Introduction

An estimated 80% of animal species have complex life cycles (CLCs) wherein metamorphosis separates two or more discrete, post-embryonic life stages (Wilbur, 1980). Although disagreements persist over diagnostic criteria, the general consensus is that metamorphosis involves an irreversible transformation in morphology that is typically accompanied by a pronounced change in ecology (Bishop et al., 2006). This change results in specialized stages optimized for distinct ecological niches (Benesh, Chubb, & Parker, 2013; Bishop et al., 2006; Ebenman, 1992; Istock, 1967; Moran, 1994). One explanation for the prevalence of CLCs is that independent adaptations at the different phases allow for optimal growth at some stages and optimal reproductive success at other stages (Bryant, 1969; Moran, 1994; Truman & Riddiford, 2019). Central to this explanation is the idea that pleiotropy creates genetic correlations across ontogeny that constrain evolution when traits beneficial for one stage are detrimental to another (Haldane, 1932).
The adaptive decoupling hypothesis (ADH) proposes that metamorphosis evolved as a mechanism for optimizing genetic correlations between life stages, thereby facilitating the independent evolution of traits when opposing selection pressures are experienced during different life stages (Ebenman, 1992; Haldane, 1932; Istock, 1967; Moran, 1994; Wigglesworth, 1954; Wilbur, 1980). A key prediction of the ADH is that traits that experience antagonistic selection across development will be genetically decoupled between distinct life stages and that more ecologically distinct life stages should have greater decoupling. To date, tests of this prediction have been mixed, with some studies supporting the ADH (Anderson, Scott, & Dukas, 2016; Blouin, 1992; Bonett & Blair, 2017; Goedert & Calsbeek, 2019; Hilbish, Winn, & Rawson, 1993; Jacobs et al., 2006; Loeschcke & Krebs, 1996; Medina, Vega-Trejo, Wallenius, Symonds, & Stuart-Fox, 2020; Parichy, 1998; Phillips, 1998; Saenko, Jerónimo, & Beldade, 2012; Sherratt, Vidal-García, Anstis, & Keogh, 2017; Wollenberg Valero et al., 2017), others refuting (Chippindale et al., 1998; Crean, Monro, & Marshall, 2011; Fellous & Lazzaro, 2011; Watkins, 2001; Wilson & Krause, 2012), and some with equivocal results (Aguirre, Blows, & Marshall, 2014; Helle, Johansson, Lederer, & Lind, 2010). However, many of these studies are limited by using only a few morphological traits and not taking stage-specific selection pressures into account when evaluating predictions of the ADH. Importantly, if metamorphosis is an adaptation for optimizing genetic independence, then the magnitude of trait decoupling should depend on the strength of antagonistic selection. Testing this prediction of the ADH will require quantifying decoupling for large and diverse collections of traits that vary in the extent to which they experience antagonistic selection.
Whole-transcriptome gene-expression data obtained from multiple life stages provide an ideal collection of traits for evaluating the extent to which patterns of genetic decoupling fit predictions of the ADH (Collet & Fellous, 2019). First, because all stages of the life cycle must be encoded by a single genome, dramatic phenotypic changes that accompany metamorphosis must be mediated by changes in gene expression. Second, transcriptomes provide a large number of quantitative traits, all measured in comparable units of gene expression, that can be readily compared across life stages. Third, the genes included in a transcriptome cover a wide range of biological functions that should vary somewhat predictably in the extent to which they experience antagonistic selection across the life cycle. Following the logic of the ADH, this variation in selection should generate predictable variation in gene-expression decoupling. For example, genes involved in basic cellular functions (i.e., housekeeping genes) should be more genetically coupled than genes that mediate ecological interactions that change across the life cycle.
Transcriptome-wide patterns of decoupling should also vary with the magnitude of the ecological change accompanying metamorphosis. The most extreme metamorphic transformations occur in holometabolous insects, whose defining characteristic is a distinct pupal stage that separates two completely different body plans (Gilbert, Tata, & Atkinson, 1996; Heming, 2003; Kristensen, 1999). Importantly, this profound transformation enables one stage to be optimized for feeding and growth (the larval stage) and a second stage for dispersal and reproduction (the adult stage). The genetic independence of larval and adult traits proposed by the ADH may explain, in part, why holometabolous insects are one of the most evolutionarily successful and diverse lineages on the planet (Ebenman, 1992; Haldane, 1932; Istock, 1967; Moran, 1994; Rainford, Hofreiter, Nicholson, & Mayhew, 2014; Truman & Riddiford, 2019; Wigglesworth, 1954; Wilbur, 1980). In some holometabolous lineages, pronounced ecological and morphological transformations also occur between successive larval instars. This phenomenon, which has been dubbed hypermetamorphosis (Belles, 2011), provides a valuable opportunity to test the prediction that transcriptome-wide levels of genetic decoupling between life stages will increase with the dissimilarity of the fitness landscapes to which they are adapting.
Looking beyond metamorphosis, the rationale underlying the ADH applies more generally to any scenario in which a single genome expresses multiple distinct phenotypes that are subject to opposing selection pressures (Collet & Fellous, 2019). Arguably the best-studied scenario of genetic decoupling evolving in response to antagonistic pleiotropy involves alternative phenotypes of a single life stage: adult males and adult females. Just as stage-limited gene expression can reduce genetic correlations across life stages of organisms with CLCs, sex-biased gene expression can enable the independent evolution of male and female traits, leading to the evolution of sexual dimorphism (Assis, Zhou, & Bachtrog, 2012; Ellegren & Parsch, 2007; Parsch & Ellegren, 2013; Perry, Harrison, & Mank, 2014; Proschel, Zhang, & Parsch, 2006). An important distinction between CLCs and sexual dimorphism is that only in CLCs must all alternative phenotypes (i.e., life stages) have non-zero fitness for a novel pleiotropic allele to spread in a population (Collet & Fellous, 2019). More generally, for the same net fitness difference between alternative phenotypes, selection against an allele with opposing fitness effects in different life stages may be stronger than selection against an allele with opposing fitness effects in different sexes. For this reason, trait decoupling should be more pronounced for ecologically distinct life stages than for different sexes.
To date, only a handful of studies have evaluated the prediction that gene-expression traits will be decoupled across metamorphic boundaries as it pertains to the adaptive decoupling hypothesis (Fellous & Lazzaro, 2011; Jacobs et al., 2006; Saenko et al., 2012; Wollenberg Valero et al., 2017). Furthermore, to our knowledge, no study has evaluated whether gene-expression decoupling varies predictably with the type of life-cycle transition or gene function, and few studies have directly compared patterns of sex-biased and stage-biased gene expression (but see (Ometto, Shoemaker, Ross, & Keller, 2011; Perry et al., 2014)). To these ends, we take advantage of a hypermetamorphic and sexually dimorphic species of insect with a well-characterized ecology and annotated genome, the redheaded pine sawfly (Neodiprion lecontei , order: Hymenoptera; family: Diprionidae) (Figure 1).
In addition to the complete metamorphic event that occurs during the pupal stage, there are two metamorphic transitions that occur within the larval stage of the redheaded pine sawfly that result in pronounced changes in coloration and behavior (Atwood & Peck, 1943; Coppel & Benjamin, 1965; Linnen, O’Quin, Shackleford, Sears, & Lindstedt, 2018) and references therein) (Figure 1A). The first metamorphic transition is a shift from a “cryptic” to an “aposematic” feeding larval morph and is less dramatic than the other transitions (hereafter, “minor metamorphosis”). The cryptic morph is lightly pigmented, ingests only the exterior of pine needles while avoiding the toxic resinous core, and retreats to the base of the needle when predators are near. By contrast, the aposematic morph is heavily pigmented, ingests the entire needle, and sequesters the toxic pine resins for use in dramatic anti-predator defensive displays. A more striking transition occurs when the aposematic morph molts into a “dispersing” final instar (hereafter, “major metamorphosis”). The dispersing morph is solitary, non-feeding, less intensely pigmented, and migrates to the litter or soil to spin a cocoon. Complete metamorphosis occurs within the cocoon. The non-feeding adult stage is dedicated entirely to reproduction. Sexually dimorphic adults are highly specialized for sex-specific tasks. Males are excellent fliers and use bipectinate antennae to detect female pheromones from considerable distances. In contrast, females remain near the cocoon eclosion site and use serrate antennae to search for suitable oviposition sites in Pinus needles (Anderbrant, 1993). Like most hymenopterans, N. lecontei females lay a combination of fertilized and unfertilized eggs that will develop into diploid females and haploid males, respectively.
The hypermetamorphic life cycle of the redheaded pine sawfly, the wealth of natural history data for this species, and the logic of the ADH enables us to make a priori predictions about how levels of genetic decoupling (inferred here from the magnitude of differential gene expression) will vary among genes categories, developmental stages, and sexes. We predict that: 1) The extent of gene-expression decoupling should increase with the ecological dissimilarity of the life-stages (Figure 1B). 2) Across the transcriptome, the most pronounced gene-expression decoupling will be observed for genes that mediate ecological changes across development. 3) Because traits expressed in different individuals (sexes) may experience weaker selection for decoupling than traits expressed in multiple life stages of a single individual, we predict trait decoupling between the sexes will be less extreme than that observed between metamorphic events (Figure 1B). To test these predictions, we generated expression data for 9,304 genes via whole-transcriptome sequencing of males of each N. lecontei life stage and adults of both sexes.