The predictably variable decoupling of some ecologically important traits
If metamorphosis facilitates optimization of genetic correlations, then patterns of trait decoupling should depend on how changing ecological selection pressures align with genetic correlations across development (Collet et al. 2019). Variation in selection pressures experienced by different traits should generate corresponding variation in decoupling patterns.  If selection optimizes gene-expression traits, the extent of gene-expression decoupling across life stages/sexes should vary even among genes belonging to the same narrow functional category depending on precise ecological roles (e.g., (Fellous & Lazzaro, 2011)). Consistent with this expectation, we observed a wide range of decoupling patterns for expression levels of different chemosensory genes (Figure 3). In some cases, antagonistic selection may have favored decoupling of gene expression traits—for example, perhaps males and females express different chemosensory genes to optimize mate- and host-finding, respectively. In other cases, selection may have favored positive genetic correlations between host-use traits expressed at different life stages (Agrawal & Stinchcombe, 2009; Vertacnik & Linnen, 2017). For example, in sawflies, each life stage interacts with the host plant and these stages may express some chemosensory genes across development (coupled expression) to monitor important host chemical cues (Figure 1A, Figure 3D). Beyond sawflies, numerous studies of stage-specific gene expression in many other insect taxa have found trends very similar to those reported here, with chemosensory genes exhibiting both high and low levels of gene-expression decoupling (Colgan et al., 2011; Etges et al., 2015; Harker et al., 2013; Koutsos et al., 2007; Lee et al., 2016; Sayadi, Immonen, Bayram, & Arnqvist, 2016; Schönbach et al., 2013; H. Yang et al., 2018; Zhang, He, & Wang, 2016). Thus, high variance among chemosensory genes in the degree of coupling/decoupling across development may be a common outcome of adaptive gene expression evolution (Vertacnik & Linnen, 2017).
Mixed decoupling patterns have also been observed for other types of ecologically important gene-expression traits. One such example comes from a 2011 study that investigated expression decoupling of two antimicrobial peptide (AMP) genes between Drosophila melanogasterlarvae and adults (Fellous & Lazzaro, 2011). Using a quantitative genetic approach, Fellous and Lazarro (2011) found that while one of these immunity genes had transcription levels that were significantly correlated across development, expression of the other gene was genetically decoupled. Although the authors interpreted the strong genetic correlation between larval and adult expression for one of the AMP genes as contradicting the ADH, they also acknowledged that they lacked a priori predictions regarding expected genetic correlations for these immune traits. One potential explanation for these contrasting decoupling patterns is that AMPs with decoupled expression are optimized for stage-specific pathogens, whereas coupled AMPs respond to pathogens that attack multiple life stages. Similarly, because some parasites and pathogens of N. lecontei are highly stage-specific (Coppel and Benjamin 1955; Wilson et al. 1992), we would expect at least some—but not necessarily all—immune genes to have decoupled gene expression. Consistent with these predictions, a manually curated set of N. lecontei immune-related genes (Vertacnik et al., in prep) had highly variable decoupling patterns across life stages (Supplemental Tables 3-5).
The final trait we consider that may frequently evolve under conflicting selection pressures across development is pigmentation. In insects, coloration is subject to diverse selection pressures—including abiotic factors (thermoregulation, UV resistance, desiccation tolerance), predation, and sexual selection—that are likely to change over the course of development (Medina et al., 2020; True, 2003). Consistent with the ADH, an analysis of larval and adult coloration in 246 butterfly species found that color evolution is strongly decoupled across butterfly development (Medina et al., 2020). Notably, whereas selection stemming from predation appears to constrain larval color evolution, sexual selection on adult males gives rise to extensive interspecific variation in adult color. Although Medina et al. (2020) did not investigate the genetic underpinnings of color trait decoupling, genetic analysis in another lepidopteran (Bicyclus anynana ) demonstrates that melanism in larvae and adults is controlled by separate loci (Saenko et al., 2012). Similarly, our results suggest that pine sawflies have pronounced decoupling of gene-expression traits related to pigmentation across development. For example, among our top differentially expressed genes among different life stages were top candidate genes identified by a QTL mapping analysis of among-population variation in melanin-based (pale) and carotenoid-based (Cameo2 ) larval pigmentation traits (Linnen et al., 2018).
Together with previous studies, our findings are consistent with the ADH prediction that selection for optimal trait decoupling across development yields strikingly heterogeneous patterns of decoupling for ecologically important traits. Additional studies are needed, however, to demonstrate that highly and minimally decoupled traits evolve under stage-specific and stage-independent selection pressures, respectively. Nevertheless, this collection of studies highlights the importance of taking ecology and stage-specific selection pressures into account when testing predictions of the ADH.