Larval habitats impose trait-dependent limits on the direction and rate of adult evolution in dragonflies

Abstract

Natural selection on juveniles is often invoked as a constraint on adult evolution, but it remains unclear when such restrictions will have their greatest impact. Selection on juveniles could, for example, mainly limit the evolution of adult traits that mostly develop prior to maturity. Alternatively, selection on juveniles might primarily constrain the evolution of adult traits that experience weak or context-dependent selection in the adult stage. Using a comparative study of dragonflies, I tested these hypotheses by examining how a species’ larval habitat was related to the evolution of two adult traits that differ in development and exposure to selection: adult size and male ornamentation. Whereas adult size is fixed at metamorphosis and experiences consistent positive selection in the adult stage, ornaments develop throughout adulthood and provide context-dependent fitness benefits. My results show that species that develop in less stable larval habitats have smaller adult sizes and slower rates of adult size evolution. However, these risky larval habitats do not limit ornament expression or rates of ornament evolution. Selection on juveniles may therefore primarily affect the evolution of adult traits that mostly develop prior to maturity.

1. Introduction

Darwin [1] hypothesized that exaggerated adult morphology arises from pressure to maximize reproduction. Indeed, in the years since his seminal work, fecundity and sexual selection on adults have proven to favour larger bodies, elaborate colours and formidable weapons [2]. Despite the strong selective pressures on adults, however, fossil records indicate that the exaggeration of adult morphology is often limited [3]. The factors restricting adult evolution are therefore a major unresolved problem in biology [4,5]. One limitation on the evolution of larger or more exaggerated adult characters could be a propensity to hinder adult survival [2]. However, viability selection on adult morphology is often much weaker than fecundity and sexual selection [4], making alternate explanations necessary. Another impediment could be selection in the juvenile stage, which can oppose the development and evolution of larger adult traits over a small number of generations (e.g. [6]). Nevertheless, given that short-term trade-offs do not always translate into long-term evolutionary constraints [3,7], it is unclear if or when selection on juveniles limits the exaggeration of adult morphology.

When selection on juveniles opposes prolonged growth and development, two factors may especially influence the subsequent effects on adult evolution: (i) the adult trait's developmental timing and (ii) the strength of selection it experiences in the adult stage [8]. For instance, some adult traits develop primarily prior to maturity (e.g. arthropod body size), whereas others continue to develop throughout adulthood (e.g. avian seasonal plumage). The selection on juveniles may particularly restrict the exaggeration of adult characters that are fixed at maturity because they have few options to compensate later in life [8,9]. Alternatively, when juveniles experience intense selection, resource trade-offs may allow taxa to only optimally develop those adult traits under the strongest selection in the adult stage [8,10]. Selection on juveniles may then limit the exaggeration of adult traits that experience relatively weak or inconsistent selection in the adult stage. By exploring which kinds of adult traits are most affected over long timescales, we can resolve the conditions wherein selection on juveniles most greatly influences the evolution of adults.

Here, I used a comparative study of North American dragonflies to examine how the evolution of adult traits is related to ecological conditions in the juvenile stage. I specifically considered the evolution of two adult traits that differ in developmental timing and exposure to selection: (i) body size, which is fixed at metamorphosis [11] and is positively favoured by fecundity, sexual and viability selection [5] and (ii) male wing ornamentation, which continues developing after metamorphosis [12–14] and provides context-dependent fitness benefits [15,16]. I compared the evolution of these traits between taxa that develop in lentic versus lotic habitats, which often differ in their conduciveness to prolonged larval growth and development. For instance, whereas lotic environments usually remain cool and well oxygenated throughout the year, lentic environments frequently reach extreme temperatures and low oxygen levels [11,17,18]—two of the greatest threats to aquatic organisms' growth and survival [19,20]. Additionally, although both lotic and lentic environments can dry each year, sedimentation shrinks lentic habitats across long timescales and causes conditions to worsen over time [18,21]. Such habitat instability within and across generations opposes lengthy juvenile periods and favours traits that improve dispersal and juvenile survival [17]. Indeed, lentic dragonflies and other aquatic insects exhibit characteristics associated with adapting to unstable habitats (e.g. higher between-population gene flow [21]; evolution of greater dispersal [22–24]).

Using the relationships between a species’ larval habitat and its adult body size and ornamentation, I then tested two competing hypotheses. If restrictions on juvenile growth and development primarily impede the exaggeration of adult traits that are fixed at maturity, then lentic species will have smaller adult bodies and slower rates of body size evolution than lotic species, but similar patterns of ornament evolution. If restrictions on juvenile growth and development instead limit the exaggeration of inconsistently selected adult traits, then body size evolution may not differ between lentic and lotic taxa, but lentic species will be less likely to possess male ornamentation and will exhibit slower rates of ornament evolution.

2. Methods

(a) . Data collection

I compiled species' traits for 320 Nearctic dragonflies. A species' larval habitat preference was taken from Paulson's field guides on Nearctic odonates [25,26]. Larval habitat use was defined for 294 of these species. Body size was taken as the mid-point of the size range listed by Paulson [25,26]. Species were scored as having male wing ornamentation if multiple field guides and iNaturalist observations showed dark melanin pigmentation within their wing cells (sensu [27]). Because size and ornamentation could be related to the major reproductive strategies of dragonflies—‘perchers’ that remain perched in reproductive territories until approached by mates or rivals versus ‘fliers’ that continuously patrol territories—I compiled each species' strategy from research articles and field guides [11,25,26,28,29]. Finally, as some species oviposit into plant stems (endophytic), which may alter abdomen evolution compared to species that oviposit into the water or soil (collectively termed ‘exophytic’) [11], I recorded each species' oviposition strategy from field guides where available [25,26]. To estimate the shared evolutionary history of the species in the analyses, I pruned a recent time-calibrated Odonata phylogeny, which includes 21% of Earth's approximately 6400 extant taxa [5,30].

(b) . Analyses

Using R v. 4.0.3 [31], I first assessed differences in adult size evolution using phylogenetic generalized least-squares (phytools [32], electronic supplementary material, table S1). Each species' log_e-transformed adult size was fit as the response and its larval habitat as an explanatory variable. To control for size differences between perchers and fliers [11], I also included each species’ reproductive strategy. Finally, I modelled a species' oviposition strategy to account for possible relationships with size or larval habitat. In this analysis, a Pagel's λ branch length transformation had more support than Brownian motion or a star phylogeny (electronic supplementary material, table S1). I assessed significance using likelihood ratio tests. I also compared the rate of adult size evolution between larval habitats using the phylogenetic simulation approach described by Adams ([33], geomorph [34]). If selection against prolonged larval growth restricts the exaggeration of adult traits that are fixed at metamorphosis, then lentic species should be smaller and/or have slower rates of adult size evolution.

I next assessed patterns in male wing ornamentation using phylogenetic logistic regression (phylolm [35], electronic supplementary material, table S2). Male wing coloration was fit as a binary response variable (present = 1, absent = 0) and larval habitat as an explanatory variable. To control for possible confounding relationships, I included a species' reproductive strategy and log_e-transformed body size. Terms are significant when 95% confidence intervals from 1000 bootstrapped replicates do not overlap zero [36]. I also tested if the rate of ornament evolution depended on a species’ larval habitat. Here, I used AIC comparisons of models from Pagel's method [37] in which evolutionary rates depend on larval habitat to models where they do not. If selection against prolonged larval growth limits the exaggeration of adult traits that are under inconsistent or context-dependent selection, then lentic species should be less likely to have male ornamentation and/or have slower rates of ornament evolution.

3. Results

I tested if larval development in lentic habitats restricts the evolutionary exaggeration of adult size and/or male ornamentation in dragonflies. Even when controlling for size differences between perchers and fliers (electronic supplementary material, table S3), lotic species are larger than lentic species (β ± s.e. = 0.060 ± 0.025, ${\chi }_1^2=5.55$, p = 0.019, figure 1). Endophytic species are similarly sized to exophytic species (electronic supplementary material, table S3), and results were similar when dividing exophytic species into water-ovipositing versus soil-ovipositing species (electronic supplementary material, table S4). Rates of adult size evolution were also faster in lotic than lentic taxa (${\sigma }_{\mathrm{lotic}}^2=0.00173$ and ${\sigma }_{\mathrm{lentic}}^2=0.00103$, p = 0.001). Conversely, when accounting for the relationships among ornamentation, size and reproductive strategy (electronic supplementary material, table S5), lotic species are actually less likely to possess male ornamentation (β ± s.e. = −0.553 ± 0.369, 95%CIs = −1.419 to −0.023, figure 2). However, supplemental analyses indicate that this overall pattern is not present when focusing only on the Libelluloidea, which typically occupy lentic habitats, or only on the other major lineages (electronic supplementary material, table S6). Moreover, rates of ornament evolution did not differ between lotic and lentic taxa (ΔAIC = 1.81, electronic supplementary material, table S7).

Figure 1.

Species that develop in lentic habitats have smaller adult sizes. (a) Phylogeny of focal species. Bars show adult size (mm) and are coloured by larval habitat (lentic = green, lotic = purple). Family abbreviations: Corduli = Corduliidae; Macrom = Macromiidae; Libell = Libellulidae; Petal = Petaluridae; Cordule = Cordulegastridae; Aeshn = Aeshnidae; Gomph = Gomphidae. (b) Estimated-marginal mean adult size (±s.e.) of lentic versus lotic species from phylogenetic generalized least-squares.

Figure 2.

Species that develop in lentic habitats are more likely to possess male wing ornamentation. (a) Phylogeny of focal species. Bars are coloured by larval habitat (lentic = green, lotic = purple/blue) and the species' possession of wing coloration (Y = dark, N = light). Family abbreviations: Corduli = Corduliidae; Macrom = Macromiidae; Libell = Libellulidae; Petal = Petaluridae; Cordule = Cordulegastridae; Aeshn = Aeshnidae; Gomph = Gomphidae. (b) Probability of male ornamentation (±s.e.) for lentic versus lotic species from phylogenetic logistic regression.

4. Discussion

A major topic in evolutionary biology concerns the conditions under which natural selection on juveniles affects the long-term evolution of adult morphology [5,7,38,39]. My results indicate that the direction and rate of adult size evolution are restricted in taxa that develop in lentic habitats, which are prone to periods of hypoxia and extreme temperatures within generations as well as gradual shrinking across generations [19,20]. By contrast, lentic habitats do not appear to limit either the production of male ornamentation or the rate of ornament evolution. Such patterns suggest that inhabiting a riskier juvenile environment is a larger impediment to the exaggeration of adult traits that are fixed at maturity than it is to traits that continue developing throughout adulthood.

When the juvenile environment is not conducive for prolonged growth, selection favours resource allocation towards traits that improve immediate survival and accelerate development [17]. In such cases, the resources available for subsequent adult traits may be limited. Though all evolutionary constraints are predicted to be circumvented over long timescales [7,38], Waller & Svensson [5] instead found that the evolution of larger adult size in odonates was associated with longer development times. Accordingly, they suggested that the evolution of large size might be impeded by the likelihood of encountering larval habitats that are dangerous for prolonged occupancy [5]. Consistent with this hypothesis, my results indicate that species with smaller adult sizes tend to complete their larval stages in habitats that are prone to hypoxia and extreme temperatures—two of the gravest threats to aquatic organisms [19,20]. If selection against prolonged juvenile growth and development commonly shapes the evolution of adult size, as suggested by their findings [5] and mine, species that prefer other high-risk juvenile habitats (e.g. predation [39]) should also evolve smaller adult sizes and at slower rates. In the future, it will be valuable to examine which juvenile adaptations are most responsible for these limits on larger adult sizes (e.g. investment in maintenance versus growth; multivoltinism). As lotic and lentic habitats could also plausibly differ in selective factors in the adult stage (e.g. temperature, habitat openness, degree of territoriality, predators [15,40,41]), future work should address how pressures on adults augment and/or counteract the physiological conditions faced by larvae. Nonetheless, my findings support the hypothesis that inhabiting larval environments that are less suitable for prolonged growth and development can restrict the exaggeration of adult traits that are fixed at maturity.

Whereas lentic larval habitats limit adult size evolution, my analyses show that they do not impede the evolution of male ornament production. Unlike body size, wing ornamentation begins to develop prior to metamorphosis but continues throughout adulthood [12–14]. Such adult traits may have more options to circumvent any early-life limits on their development, perhaps including increased reliance on adult resources and developmental processes [3,7,8,42]. Intriguingly, however, lentic dragonflies are somewhat more likely to possess male ornamentation—though this pattern across the major lineages was not recapitulated within them (electronic supplementary material, table S4). The Anisoptera-wide patterns could plausibly arise if lotic and lentic habitats differ, on average, in some other unknown selective pressures on adults (e.g. light [41], temperature [16] and predation [15]) or even on juveniles (e.g. via resource trade-offs over tyrosine with immune defense or cuticle integrity [12,13,42,43]; via other physiological costs of melanin production [14]). Undoubtedly, more research should be devoted to studying how juvenile and adult habitats interactively shape ornament evolution, as well as why patterns manifest across the major lineages but not within them [44]. However, at the very least, my findings suggest that less stable juvenile habitats do not limit the exaggeration of adult traits that continue developing after maturity.

A longstanding problem in biology concerns the forces that counteract strong selection for exaggerated adult morphologies [4,5]. As previous work has suggested that most constraints can eventually be circumvented [3,7], little consideration has traditionally been given to ecological pressures in non-adult stages. My results indicate that the juvenile habitat may be one such force that affects the evolution of adult traits that are fixed at maturity, like body size. As male ornamentation was not similarly affected, however, my results suggest that the timing of an adult trait's development may determine the evolutionary impact of selection in other life cycle stages. Moreover, since dragonflies are one of the oldest extant animal clades [5,40], and undergo a dramatic metamorphosis between life cycle stages [36], the patterns observed here may be even more pronounced in the many younger lineages that exhibit less extreme life-history transitions. In this way, my findings indicate that selection on juveniles may indeed be an overlooked constraint on the exaggeration of adult morphology across the animal kingdom.

Ethics

This work complied with requirements of all relevant authorities.

Data accessibility

Data and code are available at: https://doi.org/10.5061/dryad.2rbnzs7n6.

Competing interests

I declare I have no competing interests.

Funding

M.P.M. was supported by G. Kornblum and the Living Earth Collaborative.