The degree of adaptive phenotypic plasticity is ... - Wiley Online Library

doi: 10.1111/j.1420-9101.2007.01353.x

The degree of adaptive phenotypic plasticity is correlated with the spatial environmental heterogeneity experienced by island populations of Rana temporaria M. I. LIND & F. JOHANSSON Department of Ecology and Environmental Science, Umea˚ University, Umea˚, Sweden

Keywords:

Abstract

development rate; environmental variation; hydroperiod; local adaptation; phenotypic plasticity; pool drying; Rana temporaria.

Although theoretical models have identified environmental heterogeneity as a prerequisite for the evolution of adaptive plasticity, this relationship has not yet been demonstrated experimentally. Because of pool desiccation risk, adaptation of development rate is important for many amphibians. In a simulated pool-drying experiment, we compared the development time and phenotypic plasticity in development time of populations of the common frog Rana temporaria, originating from 14 neighbouring islands off the coast of northern Sweden. Drying regime of pools used by frogs for breeding differed within and among the islands. We found that the degree of phenotypic plasticity in development time was positively correlated with the spatial variation in the pool-drying regimes present on each island. In addition, local adaptation in development time to the mean drying rate of the pools on each island was found. Hence, our study demonstrates the connection between environmental heterogeneity and developmental plasticity at the island population level, and also highlights the importance of the interplay between local specialization and phenotypic plasticity depending on the local selection pressures.

Introduction Phenotypic plasticity is the ability of one genotype to express alternative phenotypes to match the current environment it is exposed to (Nylin & Gotthard, 1998; Agrawal, 2001; Pigliucci, 2005). Phenotypic plasticity can have large effects on individual fitness (Van Buskirk et al., 1997) and population dynamics (Miner et al., 2005; Plaistow et al., 2006). Genetic variation in phenotypic plasticity, measured as genotype · environment interactions, is often found in natural populations (Pigliucci, 2005), which implies that the degree of plasticity can evolve as a result of natural selection. Theoretical modelling indicates that the degree of environmental heterogeneity is a major factor regulating the evolution of phenotypic plasticity (e.g. Moran, 1992; Sultan & Spencer, 2002; Ernande & Dieckmann, 2004). However, Correspondence: M. I. Lind, Department of Ecology and Environmental Science, Umea˚ University, 901 87 Umea˚, Sweden. Tel.: +46 90 786 66 37; fax: +46 90 786 67 05; e-mail: [email protected]

1288

few studies have focused on the variation in phenotypic plasticity among populations, and none have been able to demonstrate the selection pressures responsible for it (Pigliucci, 2005). The evolution of phenotypic plasticity under heterogeneous environmental conditions has been extensively modelled in relation to both spatial (e.g. Via & Lande, 1985; Gabriel & Lynch, 1992; Gomulkiewicz & Kirkpatrick, 1992; Moran, 1992; Van Tienderen, 1997; Sultan & Spencer, 2002; Ernande & Dieckmann, 2004) and temporal (Gabriel & Lynch, 1992; Gomulkiewicz & Kirkpatrick, 1992; Moran, 1992; Zhivotovsky, 1996; Sultan & Spencer, 2002) environmental variations. In one of the few models considering both spatial and temporal environmental variation together, Moran (1992) identified four factors that influence whether local specialization or phenotypic plasticity will evolve in a population with access to two different environments. In short, phenotypic plasticity will be favoured if there: (a) is a high accuracy of phenotype–environment matching; (b) is low costs of plasticity; (c) are equal frequencies of the two selective environments and (d) is temporal

ª 2007 THE AUTHORS 20 (2007) 1288–1297 JOURNAL COMPILATION ª 2007 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY

Pool heterogeneity and phenotypic plasticity

rather than spatial environmental variation. An extension of the model (Sultan & Spencer, 2002) has also shown that a metapopulation structure with migration between populations will further increase the likelihood that plasticity rather than local specialization will evolve. Temporary pools are excellent systems for the study of phenotypic plasticity (e.g. Newman, 1992; De Block & Stoks, 2005). Organisms are time constrained in these systems and have to reach a critical size or age before the pool dries up. Consequently, many animals inhabiting these pools have an intensive growth period prior to maturation and/or dispersal to other habitats (Wellborn et al., 1996; Williams, 1997). Furthermore, temporary pools are not only time-constrained environments, but are also a classical example of a temporally heterogeneous environment (Newman, 1992; Loman & Claesson, 2003). Temporary pools not only dry out, but are also more variable in water level than permanent pools (e.g. totally drying in some years, whereas not drying out in others) (Brooks & Hayashi, 2002), which could select for both local specialization and phenotypic plasticity, depending upon conditions (Moran, 1992; Van Tienderen, 1997; Sultan & Spencer, 2002; Ernande & Dieckmann, 2004). However, habitat characteristics can also vary on a spatial scale. For example, a local population can have access to temporary pools as well as permanent pools, and where two environmental states are present in a localized area, phenotypic plasticity in size and age at the transition can evolve. Among developmental studies of organisms inhabiting temporary environments, amphibian metamorphosis has been well studied and an extensive literature now documents the selection pressures operating upon the developing larvae and their metamorphosis. Having a large size at metamorphosis implies fitness benefits in amphibians, and is under natural conditions selected for (Altwegg & Reyer, 2003). For example, frogs that metamorphose at a large size usually maintain this size advantage (Smith, 1987) or even increase it by having a higher growth rate (Morey & Reznick, 2001; Altwegg & Reyer, 2003). Large size at metamorphosis also increases survival during the first terrestrial year (Berven, 1990; Morey & Reznick, 2001; Altwegg & Reyer, 2003). However, because of the short duration of temporary environments, there is a trade-off between metamorphic size and development time (Abrams et al., 1996), and amphibians that show phenotypic plasticity in development time usually respond to decreased water levels with accelerated development, at the cost of metamorphosing at a smaller size (Laurila & Kujasalo, 1999). Although phenotypic plasticity in time to metamorphosis is rather common (e.g. Newman, 1992), it has been harder to identify which environmental variables are associated with its evolution. The ideal way to study the evolution of phenotypic plasticity is to study experimental populations with a common ancestry in different selective environments (e.g. Fischer et al.,

1289

2004), however, as this is seldom possible, another approach is to compare the degree of phenotypic plasticity in different natural populations, and correlate the difference in plasticity with the selective factors operating on the populations. This approach has been used, but has so far proven unsuccessful. In such studies, population differences in developmental plasticity have not been found (Loman & Claesson, 2003; De Block & Stoks, 2004) or the difference in plasticity among populations were explained by other environmental factors unrelated to pool drying (Laurila et al., 2002; Merila¨ et al., 2004). A system with populations living in locations that are isolated from one another, differing only in the temporal and spatial heterogeneity of one important environmental component, would provide an excellent system for the study of the evolution of phenotypic plasticity. The Baltic sea coast of northern Sweden, affected by extensive land uplift, offers such a system, where isolated islands, inhabited by populations of the common frog Rana temporaria Linnaeus 1758, differ in temporal and spatial heterogeneity of pool types (Johansson et al., 2005). By using a common garden approach with a simulated pool-drying treatment, it is possible to estimate the degree of phenotypic plasticity in time to metamorphosis in a population, and it has been widely documented that there is adaptive phenotypic plasticity in development time in a number of amphibian species (e.g. Wilbur, 1987; Newman, 1992; Denver et al., 1998; Loman, 1999; Laurila et al., 2002; Loman & Claesson, 2003; Merila¨ et al., 2004, but see Brady & Griffiths, 2000). The environmental cue seems to be the reduced water volume per se (Denver et al., 1998). In this study, we investigated whether frog populations from different islands exposed to different pool-drying regimes show adaptive phenotypic plasticity in development, and whether the degree of plasticity varies among the island populations. Our main aim was to investigate whether the difference in phenotypic plasticity in development time can be explained by spatial heterogeneity of pool-drying regimes within an island, and to what extent it is correlated with the mean drying rate of the pools, which indicates temporal heterogeneity of drying regimes. The second aim of our study was to examine whether local specialization towards pool drying occurs in these island populations, i.e. whether there is a correlation among development rate and pool-drying regime among islands. As maternal effects can potentially influence the life-history traits expressed, we estimated their influence in one population using a North Carolina II breeding design.

Materials and methods Study area Eggs from R. temporaria were collected from 14 islands in the Gulf of Bothnia, south-east of the city of Umea˚, in


1290

M . I . L I N D A N D F. JO H A N S S O N

Fig. 1 Locations of study populations in an island archipelago along the coast of Va¨sterbotten, northern Sweden.

Va¨sterbotten, northern Sweden (Fig. 1). All islands along a 10-km section of the coastline with an effective population size of at least 10 breeding females (based on unpublished monitoring during 2001–2004) were used in our study. The islands range in size between 9 and 38 ha and small, more or less temporary pools are common in rocky depressions on the islands. Whereas some islands harbour only one pool, others harbour ‡ 30 pools, depending upon the rainfall during the season. However, not all of the pools present have breeding frogs. Previous studies have shown that predators have a strong impact on life-history characters in tadpoles (Skelly & Werner, 1990; Relyea, 2002). However, vertebrate predators are absent and invertebrate predator abundance is very low in the pools on the islands, and no relationship between life-history characters and predator abundance was detectable when 16 islands were analysed (Johansson et al., 2005). We therefore conclude that the effect of predators, if present, is minor compared with the effect of pool drying. Field sampling Eggs were collected on 2 and 6 May 2005. From each island, samples of 20–50 eggs were taken from each of 10 clutches of eggs, and transported to the laboratory. Each clutch was assumed to represent one female, as female R. temporaria lay only one egg clump per season (Savage, 1961). If multiple pools were present, eggs were taken from all pools in which breeding had occurred. The range of pools from where eggs were collected was one to six. On some islands, fewer than 10 clutches were found, with the lowest number collected being six. While collecting the eggs, the maximum depth of each pool was also measured. To estimate the natural drying regime of the pools, the same measurements were collected on

June 26. This was used as a proxy for the hydroperiod, which was not measured directly. Experiment set-up In the laboratory, eggs from each female were kept in separate containers filled with aged tap water. The eggs were stored in thermo-constant laboratory conditions and kept cool (4 C) until all eggs were collected. Before hatching, at Gosner stage 10 (Gosner, 1960), 10 eggs from each female were placed in a Petri dish, covered with water and photographed using a Minolta Dimage 7 digital camera (Kowica Minolta Photo Imaging Inc., Tokyo, Japan). The photographs were later used to measure the egg sizes, which allowed us to control for egg size-mediated maternal effects (Laugen et al., 2002). When all eggs were collected, the temperature of the laboratory was set to 22 C for the remainder of the experiment. A light : dark cycle of 18 h : 6 h was employed, corresponding to the natural light : dark cycle in the area of egg collection. When the tadpoles had reached Gosner stage 23 (active swimming) (Gosner, 1960), four tadpoles from each female (two allocated to each treatment) were randomly chosen and placed into individual plastic containers (9.5 cm · 9.5 cm, height 10 cm). The containers were filled with 750 mL of tap water, previously aged and aerated together with dried deciduous leaves. The leaves were removed when the water was transferred to the experimental containers. From each female, three additional tadpoles were also weighed (wet weight) to the nearest 0.1 mg, to provide estimates of initial weight used for growth rate calculations. The tadpoles were fed ad libitum every fourth day on a mixture (1 : 2) of finely ground fish food and rabbit chow. The food mass was 15 mg per tadpole at the beginning of the experiment, increased to 30 mg at day



9, 45 mg at day 13, 60 mg at day 17, and 75 mg from day 21 and to the end of the experiment. Water was replaced every fourth day, prior to feeding. The tadpoles were subjected to one of two treatments, either constant water level (C) or simulated pool drying (D). In the simulated drying treatment the initial water volume of 750 mL was lowered by 33% every fourth day, starting at day 5 and continued until day 25, after which the water volume in the D treatment was kept constant at 66 mL. The water temperature did not differ between the two treatments. The experiment was terminated at Gosner stage 42 (front legs visible), after which tadpoles were weighed. Time to reach this stage was recorded as ‘development time’. Measurements To get an estimate of the natural degree of pool drying and therefore the length of the hydroperiod, a pooldrying index for each pool was estimated as 1 ) (water depth on 26 June)/(water depth on the day of egg collection), so that values below one indicate drying of the pool. The index of pool drying rate was correlated with the depth of the remaining water in the pools (t46 ¼ 6.352, P < 0.001, r2 ¼ 0.46). It could be argued that expected hydroperiod rather than pool drying rate should be used in the analysis. However, pool drying rate was used as the relationship between the sorting of the pools according to drying rate probably is stable, independent of the actual drying on a particular year. In a wet year, drying rate would still give the risk of pool drying, whereas expected hydroperiod would indicate that only a very few of the pools would dry out during this particular season. The spatial variation in pool-drying regimes within an island was estimated as the coefficient of variation of pool drying for each pool containing eggs. Coefficient of variation, rather than standard deviation, was used because it is scaled by the mean. Thus, it down weighs islands where variation in pool drying is present, but where the pools do not dry out but rather vary around the initial pool depth, because of a rising water level rather than drying in some pools and permanence in others. However, both coefficient of variation and standard deviation yield qualitatively similar results in analyses. The development time was measured as the number of days from the start of the experiment (Gosner stage 23) to the end of the experiment (Gosner stage 42). Growth rate was estimated as the change in wet weight between Gosner stages 42 and 23, divided by the development time. Plasticity in development time for each island was calculated as the mean development time for individuals in the simulated pool-drying treatment subtracted from the mean development time in the constant water level treatment. Plasticity in final wet weight was calculated in the same way, replacing development time with wet weight at Gosner stage 42. Admittedly, metamorphosis is not complete until Gosner

1291

stage 46, and mass loss during stages 42–46 can vary nonadditively between larval growth conditions in some Rana species (Van Buskirk & Saxer, 2001). However, wet weight at Gosner stages 42 and 46 is correlated in these populations (t108 ¼ 6.87, P < 0.001, r2 ¼ 0.25; F. Johansson, unpublished). Statistical analysis To examine whether the drying regimes present on each island were independent of the island’s geographical location and other physical characteristics, multiple regression analyses were performed using either drying regime, coefficient of variation in drying regime, development time and plasticity in development time as response variables, and island area, island age, number of pools present on each island, distance from the nearest population and distance from the mainland as predictor variables. The effect of the treatment on the degree of phenotypic plasticity in life-history traits present in an island population was analysed in a mixed model A N O V A using restricted estimated maximum likelihood (REML). Treatment was incorporated in the model as a fixed factor, population as a random factor and, to account for maternal effects, egg size was used as a covariate. Because random mortality sometimes reduced the two replicates from each family and treatment combination, these replicates were pooled, and mean values were used in all analyses of treatment effects. Development time, weight at Gosner stage 42 and growth rate were used as response variables. One island population was excluded from this analysis (Fja¨rdgrund), as some eggs were collected after Gosner stage 10, and therefore their size could not be measured for the majority of females. However, the analysis was insensitive to this reduction in population number. To determine the minimal adequate model, model simplification was used when necessary (Underwood, 1997; Crawley, 2002). This implied a stepwise reduction of insignificant interaction terms, carefully observing that the explanatory power of the model was not significantly reduced. As population was considered a random factor, its main effect was not estimated in the F-test by the LME4 module; however, an F-test for the effect of population was specified separately. To examine the correlation between life-history traits and the prevailing environmental conditions at the islands, linear regression models based upon population means were used, and all 14 island populations were included. Development time, weight at Gosner stage 42 and plasticity in development time were used as response variables. Treatment and either an index of mean pool drying or coefficient of variation in pool-drying index were used as explanatory variables, and egg size was incorporated in the models as a covariate. When analysing phenotypic plasticity, treatment was not used as an


1292


explanatory variable, as phenotypic plasticity was defined as the difference in the mean trait values between the two treatments. It could be argued that a nested analysis using each family as a replicate should be used, but we chose to use mean instead because our focus was on between-population variation. A nested analysis would only give extra information on the within-population variance, which was irrelevant for the hypothesis tested. All analyses were run using the R statistical software (R Development Core Team, 2005) with the BASE and LME4 packages. Maternal effects Maternal effects, mainly manifested through egg size, are a potentially important influence on the life history of offspring. However, our experimental design does not allow us to determine the effects of environmentally mediated maternal effects directly. The use of egg size to control for maternal effects is valid only if egg size is the main way by which maternal effects are mediated between generations, or if maternal effects are generally of small importance on the life history of the offspring. Ideally a North Carolina breeding design II (Lynch & Walsh, 1998) would be the best way of examining the contribution of maternal effects on total variance. However, applying such a design implies that live frogs must be brought to the laboratory and mated artificially. As many of the island populations are small (e.g. the island Svart Lass only had egg clumps from six breeding females), use of these breeding designs would entail a risk of exterminating the frog populations on the islands. However, to get a rough estimate on the importance of maternal effects in the region, we applied a North Carolina breeding design II to the frogs from Storhaddingen, which is an island that had a fairly large population size (44 eggs clumps were found). Five males and five females were collected at the onset of breeding and were transported to the laboratory. Using artificial fertilization, each of the five males was crossed with each of the five females, creating a 5 · 5 breeding design. The artificial breeding was performed following Laugen et al. (2002). Two tadpoles per cross were raised in the constant water level and the artificial drying treatment. The experiment was run as for the 14 populations described above. Mass at the start of the experiment was estimated for each individual tadpole. Development time and weight at metamorphosis were analysed with a mixed model A N O V A using the REML option in R. Start weight of tadpoles was used as covariate, treatment as a fixed factor and male and female as random factors. Variance components were obtained with REML, and the significance of the maternal effect (VM) was calculated by dividing mean squares of the dam with the sire (Lynch & Walsh, 1998). Unfortunately some of the female · male crosses failed, resulting in a final design of 4 · 3 (female · male).

Results Island characteristics Neither the mean drying regime on each island, nor the variation in drying regimes between pools within an island were correlated with island age, area, number of pools, distance from nearest frog population or distance from the mainland (Table 1). In addition, neither development time nor phenotypic plasticity in development time were affected by the habitat characteristics: all P values were larger than 0.20 for development time and larger than 0.25 for plasticity in development time. Adaptive phenotypic plasticity The R. temporaria tadpoles responded to the simulated pool-drying treatment by accelerating their development (Fig. 2a). This phenotypic plasticity was supported by a significant treatment effect (Table 2). Moreover, developmental time varied between the island populations (Table 2; Figs 2a and 3), although the degree of acceleration did not significantly differ among populations (P ¼ 0.14). However, removal of the interaction term from the model was not possible because of the reduction of the model’s explanatory power. Maternal effects, mediated through egg size, also explained part of the variation in development time, as the development time was shorter for larger eggs (Table 2). The accelerated development in the drying treatment was associated with a lower weight at metamorphosis as tadpoles from drying containers had a significantly lower weight (Fig. 2b; Table 2). As for development time, the final weight at metamorphosis was not only dependent upon treatment, but was also population specific (Table 2; Fig. 2b), and the degree of growth differed among populations (Table 2). Maternal effects mediated through egg size did not influence the final weight of the tadpoles (Table 2). Populations varied in growth rate, and in all populations growth rate was lower in the drying treatment than in the constant water level treatment (Fig. 2c; Table 2). Table 1 Multiple regression analysis of the effects of various island features on mean pool drying and coefficient of variation in pool drying within the islands. Pool drying

Intercept Island area Island age Number of pools on island Distance from next population Distance from the mainland

CV in pool drying

t

P

t

P

)0.046 )0.173 0.251 )0.193 )0.915 1.327

0.965 0.867 0.808 0.852 0.387 0.221

)0.298 )1.238 0.739 1.847 )1.281 1.368

0.773 0.251 0.481 0.102 0.236 0.208

The analysis used eight degrees of freedom for the error term. None of the main factors were significant.



0.018). Development time was more rapid and weight at metamorphosis was lower in frogs that came from islands with rapidly drying pools (Fig. 3a, b; Table 3). This pattern was similar for both treatments (nonsignificant interaction term), and thus there was no increase or decrease in plasticity in development time as a response to the mean pool-drying time on each island. Maternal effects, mediated through egg size, explained a significant part of the variation in development time but not mass at metamorphosis (Table 3). The degree of phenotypic plasticity in development time was not correlated with the mean development time present in that population (constant water level: t12 ¼ 1.26, P ¼ 0.23, artificial drying: t12 ¼ )0.75, P ¼ 0.47); however, plasticity in weight at Gosner stage 42 was positively correlated with weight at Gosner stage 42 (constant water level: t12 ¼ 5.37, P < 0.001, artificial drying: t12 ¼ 3.05, P ¼ 0.01) for both treatments.

Development time (days)

(a) 28 27 26 25 24

(b) 0.7 0.6 Weight (g)

1293

0.5 0.4

Environmental heterogeneity and phenotypic plasticity (c) Growth rate (g* day–1)

0.025

0.020

0.015

Constant Drying Treatment Fig. 2 Tadpole larval development time, weight at Gosner stage 42, and growth rate under two water level treatments: constant and drying. The lines connect mean values in the two treatments for each of the 14 populations.

Local specialization to pool drying Drying rates varied among pools within each island, and the coefficient of variation in drying rate of pools differed between islands (Figs 3 and 4; Bartlett’s test on islands 2 with more than one pool: Bartlett’s K10 ¼ 21.52, P ¼ Table 2 Mixed model A N O V A of development time, weight at Gosner stage 42 and growth rate for the effects of treatment (drying and constant water), population and egg size (covariate).

Population Treatment Egg size Population · treatment Residuals

The phenotypic plasticity in development time expressed in a population on a given island was positively correlated with spatial environmental heterogeneity, estimated as the coefficient of variation of drying regimes within each island (Fig. 4; t11 ¼ 2.23, P ¼ 0.047, r2 ¼ 0.36). The pattern was not influenced by maternal effects mediated through egg size (t11 ¼ 0.141, P ¼ 0.28). Plasticity in weight at maturation was not correlated with the variation in pool-drying regimes on an island (t11 ¼ 0.39, P ¼ 0.70, r2 ¼ 0.038) or egg size (t11 ¼ 0.52, P ¼ 0.61). In contrast to the results found under spatial heterogeneity, the phenotypic plasticity in development time was not explained by the mean drying regime on the islands (t11 ¼ 0.052, P ¼ 0.96, for drying and t11 ¼ 0.95, P ¼ 0.36, for egg size). Neither plasticity in development time, nor plasticity in final weight was correlated with population size (t11 ¼ 1.24, P ¼ 0.24 and t11 ¼ )0.07, P ¼ 0.95 for development and weight respectively). Maternal effects The result of the north Carolina II breeding design from the Storhaddingen population showed that start weight

Development time

Weight

Growth rate

d.f.

F

P

d.f.

F

P

d.f.

F

P

12 1 1 24 187

7.83 36.3 13.6 1.34

< 0.001 < 0.001 < 0.001 0.14

12 1 1 24 187

22.2 323 0.58 3.68

< 0.001 < 0.001 0.45 < 0.001

12 1 1 24 187

19.2 237 0.39 3.15

< 0.001 < 0.001 0.53 < 0.001

The interaction between population and egg size, as well as all higher order interactions, was nonsignificant and was therefore excluded from the final model. Population corresponds to an island population.


1294

(a)


(c)

28

26

Weight (mg)

Development (days)

0.60 27

25

0.50

0.40

24 23 –10 0

0.30

10 20 30 40 50 60 70

–10 0

10 20 30 40 50 60 70

Pool drying (%)

(b)

Pool drying (%)

(d)

28

26

Weight (mg)

Development (days)

0.60 27

25

Fig. 3 The relationship between development time or weight at Gosner stage 42 and pool drying among the 14 populations. Development time under constant (a) and drying water (b) treatment in the laboratory, and weight at Gosner stage 42 under constant (c) and drying (d) treatments. Pool drying is given as percentage water reduction during the season. A negative value implies an increase in the water level of the pool.

0.50

0.40

24 23 –10 0

10 20 30 40 50 60 70

0.30 –10 0

10 20 30 40 50 60 70

Pool drying (%)

Pool drying (%)

Table 3 Result from general linear models for development time and weight at metamorphoses for the effects of treatment (drying and constant water), drying rate of pools (covariate) and egg size.

Plasticity in development (days)

2.0

1.5

Development time

1.0

Intercept Treatment Drying Egg size Residuals

0.5

0.0 0

10

20 30 40 50 60 CV in pool drying

70

80

Fig. 4 The relationship between plasticity in development time and coefficient of variation in pool drying among the 14 populations.

had no significant effect on weight at metamorphosis (t1 ¼ 1.07, P ¼ 0.52) or development time (t1 ¼ 1.32, P ¼ 0.59). Weight at metamorphosis was significantly smaller in the simulated pool-drying treatment compared with the constant water level treatment (t1 ¼ 13.53, P < 0.001). In contrast, development time was not affected significantly by treatment (t1 ¼ 0.66, P ¼ 0.35).

Weight

d.f.

t

P

d.f.

t

P

1 1 1 1 24

4.65 )3.44 3.58 )3.49

< 0.001 0.002 0.002 0.002

1 1 1 1 24

)1.44 )5.74 2.33 )0.89

0.16 < 0.001 0.03 0.38

All higher order interactions were nonsignificant and were therefore excluded from the final model. r2 ¼ 0.58 for development time and r2 ¼ 0.62 for weight at metamorphosis.

Maternal effects did not affect weight at metamorphosis or development time significantly (F3,2 ¼ 1.93, P ¼ 0.35 and F2,3 ¼ 0.59, P ¼ 0.68). Variance due to maternal effects was 5% for both weight and development. Admittedly, our sample size is small, but the low percentage of the total variation explained by maternal effects suggest that these are small, and that initial size of tadpoles has no large impact on the two life-history traits in the Storhaddingen population. We used egg size to account for maternal effects in our analysis on popula-



tions; there was a correlation between egg size and initial larval size in the R. temporaria populations (t92 ¼ 4.65, P < 0.001, r2 ¼ 0.32).

Discussion Spatial environmental heterogeneity and the degree of phenotypic plasticity This study indicates that phenotypic plasticity in lifehistory traits can evolve as a response to the environmental heterogeneity that a population is subjected to. The degree of phenotypic plasticity in development time of R. temporaria populations on isolated islands was positively correlated with spatial variation in pool-drying regimes present on the island. This result matches the expectations based on theoretical models of the evolution of phenotypic plasticity (Moran, 1992; Sultan & Spencer, 2002; Ernande & Dieckmann, 2004) but has not been demonstrated experimentally until now despite several attempts (Loman & Claesson, 2003; Huber et al., 2004; Merila¨ et al., 2004). In contrast to previous studies, the island populations investigated in this experiment are all relatively close to each other on a geographical scale, and do not differ from each other in environmental variables, such as seasonal length or temperature. Moreover, the local selection pressures are not confounded in any systematic way with island area, age, number of pools present on an island or distance from the nearest frog population or from the mainland. In addition, vertebrate predators are absent and invertebrate predator abundance is very low and uncorrelated with life-history characters of island frog populations in the area (Johansson et al., 2005). Therefore, local adaptations to pooldrying regime and spatial variation in pool drying become more visible as there is no covariation with other environmental variables. One reason why this relationship is found could be that this study focuses on more or less isolated island populations. Isolation makes populations less likely to interact along the population margins, and there is scope for local adaptation to the present pool types to take place within each island population. Unfortunately, we do not know the genetic distances for neutral markers among our populations, although such measures are currently being obtained. Because of the low salinity of the Gulf of Bothnia, gene flow between island populations of R. temporaria could be relatively high, at least for islands up to 1 km away (Seppa¨ & Laurila, 1999). However, local adaptation is often detected despite a high gene flow (Merila¨ & Crnokrak, 2001; McKay & Latta, 2002). When disruptive selection acts among populations, linkage disequilibrium can develop among the underlying allele frequencies. Therefore, trait values can change substantially even with fairly small changes in the underlying allele frequencies, and differentiation in quantitative traits is

1295

hence decoupled from that of neutral markers (Latta, 1998, 2003). If we assume that the local adaptations on these islands are due to selective forces, then evolution of life-history traits has occurred within a short time span. Land is rising by isostatic uplift in this geographic area, with the youngest islands being only about 180 years old. The first islands rising up from the sea were probably colonized by frogs from the mainland. Mainland pools are very common and are of both temporary and permanent types. Rana temporaria is abundant in these pools, and preliminary studies have shown that these mainland pools also show variation in developmental plasticity among populations (Almfelt, 2005). Temporal heterogeneity and phenotypic plasticity Theoretical models have shown that the evolution of phenotypic plasticity in response to environmental heterogeneity is most likely if the environment varies on a temporal rather than a spatial scale (Moran, 1992). This pattern was not found in our study, because we found no relationship between mean pool drying rate and plasticity in development. Admittedly, the temporal variation in pool drying was not measured directly, but temporary pools are known to be more variable in drying regime than permanent pools (Williams, 1997; Brooks & Hayashi, 2002), and temporary pools often represent a temporally variable environment in studies of phenotypic plasticity of pool-living organisms (e.g. Newman, 1992; Loman & Claesson, 2003). Maternal effects Although nongenetic maternal effects can potentially influence the life-history traits of offspring (Mousseau & Fox, 1998), they were only of minor importance in this study. Using a north Carolina II breeding design, we estimated the contribution by maternal effects to metamorphic weight and development time to only 5% of the total variation, which is in accordance with the low impact of maternal effects commonly found within R. temporaria populations in Sweden (Laugen et al., 2005). As maternal effects are generally expressed through egg size in R. temporaria (Laugen et al., 2002), the effect of egg size on all life-history traits was estimated using egg size as a covariate in all analyses. Maternal effects mediated through egg size significantly affected the analyses of development time, but the treatment effects were not affected by incorporating egg size as a covariate in the model. These results are in accordance with what is generally found in other studies (Newman, 1992; Johansson et al., 2005). Neither the phenotypic plasticity in development time or weight, nor growth rate or metamorphic weight was significantly influenced by maternal effects mediated through egg size.


1296


Phenotypic plasticity and local specialization Exposed to the artificial drying treatment, individuals metamorphosed earlier and at a smaller size compared with their siblings in the constant water level treatment. This confirms findings of numerous other studies demonstrating adaptive phenotypic plasticity in life-history traits of R. temporaria populations when subjected to pools of a short duration (e.g. Loman, 1999; Merila¨ et al., 2004). However, in addition to phenotypic plasticity, specialization to local pool-drying regimes is present. Individuals from islands with a predominance of temporary pools metamorphose earlier and at a smaller size than individuals from islands with a predominance of permanent pools, as repeatedly shown in this system (Johansson et al., 2005; this study). These results are not conflicting, but can be explained by varying selection pressures acting on different islands, some with spatial heterogeneity in pool-drying regimes, and others with only one pool-drying regime present. Leimar et al. (2006) suggested a framework in which development is idealized as a switching device with genetic and environmental cues as inputs. According to this framework, the present experiment can be interpreted as follows: a genetic cue (which expresses itself as fixed development) is more reliable in populations where the environment does not vary. In accordance to this, we found that tadpoles from islands with a low environmental heterogeneity in pool drying had a relatively small response to the different environmental conditions they were exposed to in the laboratory, manifesting itself as a low level of phenotypic plasticity in development time. In contrast, environmental cues (which express themselves as phenotypic plasticity) are the most important source of information for the idealized switching device in populations where the environment is variable, as indicated in our experiment by the large phenotypic plasticity in development time of tadpoles originating from islands with a large heterogeneity of pool-drying regimes. This is probably why developmental plasticity in weight was positively correlated with the mean metamorphic weight of the island population. Populations from islands with extensive pool drying show low plasticity, but are specialized towards rapid development. As rapid development is traded off against low metamorphic weight, there will be a correlation between the metamorphic weight and the degree of phenotypic plasticity present. Based on our study, we emphasize the importance of understanding the interactions between local specialization and phenotypic plasticity, and that both mechanisms need to be considered when studying life-history adaptations of a population.

Acknowledgments Many thanks to J. Almfelt and F. Olofsson for help with egg collecting and laboratory experiments. P. Ingvarsson gave valuable statistical advice. Thanks also to B. Giles, B.

Anholt, B. McKie, M. Monroe, T. Hipkiss and two anonymous referees for valuable comments on earlier drafts of this article. The research was funded by the Swedish Research Council and the Swedish Research Council for Environment, Agricultural Sciences and Spatial Planning.

References Abrams, P.A., Leimar, O., Nylin, S. & Wiklund, C. 1996. The effect of flexible growth rates on optimal sizes and development times in a seasonal environment. Am. Nat. 147: 381– 395. Agrawal, A.A. 2001. Phenotypic plasticity in the interactions and evolution of species. Science 294: 321–326. Almfelt, J.E.M. 2005. Plastic acceleration of metamorphosis in the common frog (Rana temporaria) in response to pool drying: the effect of geographical isolation. Master’s Thesis, Umea˚ University. Altwegg, R. & Reyer, H.U. 2003. Patterns of natural selection on size at metamorphosis in water frogs. Evolution 57: 872–882. Berven, K.A. 1990. Factors affecting population fluctuations in larval and adult stages of the wood frog (Rana sylvatica). Ecology 71: 1599–1608. Brady, L.D. & Griffiths, R.A. 2000. Developmental responses to pond desiccation in tadpoles of the British anuran amphibians (Bufo bufo, B. calamita and Rana temporaria). J. Zool. 252: 61–69. Brooks, R.T. & Hayashi, M. 2002. Depth-area-volume and hydroperiod relationships of ephemeral (vernal) forest pools in southern New England. Wetlands 22: 247–255. Crawley, M.J. 2002. Statistical Computing: An Introduction to Data Analysis Using S-Plus. John Wiley and Sons Ltd, Chichester, UK. De Block, M. & Stoks, R. 2004. Life-history variation in relation to time constraints in a damselfly. Oecologia 140: 68–75. De Block, M. & Stoks, R. 2005. Pond drying and hatching date shape the tradeoff between age and size at emergence in a damselfly. Oikos 108: 485–494. Denver, R.J., Mirhadi, N. & Phillips, M. 1998. Adaptive plasticity in amphibian metamorphosis: response of Scaphiopus hammondii tadpoles to habitat desiccation. Ecology 79: 1859–1872. Ernande, B. & Dieckmann, U. 2004. The evolution of phenotypic plasticity in spatially structured environments: implications of intraspecific competition, plasticity costs and environmental characteristics. J. Evol. Biol. 17: 613–628. Fischer, M., Van Kleunen, M. & Schmid, B. 2004. Experimental life-history evolution: selection on growth form and plasticity in a clonal plant. J. Evol. Biol. 17: 331–341. Gabriel, W. & Lynch, M. 1992. The selective advantage of reaction norms for environmental tolerance. J. Evol. Biol. 5: 41–59. Gomulkiewicz, R. & Kirkpatrick, M. 1992. Quantitative genetics and the evolution of reaction norms. Evolution 46: 390–411. Gosner, K.L. 1960. A simplified table for staging anuran embryos and larvae with notes on identification. Herpetologica 16: 183– 190. Huber, H., Kane, N.C., Heschel, M.S., von Wettberg, E.J. & Banta, J. 2004. Frequency and microevolutional pattern of selection on plastic shade-avoidance traits in a natural population of Impatiens capensis. Am. Nat. 163: 548–563.



Johansson, F., Hjelm, J. & Giles, B.E. 2005. Life history and morphology of Rana temporaria in response to pool permanence. Evol. Ecol. Res. 7: 1025–1038. Latta, R.G. 1998. Differentiation of allelic frequencies at quantitative trait loci affecting locally adaptive traits. Am. Nat. 151: 283–292. Latta, R.G. 2003. Gene flow, adaptive population divergence and comparative population structure across loci. New Phytol. 161: 51–58. Laugen, A.T., Laurila, A. & Merila¨, J. 2002. Maternal and genetic contributions to geographical variation in Rana temporaria larval life-history traits. Biol. J. Linn. Soc. 76: 61–70. Laugen, A.T., Kruuk, L.E.B., Laurila, A., Ra¨sa¨nen, K., Stone, J. & Merila¨, J. 2005. Quantitative genetics of larval life-history traits in Rana temporaria in different environmental conditions. Gen. Res. 86: 161–170. Laurila, A. & Kujasalo, J. 1999. Habitat duration, predation risk and phenotypic plasticity in common frog (Rana temporaria) tadpoles. J. Anim. Ecol. 68: 1123–1132. Laurila, A., Karttunen, S. & Merila¨, J. 2002. Adaptive phenotypic plasticity and genetics of larval life histories in two Rana temporaria populations. Evolution 56: 617–627. Leimar, O., Hammerstein, P. & Van Dooren, T.J.M. 2006. A new perspective on developmental plasticity and the principles of adaptive morph determination. Am. Nat. 167: 367–376. Loman, J. 1999. Early metamorphosis in common frog Rana temporaria tadpoles at risk of drying: an experimental demonstration. Amphib.–Reptil. 20: 421–430. Loman, J. & Claesson, D. 2003. Plastic response to pond drying in tadpoles Rana temporaria: tests of cost models. Evol. Ecol. Res. 5: 179–194. Lynch, M. & Walsh, B. 1998. Genetics and Analysis of Quantitative Traits. Sinauer, Sunderland. McKay, J.K. & Latta, R.G. 2002. Adaptive population divergence: markers, QTL and traits. Trends Ecol. Evol. 17: 285–291. Merila¨, J. & Crnokrak, P. 2001. Comparison of genetic differentiation at marker loci and quantitative traits. J. Evol. Biol. 14: 892–903. Merila¨, J., Laurila, A. & Lindgren, B. 2004. Variation in the degree and costs of adaptive phenotypic plasticity among Rana temporaria populations. J. Evol. Biol. 17: 1132–1140. Miner, B.G., Sultan, S.E., Morgan, S.G., Padilla, D.K. & Relyea, R.A. 2005. Ecological consequences of phenotypic plasticity. Trends Ecol. Evol. 20: 685–692. Moran, N.A. 1992. The evolutionary maintenance of alternative phenotypes. Am. Nat. 139: 971–989. Morey, S. & Reznick, D. 2001. Effects of larval density on postmetamorphic spadefoot toads (Spea hammondii). Ecology 82: 510–522. Mousseau, T.A. & Fox, C.W. 1998. The adaptive significance of maternal effects. Trends Ecol. Evol. 13: 403–407. Newman, R.A. 1992. Adaptive plasticity in amphibian metamorphosis: what type of phenotypic variation is adaptive, and what are the costs of such plasticity. BioScience 42: 671–678. Nylin, S. & Gotthard, K. 1998. Plasticity in life-history traits. Annu. Rev. Entomol. 43: 63–83.

1297

Pigliucci, M. 2005. Evolution of phenotypic plasticity: where are we going now? Trends Ecol. Evol. 20: 481–486. Plaistow, S.J., Lapsley, C.T. & Benton, T.G. 2006. Contextdependent intergenerational effects: the interaction between past and present environments and its effect on population dynamics. Am. Nat. 167: 206–215. R Development Core Team. 2005. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna. Relyea, R.A. 2002. Local population differences in phenotypic plasticity: predator-induced changes in wood frog tadpoles. Ecol. Monogr. 72: 77–93. Savage, R.M. 1961. The Ecology and Life History of the Common Frog. Sir Isac Pitman and Sons, London. Seppa¨, K. & Laurila, A. 1999. Genetic structure of island populations of the anurans Rana temporaria and Bufo bufo. Heredity 82: 309–317. Skelly, D.K. & Werner, E.E. 1990. Behavioral and life-historical responses of larval American toads to an odonate predator. Ecology 71: 2313–2322. Smith, D.C. 1987. Adult recruitment in chorus frogs: effects of size and date at metamorphosis. Ecology 68: 344–350. Sultan, S.E. & Spencer, H.G. 2002. Metapopulation structure favours plasticity over local adaptation. Am. Nat. 160: 271– 283. Underwood, A.J. 1997. Experiments in Ecology: Their Logical Design and Interpretation Using Analysis of Variance. Cambridge University Press, Cambridge. Van Buskirk, J. & Saxer, G. 2001. Delayed costs of an induced defence in tadpoles? Morphology, hopping, and development rate at metamorphosis. Evolution 55: 821–829. Van Buskirk, J., McCollum, S.A. & Werner, E.E. 1997. Natural selection for environmentally induced phenotypes in tadpoles. Evolution 51: 1983–1992. Van Tienderen, P.H. 1997. Generalists, specialists, and the evolution of phenotypic plasticity in sympatric populations of distinct species. Evolution 51: 1372–1380. Via, S. & Lande, R. 1985. Genotype–environment interaction and the evolution of phenotypic plasticity. Evolution 39: 505– 522. Wellborn, G.A., Skelly, D.K. & Werner, E.E. 1996. Mechanisms creating community structure across a freshwater habitat gradient. Annu. Rev. Ecol. Evol. Syst. 27: 337–363. Wilbur, H.M. 1987. Regulation of structure in complex systems: experimental temporary pond communities. Ecology 68: 1437– 1452. Williams, D.D. 1997. Temporary ponds and their invertebrate communities. Aquat. Conserv. Mar. Freshw. Ecosyst. 7: 105– 117. Zhivotovsky, L.A. 1996. On the evolution of phenotypic plasticity in a spatially heterogeneous environment. Evolution 50: 1052–1061. Received 25 January 2007; revised 26 February 2007; accepted 27 February 2007
