- Top of page
- Materials and methods
When genetic influences on a phenotype oppose environmental influences, a geographical pattern referred to as countergradient variation results (Conover & Present 1990; Arnott, Chiba & Conover 2006). Countergradient variation can lead to similar phenotypes along an environmental gradient because genetic variation masks environmental variation (Conover & Schultz 1995; Gotthard 2001). Countergradient variation calls attention to what should be obvious, yet is frequently unexpected by researchers: phenotypes generally reflect the optimization of multiple traits not the maximization of a single trait. Indeed, life-history theory was formulated according to a principle of optimization, where an individual trait may be constrained from reaching its theoretical maximum due to trade-offs between traits (Conover & Present 1990; Stearns 1992; Roff 2002). In other words, natural selection maximizes fitness as a function of the entire phenotype. Countergradient variation in growth rate provides an illuminating example. If we assume that growth rate has an overwhelming effect on fitness, we might expect that maximizing growth rate would also maximize fitness. However, as Arendt (1997) emphasized, subsequent studies reinforced (Arnott et al. 2006), and theory has begun to address (Mangel & Stamps 2001), most organisms grow more slowly than their physiological maxima.
Potential constraints on growth rate can take many forms but recent work suggests that two mechanisms are particularly important. First, when rapid growth increases the risk of predation (Biro et al. 2006), natural selection can favour sub-maximal growth. Second, studies of compensatory growth have revealed that organisms sometimes divert energy to growth at the expense of other performances related to survival and fecundity. We draw these two examples from a large set of potential trade-offs that constrain growth rates (Mangel & Stamps 2001). Indeed, many trade-offs can cause the costs of rapid growth to outweigh the benefits (Metcalfe & Monaghan 2001). Such trade-offs represent the core of life-history theory, which provides a general framework for understanding how growth rate evolves in the context of the entire life cycle (Angilletta, Steury & Sears 2004; Angilletta, Oufiero & Leache 2006b).
In this article, we extend a previous study that revealed countergradient variation in embryonic growth and development of the eastern fence lizard, Sceloporus undulatus. We ask whether countergradient variation evolved because of trade-offs between embryonic and juvenile traits. Specifically, lizards from colder environments lay larger eggs that sustain faster growth and development of embryos (Warner & Andrews 2003; Niewiarowski, Angilletta & Leache 2004; Oufiero & Angilletta 2006b). Importantly, variation in growth and development persisted even after differences in egg size were eliminated by yolkectomy (Oufiero & Angilletta 2006). All else being equal, embryos from southern populations should grow and develop as rapidly as those from northern populations unless the resulting costs differ between environments (Conover & Schultz 1995; Gotthard 2001). Two, complementary explanations could account for the evolutionary divergence of growth and development between populations. First, trade-offs associated with rapid growth and development could differ among populations. Second, different environments could favour different growth rates despite identical trade-offs between embryonic physiology and other traits. In other words, the phenotypic trade-off could be driven by genetic, environmental, or interactive effects. Whatever its source, if rapid embryonic growth and development have different consequences for juvenile traits expressed in different environments, then this strategy would not be favoured in all environments. To distinguish between these alternative explanations, we must determine whether the covariation among embryonic physiology and other traits differs between lizards from cold and warm environments. If the covariation does not differ, we can reject the first explanation, at least in terms of the specific traits examined. We can then investigate whether countergradient variation evolved, because the optimal phenotype varies among environments.
If the allocation of energy to rapid embryonic growth and development limits the energy available for juveniles, a trade-off between embryonic and juvenile performances would result (Gotthard 2004; Munch & Conover 2004; Fischer et al. 2005; Royle, Metcalfe & Lindstrom 2006). Although many juvenile traits could trade off with embryonic growth and development, we restricted our attention to the growth rate and locomotor performance of hatchings. These traits integrate many physiological and morphological systems and should indicate overall quality of the juvenile phenotype (Bennett 1978; Sinervo 1990). Moreover, both measures are frequently considered correlates of fitness (see Angilletta, Hill & Robson 2002). Specifically, we asked whether rapid growth and development by embryos from three populations (New Jersey, Virginia and South Carolina) comes at the expense of juvenile performance. We incubated eggs and reared hatchlings in a common garden, while employing the techniques of allometric engineering to control for differences in egg size among populations (Oufiero & Angilletta 2006a). Our experiment enabled us to test (i) whether embryonic growth and development negatively covary with juvenile growth and locomotion, and (ii) whether similar patterns of phenotypic covariation exist within and among populations.
- Top of page
- Materials and methods
Survivorship from egg to hatching (NJ = 75%; SC = 56%; VA = 73%) differed significantly among populations (χ2 = 8·5, P = 0·015) because of the lower survivorship of eggs from SC. Nonetheless, no significant differences in survivorship were detected among treatments within populations (χ2 = 0·05, P = 0·82). In addition, survivorship from hatching to 60 days of age (NJ = 59%; SC = 76%; VA = 57%) did not differ significantly among populations (χ2 = 3·93, P = 0·14) or treatments within populations (χ2 = 0·07, P = 0·80). For the results reported below, we included only those animals that survived to the age of 60 days after hatching; including individuals that did not survive the duration of the experiment would not have changed the mean values for embryonic traits, but would have precluded analyses of juvenile traits.
Yolkectomy successfully reduced the size of naturally large eggs from NJ and VA, such that mean egg masses were similar among populations (F2,39 = 1·7, P = 0·19; Table 1). The manova yielded a significant model, in which population and treatment contributed to variation in incubation period, hatchling SVL, maximum sprint speed and growth rate (Wilks’λ = 0·25, overall model approximate F28·113 = 1·88, P = 0·0109). Population had a significant effect (F8,62 = 4·90, P = 0·0001), but treatment did not (F20,104 = 0·77, P = 0·77). Because we primarily wanted to know whether our manipulation of egg size introduced artefacts, we removed treatment from the model and refit the manova. Excluding treatment from the final model simplified interpretation of the differences in traits among populations (see below; Fig. 2).
Table 1. Characteristics of eggs, embryos and juveniles from three populations of Sceloporus undulatus. Values are means ± 1 SE. Eggs from New Jersey and Virginia included those reduced in size by yolkectomy (see text for methodological details). Sprint speed and growth in length were assayed at an age of 60 days. Unadjusted sprint speeds are given here, but size-adjusted residuals were used for statistical analyses. Statistical differences reported in the text are based on the anova for each trait after a significant difference among populations was detected with a manova
|Population||n†||Egg mass (g)||Hatchling SVL (mm)||Incubation period (day)||Sprint speed (cm/s)||Growth (mm)|
|NJ||13||0·42 ± 0·008||25·5 ± 0·20||67·3 ± 1·00||75·9 + 7·4||7·5 + 0·60|
|SC||19||0·39 ± 0·010||24·7 ± 0·34||72·1 ± 0·67||75·9 + 5·8||9·3 + 0·56|
|VA||10||0·40 ± 0·020||24·8 ± 0·28||66·7 ± 0·84||69·7 + 5·7||6·9 + 0·68|
Figure 2. Pairwise plots of principal components from a CPCA of the covariance matrix. Because principal components are orthogonal to one another, we can interpret each plot independently. Sprint speed was adjusted for body length prior to the analysis (see text for details). All populations share a common pattern of phenotypic covariation, even though mean values of several traits differ among populations (see Tables 1 and 2).
Download figure to PowerPoint
The manova revealed significant heterogeneity of means among populations (manova, Wilks’λ = 0·4, approximate F8,72 = 5·2, P = 0·0001), driven mainly by differences in incubation period and juvenile growth among populations (Table 1). These differences are reflected by the clustering of phenotypes in multivariate space: the phenotypes of NJ and VA lizards cluster together and appear relatively distinct from the phenotypes of SC lizards (Fig. 1). Subsequent analysis of individual traits (anova) exposed trends consistent with a previous experiment (Table 1; Oufiero & Angilletta 2006). Specifically, sizes at hatching were similar among populations (F2,39 = 2·2, P = 0·12), but lizards from NJ and VA hatched earlier than embryos from SC (F2,39 = 14·1, P < 0·0001). These observations imply that embryos from NJ and VA grew and developed faster than embryos from SC. Maximal sprint speed of hatchlings did not differ significantly among populations (F2,39 = 0·07, P = 0·93), but lizards from SC grew faster in length than lizards from NJ and VA (F2,39 = 4·4, P = 0·02).
Figure 1. A canonical plot showing variation in the mean values of traits among populations (NJ, VA and SC). Black squares represent the position of the centroid in canonical space (i.e. the grand mean of the vector of trait means). Blue spheres depict the 95% confidence intervals in multivariate space, as determined by manova. Green rays show original variables and their contribution to discrimination in the canonical space. Note, SC appears differentiated from VA and NJ along canonical axis 1, which mainly describes covariance between incubation period and growth rate. All populations overlap substantially along canonical axis 2, which mainly describes covariance between sprint speed and hatchling snout-vent length (SVL).
Download figure to PowerPoint
Despite significant variation in incubation period and juvenile growth among populations (Fig. 1), CPCA did not reject the hypothesis of a common covariance structure among populations (Table 2). Three principal components of the CPCA describe 26%, 25% and 34% of the total variance (note: in CPCA, components are not necessarily extracted in order of the variance described; see (Phillips & Arnold 1999). The phenotypic space depicted by the three principal components contrasted SVL at hatching with juvenile growth rate (Fig. 2a,b) and incubation period (Fig. 2b,c). For example, PC1 describes an axis defined at one end by lizards that hatched at a large size but grew slowly after hatching, and at the other end by lizards that hatched at a small size but grew rapidly after hatching. Sprint speed loaded very strongly on PC2, but none of the other variables had a major influence on this axis. Finally, PC3 described an axis defined at one end by lizards that developed rapidly (short incubation period) as embryos and grew rapidly after hatching, and at the other end by lizards that developed slowly and grew slowly. Thus, the largest common component, PC3, revealed a relationship between incubation period and juvenile growth within populations that contrasted the relationship among populations (Fig. 3).
Table 2. (a) Flury decomposition of the χ2 statistic, using a step-up model-building approach (Phillips & Arnold 1999). At the highest level, we considered a model in which covarinace matrices were the same for all populations. At the lowest level, we compared models in which the covariance matrices shared only one principal component (CPC) with a model in which the matrices have no components in common vs. being unrelated is tested : (B) Eigenvectors for the best pooled covariance matrix
|CPC (2)||CPC (1)||0·221||4||0·9943||0·055||23·933|
|Unrelated||–|| || || || ||40|
|Trait||PC1 (26%)||PC2 (25%)||PC3 (34%)|
Figure 3. The relationship between the residual incubation period of embryos and the residual growth rate of juveniles differed between lizards from northern and southern environments. The regression for northern populations (NJ and VA; open circles, dotted line) was significant (F1,21 = 9·3, P = 0·002), but that for SC (South; closed circles, solid line) was not (F1,17 = 0·01, P = 0·9). Note that a low residual incubation period translates to a relatively short developmental period; therefore, embryos that developed relatively rapidly grew relatively rapidly as juveniles. Data are multivariate residuals accounting for correlations between sprint speed and size at hatching.
Download figure to PowerPoint
- Top of page
- Materials and methods
The results of our experiment extend those of a previous experiment by Oufiero & Angilletta (2006), who concluded that countergradient variation in embryonic growth and development evolved independently in two clades of S. undulatus. In both experiments, fence lizards from cold environments grew and developed rapidly as embryos. Both experiments included genotypes from VA, NJ and SC, and controlled for variation in egg size among populations. For countergradient variation to evolve by natural selection, rapid embryonic growth and development must impose costs that arise from one or more trade-offs (Gotthard, Nylin & Wiklund 2000; Angilletta et al. 2003). Our experiment was designed to test whether a trade-off between embryonic and juvenile performances could explain why genotypes from all environments do not grow and develop as rapidly as possible. We found significant differences in traits among populations, driven mainly by the prolonged embryonic development (i.e., a long incubation period) and rapid juvenile growth of SC lizards relative to VA and NJ lizards (Fig. 1). These differences among populations are consistent with the hypothesis that embryonic performance (e.g., incubation period) trades off with juvenile performance (e.g., growth rate).
Surprisingly, our analysis of phenotypic covariances within populations revealed two results that counter our hypothesis about trade-offs between embryonic and juvenile traits (see Fig. 2). Incubation period covaried negatively with juvenile growth rate. In other words, individuals that developed rapidly as embryos also grew rapidly during the first 60 days after hatching (Fig. 3). Consequently, although the phenotypic covariance among populations suggests a trade-off that could explain countergradient variation, the phenotypic covariances within populations do not. Furthermore, we could not reject the hypothesis that all populations shared a common covariance structure (see Table 2), even though the relationships between embryonic developmental rate and juvenile growth rate seemed to differ among populations (Fig. 3); specifically, lizards from NJ and VA exhibited a significant negative relationship whereas lizards from SC show no significant relationship. Admittedly, low statistical power might account for the inability of the CPCA to reject a hypothesis of a common covariance structure (Phillips & Arnold 1999; Houle, Mezey & Galpern 2002; Mezey & Houle 2003), but direct inspection of the covariances supports the interpretation that no trade-off occurred at the individual level (Phillips & Arnold 1999; see Fig. 3 and Table 2b).
We can explain these contradictory results in several ways. First, countergradient variation might have resulted from genetic drift among populations, meaning we should expect no trade-off between embryonic and juvenile performances. While this possibility will always remain, we prefer to explore more interesting hypotheses before invoking chance as an explanation for phenotypic patterns. Second, countergradient variation might have evolved because of a trade-off between embryonic traits. For example, rapid embryonic growth might lead to a poor chance of survival to hatching. Our data fail to support this hypothesis because embryos from SC developed more slowly and were less likely to survive than were embryos from NJ and VA. Finally, our experiment might have been insufficiently designed to detect certain trade-offs between the embryonic and juvenile performances, even if these trade-offs do occur in natural environments. Of the three hypotheses, this last one seems most worthy of further discussion.
If natural selection caused the evolution of countergradient variation in S. undulatus, maximal rates of growth and development must not confer the greatest fitness in all natural environments (Arnott et al. 2006). Our findings suggest a trade-off between embryonic performances and either juvenile growth or locomotion cannot explain the evolution of countergradient variation in S. undulatus. Accordingly, an unidentified trade-off would have to constrain embryonic performance in nature. In free-ranging juveniles, rapid growth likely comes at the expense of a greater risk of predation, or some other source of mortality (Mangel & Stamps 2001). For example, Conover and colleagues (Billerbeck, Lankford & Conover 2001) showed that fast-growing genotypes of Menidia menidia fed more, swam slower, and hence suffered greater predation during staged encounters in the laboratory. Whether such trade-offs maintain countergradient variation in natural environments remains to be demonstrated for M. menidia or any other species. Given that rapid embryonic growth and development in S. undulatus entails depletion of yolk reserves (Storm & Angilletta 2007), hatchlings may emerge with less discretionary energy and greater metabolic demands. This need for intensive foraging would have imposed no risk in the artificial environment of our experiment, but would likely do so in a natural environment. Only a field experiment can definitively test this hypothesis for the maintenance of countergradient variation.
Even if we can identify a plausible trade-off, we must also explain why the inter-populational covariance between embryonic and juvenile traits differed from the inter-individual covariance. The relatively rapid juvenile growth of SC lizards could have arisen from compensatory mechanisms triggered by slow embryonic growth. In our experiment, eggs were incubated at the temperatures of nests in NJ and VA. Hatchlings were reared in a common environment, in which each lizard could behaviourally thermoregulate. Furthermore, energy during the embryonic stage was limited by egg size, but food after hatching was unlimited. Consequently, we did not equally constrain rates of embryonic and juvenile growth. Our incubation temperatures, although characteristic of NJ and VA, might have been relatively low for SC embryos. If so, we might have observed compensatory growth by SC lizards during the juvenile stage. Compensatory growth commonly follows periods of food deprivation in a diversity of ectotherms (Metcalfe & Monaghan 2001), but may also follow periods of slow growth at low temperatures (Metcalfe & Monaghan 2001; Hurst et al. 2005). Several complementary mechanisms could have produced compensatory growth, including greater feeding rates and more effective thermoregulation of juveniles (sensu Hertz, Huey & Stevenson 1993). The potential costs of such behaviours would not necessarily be manifested in the laboratory, which was free of predators, parasites, and other sources of mortality. Therefore, compensatory growth could have produced the negative covariation between embryonic and juvenile performances that we observed among populations (Fig. 1).
Field experiments (e.g., reciprocal transplants) will likely reveal more complex patterns of phenotypic covariation than those anticipated by classical theory, pointing the way towards better models of life-history evolution. In general, the rate at which empiricists have uncovered behavioural and physiological mechanisms underlying life-history strategies has outpaced the development of theory (Angilletta et al. 2003). For example, recent work on animals with complex life cycles suggests the phenotypes of adults depend on trade-offs that span multiple ontogenetic stages (De Block & Stoks 2005; Fischer et al. 2005; Ficetola & De Bernardi 2006; Stoks, De Block & Mcpeek 2006). Yet most models of life-history evolution compartmentalize rather than integrate these stages. This compartmentalization has likely limited our ability to predict how life histories evolve along environmental clines. Although we failed to find evidence for specific trade-offs between embryonic and juvenile performances, some remaining hypotheses for the evolution of countergradient variation concern the integration of phenotypes among life stages. Thus, broadening our focus from the independent evolution of traits to the evolutionary integration of traits should result in a more robust theory (Pigliucci 2003; Pigliucci & Preston 2004).