure 2 are observed, predicted (Eqns. 2–5) and SPME measurements-derived models are significantly different for the EC10 (p = 0.03) and the EC50 (p = 0.002) in a pair-wise t test. Equations based on measured concentrations in interstitial water are
Table 3. Polyacrylate fiber-water partition coefficients (logKfiber) for selected chlorobenzenes and measured concentrations in interstitial water of LUFA2.2 soil at effective concentration for 10% (EC10) and 50% (EC50) levels for the effect on the reproduction of Folsomia candidaa
|µmol L−1||µmol L−1|
|1,2,3-trichlorobenzene||3.91 (0.03)||14.2 (1.12)||28.54 (0.94)|
|1,2,4-trichlorobenzene||3.82 (0.03)||7.89 (0.49)||14.48 (0.68)|
|1,3,5-trichlorobenzene||3.83 (0.04)||13.5 (2.02)||15.84 (1.06)|
|1,2,3,4-tetrachlorobenzene||4.51 (0.02)||3.20 (0.21)||4.75 (0.17)|
|1,2,3,5-tetrachlorobenzene||4.39 (0.04)||1.99 (0.03)||11.69 (0.24)|
|1,2,4,5-tetrachlorobenzene||4.35 (0.03)||5.80 (0.15)||7.17 (0.37)|
|Pentachlorobenzene||4.96 (0.01)||1.44 (0.04)||1.67 (0.06)|
Figure 3. Relationship between polyacrylate fiber-water partition coefficients (logKfiber) and octanol-water partition coefficients (logKow) for chlorobenzenes.
Download figure to PowerPoint
Figure 4. Comparison of the relationships between estimated (Koc-derived) effective concentration for 10% (EC10) (a; grey circles) and 50% (EC50) (b; grey circles) values and measured Solid-phase microextraction (SPME)-derived EC10 (a; black circles) and EC50 (b; black circles) values for the toxicity of chlorobenzenes to Folsomia candida in LUFA2.2 natural soil.
Download figure to PowerPoint
Interestingly, the obtained correlations based on measured concentrations are less significant than the QSARs based on estimated concentrations, even though one would expect accurate measurements to improve the significance. Measured compound-specific concentrations were all lower than corresponding estimated concentrations, with the exceptions of the EC50 values for 1,2,3- and 1,3,5-trichlorobenzene and 1,2,3,5- and 1,2,4,5-tetrachlorobenzene, which were similar. Lower measured concentrations may be due to loss through volatilization or biodegradation, or stronger sorption. However, the latter is most likely because the SPME measurements started already 12 h after soil preparation and were performed in closed systems to which a biocide was added. Assuming that the SPME-derived concentrations reflect actual exposure concentrations, the literature Koc values may provide poor reflections of actual sorption, as also observed for other chemicals 17. This hypothesis is supported by the observation that generally the distance between the estimated and measured EC10 and EC50 values increased with increasing logKow (Figs. 4a and 4b).
Perspectives for REACH
Folsomia candida is generally more sensitive to chlorobenzenes than other soil invertebrates, such as the earthworms E. andrei and L. rubellus7. Thus, tests performed with this standard test organism should be preferred over earthworm tests under the perspective of REACH.
The observed toxicity of chlorobenzenes correlated with their logKow, which agrees with prior findings, with two major exceptions. Hexachlorobenzene, being the most hydrophobic chemical tested, demonstrated only reduced toxicity, but should be considered in future tests for superlipophilic compounds. However, mixture effects in the presence of other compounds are possible and should be the focus of future studies. Furthermore, the prominent pattern of 1,2,4,5-tetrachlorobenzene, having effects on reproduction only and not on survival, agrees with findings by Hurdzan et al. 7. This indicates that the prediction of potential toxicity of compounds, with a classification based only on hydrophobicity, is not sufficient. However, replacing classical approaches by alternative descriptors is risky. This is illustrated by the fact that QSARs based on melting points cannot predict the present soil toxicity data. The melting point may explain outlier behavior of certain compounds, even groups of isomers, but is so far more an asset than an alternative. More data, also for heterocyclic and polar compounds are therefore required.
The QSARs based on EC10 and EC50 values developed in the present study follow baseline toxicity, as they are proportional to logKow. The differences observed between the natural LUFA2.2 soil and the OECD artificial soil, however, reveal a major problem in soil toxicity modeling. The estimated effect concentration values depend on the applied logKoc to such an extent that the same physiological response appears to occur at different concentrations in the interstitial water. This makes QSARs less accurate for predicting toxicity over a wide range of soils differing especially in the nature and properties of the organic matter fraction. Nevertheless, QSARs remain a very efficient tool for assessing potential risks of individual compounds and compound series. Solid-phase microextraction offers a simple and fast technique to measure concentrations in interstitial water and can help optimizing QSARs for soil ecotoxicity, as presented in the present study.