Author's personal copy

Transcription

1 Science of the Total Environment 408 (2010) Contents lists available at ScienceDirect Science of the Total Environment journal homepage: Understanding toxicity as processes in time Jan Baas, Tjalling Jager, Bas Kooijman Vrije Universiteit of Amsterdam, Department of Theoretical Biology, De Boelelaan 1085, 1081 HV Amsterdam, The Netherlands article info abstract Article history: Received 25 June 2009 Received in revised form 9 September 2009 Accepted 26 October 2009 Available online 6 December 2009 Keywords: Effects in time Mechanistic modeling Mixtures Toxicokinetics Toxicodynamics Studies in ecotoxicology usually focus on a single end point (typically mortality, growth, or reproduction) at a standardized exposure time. The exposure time is chosen irrespective of the properties of the chemical under scrutiny, but should depend on the organism of choice in combination with the compound(s) of interest. This paper discusses the typical patterns for toxic effects in time that can be observed for the most encountered endpoints growth reproduction and survival. Ignoring the fact that toxicity is a process in time can lead to severe bias in environmental risk assessment. We show that especially EC x values for sublethal endpoints can show very distinct patterns in time. We recommend that the test duration for survival as an endpoint should be extended till the incipient LC 50 is observed. Given the fact that toxicity data for single compounds show clear patterns in time, it is to be expected that effects of mixtures will also be strongly dependent on time. The few examples that have been published support this statement Elsevier B.V. All rights reserved. 1. Introduction Studies in ecotoxicology usually focus on a single end point (typically mortality, growth, or reproduction) at a standardized exposure time, which depends on the organism (see Table 1). Exposure time is chosen irrespective of the properties of the chemical under scrutiny. At first glance, it seems to make sense to standardize test duration as it allows a comparison between different chemicals. However the pattern of toxic effects in time differs between chemicals. Already in 1969, Sprague recognized that a single exposure time for all chemicals cannot be given and advised to continue acute tests until mortality ceases (Sprague, 1969). A standard test statistic such as the LC 50 or EC 50 therefore cannot be properly compared between compounds, which even hampers toxicity ranking of chemicals. Furthermore the time patterns differ between species and the endpoint of interest (Alda Alvarez et al., 2006). This implies that an EC 50 for a Daphnia magna reproduction test is incomparable to an EC 50 for fish growth or an EC 50 for algal population growth. In order to improve risk assessments we need to understand the time pattern of effects, which requires a better understanding of the mechanisms leading to toxic effects. This would also enable us to explain why a pulsed exposure could have a different effect in an early life stage than in a later life stage. Measurements of effects at different points in time, is a first step in this direction. While external concentrations may often be all we have they are not sufficient to Corresponding author. Tel.: ; fax: address: Bas@Bio.vu.nl (B. Kooijman). understand or predict effects, so we can either measure the internal concentrations or use the best possible knowledge to estimate them. We also need to understand how effects change in time and why results can differ between endpoints in the same test. Especially in mixture toxicity there are very few studies published that follow endpoints in time. Examples are the studies by Van Gestel and Hensbergen (1997) and Khalil et al. (1996) which both followed sublethal endpoints in time and Baas et al. (2007) where mortality was followed in time. VanGestel and Hensbergen (1997) and Khalil et al. (1996) show that (statistically derived) interactions differ for different endpoints, this was later also shown by Cedergreen et al. (2005). All studies observed different interactions at different exposure times. Only Baas et al. applied a process based model to analyze all data at all points in time in an integrated way, instead of treating the data at each time step as a new data set. Interestingly with the integrated approach no or very minor interactions were found when the dataset was analyzed. These studies emphasize the need to incorporate (mixture) effects at different points in time in (eco-) toxicological studies. In this paper we will discuss the typical patterns for toxic effects in time that can be observed for the most frequently encountered endpoints of growth reproduction and survival. 2. Time dependence of toxic effects, general aspects 2.1. Toxicokinetics and toxicodynamics Part of the reason why toxic effects change in time is related to toxicokinetics. It takes time for the organism to build up body residues of the toxicant, and it is the internal concentration that is the cause of /$ see front matter 2009 Elsevier B.V. All rights reserved. doi: /j.scitotenv

2 3736 J. Baas et al. / Science of the Total Environment 408 (2010) Table 1 Standard test duration for different endpoints and organisms, based on OECD test protocols. Species Endpoint Test duration (day) Daphnia magna Survival/immobility 2 Daphnia magna Total reproduction 21 Several fish species Survival 4 Several fish species Body size 28 Green algae Population growth 4 Folsomia candida Total reproduction 28 Enchytraeus albidus Survival 14 Enchytraeus albidus Total reproduction 42 Eisenia andrei/fetida Survival/growth 28 Mussel (freshwater) Survival (embryo or larvae) 2 the observed effects. Note that when organisms grow during the test, toxicokinetics will be affected due to the influence of dilution by growth and the fact that the surface area to volume ratio will change (Kooijman and Bedaux 1996a). Therefore, it is to be expected that large organisms require longer test durations to reach steady state with their environment than smaller organisms with a comparable shape. However the OECD fish test protocols allow freedom to choose a particular species and fish size, all with the same test duration. However if this was the full story, we would only see toxic effects increase in time until they reach a constant level, when the organism is in steady state with its environment. This is not what is observed, as toxic effects may show unexpected patterns in time and different endpoints of the same organism can show entirely different patterns of effects in time, even with a constant external concentration (Alda Alvarez et al., 2006). The link between the internal concentration and the effect in time (toxicodynamics) therefore also needs to be considered Kinetics Any form of mechanistic modeling starts with a description of uptake and elimination. The simplest way in describing uptake and elimination is by the one-compartment model. In this model the organism is represented by one well mixed compartment and takes up toxicants from its environment with an uptake rate and excretes with an elimination rate. There are two ways to get hold of the kinetic parameters. The first and most widely used method makes use of measured internal concentrations at different points in time from which the kinetic parameters are derived, usually assuming the validity of the onecompartment model. In the second method the kinetic parameters are derived from the development of the toxic effects in time (Zitko 1979, Kooijman and Bedaux 1996b). These two methods do not necessarily lead to the same parameter values. Deriving kinetics from the development of toxic effects in time necessarily depends on the effects at the target site (which can be some specific organ and perhaps even a metabolite). When this is compared with kinetic parameters that are derived from whole body residue concentrations this may lead to a different result as was already pointed out by Zitko (1979). The advantage of deriving the kinetic parameters from the development of toxic effects in time, is that the most sensitive pathway is automatically shown in the experimental data, rather than measuring whole body residues, which do not contain information on the most sensitive pathway. For larger organisms it is feasible from an experimental point of view to distinguish different compartments in the organism and then uptake and elimination can be described for different tissues in the body, as is done in physiologically based pharmaco-kinetic modeling (e.g. Clewell and Andersen, 2004, Krishnan et al., 2002). This comes at the expense of a large number of parameters and therefore a major effort in time and resources is required to obtain all relevant data for the different compartments and their interactions. In practice, this multi compartment modeling is mainly found in mammalian toxicology, where the one-compartment model might be too crude. In these compound specific models several compartments are distinguished with different interactions. Note that if one of the steps is rate limiting the whole system can again be described by the one-compartment model. 3. Time dependence of toxic effects for different endpoints 3.1. Survival For survival it has long been recognized that LC 50 s tend to decrease in time, approaching an asymptote called the incipient LC 50 (e.g. Sprague, 1969). An example for the effect of cadmium on the survival of Folsomia candida is shown in Fig. 1.The reason that LC 50 s follow a monotonously decrease in time lies in the nature of the endpoint and the way in which it is expressed. Survival is a so called quantal response, which means that the number of surviving individuals is counted. The effect is then generally expressed as the fraction of the individuals that responded to the treatment (dead or alive). Because death is irreversible, the LC 50 will strictly decrease in time. The most popular explanation for the observed decrease is that the LC 50 time curve directly reflects the build up of the internal concentration in time. It is assumed that an organism dies when its internal concentration exceeds a certain threshold and that this threshold differs between the individuals in the test population. If this is true the LC 50 shows an exponential decay in time with a rate constant that equals the elimination rate constant of the chemical from the body. However this view on the cause of mortality is debated. The hazard approach assumes that survival is a stochastic process on the level of the individual, implying that mortality is a chance process, instead of instant death when a threshold is exceeded. The truth is probably somewhere between these two views, although fish studies showed that the stochastic component dominated for mortality (Newman and McCloskey, 2000). Under the assumptions of the hazard model, both the toxicokinetics and the toxicodynamics (the increase in the probability to die with an increasing internal concentration) affect the time course of the LC 50 s (see the paper by Ashauer and Brown, 2008). Note that because LC 50 s change in time in a manner that depends on the chemical properties, this implies that great care should be taken when LC 50 s between different compounds are compared. Still Quantitative Structure Activity Relationships (QSARs) for LC 50 s are quite successful, at least for narcotic compounds (e.g. Russom et al., 1997). Often QSARs are based on 4-day fish tests and if the fish are not too big 4 days is enough to reach the incipient LC 50 at least for Fig. 1. Example of the course of LC 50 values in time for Folsomia candida exposed to cadmium (data from Baas et al. (2007)).

3 J. Baas et al. / Science of the Total Environment 408 (2010) compounds with a log octanol/water partition coefficient less than 5 (Jager and Kooijman, 2009). This is caused by the fact that the incipient LC 50 is a metric that does not change in time anymore and thus can be compared between chemicals Sublethal endpoints For sublethal endpoints the EC 50 does not necessarily decrease in time. The EC x -time course is less straightforward than for effects on survival. In contrast to the quantal form for lethal effects, for sublethal effects one does not report the fraction of the test organisms that respond, but rather the degree of the response (body size or reproductive output). Furthermore, one should realize that the value of the EC x, but also its time course, depends on the way that the effect is expressed. An EC x for body length or volume will be different from that of body weight. Similarly, an EC x for cumulative reproduction will be different from that of the reproduction rate. Similar to what was shown by Nyholm (1985) for algae, an EC x based on growth rate is different from an EC x based on biomass. In the figures below some examples are given for different end points and different organisms and the accompanying EC x -time course. In the figures examples of the time course of the effect on reproduction and growth in time are shown. Growth might be slowed down by the toxicant, but the organism still reaches the same maximum (Fig. 2). The maximum length may also be reduced because of the toxic effect (Fig. 3). In Fig. 4 there is no difference in the growth curves for the different dosages for the parent, but the number of juveniles depends strongly on the applied dose. More subtle differences can be found in the reproduction data. The time to first reproduction can be influenced by toxic effects, as illustrated by Figs. 2 and 3. It is clearly shown that EC x values can show distinct effects if plotted against time, even so that the EC x for growth shows an opposite pattern to the EC x for reproduction (Fig. 3) or that the EC x does not depend on time at all (Fig. 4). It is also clear that an incipient EC x does not necessarily exist Population endpoints For tests with algae the effect on the population as a whole is measured. Single celled algae are so small that they are expected to be in equilibrium with the exposure medium rapidly. Therefore the influence of toxicokinetics on the pattern of toxic effects in time is expected to be small. When both the control and the exposed populations grow exponentially, but at a different rate the EC x for the population growth rate and only the EC x for population growth rate will be constant over time Effects of mixtures In mixture studies, usually the observed toxic effects are related to the external concentrations and then compared to the results expected from the reference models of concentration addition (CA) or Effect Addition (EA). Interactions between the mixture components are inferred from how well the observed effects compare with the expected effects for one endpoint at one single point in time. However, if the results do agree with CA, can we be confident that no interactions are occurring? There may be several mixtures for which the effects of individual chemicals or interactions may Fig. 2. Example of the time course of the effect on reproduction and body length of Caenorhabditis elegans exposed to pentachlorobenzene and the resulting time course of EC 5,EC 10 for body length (solid line) and EC 5,EC 10 and EC 50 for cumulative reproduction (dashed lines). Concentrations are in mg/l agar. Data are taken from Alda Alvarez et al. (2006).

4 3738 J. Baas et al. / Science of the Total Environment 408 (2010) Fig. 3. Example of the time course of the effects on body length and cumulative reproduction of Folsomia candida exposed to triphenyltin and the resulting time course of EC 5,EC 10 for growth (solid line) and EC 5,EC 10 and EC 50 for cumulative reproduction (dashed lines). Concentrations are in mg/kg food. Data are taken from Jager et al. (2004). cancel each other out, particularly for mixtures with a large number of compounds. Even if we observe a certain deviation from CA (say antagonism or a ratio-dependent interaction ), what is the underlying cause? Very simple kinetic or dynamic processes may already result in very complicated dynamic behavior. Exposure to a mixture of two compounds with constant environmental concentrations will generally result in a time varying mixture concentration within the body that may show specific patterns in the dose response surface (even without any actual interaction between the mixture components). 4. Conclusions, recommendations and outlook Toxicity is a process in time, and ignoring this can lead to severe bias in environmental risk assessment. The use of process based models in which toxic data are interpreted in an integrated manner can help in interpreting these patterns and will help improve our understanding of toxic effects and will facilitate extrapolation procedures, which are of special importance in effects of mixtures as it is impossible to experimentally assess all mixtures. It is important to build up more knowledge about the time dependence of effects of mixtures. Given the fact that toxicity data for single compounds show clear patterns in time, it is to be expected that effects of mixtures will also be strongly dependent on time. The few examples that have been published support this statement. For survival the properties of the chemical should be determined for how long the experiment should take. Especially narcotics with large K ow values take a long time to get in equilibrium with the organism of choice. The standardized time for the exposure might not be long enough to understand the toxicological mechanisms (Laskowski 1995, Baas et al., 2009). So, the test duration should be extended till the incipient LC 50 is observed, or can be estimated with sufficient accuracy. For narcotics the K ow value can be used to estimate how long an experiment would take, however for other modes of action the incipient LC 50 does not depend on the K ow (Jager et al., 2005). For sublethal endpoints, the test duration and observation times should depend strongly on animal properties. The optimal choice here is a (partial) life-cycle test that starts with juveniles and continues until the maximum size is clear. If possible it should also include standard test duration to enable comparison with earlier experiments. For aquatic organisms it is usually possible to test the organism at intermediate points in time. For effects on mortality or reproduction, it is part of the standard protocol, but normally these data are not further used in the data analysis. It is a waste of resources not to use these data any further as these data do contain information. For soil dwelling organisms it is usually more difficult to get data at intermediate points in time and still have the possibility to continue the experiment. Sometimes the organisms can be kept on compacted soil and in doing so forced to live on top of the soil which makes it very easy to follow them in time (Baas et al., 2007). Another possibility might be to keep the organisms on filter paper. In this way exposure and effect are separated which enable the unraveling of the mechanisms behind the toxic effects. The ecological relevance of the experiment might be subordinate to this wish. Another possibility is to start the experiment with additional doses and sacrifice these at intermediate points in time. Acknowledgements The study was supported by the European Union (EU) integrated project No Miracle (Novel Methods for Integrated Risk Assessment of Cumulative Stressors in Europe; contract

5 J. Baas et al. / Science of the Total Environment 408 (2010) Fig. 4. Example of the effect on body length and reproduction of Folsomia candida exposed to chloropyriphos and the resulting time course of EC 5,EC 10 and EC 50 for cumulative reproduction (dashed lines). Concentrations are in mg/l food. Data are taken from Jager et al. (2007) under the EU theme Global Change and Ecosystems topic development of risk assessment methodologies, coordinated by H. Løkke at the National Environmental Research Institute (DK-8600 Silkeborg, Denmark). References Alda Alvarez O, Jager T, Nunez Colao B, Kammenga JE. Temporal dynamics of effect concentrations. Environ Sci Technol 2006;40: Ashauer R, Brown CD. Toxicodynamic assumptions in ecotoxicological hazard models. Environ Toxicol Chem 2008;27: Baas J, van Houte BPP, van Gestel CAM, Kooijman SALM. Modelling the effects of binary mixtures on survival in time. Environ Tox Chem 2007;26: Baas J, Jager T, Kooijman SALM. Estimation of no effect concentrations from exposure experiments when values scatter among individuals. Ecotoxicol Mod 2009;220: Cedergreen N, Andersen L, Olesen CF, Spliid HH, Streibig JC. Does the effect of herbicide pulse exposure on aquatic plants depend on K-ow or mode of action? Aquat Toxicol 2005;71: Clewell HJ, Andersen ME. Applying mode-of-action and pharmacokinetic considerations in contemporary cancer risk assessments: an example with trichloroethylene. Crit Rev Toxicol 2004;34: Jager T, Kooijman SALM. A biology-based approach for quantitative structure activity relationships (QSARs) in ecotoxicity. Ecotoxicol 2009;18: Jager T, Crommentuijn T, Van Gestel CAM, Kooijman SALM. Simultaneous modeling of multiple end points in life-cycle toxicity tests. Environ Sci Technol 2004;38: Jager T, Alda Alvarez O, Kammenga SALM, Kooijman JE. Modelling nematode life cycles using dynamic energy budgets. Funct Ecol 2005;19: Jager T, Crommentuijn T, van Gestel CAM, Kooijman SALM. Chronic exposure to chlorpyrifos reveals two modes of action in the springtail Folsomia candida. Environ Poll 2007;145: Khalil MA, AbdelLateif HM, Bayoumi BM, vanstraalen NM, vangestel CAM. Effects of metals and metal mixtures on survival and cocoon production of the earthworm Aporrectodea caliginosa. Pedobiologia 1996;40: Kooijman SALM, Bedaux JJM. Analysis of toxicity tests on fish growth. Water Res 1996a;30: Kooijman SALM, Bedaux JJM. Analysis of toxicity tests on Daphnia survival and reproduction. Water Res 1996b;30: Krishnan K, Haddad S, Béliveau M, Tardif R. Physiological modeling and extrapolation of pharmacokinetic interactions from binary to more complex chemical mixtures. Environ Health Perspect 2002;110: Laskowski R. Some good reasons to ban the use of NOEC, LOEC and related concepts in ecotoxicology. OIKOS 1995;73: Newman MC, McCloskey JT. The individual tolerance concept is not the sole explanation for the probit dose effect model. Environ Toxicol Chem 2000;19: Nyholm N. Response variable in algal growth inhibition tests biomass or growth rate? Water Res. 1985;19: Russom CL, Bradbury SP, Broderius SJ, Hammermeister DE, Drummond RA. Predicting modes of toxic action from chemical structure: acute toxicity in the fathead minnow (Pimephales promelas). Environ toxicol chem. 1997;16: Sprague JB. Measurement of pollutant toxicity to fish. I. Bioassay methods for acute toxicity. Water res 1969;3: Van Gestel CAM, Hensbergen PJ. Interaction of Cd and Zn toxicity for Folsomia candida Willem (Collembola:Isotomidae) in relation to bioavailability in soil. Environ Tox Chem 1997;16: Zitko V. An equation of lethality curves in tests with aquatic fauna. Chemosphere 1979;2:47 51.