Mitchell Max Planck Institute for Limnology, Postfach 165, Plön, D-24302, Germany. Tel.: +49 4522 763282; fax: +49 4522 763310; e-mail: firstname.lastname@example.org
Evidence for temperature adaptation in Daphnia magna was inferred from variation in the shape of temperature reaction norms for somatic growth rate, a fitness-related trait. Ex-ephippial clones from eight populations across Europe were grown under standardized conditions after preacclimation at five temperatures (17–29 °C). Significant variation for grand mean growth rates occurred both within populations (among clones) and between populations. Genetic variation for reaction norm shape was found within populations, with temperature-dependent trade-offs in clone relative fitness. However, the population average responses to temperature were similar, following approximately parallel reaction norms. The among-population variation is not evidence for temperature adaptation. Lack of temperature adaptation at the population level may be a feature of intermittent populations where environmentally terminated diapause can entrain the planktonic stage of the life-history within a similar range of temperatures.
Temperature is one of the major abiotic factors affecting organisms ( Begon et al., 1990 ; Hoffman & Parsons, 1991; Johnston & Bennett, 1996) and is a major regulating factor in aquatic ecosystems ( Goss & Bunting, 1983; Moore et al., 1996 ) through impacts on biological rate processes, including biochemical and physiological processes that influence organismal and population-level traits ( Hochachka & Somero, 1984; Johnston & Bennett, 1996). Recent focus on global change has increased interest in the effects of temperature on the ecology and evolutionary potential of biological populations and communities, as average temperatures are expected to increase by a few degrees with associated large increases in short-term extreme temperatures ( Houghton et al., 1990 ). The extent to which individual populations within a species are locally adapted to the prevailing temperatures may have implications for the population survival and genetic composition under a potentially rapidly changing thermal regime.
Geographical and altitudinal clines of species distributions and morphological or fitness-related traits have been interpreted as temperature adaptation (e.g. Kimura et al., 1994 ; Bischoff & Wiencke, 1995; Karan et al., 1998 ; Oleksyn et al., 1998 ), as temperature is one of the major environmental variables among the sites. Elevated temperatures during summer are considered to be one of the factors driving changes in zooplankton community structure, particularly through the reduction in body size ( Moore & Folt, 1993; Moore et al., 1996 ). Intraspecific temperature adaptation has been shown by the rapid response of thermal sensitivity to laboratory selection in ectotherms ( Huey et al., 1991 ; Lenski & Bennett, 1993). A latitudinal cline was found for growth rate in copepods on the east coast of N. America, where northern populations of Scottolana canadensis were found to be faster growing at cooler temperatures than southern populations ( Lonsdale & Levinton, 1985). Seasonal temperature adaptation was reported for a permanent population of the Cladoceran, Daphnia magna Straus ( Carvalho, 1987), where clones isolated in winter had higher fecundity and survival at cold temperatures than clones isolated in summer, which did better at warm temperatures.
Small, fast-growing organisms might be expected to respond to temperature change in a short time by genetic selection on standing variation or variation introduced by migration and thus are potentially less directly at risk from global warming ( Huey et al., 1991 ). By the same token, one would expect greater temperature adaptation to occur for such species, especially in populations with little geneflow, such as zooplankton in isolated freshwater ponds where local adaptation can occur ( De Meester, 1996) and migration is rare but possible (K. Schwenk, personal communication). The cladoceran component of zooplankton communities, particularly members of the genus Daphnia, can have a significant influence on aquatic productivity, as they are effective filter feeders on phytoplankton ( Lampert, 1987; Sterner, 1989) and are an important food source for planktivorous fish and other predators ( Gliwicz & Pijanowska, 1989). Ecological divergence for thermal tolerance is likely to be greater in Daphnia from shallow, small habitats compared with deep thermally buffered lakes ( Carvalho, 1987) where a temperature refuge, approximately 4 °C, is usually present. D. magna has a wide thermal tolerance ( Goss & Bunting, 1983) and inhabits shallow ponds across a large geographical range, including Europe, western USA and parts of Africa ( Hebert, 1978, 1995). Intraspecific variation and genotype by environment interactions have been demonstrated for a number of traits with temperature in Daphnia (e.g. Loaring & Hebert, 1981; Korpelainen, 1986) but it is not known to what extent temperature adaptation occurs among populations.
Two scenarios may be hypothesized. One is that populations from southern warmer climates should show temperature adaptation for higher temperatures compared with populations from northern colder climates. Alternatively, populations from all climates could show adaptation for an approximately similar temperature range as the life-cycle of D. magna controls the environment that the planktonic stage experiences. Temporary populations survive part of every year in diapause and, owing to control by environmental cues ( Schwartz & Hebert, 1987), emergence is entrained to occur only at the onset of suitable limnetic conditions. Despite extreme differences in annual climate across Europe, water temperatures usually fall within a similar range when D. magna are present. For example, D. magna tend to appear after water temperatures reach 15 °C in Moscow, experiencing a summer maximum of around 20–25 °C and avoiding the frozen winter (Yan Galimov, personal communication). At the opposite extreme, in southern Spain, the D. magna appear between October and April, experiencing a temperature range between usually 8 and 20 °C, with a maximum of 25 °C ( Lopez et al., 1991 ; Serrano & Toja, 1998), and avoid the onset of increased temperatures and desiccation during the summer. In northern Germany, D. magna start to hatch at <4 °C in years with mild winters when the ponds are not frozen, although temperatures rise rapidly to 10–15 °C with a summer maximum of approximately 20–23 °C ( Mitchell et al., 1998 ).
In order to distinguish whether a temperature response implies adaptation, the beneficial impact needs to be estimated explicitly in the currency of fitness or a close correlate for individuals with different histories of thermal selection ( Huey & Berrigan, 1996). In this study, we established temperature reaction norms for somatic growth rate under common experimental conditions for clones of D. magna originating from geographically separated populations. The specific somatic growth rate correlates with the intrinsic rate of population increase r ( Lampert & Trubetskova, 1996). This is a useful indicator of fitness as Daphnia have life-history characteristics that allow them to rapidly exploit favourable environmental conditions as well as ensuring that a small population can recover from mass mortality ( Korpelainen, 1986). We estimated the true reaction norm for genetically identical clone replicates that were physiologically pre-adapted to test temperatures. Differences among genotypes in the reaction norm shape, including the slope, breadth, temperature of maximum response and maximum value ( Gabriel & Lynch, 1992), may be indicated by genotype by environment interactions, and would be evidence for genetic variation with respect to thermal plasticity, which has rarely been documented ( Huey & Berrigan, 1996).
However, evidence for temperature adaptation requires a change of the reaction norm in response to a thermal selection regime, and we therefore investigated variation among populations. Reaction norms of fitness-related traits, or performance curves, of ectotherms tend to follow a characteristic asymmetrically skewed curve ( Huey & Kingsolver, 1989, 1993), in which performance initially rises linearly or geometrically with increasing temperature, reaches a maximum, at what is usually assumed to be the optimal temperature (e.g. Gabriel & Lynch, 1992), and then declines rapidly at higher temperatures. The position of the metabolic optimum only a few degrees below the critical maximum tolerance is a consequence of temperature effects on underlying biological processes such as enzymatic reactions ( Hochachka & Somero, 1984). Several outcomes could support the hypothesis that southern populations should be adapted for higher fitness, and therefore higher growth rates, at higher temperatures compared with northern populations, depending on whether genetic correlations occur among growth rates at different temperatures and whether they are positive (no trade-offs) or negative (trade-offs) ( Gomulkiewicz & Kirkpatrick, 1992; Huey & Kingsolver, 1993). If no genetic correlations occur, southern populations could have a higher maximum temperature limit but the temperature of maximal response and the response at lower temperatures would be similar to other populations. If negative genetic correlations occur, southern populations could have a similar shaped reaction norm shifted towards higher temperatures, including higher temperatures for maximum response and thermal limit ( Huey & Kingsolver, 1993). If positive genetic correlations occur, the southern populations might have reaction norms with higher growth rates at all temperatures than other populations ( Gomulkiewicz & Kirkpatrick, 1992).
Materials and methods
D. magna reproduces by cyclic parthenogenesis ( Hebert, 1987), where female neonates are normally released asexually into the water column. Sexual reproduction is induced by environmental cues such as photoperiod, temperature or crowding ( Banta, 1939; Stross, 1987), when male neonates and diapausing eggs that require fertilization are produced. These sexual eggs are protected in a chitinous case, an ephippium, developed from a thickened part of the female’s carapace, and can survive extreme conditions of temperature or desiccation. In temporary populations, sexual reproduction often occurs early in response to crowding and food stress ( Korpelainen, 1987; Innes, 1997; Mitchell et al., 1998 ) as well as at the end of the season. Genotype–environment interactions occur in the sex induction response ( Deng, 1996) and strongly competitive clones may delay ephippial production while poorer competitors invest more in early sexual reproduction ( Loaring & Hebert, 1981). Hatching is induced by environmental cues ( Schwartz & Hebert, 1987) when conditions are favourable. Not all ephippia hatch at the first opportunity, leading to the development of a diverse ‘egg bank’ in the sediments that may include genotypes from several seasons or years ( Hairston et al., 1996 ; Hairston et al., 1999 ).
Ephippia were isolated from sediment samples collected at seven geographically widespread locations in Europe ( Fig. 1) ranging from Finland (60°N) with a northern climate of severe winters and brief summers to southern Spain (37°N) with a Mediterranean climate of mild winters and dry hot summers. Hatching was induced at various temperature and light : dark regimes (including 15 °C 12 : 12 h L : D, 18 °C 12 : 12 h L : D or 14 : 10 h L : D, 22 °C 16 : 8 h L : D, and 8–12 °C or 22 °C with natural increasing Spring daylength) to ensure as nonselective a regime as possible. All hatchlings were isolated to initiate clone lineages, with a success rate of >95%, and eight clones per population were picked randomly for inclusion in the experiment. It was assumed that each hatchling was genetically distinct as a result of sexual reproduction and subsequent allozyme analysis confirmed that most lineages were electrophoretically distinct. Standard cellulose acetate allozyme electrophoretic techniques ( Hebert & Beaton, 1993) were used to screen enzymes.
An eighth population was sampled in Sicily from which live females were isolated and grouped according to eight distinct electrophoretic genotypes at the two polymorphic loci found. One clone was picked at random from each group for inclusion in the experiment to avoid inadvertent replication of one clone.
Life history experiment
The experiment was performed at six temperatures (17 °C, 20 °C, 23 °C, 26 °C, 29 °C, 32 °C) that should characterize the upper tail of a temperature reaction norm and include the maxima for both cool and warm adapted clones ( Carvalho, 1987). Daphniids were fed exclusively the green alga Scenedesmus obliquus Meyen, that has proven to be a good food for Daphnia ( Lampert, 1977), at high concentration, 1.5 mgC L–1, as growth rate response to temperature is minimized under food stress ( Orcutt & Porter, 1984). Owing to the large numbers involved (8 populations × 8 clones × 3 replicates × 6 temperatures = 1152 vessels), the experiment was repeated over two trials, with four clones from each population randomly assigned to each block in time.
To remove the influence of maternal effects ( Lynch & Ennis, 1983), clones were acclimated over two complete generations (grandmother and mother) at the experimental conditions. The grandmothers for each clone were isolated from the second brood of one female that had been cultured since birth at 23 or 26 °C under experimental food conditions. The great grandmother generation was cultured at 23–26 °C as reduction in temperature had little impact on survival, but a relatively rapid increase to 29 °C or 32 °C was often detrimental (Mitchell, unpublished data). To improve survival, the grandmother neonates had a gradual change in temperature (lasting 1–2 days) until experimental conditions were achieved. Mothers and experimental neonates were taken from the second brood of the previous generation female. On the rare occasions when neonates in the second brood were male, females were taken from the third brood.
Three replicate neonates in the first trial, and at least four in the second, were isolated at each temperature. The remainder of neonates in the brood were used to estimate weight at birth. Ten neonates, or as many as remained, were gently rinsed with distilled water on a 100-μm mesh, pooled in a preweighed aluminium boat and dried overnight at 60 °C. They were cooled and stored in a desiccator until weighed with an electronic balance to the nearest 0.1 μg to determine the dry mass.
Experimental animals were raised in 100 mL filtered lake-water that had been aged for 24 h and was changed every 2 days. Algae were added daily. Observations were made every 12 h, and times of birth and maturity were recorded. On reaching maturity, the first instar in which eggs were present in the brood chamber, the clutch size was counted under a binocular dissecting microscope and the adult female was rinsed and prepared for weighing as above.
Specific somatic growth rates (g) were calculated from dry masses (W1 and W2) at successive times (t1 and t2) according to: g=(ln W2 – ln W1)/(t2 – t1). This measurement of g includes the first batch of reproductive products, and has been shown to be strongly correlated with the intrinsic rate of increase, r ( Lampert & Trubetskova, 1996).
The effect of temperature on growth rate at the intra- and interpopulation level was analysed by ANOVA using STATISTICA v5.0 (1995). As there was significant variation between the two trials, care was required to ensure that intertrial variation did not increase apparent intrapopulation variation. Missing data cells, owing to death of all replicates in the experimental or mother generation, prevented a properly balanced hierarchical ANOVA including clone, population and trial. As the question of interest for within-population variation was whether (a) it existed and (b) it was present in all populations, each population was analysed separately for each trial (16 independent tests) using a two-way mixed model ANOVA on clone (random effect) and temperature. Interpopulation variation was tested using the clone mean growth rates at each temperature by repeated measures ANOVA for temperature with population and ‘trial’ as fixed factors, as this removed the need for hierarchical testing including all replicates for clones. The interaction terms between clone or population and temperature in the above analyses gave an indication of the genetic variation for the shapes of the reaction norms within populations (among clones) or among populations. Further analysis of the shape of the response to temperature included trend analysis using polynomial contrasts in ANOVA and genetic correlations.
Cross-environment genetic correlations of clone mean growth rates were calculated among each pair of temperatures to test for genetic constraints on growth rate among temperatures, which affect the possibility for a change in the relative values of two character states ( Via, 1994a). When the genetic correlation is zero the genetic variance available for a change in the relative value of phenotypes in each environment, and thus the shape of the reaction norm, is maximal ( Via, 1994a). Pairwise genetic correlations between the same character expressed in different environments may be calculated using product moment correlations of family means, but variance component correlations are less conservative as the use of genetic variances reduces contamination by environmental variance in the denominator ( Via, 1991; Fry, 1992). Clone mean growth rates at each temperature were used to calculate both product moment and variance correlations, the latter using genetic variances calculated from ANOVA tables ( Sokal & Rohlf, 1995). Genetic correlations at the interpopulation level were calculated based (i) on the entire set of clones (n=56) that had data at all temperatures and (ii) on population means (i.e. mean of clone means, n=8) as a check against pseudo-replication and for a test with a similar sample size as the within-population correlations. Genetic correlations were calculated for each trial separately and for the combined data set, although the latter includes inflated environmental variation due to trial. Confidence intervals were obtained for correlations using the z-transform ( Zar, 1984).
The average genetic correlation across environments was additionally tested for significant difference from zero using a second F-test in the two-way ANOVAs ( Fry, 1992; Via, 1994b), which should not be confused with the usual test of main effects and interactions. For the latter, the appropriate F-test is determined by the model used, in this case a mixed model where the F-ratio, ‘MS (clone)/MS (error)’, is a test of whether clones differ in marginal mean growth rate ( Via, 1994b). Fry (1992) showed that it is possible to use an alternative F-ratio, ‘MS (clone)/MS (interaction)’, to test whether growth rates of clones are positively correlated among temperature environments due to the relationship between the variance due to clone (family) and the covariance of growth rates of a clone among each pair of environments. This assumes that correlations are equal where more than two environments are included, and is approximate in this experiment as the data are slightly unbalanced ( Fry, 1992).
As latitudinal clines are often associated with temperature adaptation, the product moment correlation between population mean growth rate and latitude or longitude was calculated for each temperature separately.
No clone was successful at 32 °C. Although grandmothers often survived to produce a mother generation (viable neonates from a second brood), the mother generation either did not survive, remained barren, or released undeveloped eggs or deformed neonates on moulting. One female started to show male secondary characteristics, as well as a fully developed brood pouch, after two instars with ovaries. The experiment and data analyses were continued at five temperatures.
The results followed the same pattern over both trials (Table 1), although there was a significant variation between trials for the average growth rate and also response to temperature (Table 2) due to slower growth rates in the second trial at 17 °C and 20 °C. Temperature, as expected, had highly significant impact on growth rate, with the maximum occurring on average around 23–26 °C ( Figs 2 and 3) although some clones showed no significant difference between 26 and 29 °C.
Table 1. Two-way ANOVA results for growth rates with temperature for clones in each population, tested separately for each trial. Two F-tests are shown for factor 2 ‘Clone’: the first is the usual main effects test; the second tests for significance of genetic correlations (see text for details). Asterisks indicate whether P level is significant after Bonferroni correction for 16 tests. ****P < 0.0001, ***P < 0.001, **P < 0.01, *P < 0.05. Additionally, Fishers test was conducted to investigate the overall significance of the 16 ‘clone–temperature interaction’ terms and tested against a χ2 distribution with 32d.f.: χ2 = − 2Σ ln(pi) = 407.8, P < 0.0001.
Table 2. Repeated measures ANOVA on clone mean growth rates, with population and trial as fixed factors and temperature repeated at five levels. Univariate F-statistics are shown for the repeated factor, and the results remain qualitatively the same when multivariate statistics are applied. F1 = main effects F-test, F2 = test for significance of genetic correlations (see text for details).
Within population (among clone) variance in growth rate with temperature was significant for all populations. Clones had different grand mean growth rates (averaged across all temperatures) and also different shaped reaction norms of growth rate with temperature, as evidenced by significant interactions in ANOVA tests (Table 1) and cross-overs between clone reaction norms ( Fig. 2). Trend analysis indicated that individual clone reaction norms usually contained significant linear and quadratic components, and often cubic components.
Within most populations no significant correlation was found between growth rates at different temperatures using product moment correlations (Table 3) and the 2nd ‘F-tests’ (Table 1). However, in one population, Doñana, six of the 10 pairwise cross-environment correlations of growth rate were significant at P < 0.05, although, similar to the Moscow and Finland populations, only one correlation was significant at P < 0.05 after Bonferroni correction for 10 simultaneous tests. The variance component genetic correlations could not be calculated owing to negative genetic variance for some clones.
Table 3. Pairwise genetic correlations with 95% CI (min max). *Correlation is significantly different from zero, P < 0.05 after Bonferroni correction for 10 tests, for product moment tests only.
Among population variation was significant for the grand mean growth rate (Table 2). However, the response to temperature was similar in all populations, with no significant difference for the shape of reaction norms ( Fig. 3, Table 2: interaction population × temperature). All population reaction norms exhibited significant linear and quadratic, but not cubic, components. Growth rates strongly correlated with temperature in tests including all clones or population mean growth rates (Table 3). Results for each trial considered separately were qualitatively the same, with the same pattern for significance and sign of correlations, and are not shown to save space. The variance component calculation for genetic correlation always fell well within the 95% CI of the product-moment correlation.
The population mean growth rate at each temperature showed a slight but significant correlation with latitude ( Fig. 3, Table 4), but no relation with longitude. The faster growing populations tended to be more southern.
Table 4. Correlation between latitude or longitude and mean population growth rates at each temperature. n = 8 for all tests.
Reaction norms for somatic growth rate at five temperatures were constructed for D. magna clones to investigate population-level adaptation to temperature. Clones within populations had variable reaction norms with temperature, exhibiting significant linear and quadratic or cubic components, genotype by environment interactions, and different grand mean growth rates (the average growth rate over all temperatures tested). Among populations, however, although grand mean growth rates varied significantly, there was no significant variation for shape/breadth of reaction norm, or temperature of maximum value or thermal limit, as more-or-less parallel quadratic reaction norms had maxima occurring between 23 and 26 °C.
This study focused on the upper tail of the reaction norms, as the temperature of maximum trait response is often considered indicative of temperature adaptation. While the slope of the lower tail of the reaction norms is also of interest, particularly as relative performance of clones at lower temperatures may be most important in natural conditions, space and time constraints prevented estimation over a wider range of temperatures. Pre-acclimation at 17 °C required 5–6 weeks. As growth at the high temperature of 32 °C did not continue into the third generation, it appears that the upper limit for the true genetic reaction norm was reached for all populations, although the growth rate for many clones remained similar or slightly increased between 26 °C and 29 °C. However, during the period that maternal effects operated, which is usually about two generations ( Lynch & Ennis, 1983), somatic growth and limited reproduction was possible at 32 °C. Maternal effects, resulting from environmentally determined reproductive provisioning, are probably ecologically important, particularly for increased survival through relatively short periods of high temperature (diurnal fluctuations or short heat-waves of approximately one generation time, or less than 1 week), such as are expected under global warming scenarios. It would be constructive to quantify the influence of maternal effects as the constant environmental conditions required to estimate the true genetic reaction norm are highly unnatural.
The within-population variation of temperature reaction norms was observed consistently in both trials for all populations (Table 1). Although the maximum growth rate for all clones occurred between 23 °C and 29 °C, there were significant differences among clones from the same population for growth rate at these as well as lower temperatures. Additionally, cross-overs were observed ( Fig. 2): many clones that had significantly different growth rates at one temperature either had the same growth rates at another temperature or reversed their relative fitness rank (significant in LSD post hoc tests, not shown to save space but see SE intervals in Fig. 2). Furthermore, it can be seen from Fig. 2 that several clones tended to have broader reaction norms with higher growth rates over a wider range of temperatures while others had a more defined peak growth rate, typically at 26 °C. These interactions indicate that, as growth rate correlates strongly with fitness, the relative fitness of clones from the same population is temperature-dependent and genetic variation for thermal plasticity occurs within populations.
Genetic correlations of growth rates between each pair of environments were calculated to establish whether the pattern of reaction norms found in this experiment indicates a presence or lack of temperature adaptation at the population level. If the genetic correlation ( Fig. 3, Table 3) detected at the population level should indicate temperature adaptation through selection for higher performance at one temperature resulting in correlated higher performance at all temperatures ( Gomulkiewicz & Kirkpatrick, 1992), one would expect to find such a constraint at the individual clone level. However, only two of the six populations showed consistently positive genetic correlations for growth rate with temperature among clones. More generally within populations, genetic correlations were both positive and negative and mostly nonsignificant, suggesting that there is no inherent constraint ( Via, 1994a) on the response of growth rate to temperature among clones that have presumably the same selection history. In other words, increased growth rate at one temperature does not necessitate increased growth rate at other temperatures. Thus one would not expect such a constraint to occur at the population level in response to temperature alone, indicating that the variation among populations detected here is not evidence for temperature adaptation.
The lack of significant genetic correlation within populations is likely to be real, although the more accurate variance method ( Via, 1991; Fry, 1992) was not estimable owing to negative genetic variances for some clones at some temperatures. This problem does not appear to be unique; Via (1991) reports the same pattern with aphid clone populations. Although the product-moment correlation coefficients are conservative, in the interpopulation calculations the 95% confidence interval always included the coefficient calculated using genetic variances (Table 3). Neither is it likely that the lack of significance is due only to the lower power of tests with a smaller sample size since strong correlations remain significant. For example, the correlation between growth rates at 20 °C and 17 °C using the variance method (n: 56, r: 0.65) had a power of 0.99 whereas the correlation among population mean growth rates between these two temperatures (n: 8, r:0.86) had a power of 0.83. Furthermore, in the two-way ANOVAs calculated for within-population variation of growth rates with temperature, the 2nd F-ratio (MS (clone)/MS (interaction)) was usually nonsignificant (Table 1), indicating that, on average, genetic correlation across environments was not significantly different from zero. Although the underlying assumption that correlations should be the same for every pair of environments ( Fry, 1992) was violated (Table 3), the results of these F-tests support the detection of low correlations within most populations, often of mixed direction, contrasting with the strong and consistently positive genetic correlations for population mean growth rates among temperatures (Tables 2 and 3).
The two major exceptions were populations from the extremes of the sampling range, Moscow, Russia, and Doñana, southern Spain, that showed consistently positive genetic correlations (Table 3) with few cross-overs in the reaction norms ( Fig. 2). Further investigation into whether this results from stress resistance ( Huey & Kingsolver, 1993) may be instructive. It may be relevant that the Doñana sample probably contained ephippia from many ponds that were drained into the sampled pond at the head of the marsh after unusually high floodwaters earlier in the season.
Maximum growth rates occurred in this study at temperatures (23–29 °C) experienced rarely or during only a brief part of the growing season, although under the optimization principle a maximum trait response should occur at the optimal temperature ( Gabriel & Lynch, 1992; Stearns, 1992). Such a phenomenon appears to be common. For example, the winter-adapted genotypes of D. magna recorded by Carvalho (1987) showed maxima at 15 °C which exceeds winter temperatures in the pond ( Carvalho & Crisp, 1987). Similarly, Drosophila melanogaster populations showed maxima at temperatures unrelated to the average environmental temperatures in the original location ( Delpuech et al., 1995 ). Possibly, relatively rare selection maintains the requirement for survival at high temperature, and, owing to the enzymatic constraint for the maximal response usually lying close to the limit for survival ( Hochachka & Somero, 1984), pushes the maximal response to occur at higher than average temperatures. The lack of clones in this experiment with a lower temperature of maximal response does not necessarily mean they were absent from the populations. Such clones would only be favoured in a population where there was a predictable niche, such as over winter in a permanent population, but low population densities and growth rates would reduce their input to the ephippial pool. Common genotypes have a greater probability than rare ones of being included in this study, due to being picked at random from the ephippial pool, but may be more indicative of the average level of temperature adaptation for the population.
Although temperature has unquestionably an important effect on growth rate, the data from this experiment indicate that differences among populations of D. magna were not due to temperature adaptation for increased growth rates at higher temperatures, supporting our alternative hypothesis. Since a similar range of temperatures occurs among populations during the planktonic period of D. magna, the tendency for higher growth rates in populations from lower latitudes (Table 4) is unlikely to reflect temperature adaptation, but rather some other aspect of variation that also correlates with latitude, although temperature may have some indirect effect through influence on the local environment. The approximately parallel population mean reaction norms suggest adaptation to factors not included in the experiment, which would result in increased growth rates at all test temperatures in those populations where a faster growth rate is favoured. Selective factors could include, in addition to temperature, a combination of abiotic and biotic factors such as the length of the growing season and the presence and abundance of competitors and predators. Biotic factors were found to be more important than temperature for excluding D. magna from tropical ponds, as despite having highest r at the tropical pond temperature (27 °C) compared with temperate pond temperatures, D. magna was more susceptible to predators and was a poorer competitor than a tropical species, D. laevis, that had a higher r ( Foran, 1986).
The relationship detected among populations is based on mean growth rates at different temperatures under uniform laboratory conditions, which is necessary for a standard comparison, but does not necessarily represent the relative hierarchy if growth rates were measured for populations in situ. The populations appeared to contain sufficient variability to withstand increased temperatures in the environment. The maximum growth rates occurred at relatively high temperatures (23–29 °C) and survival at 32 °C was possible while maternal effects lasted, which is important as increased variability in short-term extreme temperatures is predicted under global warming ( Houghton et al., 1990 ). Thus, although global warming may favour different genotypes within a population, a change in temperature regime would not necessarily favour introduced novel genotypes.
We thank S. Flöder, J. Havel, K. Schreiber and J. Sitzler for practical assistance; T. Berendonk, H. Boriss, C. Clabby, Y. Galimov, S. Romo, R. Tollrian, J. Vanoverbeke for providing ephippia and sediment samples; L. Serrano and the Biological Station of Doñana for permission and assistance with field collections in Doñana National Park, Spain; and two anonymous reviewers. The study was supported by the Deutsche Forschungs Gesellschaft. S.E.M. was supported by a Max-Planck postdoctoral fellowship.