SEARCH

SEARCH BY CITATION

Keywords:

  • aquatic ecotoxicology;
  • Chaoborus crystallinus;
  • immunomodulation;
  • independent action;
  • intrinsic rate of natural increase;
  • joint effects;
  • natural antagonists;
  • non-additive interactions;
  • Pasteuria ramose;
  • population growth rate

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  • 1
    Predation and parasitism are important factors in the ecology and evolution of natural populations and may, along with other environmental factors, interact with the impact of anthropogenic pollutants.
  • 2
    Our study aimed at identifying potential interactions between three stressors (predation threat, parasitism and pesticide exposure) and at exploring the predictability of their joint effects by using the model of independent action. We assessed in a full-factorial design the impacts of these stressors on key life-history traits and population growth rate of the water flea Daphnia magna.
  • 3
    When applied as single stressors, predation threat and parasite challenge induced varying stressor-specific adaptive responses. The pesticide carbaryl was applied at a generally sublethal concentration, which caused low mortality only in first-brood offspring.
  • 4
    Pesticide exposure interacted synergistically with parasite challenge regarding survival, which suggests immunomodulatory activity of the pesticide. Predation threat by phantom midge larvae showed antagonistic interactions for amount of first-brood offspring with both parasite challenge and carbaryl exposure. All stressors additively affected age and size at maturity, which added up to a considerable delay in the onset of reproduction in the three-stressor combination. The intrinsic rate of natural increase, r, reflected the non-additive and additive effects on single endpoints and showed significant synergistic interactions for all two-stressor combinations. The combination of all stressors resulted in a dramatic reduction of r compared to the stressor-free control.
  • 5
    The model of independent action proved useful in quantitatively predicting effects of additively acting stressors, and in visualizing the occurrence and magnitude of non-additive effects in accordance with results of analysis of variances.
  • 6
    Synthesis and applications. Cumulative additive effects and non-additive interactions of natural antagonists and pollutants are shown to result in considerable impacts on ecologically relevant parameters. As a starting point for an environmentally more realistic risk assessment of chemicals, it may be a valuable strategy to screen for non-additive effects among many stress factors simultaneously in simplified experimental designs by using the model of independent action.

Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

The ecological risk assessment (ERA) of anthropogenic pollutants aims to protect natural populations by extrapolating effects assessed mostly at the level of individuals under standardized laboratory conditions to effects on populations in field conditions. The impact of pollutants may interact with environmental factors as shown for various abiotic and biotic stressors (Heugens et al. 2001; Jonker et al. 2004; Vinebrooke et al. 2004). Consideration of multiple-stressor scenarios has consequently been recognized as one step to improve the effect assessment of toxicants and thereby reduce the uncertainty due to extrapolations inherent to the ERA procedure (Hope 2006).

The observation that macroinvertebrate community structure was affected at low pesticide levels previously considered to be protective (Schäfer et al. 2007) might represent an example of a combined synergistic impact of multiple natural stressors and anthropogenic pollution in the field. The common biotic stress factors, predation and competition, interact in various ways with each other regarding their impacts on life-history traits of many organisms (Bolnick & Preisser 2005; Beketov & Liess 2007). Likewise, biotic stress factors such as predation may also show interdependence with the effects of toxicants (Vinebrooke et al. 2004; Beketov & Liess 2006). One example is the observation of increased lethality of carbaryl in tadpoles additionally stressed by the presence of predators (Relyea & Mills 2001). The study of Heugens et al. (2006) is one of the few investigations into joint effects of three stressors, and reported non-additive effects between all pairs of stressors (food limitation, temperature and cadmium) in the crustacean Daphnia magna Straus.

Joint effects of various chemicals applied together have been investigated increasingly in the last years, and the models of concentration addition (CA) and independent action (IA) are recommended for the prediction of their joint effects (Altenburger, Nendza & Schüürmann 2003; Jonker et al. 2004). Both models assume additivity and denote non-additivity by deviations of the measured from the predicted (reference) response. CA is generally recommended for similarly and IA for dissimilarly acting substances, although it is still under debate how ‘similar’ the substances actually must behave in order to justify the choice of CA (Borgert et al. 2004). Simple and often intuitively applied effect summation (ES) is only appropriate under the condition of a linear relationship between the intensity of the single stress factors and their effects (Berenbaum 1989). Chemicals do not meet this assumption due to their typically non-linear dose-response curves.

Our study aimed at experimentally evaluating joint effects on single life-history traits, as well as population growth rate of a three-stressor combination involving two of the most relevant biotic stress factors, predation threat and parasitism, in combination with one model toxicant, the pesticide carbaryl. In addition, we explored the use of a concept commonly applied for chemical mixtures to quantitatively predict the joint effects for this three-stressor combination. If stressors act additively, the prediction of the impact of a multiple-stressor environment based on the observed effects of single stressors would be a pragmatic approach in considering environmental factors in the ERA without the need to test all possible multiple-stressor combinations.

As a model organism, we used the water flea D. magna, a standard test species in ecotoxicology and a keystone species in lentic freshwaters (Peters & De Bernardi 1987). Presence of predators (detected by the prey through chemicals produced by the predator, so-called kairomones) can induce predator-specific adaptive shifts in morphology, life history and behaviour of Daphnia (Harvell 1990). Daphnids host a broad range of endoparasites (Stirnadel & Ebert 1997), among them the bacterium Pasteuria ramosa. The likelihood of a P. ramosa infection and the speed of disease development depend on encounter rates between host and parasite (Decaestecker, De Meester & Ebert 2002) and on condition-based factors such as genetically or environmentally determined host immunity (Little & Ebert 2000). Both predators and parasites are important factors driving ecology and evolution of natural Daphnia populations (Kerfoot & Sih 1987; Minchella & Scott 1991), and shape communities of eutrophic freshwater ponds and shallow lakes, the natural habitat of D. magna. Predators and parasites of Daphnia often co-occur (Pulkkinen & Ebert 2006); hence, the investigated combination reflects a rather common natural situation. Given the fundamental biological differences between predation, parasitism and the specific neurotoxin carbaryl as an inhibitor of acetylcholinesterase, we assumed dissimilar modes of action for the three stressors. Even if these stressors affect the same target organ, they will probably do so by different biochemical pathways. As a consequence, we chose the model of IA to calculate the reference value for additive effects of combined stressors based on measured single-stressor effects.

Materials and methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

experimental set-up

We conducted a full-factorial experiment in which each stressor was tested at two levels (absence and presence), resulting in eight treatments, with three replicates per treatment and six D. magna per replicate vessel. The experiment was started with the second brood (< 24-h old) of stock culture females kept individually at a high food level for more than two generations, and was conducted for 21 days at 19 °C ± 1 °C under diffuse light (16:8 h light:dark cycle). The exposure medium (ADaM, Klüttgen et al. 1994) was renewed every second day and animals were fed daily with Scenedesmus obliquus at a level of 100 000 cells mL−1 (day 0 to day 3) and 156 000 cells mL−1 (equal to 1·2 mg C L−1) from day 4 onwards. We recorded survival, age at maturity (defined as the release of the first clutch of eggs into the brood pouch and checked every 12 h), size at maturity, daily offspring production (dead and living offspring were removed daily and counted separately), and the body length of first-brood offspring (12–16 individual measurements per treatment, representative for all days of offspring release). Daphnid size was measured from the base of the spine to the top of the head. Oxygen content (range 6·7 to 8·9 mg L−1, mean 8·2 mg L−1) and pH (range 7·1 to 8·3, mean 7·8) in fresh and aged experimental medium did not differ systematically among treatments.

parasite challenge as stressor

The obligate endoparasite Pasteuria ramosa castrates its host Daphnia within 5 to 15 days after infection, presumably by hormonal control (Ebert et al. 2004). Transmission of the parasite occurs horizontally by endospores released only from dead hosts, which excludes infections from living hosts in grouped exposures during experiments. The experimental challenge with parasite endospores, as well as the procedure to obtain these spores, followed a protocol previously established (Coors et al. 2008), requiring exposure of daphnids during the first 4 days in volumes of 20 mL. Medium with parasite challenge was prepared at day 0 and day 2 by adding 1 mL of a solution containing 106 mL−1 mature endospores (previously obtained by grinding infected daphnids) to 19 mL of medium (which, depending on the treatment, contained the other stressors; see below). All parasite-free treatments received 1 mL of placebo solution prepared by grinding healthy stock culture daphnids. From day 4 onwards, parasite challenge stopped and test vessels contained 900 mL medium.

predation threat as stressor

Predation threat also followed a previously established method (Coors, Hammers-Wirtz & Ratte 2004) and was based on direct water contact by keeping Chaoborus crystallinus larvae (third and fourth instar) in net enclosures (nylon gauze, mesh size 0·25 mm) of about 100 mL volume within the experimental vessels (900 mL). This set-up allows diffusion of kairomones but prevents real predation. All vessels with predator presence contained five C. crystallinus larvae. The larvae were fed daily within the vessel with 25 neonate (< 24 h) D. magna, which were so quickly consumed that their impact as food competitors can be considered negligible. Dead and pupated larvae were replaced daily. Empty net enclosures were deployed in all treatments with absence of predator. Exposure with direct water contact to the predator was not possible during the first 4 days due to the small volume of medium (see above, parasite challenge). Therefore, we used medium conditioned for 24 h by C. crystallinus larvae (5 larvae per 900 mL, fed daily with D. magna neonates) to prepare the medium on days 0 and 2.

pesticide exposure as stressor

Carbaryl (1-naphthyl methylcarbamate, CAS 63-25-2, 99·8% purity, Sigma-Aldrich, Germany) stock solutions were prepared in ethanol (p.a.). The final nominal and, based on previous results (A. Coors, unpublished data), sublethal concentration in carbaryl treatments was 5·6 µg L−1. All carbaryl-free treatments contained the same amount of the solvent ethanol as the carbaryl treatments (50 µL L−1). Three 500-mL samples were taken from carbaryl treatments and one from the solvent control, stored at −20 °C and later analysed by an external laboratory according to DIN EN ISO 11369 (F12). Chemical analysis confirmed absence of carbaryl in the solvent control (< 0·01 µg L−1). Measured carbaryl concentration in freshly prepared medium (used for all carbaryl treatments) was 0·72 µg L−1, whereas measured concentrations in 48 h-old medium in absence and presence of Chaoborus were 0·54 µg L−1 and 0·46 µg L−1, respectively. Although the measurements do not allow statistical evaluation due to the lack of replication, the results do not suggest a strong impact of predator presence on carbaryl stability in the exposure medium. Carbaryl is a rapidly dissipating substance (EPA 2003), which might explain the approximate factor 10 between measured and nominal carbaryl concentrations even in freshly prepared medium. The difference between the nominal and measured concentration of carbaryl does not interfere with the interpretation of our results, as we aimed at treating carbaryl exposure as a stress factor and not at establishing a toxicity threshold.

A pre-test with C. crystallinus using daily-refreshed exposure medium and performed under conditions otherwise identical to the full-factorial experiment confirmed that the carbaryl level of 5·6 µg L−1 was not toxic to the predator. After 96 h exposure, 100% of larvae in the solvent control were still alive, and survival in the highest-tested carbaryl concentration (which was 8·0 µg L−1 in this pre-test) and the blank control was both 86·7% (three replicates per treatment with five larvae each).

population growth rate

We calculated the intrinsic rate of population increase, r, iteratively according to the Euler–Lotka equation (Lotka 1913)

  • image(eqn 1)

where lx represents the proportion of survivors at age x, and mx the number of offspring released at day x. We calculated r separately for each replicate for x= 0 to 21 days.

prediction of joint effects

Joint effects in all two-factor combinations and the three-factor combination were predicted from the respective single-stressor effects (defined as deviation from control) according to equation 2 which is based on the model of IA first introduced by Bliss (1939).

  • image(eqn 2)

Emix represents the effect of the joint stressors, and Ei the effect of each single stressor i. Observed effects (ei) in absolute units (e.g. hours) must be transformed to proportional effects according to equation 3, and the calculated Emix was rescaled afterwards in absolute units to allow comparison with observed joint effects (Payne, Scholze & Kortenkamp 2001).

  • image(eqn 3)

eCONTROL is the value of the stressor-free control. For the endpoints survival, number of offspring and castration rate, the maximum possible effect (emax) was set as zero. For endpoints where a maximum effect of ‘zero’ is of no biological meaning, such as body size, the maximum possible effect was defined based on the range of the experimental data. In the case of age at maturity, we assumed no maturation within the whole experiment as maximum possible effect, thus, an emax of 504 h. For body length measurements, the maximum individual length observed in the experiment was set as maximum possible effect (0·898 mm for offspring size and 3·2 mm for size at maturity, respectively). Deviations of observed from predicted joint effects are considered as significant and, thus, stressors as acting non-additive, if the predicted effect lies outside the 95% confidence interval of the observed effect. Synergistic interaction between stressors is indicated by a significantly stronger observed effect of the combined stressors than that predicted from the single stressors, whereas an antagonistic interaction is indicated by a significantly weaker effect of the combined stressors than predicted.

statistical analysis

All parameters were analysed by two- or three-way anova tests. Of total mortality, 92·3% occurred during the juvenile phase, and statistical tests were therefore performed with data on survival until day 10 (end of juvenile phase). High mortality caused the loss of some replicates, which resulted in an unbalanced design for the other endpoints and reduced statistical power. Data on survival and castration were arcsine-transformed to comply with statistical assumptions, whereas anova tests could be performed on raw data of age and size at maturity, amount and size of offspring and r. Assumptions for anova were checked by inspection of residuals and by Bartlett's test.

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

survival

All test daphnids in the single-stressor treatments and the stressor-free control survived, except for one out of 18 in the carbaryl-only treatment (Fig. 1a). Assuming additive effects, no relevant mortality was therefore to be expected in the treatments with combined stressors. However, high mortality was observed in the two treatments including both carbaryl exposure and parasite challenge, indicating a synergistic interaction between these stressors. This synergistic interaction is illustrated in Fig. 1a by the deviation of the IA-predicted from the measured joint effects, and was substantiated by a statistically significant two-way interaction between carbaryl and parasite (Table 1). This strong interaction must also be the cause of the significant main effects of carbaryl and parasite, since survival in all single-stressor treatments and in the control did not differ from each other (Bonferroni post hoc test, P > 0·05). All other interactions between stressors were non-significant with regard to survival.

image

Figure 1. Survival (a), castration rates at the end of the experiment (b), age at maturity (c), size at maturity (d), amount (e) and size (f) of living first-brood offspring per female (no castrated females at this stage) in Daphnia magna exposed to no stressors (squares), single stressors (triangles), and combined stressors (circles). Filled symbols denote observed responses (means and 95% confidence intervals) and empty symbols represent additive effects predicted by the model of independent action.

Download figure to PowerPoint

Table 1.  Results of three-way anova tests for key life-history traits and intrinsic rate of natural increase in Daphnia magna exposed to multiple stressors.
 d.f.MSP
  • d.f., degrees of freedom; MS, mean square.

  • power < 0·1.

Survival
 Carbaryl (C) 10·6810·002
 Predator (Pr) 10·0240·491
 Parasite (Pa) 10·6810·002
 C × Pr 10·0660·259
 C × Pa 10·5470·004
 Pr × Pa 10·0050·758
 C × Pr × Pa 10·0240·491
 Error160·048 
Age at maturity
 Carbaryl (C) 1632·0< 0·001
 Predator (Pr) 1535·70·001
 Parasite (Pa) 1411·80·003
 C × Pr 128·40·357
 C × Pa 136·90·296
 Pr × Pa 120·00·437
 C × Pr × Pa 15·40·682
 Error1331·1 
Number first-brood offspring
 Carbaryl (C) 129·90·007
 Predator (Pr) 111·70·07
 Parasite (Pa) 1141·8< 0·001
 C × Pr 117·90·029
 C × Pa 10·070·879
 Pr × Pa 120·40·021
 C × Pr × Pa 10·50·689
 Error133·0 
Size at maturity
 Carbaryl (C) 10·00450·508
 Predator (Pr) 10·05310·0357
 Parasite (Pa) 10·15820·0014
 C × Pr 10·00350·558
 C × Pa 10·00500·483
 Pr × Pa 10·00950·3399
 C × Pr × Pa 10·00050·8222
 Error130·0097 
Intrinsic rate of natural increase
 Carbaryl (C) 10·00350< 0·001
 Predator (Pr) 10·00852< 0·001
 Parasite (Pa) 10·000640·016
 C × Pr 10·000640·016
 C × Pa 10·001130·003
 Pr × Pa 10·001190·002
 C × Pr × Pa 10·000230·119
 Error130·00008 

parasite-induced castration

Hosts were only castrated upon parasite challenge but never in placebo treatments, and all castrated individuals were later identified as being infected with P. ramosa by microscopic inspection for endospores. Castration occurred at the earliest after release of one brood. Parasite challenge as the only stressor resulted in a castration rate of 16·7% at the end of the experiment, whereas castration rates tended to be slightly higher in combination with either predation threat or carbaryl exposure (both 20%). A castration rate of 23% was predicted by IA for the three-stressor combination, but the actual percentage of castrated females was much higher (37·5%; Fig. 1b). The expected value was clearly outside the confidence interval of the observed castration rate, which points to a synergistic interaction. Yet, neither main effects of carbaryl and predator nor their interaction were significant in a two-way anova within parasite-challenge treatments (all P values > 0·46, always with power < 0·1).

onset of reproduction and size at maturity

Females in the stressor-free control treatment matured between day 6 and day 8, whereas maturation was on average delayed to different degrees in all treatments with one single stressor (Fig. 1c). The strongest delay of about 13 h was induced by carbaryl. The effects of all single stressors were statistically significant (Table 1). For all combinations of multiple stressors, predicted ages at maturity agreed well with observed values. This additive action of stressors was confirmed by non-significant interaction terms in the anova. Size at maturity was significantly influenced by predation threat and parasite challenge, but not by carbaryl exposure, and no significant interactions were detected (Table 1). Predicted values for size at maturity agreed well with observations in all multiple-stressor treatments (Fig. 1d).

reproductive output

The amount of living first-brood offspring per female was enhanced in parasite-challenged females, reduced in carbaryl-exposed females, and was no different from the stressor-free control in daphnids under predation threat (Fig. 1e). The comparison of IA-predicted number of first-brood offspring with observed values revealed smaller effects than expected, and thus antagonistic interaction in the stressor combinations predation threat–parasite challenge and predation threat–carbaryl exposure. These non-additive interactions were statistically significant in anova (Table 1). The sum of living and dead offspring per female totalled 14·2 in daphnids exposed only to carbaryl, which did not differ significantly from the 16·3 produced in the control (t-test, P = 0·224). Dead offspring was almost exclusively recorded in the carbaryl treatments. Together, this suggests that the effect of carbaryl as single stressor was not due to reduced reproductive effort but to offspring mortality caused by acute toxicity of carbaryl. This effect was only seen in first-brood offspring but not in later broods (data not shown). Regardless of other stressors, daphnids always produced much larger-sized offspring in the presence of Chaoborus (Fig. 1f). Chaoborus was the only factor significantly influencing offspring size (three-way anova, P < 0·0001), whereas all other factors and interactions were non-significant (all P-values > 0·54).

population growth rate

The intrinsic rate of natural increase integrates effects on survival and fecundity, more specifically on developmental rate (inverse value of time needed until maturation), number and viability of offspring, and parasite-induced castration. Compared to the stressor-free treatment, r decreased in the two-stressor combinations by 4 to 12% (Fig. 2). In the three-stressor combination, the reduction of r amounted to 27·6% and ranged in-between the changes observed for single endpoints, which were 9·2% (offspring size), 19·9% (age at maturity), 28·4% (living first-brood offspring), and 72·2% (survival). Survival was only impacted in two treatments; in all other treatments, r therefore solely reflected effects on fecundity. Whereas none of the recorded life-history parameters showed significant interactions for all possible combinations of two stressors, r exhibited significant interactions for all three two-way interactions. The three-way interaction was not significant (Table 1); thus, the interaction between pairs of stressors was not influenced by the presence of the third stressor.

image

Figure 2. Intrinsic rate of natural increase of Daphnia magna exposed to no stressors (square), single stressors (triangles), and combined stressors (circles). Filled symbols denote observed responses (means and 95% confidence intervals) and empty symbols represent additive effects predicted by the model of independent action.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

complex effects of single stressors

In concordance with previous results (Coors et al. 2004), predation threat imposed by Chaoborus crystallinus larvae induced delayed maturation at a larger body size and increased offspring body size in D. magna. Because fecundity was only marginally, and survival not at all, affected by predation threat, the reduction in r to 93 % of r in the control treatment reflects mainly the contribution of developmental rate. Larger-sized offspring has the advantage of decreased risk of predation by a gape-size limited predator such as Chaoborus sp. (Swift 1992). Producing larger neonates can therefore be interpreted as an adaptive response to the presence of these predators. The adaptive value of this response is not incorporated in r, but it would clearly be relevant in cases of real predation pressure.

Changes in reproductive strategy of P. ramosa-challenged D. magna have been reported previously (Ebert et al. 2004) and were expressed in the present experiment by delayed host maturation, increased host size at maturity, and an increase in host offspring number. This shift in reproductive strategy can be seen as an individual-level compensation for imminent, parasite-induced castration, and paid off with regard to r, which was slightly greater than in the stressor-free control. In a previous experiment (Coors et al. 2008), much stronger infection conditions were created with a higher parasite infection rate and a faster development of the disease. The strategy of reproducing later might not be advantageous under such circumstances, because it will increase the risk of being castrated before producing any offspring.

Carbaryl exposure resulted in the greatest delay of maturation, but was not accompanied by an increase in size at maturity. There was also no compensation of later maturation by either more or larger offspring. Such an impact on developmental rate in Daphnia has been observed before for carbaryl (Hanazato 1991) and might be due to a reduced filtration rate as reported for the pyrethroid fenvalerate (Reynaldi et al. 2006). Substantial mortality caused by carbaryl was only observed in first-brood offspring, which indicates that, overall, exposure concentrations were sublethal. It probably also reflects size-dependent acute toxicity, because neither the larger-sized neonates of subsequent broods were affected nor the test animals themselves. This conclusion is supported by Hanazato (1991), who found that Daphnia ambigua was most sensitive to carbaryl when exposed during the first instar. Likewise, Enserink, Luttmer & Maas-Diepeveen (1990) reported higher susceptibility to toxicants in smaller Daphnia offspring. Lethal concentrations at target organs can be reached faster in smaller animals, because the water-borne toxicant needs to diffuse a shorter distance. For rapidly dissipating substances such as carbaryl (EPA 2003), this phenomenon may be especially relevant.

Overall, the observed complex and specific effects of single stressors on D. magna life-history traits and reproductive strategy complied with expectations. These single-stressor impacts were reflected in the intrinsic rate of natural increase, which was either decreased (carbaryl and predation threat) or increased (parasite challenge) compared to the stressor-free control treatment.

additive, antagonistic and synergistic effects of combined stressors

All possible outcomes of joint effects – additive, synergistic or antagonistic – were observed, and the nature of the joint effects depended both on the stressor combination and the measured endpoint.

Clear additive effects were detected for age and size at maturity as well as offspring body size. The shifts along the trade-off between investment in growth (age and size at maturity) and in reproduction (offspring size and number) differed among the two biotic stressors, whereas no shift was induced by pesticide stress. This strongly suggests that stressors influenced the regulation of these life-history traits independently by different mechanisms and supports the choice of the IA over the CA model to calculate joint effects of these stressors. Additivity of joint effects appears at first sight less dramatic compared to synergistic interactions. However, rather small impacts of the single stressors on age at maturity added up to a delay in maturation of more than 1 day in the three-stressor combination. Given the importance of age at maturity for population growth, such an additive effect is likely to be of great biological relevance.

Whereas predation threat alone did not result in a change in the amount of living first-brood offspring, this endpoint was influenced in opposite directions by the other two stressors (Fig. 1e). Yet, when combined with predation threat, the effect of both stressors was antagonistically overwhelmed to a degree that the resulting trait value in the two-stressor treatments almost equalled that of the control value. The antagonism between predation threat and pesticide exposure probably reflects a higher survival of first-brood offspring in the presence of predator threat because of their larger size. In this case, the antagonistic joint effect is actually based on stressors influencing different traits (offspring survival and offspring size), which were combined in one measure (amount of living offspring). Assuming that it is indeed the larger size of the neonates that protects them from being killed by carbaryl, it can be concluded that the adaptive responses to a biotic antagonist can pay off with regard not only to that stressor but also with regard to the impact of an anthropogenic pollutant. The observed antagonism between predation threat and parasite challenge must be due to a different mechanism, as parasite challenge did not affect offspring size. We hypothesize that this antagonism reflects the importance of offspring size as a key response to Chaoborus predation threat, which dominated over the response to the Pasteuria challenge.

A significant synergistic interaction was detected for the effects of parasite challenge and carbaryl exposure on survival, which is in agreement with earlier results (Coors et al. 2008). In this previous study, it was shown that exposure to carbaryl accelerated the development of the P. ramosa infection, which strengthens the non-significant trend of increased parasite infection rate in additionally stressed hosts in the present study. The synergistic interaction between parasite challenge and carbaryl points to an immunomodulatory activity of carbaryl, leading directly or indirectly to an impairment of the immune response. In consequence, this means that the two stressors act not independently but that one stressor (carbaryl) affects the mechanism that is important to cope with the other stressor (parasite infection).

No significant synergistic interaction between carbaryl exposure and predation threat was detected on any single life-history trait recorded in this study. Thus, previous reports of increased lethality of neurotoxic pesticides upon exposure to predator kairomones (Relyea & Mills 2001; Maul, Farris & Lydy 2006) could not be confirmed. Relyea & Mills (2001) tested carbaryl in an amphibian predator–prey system, whereas Maul et al. (2006) assessed the organophosphorus insecticide malathion in Ceriodaphnia dubia exposed to homogenized fish tissue. Likewise, the synergistic interaction between exposure to Chaoborus flavicans kairomones and carbaryl on age at maturity reported for Daphnia pulex (Hanazato & Dodson 1992) was not confirmed by our study for D. magna. Differences in the experimental set-ups, in the mode of action of the pesticides, and interspecific differences in the biochemical signalling pathways in the various predator–prey systems might explain the differences between our results and these previous studies. Although an alternative explanation might be that the presence of the predator reduced the bioavailability of carbaryl in our experimental set-up, the results on measured carbaryl concentrations do not suggest a strong influence of predator presence on actual carbaryl exposure concentrations. Because the key response to Chaoborus, the production of larger-sized offspring, was expressed in all predator treatments, it is also unlikely that carbaryl caused a reduction of kairomone production and thereby indirectly lowered the stress level imposed by the predator.

population growth rate as integrative measure

In order to increase ecological relevance, Van Leeuwen, Luttmer & Griffioen (1985) suggested the use of the intrinsic rate of natural increase r in ecotoxicological studies. r integrates information on developmental rate, which is not used as a standard ecotoxicological endpoint but is highly relevant (Calow, Sibly & Forbes 1997). Ratte (1996) demonstrated that even an increase of 30% in offspring production in D. magna cannot compensate for a 1-day delay in maturation. We here show that age at maturity, and thereby population growth, can indeed be affected both by chemicals and by natural stressors, and that such effects can add up considerably.

The observed non-additive and additive joint effects among stressors on different life-history traits were integrated in r in a way that significant interactions were found for all three combinations of stressor pairs. Thus, interaction effects detectable on various individual-level endpoints were not ameliorated but all were reflected by r. Our experimental set-up did not allow for large changes in daphnid density. The number of survivors per vessels ranged from 4–6 (in the case of the three-stressor treatment achieved by pooling two replicates). This minimized the potentially confounding influence of daphnid density on the responses to the investigated stressors. Interactions between toxicants and density-dependent factors on population growth rate have often been observed and can be antagonistic as well as synergistic, thereby complicating the extrapolation of effects on r to effects on usually resource-limited natural populations (Forbes, Sibly & Calow 2001). Examples are compensation reactions due to toxicant-induced mortality leading to reduced intraspecific competition (Moe, Stenseth & Smith 2002; Beketov & Liess 2005) and increased population extinction risk under a simulated predation regime after short-term pesticide exposure (Beketov & Liess 2006).

Whereas single stressors and pairs of stressors resulted in rather small decreases of r, the three-stressor combination had an impressively strong impact with a reduction by 27·6%. Nevertheless, changes in single classical endpoints in this three-stressor treatment were all larger; thus, even the great reduction of r in the three-stressor treatment can be considered as a more conservative estimate than that based on endpoints traditionally used in ecotoxicology. The strong reduction was by no means to be expected from observed effects of stressors applied alone, and not obvious from the complicated interaction patterns of pairs of stressors. In consequence, such effects are not accessible by investigating only single stressors.

prediction of joint effects

For some of the recorded endpoints, the observed value in the three-stressor treatment greatly deviated from the one predicted by the IA model. While this clearly denotes the occurrence of interaction, it does not provide information on which stressors were actually interacting. Because the three-way interaction term in anova was never significant, the deviations observed by IA can be fully explained by non-additive interactions between pairs of stressors. The strong effect observed in the three-stressor treatment for some endpoints was therefore caused either by additive effects, non-additive interaction between pairs of stressors, or by the combination of the two phenomena, but was not related to complex interactions among all three stressors.

In general, non-additivity detected by IA and anova were in good agreement with each other, with castration rate as only an exception. The difference between the observed castration rate in the three-stressor treatment and the IA-predicted value suggested involvement of synergistic effects, whereas no significant interaction terms were revealed by anova. In previous experiments, the parasite-induced castration was accelerated (Coors et al. 2008) or slightly increased (A. Coors, unpublished data) in carbaryl-exposed D. magna, corroborating the synergism indicated by IA. In the present experiment, missing replicates due to mortality in exactly those treatments involving these two stressors have probably reduced the power of anova too much to detect significant interaction between these stressors. In the case of r, IA-predicted effect values for some two-stressor combinations were at the border of the confidence limit of the observed effect values. Yet, anova clearly showed non-additive interactions, which illustrates the much higher statistical power of anova compared to the rather simple statistics of using confidence intervals.

Concluding, we demonstrated the applicability of IA as a valuable tool to quantitatively predict joint effects of a set of diverse stressors (under the assumption that their impacts are additive), and further, as a tool to visualize non-additive effects. Although a full-factorial test design with subsequent anova for statistical verification is the method of choice for elucidating complex interactions among various stressors, using a method such as IA has the advantage that it can also point to the occurrence of interaction effects in less balanced designs. This can be a major asset when analysing multiple stressors, because testing only the single stressor and the all-stressor treatments would already be sufficient to indicate interaction effects by comparing predicted and observed effects in the all-stressor combination. This much simpler test design could be used as a starting point to screen for interaction effects among many stressors, and would have the advantage of allowing investigation of many more stressors within one experiment compared to a full-factorial anova design. In case that the IA-predicted effect in such a screening equals the observed joint effect, additivity of all stressors is indicated (unless contrasting interaction effects of the stressors balance each other). If prediction and observation deviate considerably, a follow-up test using a full-factorial design can be carried out to identify the interacting stressors. As demonstrated here, an IA-based screening approach for interactions among multiple stressors can be applied for diverse stressors, including natural antagonists and toxicants, and for a variety of endpoints. Our results indicate that synergistic as well as antagonistic interactions can occur with strong impacts on ecologically relevant endpoints, which illustrates the necessity of considering such effects when aiming at a more ecologically realistic risk assessment of environmental pollutants.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

This research was supported by a fellowship grant from the K.U. Leuven to A.C. and by projects OT/04/23 and GOA/2008/06 of the K.U. Leuven Research Fund and EU IP project ALARM (GOCE-CT-2003-506675).

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References