Organisms inhabiting heterogeneous and/or seasonal environments often show phenotypic plasticity in which a single genotype yields different phenotypes in response to biotic and/or abiotic aspects of the environment (Pigliucci 2001). The adaptive value of phenotypic plasticity is trait specific; for some traits expressing different phenotypes under heterogeneous environments maximizes fitness, whereas for others maintaining the trait value over a range of conditions (phenotypic canalization) is a more beneficial mechanism (Stearns and Kawecki 1994). In order for phenotypic plasticity to evolve, genetic variation is required in the form of genotype–environment interaction (Via and Lande 1985). Furthermore, understanding the evolutionary potential of organisms requires quantifying the amount of genetic variability expressed for the traits of interest that are under selection. Therefore, to understand the evolution of phenotypic plasticity and to assess the potential for future evolutionary change, it is essential to determine the structure of genetic variation for a suite of traits within and across environments.
Genetic variances depend on allele frequencies and are thus specific to populations and environments (Scheiner 1993; Falconer and Mackay 1996). Heritability of a trait are typically affected by environmental conditions (e.g., Hoffmann and Merilä 1999; Charmantier and Garant 2005; Hallsson and Björklund 2012). This variation in heritability estimates across environmental conditions can be due to changes in the additive genetic variance (VA) and/or the environmental variance (VE), and may or may not entail a low genetic correlation between the expressions of the trait across the environments (Hoffmann and Merilä 1999; Charmantier and Garant 2005). A priori predictions about the influence of environmental conditions on the different components of heritability estimates are, however, hampered by discrepancies across studies (e.g., Hoffmann and Parsons 1991; Hoffmann and Merilä 1999; Merilä and Sheldon 1999; Charmantier and Garant 2005). The meta-analyses of studies comparing heritability estimates under favorable versus unfavorable conditions using data from wild populations by Charmantier and Garant (2005) suggested that, in general, estimates of both VA and VE are decreased under unfavorable conditions. The general trend in experimental research on Drosophila and some other insects under laboratory conditions, on the other hand, is that heritability estimates, including VA, are increased under more stressful conditions (e.g., Hoffmann and Merilä 1999). It has been suggested that some of these differences may be because research under laboratory conditions often uses more extreme and/or more novel environmental stressors (Charmantier and Garant 2005). Importantly, the additive genetic variance has commonly been shown to increase under novel environmental conditions (independent of whether these conditions are favorable or unfavorable), possibly due to expression of genes that have not been under selection in the more common environment (Holloway et al. 1990). Whereas organisms in the wild may typically experience stressful conditions, in the laboratory they do not. Therefore, once adapted to the optimal lab environment, introducing suboptimal conditions may in fact create novel environments. This highlights the importance of choosing a study system that facilitates disentangling these potentially confounding factors. One promising route might be to use a species for which it has been shown that adaptive phenotypic plasticity is an integral part of its natural life history. Using such a study organism, the value and responses of genetic variance estimates of key traits in the predictable environment that drives the phenotypic plasticity and to an unpredictable stressful environment can be determined. Thus, within one biological system, naturally perceived and more unpredictable environments can be contrasted allowing the interpretation of the effects on genetic variation estimates in the light of past and future evolution. Here, we report on this approach using the tropical butterfly Bicyclus anynana.
In B. anynana, phenotypic plasticity is a crucial component of the life cycle as it lives in highly seasonal environments for rainfall and temperature, and exhibits two very distinct seasonal forms that differ in wing pattern and many other traits (Brakefield et al. 2007). Environmental conditions during development, specifically those related to the thermal environment, are used as a cue for the future environment, and, subsequently, strongly influence hormone dynamics, juvenile growth, and the resulting adult life-history trajectories (e.g., Bauerfeind and Fischer 2005; Saastamoinen et al. 2010; Oostra et al. 2011). Alterations in a suite of life-history and morphological traits in B. anynana represent adaptive responses to seasonal differences in reproduction and survival as the wet and dry seasons are associated with favorable and more stressful environmental conditions, respectively (seasonal polyphenism; Brakefield and Larsen 1984; Brakefield et al. 2007, 2009). More specifically, individuals of the wet season form experience warmer ambient temperatures, and as a result have a shorter development time, become smaller as adults, reproduce at faster rate (higher investment to fecundity), and allocate less resources to body maintenance (i.e., fat reserves) resulting in reduced life span compared with the dry season form (e.g., Brakefield and Reitsma 1991; Brakefield and Kesbeke 1997; Pijpe et al. 2006). Dry season forms, on the other hand, tend to experience cooler ambient temperatures in the wild. Ambient temperature during the final larval instar is the main determinant of the two seasonal forms (Oostra et al. 2011). Recently, we have assessed in laboratory experiments how larval resources, which will also vary in nature, influence adult life-history traits in the wet season environment (Saastamoinen et al. 2010). Crucially in the context of our present study, even though developmental nutritional limitation generally reduces body mass and fitness (Bauerfeind and Fischer 2005; Saastamoinen et al. 2010), individuals also changed their body allocation in ways likely to reflect an adaptive response to deteriorating environmental conditions (Van den Heuvel et al. 2013).
Given that we have shown that phenotypic responses in B. anynana for a variety of traits work as adaptations allowing individuals to cope more effectively with one or other of the two alternating seasonal environments, responses to food limitation may also be season dependent. For instance, the increased thorax-to-abdomen ratio that occurs in females in response to larval food limitation in the wet season (Saastamoinen et al. 2010) may be less pronounced (or absent) in the dry season, as females in this season already allocate very little to fecundity (i.e., resulting in higher thorax ratio; see Oostra et al. 2011). It is therefore relevant to study plastic responses of the traits, and their interrelationships, under both wet and dry season environments and to determine whether the potential for evolutionary change varies across seasons and environments. In particular the aim of this study was to test the following. First, as genotypes may be more constrained in reaching their potential under harsher environmental conditions (Gebhardt-Henrich and van Noordwijk 1991; Charmantier and Garant 2005), do we observe reduced heritability of performance traits in the laboratory-induced dry season form? Second, as nutritional limitation in the wild is less predictable and hence represents a more novel and less anticipated condition compared with the predictable thermal variation, do we observe increased heritability under conditions of nutritional limitation? Finally, as the levels of additive genetic variances can depend on their relative importance to fitness (Stearns and Kawecki 1994) and the strength or opportunity for selection, are the genetic variance patterns consistent across the phenotypic traits? We implement an information-theoretic approach into a quantitative analysis to arrive at an unbiased estimation of variance components within and across environments that is insensitive to variation in trait means and measurement error. Thus, our study provides comprehensive estimates of genetic variation within and across environments in a species for which adaptive phenotypic plasticity is an integral part of its life history. We will discuss the results in the light of the effects of natural seasonal and unpredictable stressful environments and we will relate our findings to the ecology of Bicyclus. We conclude that although significant variation in genetic variance estimates exists between environments and traits, including for the strength of the across-environment genetic correlation, no uniform pattern can be observed. The interpretation of heritability both for this study and in previously published literature will be greatly helped by studying variation in gene expression to allow a direct estimate of the absolute and relative contribution of these genes to the composite estimate that heritability is.