Correspondence/Rebuttal pubs.acs.org/est
Response to Comment on “Application of the Activity Framework for Assessing Aquatic Ecotoxicology Data for Organic Chemicals” In their comments1 of our recent ES&T paper,2 Goss and Endo express the opinion that the described relationship between chemical activity and toxicity is a “step backward” rather than a “step forward”. We fail to see how an extensive effort to compile, critically evaluate data quality and interpret aquatic toxicity results and examine the data using the activity concept is a “step backwards”. One crucial aspect of this ES&T paper2 was the extensive data evaluation carried out on the studies ultimately used.2 Well over one-third of the studies initially collected for the data analysis were rejected as unfit for purpose due to identifiable errors or uncertainties in the testing data. A second important distinction is that the ES&T publication2 discussed chemicals with putative Mode of Action (MoA) 1 (baseline toxicant) classifications only, but the ES&T paper2 did not include a data analysis for other MoA classifications. A technical report (ECETOC3) included full data sets (including studies evaluated as fit for use and unfit for use) for MoA 1 substances and other putative MoA classifications. It is from the technical report,3 not the ES&T publication,2 that Goss and Endo conduct their analysis and include comparisons of the chemical activity approach and the Critical Target Membrane Concentration (CTMC) concept. Goss and Endo1 ignored the data quality assessment and apparently used the complete databases from the technical report3 for putative MoA 1 and 2 substances. Their calculations1 thus appear to include many data that were not included in the ES&T publication.2 When these test results are removed as they were in the ES&T publication,2 some points in the Goss and Endo comments become redundant, notably the comment on the estimation method for solubility of miscible compounds as no data on miscibles were used in the validated data set in the ES&T paper.2 Despite the lingering ignored data quality issues, the results from the chemical activity and CTMC approaches for MoA 1 substances (Figure 1 in1) are largely consistent for reasons discussed in related work elsewhere.4,5 Other MoA classifications were not addressed in the ES&T publication but were one of the subjects of an international ECETOC expert workshop on “Defining the role of chemical activity in environmental risk assessment within the context of mode of action: Practical guidance and advice” held in October 2015 prior to the SETAC NA meeting at Salt Lake City. Within the activity framework, the difference in toxicity between MoA 1 and 2 substances can be accounted for using alternative partition coefficients more appropriate to membrane partitioning, log Kdmpc for instance, which accounts for apparent differences between toxicity of MoA 1 and 2 substances by realigning to membrane lipids in a similar way to the CTMC concept. We feel compelled to respond to these unbalanced criticisms to avoid leaving ES&T readers with the impression that there is little merit in the activity approach. The basis of both the activity and CTMC concepts is that lipid membranes in living organisms are the site of toxic action for MoA 1 substances. This is generally accepted, as is the view that aquatic toxicity © XXXX American Chemical Society
testing is a complex dynamic process in which concentrations in membranes and other tissues rise approaching equilibrium ultimately causing death. The methods differ in their technique for estimating that membrane concentration. According to the method we have used in the ES&T paper,2 activity assumes that the membrane concentration is related to the chemical activity in the external aqueous medium which is essentially the fraction of a specified aqueous solubility, whereas CTMC employs an estimated equilibrium membrane−water partition coefficient. There are advantages to both methods. Obviously the optimal scientific path forward is to explore the merits and limitations of both approaches and various others, e.g. Critical Body Residues and LC50s. Goss and Endo focus on the uncertainties and limitations of the chemical activity approach but fail to address the corresponding uncertainties and limitations of the CTMC concept. Generating an artificial membrane that perfectly mimics biological membranes is challenging and beyond the capabilities of most laboratories. Different artificial membranes have been suggested and used by different research groups, but little is known about the partitioning differences between artificial and natural actively functioning membranes. Further, the experimental determination of partition coefficients is always associated with experimental and analytical error, which for some difficult-to-test compounds can be considerable. We question if membrane−water partition coefficients are “easier to measure than water solubility”. In most cases, such experimentally determined partition coefficients are not even available and must be estimated using, for example, poly parameter linear free energy relationships (PP-LFERs), which introduces additional error, uncertainty and often restrictions since each PP-LFER has its domain of applicability. Goss and Endo used estimated rather than experimental partition coefficients in their calculations for Figure 1,1 but did not address the uncertainties and errors associated with this procedure. Further, the correct application of partition coefficients implies that they are determined and applied at the same concentration range or that a linear isotherm can be assumed. These conditions are not demonstrated to be satisfied for all the CTMC calculations that lead to Figure 1, since the lower and intermediate membrane concentration range is within the Henry’s law domain whereas the highest concentrations (>1 M) extend into the Raoult’s law domain. Overall, the CTMC concept is exactly like other approaches and has associated uncertainties, assumptions and error sources. Goss and Endo recalculated our data to obtain critical membrane concentrations and obtained MoA 1 results that largely cross-validate rather than challenge our data analysis. We do concur with Goss and Endo that the CTMC concept has a lot to offer within this area, but find their claim that it is a superior method unjustified.
A
DOI: 10.1021/acs.est.6b00864 Environ. Sci. Technol. XXXX, XXX, XXX−XXX
Environmental Science & Technology
Correspondence/Rebuttal
(4) Mackay, D.; Arnot, J. A.; Petkova, E. P.; Wallace, K. B.; Call, D. J.; Brooke, L. T.; Veith, G. D. The physicochemical basis of QSARs for baseline toxicity. SAR QSAR Environ. Res. 2009, 20 (3−4), 393−414. (5) Smith, K. E. C.; Schmidt, S. N.; Dom, N.; Blust, R.; Holmstrup, M.; Mayer, P. Baseline Toxic Mixtures of Non-Toxic Chemicals: “Solubility Addition” Increases Exposure for Solid Hydrophobic Chemicals. Environ. Sci. Technol. 2013, 47 (4), 2026−2033. (6) Ferguson, J. The use of chemical potentials as indices of toxicity. Proc. R. Soc. London, Ser. B 1939, 127, 387−404. (7) Mayer, P.; Holmstrup, M. Passive dosing of soil invertebrates with polycyclic aromatic hydrocarbons: limited chemical activity explains toxicity cutoff. Environ. Sci. Technol. 2008, 42 (19), 7516− 7521.
The principal advantage of the activity concept as advanced by Ferguson in 19396 is that conversion from exposure concentrations to unit-less chemical activity reduces the variation between chemical concentrations by many orders of magnitude for various species. The vast majority of the toxicity data were within or near the expected activity range of 0.01− 0.1. Chemical activity has also a more practical side that is worthy of mention. In recent years, novel dosing methods have been developed which allow chemical activity to be controlled precisely in laboratory tests. With these novel techniques it is possible to link toxicity to chemical activity directly, which has already been done in several studies and led to novel results that could not be obtained with membrane water partition coefficients.5,7 The contributions of the chemical activity framework within the broader issue of environmental risk assessment of chemicals are still in its infancy, and it will likely take some years to fully explore its potential and limitations. The activity concept has the potential to contribute to a more rigorous and transparent evaluation of the proximity of monitored environmental concentrations to levels at which adverse effects are likely. In conclusion we do not agree with Goss and Endo that the activity concept represents a “step backwards”.
Paul Thomas† Donald Mackay‡ Philipp Mayer§ Jon Arnot∥ Malyka Galay Burgos*,⊥ †
■
CEHTRA/KREATiS, ZAC de Saint Hubert, 23 rue du Creuzat, 38080 L’Isle d’Abeau, France ‡ Trent University, 1600 West Bank Drive, Peterborough, Ontario K9J 7B8, Canada § Technical University of Denmark, Department of Environmental Engineering, Lyngby, Denmark ∥ ARC Arnot Research & Consulting Inc, 36 Sproat Avenue, Toronto, Ontario M4M 1W4, Canada ⊥ European Centre for Ecotoxicology and Toxicology of Chemicals (ECETOC), Avenue E. Van Nieuwenhuyse 2, B-1160 Brussels, Belgium
AUTHOR INFORMATION
Corresponding Author
*Phone: +32 2−663 3812; e-mail: malyka.galay-burgos@ ecetoc.org. Notes
The authors declare no competing financial interest.
■
REFERENCES
(1) Goss, K.-U.; Endo, S. Comment on “Application of the Activity Framework for Assessing Aquatic Ecotoxicology Data for Organic Chemicals”. Environ. Sci. Technol. 2016, DOI: 10.1021/acs.est.5b05534. (2) Thomas, P.; Dawick, J.; Lampi, M.; Lemaire, P.; Presow, S.; van Egmond, R.; Arnot, J. A.; Mackay, D.; Mayer, P.; Burgos, M. G. Application of the Activity Framework for Assessing Aquatic Ecotoxicology Data for Organic Chemicals. Environ. Sci. Technol. 2015, 49 (20), 12289−12296. (3) ECETOC. Activity-Based Relationships for Aquatic Ecotoxicology Data: Use of the Activity Approach to Strengthen MoA Predictions, Technical Report No. 120; European Centre for Ecotoxicology and Toxicology of Chemicals: Brussels, Belgium, 2013. B
DOI: 10.1021/acs.est.6b00864 Environ. Sci. Technol. XXXX, XXX, XXX−XXX