Food mediated life history strategies in Daphnia magna : their relevance to ecotoxicological evaluations

The waterflea <em>Daphnia magna</em> is a widely used test organism in ecotoxicological studies. Acute and chronic laboratory tests yield basic information for the development of water quality standards, assessment of potential hazards of (new) chemicals, waste water licences and sanitation measures for contaminated sediments. Environmental risk assessment also includes extrapolation from laboratory to field, for which theoretical models are applied. Reliable results can only be obtained if:<UL><LI>toxicity test results are both accurate and reproducible, and</LI><LI>extrapolation models take account of major ecological processes.</LI></UL>It has been recognized that under current international guidelines for <em>Daphnia</em> toxicity tests interlaboratory variation of test results is disappointingly large (Cowgill, 1987; Baird <em>et al.,</em> 1989a). Differences in test conditions and culture techniques are regarded as a main source of variation, but their relative contributions are largely unknown.The distance between a single-species test under laboratory conditions and the response of an aquatic ecosystem is enormous, in terms of complexity. A first step has been made by Kooijman, who developed a model to predict effects of toxicants on <em>Daphnia</em> populations from effects on individuals (Kooijman, 1986; Kooijman <em>et al.,</em> 1989). From comparisons between experimental and simulated population dynamics research questions regarding critical model assumptions were derived.The present study focuses on food availability, which plays a central role in the life history of <em>D.</em><em>magna.</em> Its development and reproduction are dependent on food level, as well as its sensitivity to toxic stress. The importance of food for the culture of test animals was investigated, as a contribution to standardisation of toxicity tests. Besides, assumed relationships between food level and <em>Daphnia</em> growth and reproduction are validated, in order to support development of extrapolation models. The experimental approach has been chosen to address the research questions related to both toxicity testing and extrapolation models. New techniques were developed for that purpose. Cadmium, chromium and lead, which are designated as priority pollutants by the International Rhine Committee (1987), have been applied as model substances. The investigations have been carried out at the Instute for Inland Water Management an Waste Water Treatment in Lelystad.<strong>New experimental methods</strong>Length measurements are basic to fife history research. Therefore, electronic sizing of living <em>D.</em><em>magna</em> was pursued by development of a computer program, which was based on an existing image processing system. This method, which is described in <em>chapter 2</em> , was much more rapid than manual measurements with a microscope, although it was less precise. Image analysis is recommended for determination of large samples, e.g. length frequency distributions of populations. Debris in the size range of the objects has to be removed. When it comes to precision, for instance determination of growth curves, the manual method should be used.Competition mechanisms within a <em>Daphnia</em> population were investigated with a newly developed culture system <em>(chapter 6</em> ). it consisted of two interconnected culture vessels, each holding a cohort. An air-driven flow of medium plus algae circulated between these vessels, hence allowing the cohorts to share the same food conditions. The system proved to be a useful tool for competition studies. It enables detailed studies of the exploitation of a common food source by competitors that are similar in morphology, or should be kept apart for other reasons. Different, externally driven food dynamics can be simulated. With a few adaptations the system can easily be extended to hold three or more competing groups.<strong>Standardization</strong>Current international guidelines predominantly aim at standardisation of conditions during toxicity tests. Therefore, culture methods, i.e. pre-test conditions, differ among research laboratories. As an example, methods practised at the Institute for Inland Water Management and Waste Water Treatment (RIZA) are described in <em>chapter 2</em> . The experiments in the present study were conducted with <em>D. magna</em> cultered in natural water from Lake Ijssel. The green alga <em>Chlorella pyrenoidosa</em> was used to feed the <em>Daphnia.</em> Transfer cultures were supplied with constant daily food rations, and young were removed three times a week. Hence, the density of algae varies with cumulative filtration rates of the mother cohort and the progeny which has not yet been removed. This may carry over into the results of ecotoxicological studies, as maternal feeding conditions can affect the quality of the young. Cowgill <em>et al.</em> (1985) showed an inverse relationship between brood size and weight of newborn <em>D. magna</em> in a laboratory population. They suggested the existence of a reproductive strategy, which earlier had been proposed by Hutchinson (1951) on the basis of field observations. According to this strategy females spread their genes around under favourable conditions by producing many small, 'cheap' neonates, while heavy, stress-resistent young are born when food is sparse. An important objective of the present thesis was to test the validity of this strategy and to explore ecotoxicological and ecological consequences by experimentation <em>(chapters 3 and 4</em> ).It could be demonstrated in our studies that reproduction of females, which were exposed to an abrupt increase or decrease of food ration between the third and the fourth brood, was fully adapted to the new situation in three instars, or c. eight days at 20°C (standard laboratory temperature). Brood sizes ranged from 7 to 76 and neonate carapace lengths from 1.1 to 0.92 mm. The smallest young (0.80 mm) were observed in the first brood. The latter is a well-known phenomenon that is accounted for in toxicity test guidelines (e.g. OECD, 1984) by discouraging the use of the first brood. The main reason for this advice is reduced survival in controls, which may invalidate the test (e.g. Cowgill <em>et al.,</em> 1986). The present study shows that large neonates contained more lipid reserves than small ones. Therefore we may expect a positive correlation between survival at starvation and body size. This was confirmed in the starvation experiments described in <em>chapter 5</em> and Enserink (1989), but not in similar experiments in <em>chapter 4</em> . In connection with unfed 48 h toxicity tests it is worth noting that median survival times never fen below 4 d, even for the smallest (0.77 mm) neonates.An even more important reason to pay attention to neonate size is that small neonates can be more sensitive to toxicants than large young in acute (48 h) tests ( <em>chapter 3</em> ). When exposed to cadmium there was a threefold difference between the LC _{<font size="-2">50</font>} 's for small and large animals. However, no such effect was found in toxicity tests with chromium(VI), i.e. LC _{<font size="-2">50</font>} 's were similar in all tests, which probably relates to different modes of action of these two metals.In chronic (21 d) toxicity tests the initial size of test animals appeared to be less important ( <em>chapter 4</em> ). Whereas the trends observed in acute tests were consistent in several trials (e.g. Enserink, 1989), this was not the case in chronic tests. The results suggest that environmental conditions during the test mask subtle differences in neonate size and lipid reserves. The variation in effect concentrations caused by non-simultaneous replication was in the same order of magnitude, i.e. a factor of 2, as the effects of initial body size. Probably, food supply during these tests was inadequately controlled.The most conspicuous conclusion of the present studies is that maternal food ration is of paramount importance to neonate fitness and hence to the results of acute toxicity tests, at least for some chemicals (Enserink, 1989; Enserink <em>et al</em> ., 1990). Similar results were obtained independently by Baird <em>et al</em> . (1989b). More recently, a number of other investigators have confirmed these findings.The reproductive strategy in <em>D.</em><em>magna</em> was also observed by Cox <em>et al</em> . (1992), Naylor <em>et al</em> . (1992) and Viganò (1993). According to Naylor <em>et al</em> . (1992) the inverse relationship between maternal food level and neonate size also holds for dry weight, although this parameter appeared to be more variable than body length. Even under normal culture conditions, when food level is not purposely varied, brood to brood oscillations of neonate weight and length were inversely related to oscillations of mean clutch size (Viganò,1993; Lazorchak & Waller, 1993). Such variation can be caused by unstable quality or quantity of food supply, increased grazing capacity owing to body growth and periodic presence of neonates.The results of the current studies suggest that the influence of maternal food conditions on the results of acute (48 h) toxicity tests depends on test substance. Cadmium toxicity was modified by neonate size, but the toxicity of chromium remained unaffected. The largest influence was reported for 3,4-dichloroaniline, i.e. a factor of 6 (Baird <em>et al</em> ., 1991). A significant positive correlation between neonate size and LC _{<font size="-2">50</font>} was found for sodium bromide and 3,4-dichloroaniline (Naylor <em>et al</em> ., 1992), and copper (Lazorchak & Waller, 1993). Viganò (1993) found no effect of neonate size on LC _{<font size="-2">50</font>} 's for ethylbenzene and <em>n</em> -butylbenzene, but the variation of test results was very limited, as was the size range of the test animals. For cadmium however, the effect of maternal ration was confirmed (Baird <em>et al</em> ., 1991), but neonate length showed no correlation with LC _{<font size="-2">50</font>} in the experiments of Naylor <em>et al</em> . (1992), which is inconsistent with the results of the present studies. Several differences between our study and that of Naylor <em>et al.</em> (1992) might account for these results, e.g. the presence of food during the test, the use of another clone and a different test medium. It is of interest to note that both studies were carried out according to standard, but different, test protocols. Whereas maternal food ration, which was purposely varied, accounted for a within-laboratory variation of a factor 3 in the present study and the experiments of Baird <em>et al.</em> (1991), who worked at the same laboratory as Naylor <em>et al.</em> (1993), a 20-fold difference occurred between the laboratories. From the viewpoint of standardization, sources of variation within and between laboratories are equally important.The significance of maternal food conditions for toxicity tests has been clearly demonstrated above. It is therefore recommended to include pre-test conditions in standardization programmes and test guidelines. Several other factors are known to be relevant, for instance feeding conditions during the test (Winner <em>et al.,</em> 1977; Chandini, 1988a, b; Soares, 1989; Lazorchak & Waller, 1993; Sims <em>et al.,</em> 1993; Klüttgen & Ratte, 1994), genotype (Soares, 1989; Baird <em>et al.,</em> 1991), medium (Winner, 1985) and even statistical evaluation of test results (Hoekstra, 1993). However, very little is known about their relative contributions and interactions. Ring-tests are excellent instruments to investigate such combined effects. At present, an international ring-test is conducted to improve the OECD guidelines (1984) for chronic toxicity tests, with respect to genotype, medium and feeding during the test. This investigation is coordinated by the University of Sheffield (UK), Department of Animal & Plant Sciences. In order to obtain a sufficient overview of sources of variation within and between laboratories and to identify the most important factors, further investigation is needed. Current guidelines do not garantee standardized test results in the strict sense of the word, which decreases the reliability of safe levels for water management and of bioassays conducted for regulatory purposes. Therefore, a collective decision should be made on:<UL><LI>the desired quality of toxicity test results for water management and</LI><LI>the amount of detail in culture and test protocols that is required to meet this quality.</LI></UL><strong>Extrapolation</strong>The phenotypic plasticity of <em>D. magna</em> in response to its food source is impressive. All major life history traits can adapt rapidly to food availability, which undoubtly has evolutionary significance as <em>Daphnia</em> experiences a nutritionally variable environment during its lifespan. The challenge to designers of simulation models is to describe and incorporate those relationships which areindispensable for attaining the goals of the model.In the model of Kooijman (1986) for growth and development of individual <em>Daphnia</em> a balance between mathematical simplicity and biological realism was pursued. Modelled individuals were aggregated into simulated single-species populations (Kooijman <em>et al.,</em> 1989; Van der Hoeven, 1991). In this way, physiological effects of toxic chemicals can be translated into population dynamics, which was regarded as a step towards modelling the response of ecosystems to toxic stress. Ecosystems contain many interacting populations. In order to avoid a complex tangle of detailed sub-models a collection of assumptions was produced, partly based on conceptions of physical mechanisms, partly on the premise that everything is extremely simple unless it proves to be more complex and partly on empirical data. The idea was to strip details from the sub-models in the process towards the ecosystem model. The <em>Daphnia</em> model was regarded as a test case for a more general model, which should be applicable to many species (Kooijman <em>et al</em> ., 1987). Nevertheless, a certain amount of detail is nescessary for meaningful extrapolations to higher levels of organisation.A comprehensive list of Kooijman's (1986) model assumptions at the level of the individual is given by Van der Hoeven (1991). In the present thesis a number of these assumptions was evaluated experimentally:<UL><LI>neonate size is fixed;</LI><LI>size-specific storage of a neonate is identical to that of its mother at the moment of egg formation;</LI><LI>a minimum size is required for reproduction;</LI><LI>a fixed portion of utilized energy is spent on reproduction and the remainder on growth and maintenance;</LI><LI>ingestion rate is proportional to body surface;</LI><LI>energy costs for maintenance are proportional to body weight.</LI></UL>An attempt was made to assess their relevance for extrapolation.In <em>chapters 3 and 4</em> a reproductive strategy for <em>D. magna</em> in reponse to food availability is described. Body size and lipid content of neonates appeared to be inversely related to maternal food ration. Cox <em>et A</em> (1992) observed no further decrease of neonate size when food was no longer restrictive, i.e. above the incipient limiting level. These findings do not agree with two model assumptions: size at birth is independent of food level and size-specific energy storage of a neonate is identical to the size-specific storage of its mother. The reproductive strategy mentioned above has important consequences for survival and development of neonates under low food conditions. Survival time at starvation can increase from 4 to 9 days with increasing neonate size <em>(chapter 5</em> ). In addition, large and fat young can develop into their third instar without food <em>(chapter 4</em> ). This may decrease time to first reproduction, as will be shown below.In the model, deposition of the first clutch, i.e. the primiparous instar, is initiated as the female attains a certain body size, which is independent of food level. In <em>chapter</em> 5 however, body sizes of primiparae differed significantly among food rations, with minimum sizes at intermediate rations. A threshold body length for maturation was found two instars earlier, i.e. just before the preadolescent instar, in concordance with Ebert (1992). However, this threshold was not completely independent of food. At low food rations trade-offs occurred between growth and reproduction. One group of animals delayed their first brood in favour of body growth. The young of these delayed broods were significantly larger than young that were born earlier at the same maternal food ration, probably owing to increased instar duration. It was concluded that a minimum energy requirement for reproduction was met at this food ration.A remarkable phenomenon was observed at low food rations. Allocation of energy reserves, i.e. yolk, to the ovaries appeared reversible. Accumulation during the first half of an instar and redistribution towards the end of the instar suggested that energy needs for the production of a new carapace have priority over egg production. As yet mathematical models of <em>Daphnia</em> do notallow redistribution of material allocated to reproduction (Hanstveit <em>et al.,</em> 1987; Gurney <em>et al.,</em> 1990).From the distribution of biomass during the production of the first brood it was concluded that the proportion of energy allocated to reproduction increases with food availability. Similar results were obtained by McCauley <em>et al.</em> (1990), who improved the simulations of their <em>Daphnia</em> model by incorporating a food-dependent fraction to reproduction (Gurney <em>et al.,</em> 1990). In the model of Kooijman, however, the fraction of assimilated energy allocated to reproduction is assumed constant, except for starvation conditions.Combined effects of food ration and exposure to lead on maturation were also investigated in <em>chapter 5</em> . A significant interaction between food and lead concentration was observed, which means that toxic effects were dependent on food level. Body growth was reduced at both food levels, leading to smaller primiparae with smaller progeny at abundant food, whereas delayed maturation and egg mortality were observed at the low food ration. Except for egg mortality, the effects of lead on maturation resemble increased food stress. Similar observations have been made in experiments with copper (Winner <em>et al.,</em> 1977) and cadmium (Chandini, 1989; Klüttgen & Ratte, 1994). Extrapolation from chronic toxicity tests, which are normally carried out with large food supply, to low food environments can therefore generate wrong conclusions.Density oscillations in laboratory populations with constant food supply are a common phenomenon. However, it is not clear whether such behaviour is mainly caused by intrinsic properties of the populations themselves or by experimental irregularities. On the basis of an analysis of published population behaviour, Van der Hoeven (1989) suggested that the latter option is very probable. Therefore, the simulations of the <em>Daphnia</em> model, which show regular, persistent oscillations, have been questioned (Kooijman <em>et al.,</em> 1989). The major driving force of these oscillations is synchronisation of life cycles during the decline phase of a population. This prediction mainly follows from the assumption that food intake is proportional to surface (-length <sup><font size="-2">2</font></SUP>) and maintenance costs are proportional to weight (-length <sup><font size="-2">3</font></SUP>). As a consequence, the rate of body growth declines as the animal approaches an ultimate size, which in turn is depressed at low food levels. Individuals which are smaller than the ultimate size continue to grow, whereas body growth is suspended by those which are larger than this size. This mechanism leads to convergence of body lengths. The assumed growth model is referred to as Von Bertalanffy's type I model (1969). Reproduction is closely synchronised owing to another assumption, i.e. first reproduction occurs at a fixed body size (cf. <em>chapter 5</em> ).Convergence and synchronisation among two juvenile cohorts was evaluated in an experimental set- up that simulated food dynamics and competition during a density peak and succeeding decline phase of a laboratory population <em>(chapter 6</em> ). The cohorts, which came from consecutive broods of a common mother cohort, were separated while they shared the same food source. Two feeding regimes were used, i.e. constant food input and daily pulses, to examine effects of resource variability. Under the experimental conditions convergence was very slow or absent and no synchronisation of reproduction occurred. The larger animals performed best at constant low food levels, whereas the smaller cohort was abetter competitor at fluctuating food. Modelled convergence rates, using the Von Bertalanffy model, were unrealistically high. Therefore, the suitability of this model under the present conditions was discussed in comparison with other growth models. Parallel growth curves suggested a linear growth model for juvenile <em>D.</em><em>magna,</em> although it was recognized that the shape of the growth curve is extremely dependent on the specific conditions of the study.From the mechanisms described above it is concluded that close synchronisation of fife cycles in laboratory populations and hence severe oscillations are not expected. Parallel growth, increased individual variation at low food level (cf. Cox <em>et al.,</em> 1992) and trade-offs with respect to first reproduction have a stabilizing effect. It has been shown that lipid reserves of mother and progeny are inversely related, which decreases the population growth rate and enhances the survival probability of neonates under resource limitation. In general, phenotypic plasticity, especially adaptation to low food, is underestimated in the present model assumptions. Simple models may perform satisfactory at the level of the individual, but they can lead to erroneous population dynamics. Incorporation of some more biology will undoubtly improve the model simulations and thereby the extrapolation of toxic effects from the individual to the population level. However, the balance between mathematical elegance and the full complexity of <em>Daphnia</em> is precarious. A stepwise incorporation of the most relevant life history strategies in combination with sensitivity analysis of the model could improve our understanding of population dynamics.The results of <em>chapter 5</em> and adverse effects on growth and reproduction, which are observed for many toxicants (cf. <em>chapter</em> 4 and Enserink <em>et al.,</em> 1991), suggest that toxic substances impair <em>Daphnia's</em> capacity to efficiently exploit its food source. This could lead to increased densities of (edible) algae. However, effects on ecosystem level depend on many factors, e.g. relative susceptibility of competing zooplankton, predators and food organisms. For instance, Marshall & Mellinger (1980) observed both decreased and increased phytoplankton production in cadmium spiked enclosures of plankton communities. In the first case primary production was directly affected by the test substance and in the second case reduced zooplankton abundance compensated for toxic effects on the algae. In a review on pesticide stress in freshwater ecosystems Brock & Budde (1994) concluded that primary effects can be predicted from laboratory tests, if exposure of field populations can be estimated. However, secundary effects are unpredictable in most cases. Recently, Scholten <em>et al.</em> (1994) stated that eutrophication problems are enhanced by impaired zooplankton grazing, owing to the presence of toxic substances in surface water. Although their conclusions raised great controversy in The Netherlands, the above statement could serve as a valuable hypothesis in future investigations. At present, the influence of toxicants on aquatic ecosystems is largely unknown. Research on the combined effects of nutrients and micropollutants is certainly worth pursuing. Besides, it could stimulate further cooperation between ecologists and ecotoxicologists.