Understanding phenotypic evolution for a trait requires an understanding of the indirect selection pressures it experiences via genetic correlations. Those indirect pressures are a consequence of both the selection exerted on correlated traits, and the mechanisms of correlation by which phenotypic traits can be connected. We have experimentally evaluated the correlated responses to clonal selection on body size in populations that differ in their history of sexual recombination with the aim of understanding the mechanisms of connection. These differences allow us to potentially distinguish among pleiotropy, chance linkage, and selective advantage as mechanisms underlying genetic correlations. Furthermore, sexual recombination itself remains a significant puzzle in evolutionary biology because it simultaneously has the capacity to demolish and assemble genomes with high fitness (Lehtonen et al. 2012; Meirmans et al. 2012); this puzzle is partly about the effects of recombination on genetic correlations. Our results show that the magnitude of correlated responses to selection for some critical life history traits is higher in populations whose genetic composition is the result of frequent sex, despite the observation that the slope of the associations between life history traits and body size do not differ among populations according to historical frequency of sex. Together, these results provide a novel line of support for the hypotheses that selective processes are substantial contributors to life history correlations and that sexual recombination promotes adaptation by natural selection.
Mechanisms of correlation
We observed significant correlated responses to selection on body size at maturity in one or more populations for most traits (specifics given in Tables 3 and S4). Although details for particular traits differ, these generally confirm the observation that life history traits are associated with size at maturity.
For three traits (body size at the release of the fourth clutch, total reproduction, and adult specific growth rate), the magnitude of divergence was clearly greater in populations with a high historical frequency of sex, and a similar trend was evident in two other traits (size at birth and r). Of the potential mechanisms contributing to trait associations, this was predicted only when there was both selection for a particular correlation and the net effect of the higher frequency of sex was to promote adaptation through the creation of high fitness clones. If this is true, we would predict that the strengths of the correlations are tighter in the high-sex populations than in the low-sex populations. However, we cannot evaluate this prediction effectively because we did not conduct the final life table assay on the full set of clones initially isolated. Although our clonal selection procedure allows us to estimate the slope of the relationship efficiently, the absence of the middle-value clones from the life table means that we do not know how tight that relationship is.
Only the inter-clutch interval (equivalent to instar duration) suggested greater divergence in low-sex populations. This could be consistent with a role for chance linkage or selection, if the net effect of sex was deleterious. There was no statistical support for differences among populations in the slope of the relationship (Table S5), which is consistent with the selection in the context of deleterious sex. However, none of the observed divergences were statistically significant, so it is unreasonable to lend much weight to this interpretation.
We found little support for the notion that the association between body size and other life history traits differed among populations, except for size at birth. In principle, this can arise under any of the mechanisms contributing to genetic correlations. There was no significant difference between high and low-sex populations, but the trend was for high-sex populations to have a larger difference than the low-sex populations. This is opposite of what is expected under chance linkage. Under the mechanism of pleiotropies arising from different genes in different populations, there should be no effect of sex. However, even though the trend is insignificant, it is worth considering the possibility that selection is contributing to the correlation between size at maturity and size at birth, but that the optimal correlation differs among populations (discussed below).
If genetic correlations among life history traits were largely driven by pleiotropies, recombination frequency should have no effect on the divergence of correlated traits when selection is applied directly to a key trait. In this case, we would expect to see the same degree of correlated divergence in all populations if each population harbored the same distribution of allelic variation, but this occurred for none of the traits examined. The magnitude of correlated divergence in age at maturity was idiosyncratic with respect to sex, with no evidence that the association slope differed among populations. This is consistent with pleiotropies that emerge from the same genes in each population, if allelic variation differs among populations.
Given that the traits we measured differ in their similarity to size at maturity, the directly selected trait, we expected some traits, particularly body size at later adulthood, to be driven by pleiotropy. For example, it seems reasonable that the shared genetic basis of size at other ages would be strong, of reproductive output (which in Daphnia is constrained by brood chamber size) would be moderate, and of timing traits would be small. Instead, traits from all three classes showed a strong effect of historical frequency of sex. This should not be interpreted to mean that pleiotropy is absent from these correlations, as it is possible for there to be a shared pleiotropic basis for a correlation that is common to all populations onto which selection builds the correlation further.
Collectively, our results show that the relative contribution of alternative mechanisms of correlation may differ among traits, but that selection is a key driver of life history correlations with size at maturity in Daphnia pulicaria. Our data should not be interpreted to mean that selective correlations are solely responsible for trait correlations. Rather, the complexity of our results shows that no single mechanism behind genetic correlations among complex traits provides satisfactory explanations for variation of the traits. Linkage and pleiotropies are likely to also contribute to the variation, and research aimed at identifying specific pathways that drive extant life history variation, coupled with studies that describe the physical relationships of key genes within genomes in the context of recombination frequency will be necessary.
In comparing our study to others that have sought to distinguish among correlation mechanisms, we do not see common patterns emerging, but the relevant studies have been few. Conner (2002) conducted an experiment in which he studied floral traits in large experimental populations of wild radish. He concluded that correlations among six floral traits were pleiotropic, because nine generations of enforced random mating failed to diminish the strength of genetic correlations. In an analysis of QTLs associated with the response to selection on bristle number in Drosophila, Nuzhdin et al. (1999) concluded that responses in correlated bristle traits were likely due to linkage, rather than pleiotropy, but did not attempt to distinguish between adaptive and chance correlations. Delph et al. (2011) investigated the basis of the genetic correlation between male and female flower size in Silene latifolia in a selection experiment where they sought to break the correlation. Despite prior evidence suggesting a role for pleiotropy (Delph et al. 2004), in one selection line they were able to eliminate the genetic correlation almost entirely in a few generations of selection, and concluded that selection was likely to be the fundamental cause of the correlation in nature.
We note that these other studies have focused on morphological traits rather than life history. Before generalizations can emerge, further analyses of this sort are necessary, particularly in organisms such as Raphanus, Silene, and Daphnia where the ecological consequences of trait variation can be evaluated.
Selection on correlations
We have argued that our four populations are ecologically similar, and in the broader context of potential habitats, they are. From the perspective of zooplankton like Daphnia, the most significant ecological differences occur along a gradient of waterbody size (Wellborn et al. 1996; Tessier and Woodruff 2002a) and a gradient of productivity. These differences include the shift from temporary to permanent waters, the shift to where a lake can support planktivorous fish and, at even larger sizes, piscivorous fish, and finally fully stratified lakes with a hypolimnion. It is not clear whether Daphnia pulicaria is a recently derived species or an ecologically defined variant of D. pulex and thus it is hard to define the niche breadth of D. pulicaria. There is ample evidence of natural hybridization (reviewed in Heier and Dudycha 2009), gene flow across the major habitat transitions (Dudycha 2004; Crease et al. 2011; Cristescu et al. 2012), and the absence of reproductive isolation in experimental crosses between stratified lake populations and temporary pond populations (Heier and Dudycha 2009). All four of the populations in this study are stratified lakes with planktivores (mainly bluegill) and piscivores (smallmouth bass), and other species of Daphnia (mainly D. dentifera) as a primary competitor. In addition, all four lakes have similar productivity, near the upper part of what is considered oligotrophic, as estimated from Spring measures of total phosphorus (Cáceres and Tessier 2004).
Nonetheless, the lakes are not identical, and this creates the potential for selection for differences in the precise correlations among traits. The predictions we described with regard to similarity or dissimilarity of correlations across populations are based on the assumption that the same correlations are selected for in each population. For the traits where we have inferred an important role for selection due to greater correlated differences in the higher sex populations, it is hard to see what could lead to similar correlations across populations if there were in fact selection for different correlations. Even if there were real differences in slope that we did not have the power to detect, that would still implicate selection as the driving mechanisms behind the correlations. In fact, the slope of the association with size at birth was different across populations, and this may be due to different selection pressures.
We have detected a quantitative genetic signature of selection on some correlations, but we do not have data that test the fitness consequences of the correlations in the wild. Although we estimated r, this is not indicative of fitness of Daphnia in lakes, where seasonal succession and other environmental variation may have strong impacts on what traits and trait combinations are advantageous at different times. Long-term fitness in Daphnia in the field will depend on the contribution of an individual to future generations via both asexual and sexual reproduction, and genotypes may differ in the reproductive mode in which they have an advantage. It would be very interesting to compare the field performance of clones that differ in body size and degree of deviation from the correlation in their population.
The creative power of sex
Although genetic correlations of life history traits and the evolutionary consequences of recombination have been studied extensively, few attempts have been made to juxtapose them in a single study. This is somewhat surprising, because an essential element of debate about the evolutionary role of recombination is the direction and extent to which it influences favorable combinations of traits. Relatively few study systems offer an opportunity to consider populations that vary in recombination frequency, and most considerations of the effect recombination has on correlations have focused on examining genetic correlations before and after a limited number of rounds of recombination in exclusively sexual species. In our system, the only situation that is expected to yield stronger correlated responses to selection in high-sex populations compared with low sex is when genetic correlations result from selection for the correlation and the creative power of sex outweighs its destructive power. For Daphnia, capitalizing upon the power of sex to create advantageous combinations of alleles probably depends on the opportunity for fit clones to expand through asexual reproduction. This raises the question of what constraints operate to determine an optimal frequency of sex in cyclic parthenogens, as it is easy to imagine that as the frequency of sex increases, the expansion of fit clones would decrease.