Sexual dimorphism in a top predator (Notophthalmus viridescens) drives aquatic prey community assembly

Abstract

Intraspecific variation can have important consequences for the structure and function of ecological communities, and serves to link community ecology to evolutionary processes. Differences between the sexes are an overwhelmingly common form of intraspecific variation, but its community-level consequences have never been experimentally investigated. Here, we manipulate the sex ratio of a sexually dimorphic predacious newt in aquatic mesocosms, then track their impact on prey communities. Female and male newts preferentially forage in the benthic and pelagic zones, respectively, causing corresponding reductions in prey abundances in those habitats. Sex ratio differences also explained a large proportion (33%) of differences in the composition of entire pond communities. Ultimately, we demonstrate the impact of known patterns of sexual dimorphism in a predator on its prey, uncovering overlooked links between evolutionary adaptation and the structure of contemporary communities. Given the extreme prevalence of sexual dimorphism, we argue that the independent evolution of the sexes will often have important consequences for ecological communities.

1. Introduction

Community ecology is centrally focused on how the abiotic and biotic environment can shape the distribution of species and the structure of communities [1–3]. A common approach is to use species' traits to predict habitat distributions and species interactions [4]. For instance, well-defended species tend to be consumed less and are therefore more frequently found in communities with heavy predation [5]. However, research linking traits to species’ distributions and community structure generally ignores intraspecific trait differences [6–8]. Far from being inconsequential, intraspecific variation can exceed differences among species [7]; individuals can differ by more than an order of magnitude in body size across ontogeny [9,10], and similar trait differences can arise through genetic or non-genetic mechanisms unrelated to ontogeny [6,11]. Recently, intraspecific differences have gained interest among community ecologists, with some studies showing that intraspecific differences can have equal or larger effects on ecological processes compared with equivalent differences among species [9–11].

Sexual dimorphism, with the exception of ontogenetic shifts, is perhaps the most common and striking form of intraspecific variation in sexually reproducing species [6,9], arising from opposing selection acting on males and females. Crucially, sexual dimorphism is present in virtually all sexual species and populations [12,13], and there is a longstanding appreciation that these differences can be ecologically meaningful [14]. For instance, sex differences in habitat and resource use can evolve via direct ecological causes, such as character displacement between the sexes [15], or as an indirect outcome of selection more directly related to sex roles [14,16]. Regardless of their ultimate cause, such ecological sex differences suggest that males and females can evolve to use different resources (e.g. prey [14,17–20]). For example, Schoener [17] showed the male and female anoles occupy different habitats and feed on different sized prey. From a community perspective, we may then expect that variation in the frequency of males and females (which is common in nature [21]), or the continued evolution of sexual dimorphism, would affect some resources more than others. Shifting impacts on resource abundances are then likely to have ultimate consequences for community structure and composition. Despite the clear links between the evolution of sexual dimorphism and community ecology, the widespread nature of sexual dimorphism [14], and a growing number of correlative studies suggesting a role for sexual dimorphism in shaping community structure [19,22], to our knowledge there are yet no direct empirical examples demonstrating the role of sexual dimorphism in structuring local communities.

Sexual dimorphism, and intraspecific trait variation in general, are likely to be more consequential for communities when occurring in particular types of species. In particular, individual differences probably have larger effects on community composition when arising in keystone species, including predators [6,8,23,24]. For instance, Post et al. [23] demonstrate that morphological differences in feeding structures in anadromous and land-locked alewives cause the restructuring of prey communities. We may then expect the evolution of sexual dimorphism to be particularly important when occurring in predator populations, particularly if dimorphism arises in feeding structures or habits.

Here, we analyse the community-level consequences of sexual dimorphism in a keystone predator, the red-spotted newt (Notophthalmus viridescens). In this system, resource competition has contributed to sexual divergence in morphology, such that females are better suited to benthic habitats [20]. We therefore aim to test the hypothesis that sexual dimorphism may impact prey community composition in predictable ways, ultimately demonstrating links between already characterized patterns of sexually antagonistic selection, sexual dimorphism and community ecology. We hypothesize that female-biased populations should (i) reduce the population sizes of benthic prey taxa, (ii) generally reduce the total abundance of benthic prey, and (iii) shift patterns of overall pond community composition relative to male-biased populations.

2. Material and methods

(a). Study system

Red-spotted newts are a generalist predatory salamander that exhibit subtle sexual dimorphism. Males and females differ in morphology [20], with females having dorsoventrally flattened heads, wider gapes and shorter tails. These morphological differences make females better suited to foraging in the benthos (larger gapes can likely capture larger benthic prey) and males more adept at feeding in the pelagic zone (taller tails are likely to confer increased swimming ability). Past research [20] has demonstrated that these trait differences are reflected by the ecology of males and females, with males spending threefold more time in the pelagic zone and consuming more pelagic prey relative to females. These differences are shown to have arisen at least in part through ecological character displacement between the sexes [20]. Once more, red-spotted newts have been extensively studied as a classic experimental system [25–28] and can act as keystone species structuring ecological communities [26], although this effect has been primarily studied with respect to larval anurans. Despite their known sexual dimorphism and their dominant role in structuring aquatic communities, the link between these facets of their biology has not been considered.

(b). Frequency experiment

We manipulated sex ratios (frequencies) in 10 mesocosms to test for the effects of sex-linked intraspecific variation on prey communities (a full description of the experiment can be found in [20]). The current manuscript reports data from the same experiment, but makes use of community-level samples rather than measures of newt morphology, fitness or diet. Briefly, we created 2650 l aquatic mesocosms at the Koffler Scientific Reserve (King, Ontario, Canada). Circular polyethylene tanks were arranged in a grid and filled with filtered pond water in April 2013, and then each stocked with 40 l randomly assigned aliquots of leaf litter to provide a benthic substrate. Mesocosms were then stocked (3×, sampled from two nearby natural ponds via repeated tows (greater than 10×) of a plankton net through the pelagic zone) with zooplankton combined into a homogeneous solution and allocated evenly and randomly among tanks. Macroinvertebrates were collected by dipnet and allocated randomly (2×) in a similar fashion. Each mesocosm was fitted with a 70% shade cloth lid to prevent colonization. After allowing communities to assemble for ∼one month, we introduced 25 newts to each pond, with half of all ponds receiving newts in a 9M : 16F sex ratio, and half the ponds receiving the opposite treatment (n = 10 ponds, five replicates per treatment). Treatments were assigned randomly to tanks, and randomizations with a poor interspersion of treatments were rejected; previous analyses of a Latin square design conducted in the same tanks in the same location indicated no blocking effects in two orthogonal spatial directions [29]. Individual newts were assigned to treatments/experimental units in a stratified randomization to achieve similar male and female phenotypic distributions, but differing total phenotypic distributions, across replicates. While treatments reflected sex ratio differences, sex ratios ought to only have an effect on ecological communities when the sexes diverge in some aspect of their ecology. As such, we can see these treatments as sex ratio differences that create different frequency distributions of ecologically important traits. Note that our manipulations represent realistic newt densities in natural ponds [30], and are consistent with naturally occurring sex ratios [25].

We allowed natural trophic dynamics to affect community composition for two months before ending the experiment. At the conclusion of the experiment, we sampled aquatic invertebrates using a pipe sampler (approximately 12 cm diameter × 30 cm length). One sample was taken from each quadrant near the edge (wall), and another sample from the centre of each pond (5 total samples for approximately 17 l sampled); samples were thus taken from the same locations in all tanks. Note that the pipe samples included a full sample of the water column and the upper layer of benthic material and macrophytes. This was achieved by fixing a cap to the bottom of the sampler after submergence. As a consequence, this technique more extensively samples pelagic habitat, although such a sampling approach is representative of the contribution of each habitat type to the total community. We stress that while pipe sampling is a standard technique, it is not ideal for highly motile organisms. In a mesocosm setting where many species can occupy the ‘wall’ habitat, pipe sampling away from the wall may misrepresent the actual abundances of these taxa, although 4/5 of all samples from each tank were taken near the wall (within approximately 8–12 cm). We address these limitations by (i) sampling identically from each tank, meaning that any observed patterns could not have occurred because of biased sampling (i.e. these issues add error not signal), and (ii) by repeating all analyses (below) while excluding highly motile organisms. We further reflect on these limitations in the discussion. Note also that there is no evidence that newt males and females differ in their propensity to reside near the walls of the tank (as this was not observed during the observations made in [20]). Samples were pooled then preserved in ethanol for later identification.

We classified prey into taxonomic categories that are likely to also reflect differences in habitat affinity [20]. We considered chironomid larvae, coleopteran larvae, odonate nymphs, larval ephemeropterans, megalopterans larvae, ostracods, chydorid cladocerans and harpacticoid copepods to be benthic prey, while we classified other copepods, other cladocerans, notonectids and rotifers as pelagic prey. Note that while cyclopoid copepods can occur in benthic and pelagic environments, we observed high densities of cyclopoids in the water column during our experiment and so chose to categorize cyclopoids as pelagic prey. Finally, we chose not to classify corixids or adult coleoptera as either benthic or pelagic, because they frequently associated with both habitat types. Note that our classification of organisms into benthic versus pelagic habitats is consistent with the extensive literature investigating habitat selection in aquatic organisms (e.g. [31]), and community assembly in freshwater habitats (e.g. [32]). A small portion of all individuals was unidentifiable (less than 1%), and as such were not assigned to either habitat.

(c). Statistical analyses

We tested for the effects of sexual dimorphism in predacious newts on patterns of prey abundance and community composition. We began by comparing the abundances of each taxon in male-biased and female-biased tanks. Specifically, we used a series of generalized linear models (GLMs) with Poisson error distributions to estimate the abundance of a given taxa using treatment as a main effect. Note that while analyses used count data, patterns are shown as proportion data owing to the difficulty of plotting grossly different abundances.

We next investigated patterns of abundance at the habitat level. We pooled the tank-level abundances of organisms considered to be benthic or pelagic, excluding from the analysis taxa that do not specialize on either habitat type. We then used a GLM to predict the abundance of organisms, including habitat type of the prey and treatment as main effects. Note that we repeated this analysis while excluding Corixidae, Notonectidae, Coleoptera, Odonata, Ephemeroptera and Megaloptera, taxa that may be difficult to sample using pipe-sampling techniques or that may make use of ‘wall’ habitat in mesocosms.

Beyond investigating changes in the abundances of taxa across habitats, we next linked differences in sex ratio to multivariate change in community composition. We used the ‘adonis’ function in R, which is essentially a permutational MANOVA that allows for the partitioning of distance matrices among sources of variation. Specifically, we used a Bray–Curtis metric to create a pairwise distance matrix, then tested for differences in treatment centroids across this multivariate space. This test allows us to test for multivariate differences in community composition between male- and female-biased tanks. We then tested for differences in multivariate dispersion between treatments. Specifically, we used a Bray–Curtis dissimilarity matrix to test for differences in variance using the ‘betadisper’ function in R. Again, we repeated this while excluding Corixidae, Notonectidae, Coleoptera, Odonata, Ephemeroptera and Megaloptera, taxa that may be difficult to sample using pipe-sampling techniques or that may make use of ‘wall’ habitat in mesocosms. Note that we opted not to normalize data because (i) pelagic prey were not enormously more common than benthic prey, and (ii) later analyses confirm that no single taxa drove the majority of differences in community composition (immediately below).

To relate multivariate changes in community composition to changes in the abundance of individual taxa, we used ordination techniques as an additional test of treatment differences in community composition, and to identify key species driving any potential patterns. In particular, we used the canonical correspondence analysis (CCA) to predict multivariate community composition using treatment as a main effect. This test allows us to assign loading values (their importance in explaining shifts in community composition) to each species, linking shifts in multivariate community composition to analyses of prey responses (see GLM analyses above).

All analyses were qualitatively robust to normalizing data, but we chose to report results for raw abundances because normalizing data can disproportionately inflate the importance of rare species that were likely to be missed by our sampling technique (e.g. those that were highly motile). Because this test and all others are abundance based, the possible inclusion/exclusion of rare species from our samples should have very minor consequences for our analyses. However, because extremely abundant taxa such as zooplankton can drive community-level differences without any difference to rarer yet important taxa (e.g. benthic invertebrates), we repeated the permutational MANOVA separately for benthic and pelagic species. All analyses were conducted in R [33] using the ‘lme4’ [34] and ‘vegan’ [35] packages.

3. Results

Population sex ratio affected the abundances of individual taxa, the abundances of organisms in benthic versus pelagic habitats, and the composition of entire pond communities. Taxa associated with benthic habitats, namely chironomids (p < 0.001), coleopteran larvae (p < 0.001), odonates (p = 0.004) and ostracods (p < 0.001), were more common in tanks with few females (figure 1; electronic supplementary material, table S1). Conversely, pelagic taxa such as cyclopoid copepods (p < 0.001) and some types of cladocerans (p < 0.001) were more common when males were rare (figure 1; electronic supplementary material, table S1). Note that the rarity of some taxa could have made detecting differences in abundance between male- and female-biased tanks difficult. However, we detected significant treatment effects on the abundances of all taxa with R² values greater than 0.1, suggesting a tight correspondence between the effect of sex ratio manipulations and our ability to detect significant differences in prey abundances. We also urge caution when interpreting the results for Corixidae, Notonectidae, Coleoptera, Odonata, Ephemeroptera and Megaloptera, because these taxa may be difficult to sample using pipe-sampling techniques. Overall, consistent with past work demonstrating diet differences between the sexes [20], some but not all taxa responded to differences in newt sex ratios, suggesting that sexual dimorphism among predators may shape the dynamics of particular prey taxa.

Figure 1.

The relative abundances of benthic (green) and pelagic (blue) taxa in female versus male-biased ponds. Black points show prey species that have no clear habitat preference. Values of one occur when all individuals of the taxa were found in female-biased tanks, while values of zero mean that all individuals of the taxa were in male-biased tanks. The null expectation is that individuals should be equally common in both treatments (proportion of 0.5 at the horizontal dashed line). Values below this line mean that females are having a larger negative impact on those taxa, whereas values above the line suggest that males have a larger effect on reducing prey abundance. Chironomids, coleopteran larvae, odonates and ostracods were more common in ponds with male-biased newt populations. Cyclopoid copepods and some cladocerans were more common when newt populations were female-biased. We stress that, while sampling was consistent across treatments, results for notonectids, corixids, coleopterans, odonates, ephemeropterans and megalopterans be interpreted cautiously because of issues related to capturing highly motile species using pipe sampling. The area of the circle represents the total abundance of the taxa, and those with asterisks were significantly more or less common in female-biased tanks when tested in a GLM. The dashed line represents equal abundance in ponds with male- and female-biased newt populations. Note that statistical tests were conducted on abundances, but for plotting purposes, results are represented as proportions. chi = chironomid, coleo larv =coleopteran larvae, odo = odonate, ephem = ephemeropteran, mega = Megalopteran, harpa = harpacticoid copepod, ostra = ostracod, chydo = chydorinan cladoceran, cyclo = cyclopoid copepod, cal = calanoid copepod, clad = other types of cladoceran, rot = rotifer, noto = notonectidan, cor=corixid, coleo zdult = adult coleopteran.

These taxa-level differences were reflected in overall habitat-level abundances. As a whole, benthic organisms were 48% more common when females were rare, but pelagic organisms were 25% less common (figure 2; significant habitat by treatment interaction: p = 0.002). While these results are when excluding motile taxa, the same pattern was observed when these taxa were included (p < 0.001). Beyond simply altering the abundances of organisms in certain habitats, the sex ratio of newt populations shifted patterns of multivariate community composition (figure 3; p = 0.01), but had no significant effect on multivariate dispersion (i.e. within-treatment β-diversity; p = 0.16). The significant effect of treatment on multivariate community composition was robust when separately considering pelagic (p = 0.04) and benthic (p = 0.02) species. All results for multivariate community composition were robust when including motile organisms (all p < 0.05). Overall, our treatments were able to explain 33% of all variation in community composition among ponds (R² calculated from permutational MANOVA; 34% when including motile organisms). CCA analyses agreed with analyses of individual species and multivariate community composition. The dominant axis described 30% of all variation in community composition (p = 0.016), and species that were significantly associated with one or the other habitat type in GLM analyses had the same association when analysing CCA loadings (electronic supplementary material, table S1). Once more, zooplankton species did not have exceptional loadings (electronic supplementary material, table S1), further demonstrating that zooplankton alone were not responsible for driving shifts in community composition. On the whole, sexual dimorphism appears to play an important role in mediating patterns of abundance and composition in prey communities.

Figure 2.

The total sample abundance of benthic (green circle) and pelagic (blue square) prey in ponds with female- and male-biased newt populations. Benthic organisms were more common when newt populations were male-biased, but the opposite was true of pelagic organisms. The plot shows data excluding motile organisms, although results were qualitatively identical regardless of their exclusion. Circles are mean values and error bars show 95% confidence intervals.

Figure 3.

An NMDS plot showing multivariate differences in prey communities in ponds with male- (red) and female-biased (blue) newt populations. Newt sex ratio explained 33% of all variance in community composition. In statistical terms, multivariate centroids were significantly different but dispersion about those centroids was not significantly different. Lines are minimum convex hulls connecting points, with each vertex representing one pond community. The plot shows data excluding motile organisms, although results were qualitatively identical regardless of their exclusion. Note that both treatments were replicated five times, but that one of the points for female-biased pond communities lies inside the convex hull.

4. Discussion

Our study demonstrates how intraspecific variation in the form of sexual dimorphism can ultimately structure ecological communities. Male and female newts differ in functional morphology, conferring corresponding differences in habitat preference where females forage in the benthos and males in the pelagic zone [20]. As a result, female-biased populations had larger impacts on benthic taxa, while male-biased populations disproportionally reduced the abundances of several pelagic prey groups (figure 1). Beyond affecting individual taxa, benthic prey were generally more common when females were rare (figure 2), and male- and female-biased newt populations created consistent and large (33% of all variation explained) differences in prey community composition (figure 3). Our results suggest that intraspecific variation in the form of sexual dimorphism can have important consequences for community composition. More broadly, we demonstrate relationships between sexual dimorphism and the structure of ecological communities, which, together with previous work showing sex-specific natural selection [20], highlights a previously ignored link between ecological and evolutionary processes.

In our study, the effects of newt sexual dimorphism on prey populations and communities arose because males and females have different habitat and prey preferences. Past work has demonstrated that relative to females, male newts more frequently occupy pelagic habitats and consume pelagic prey [20]. Our work corroborates and expands on this result by demonstrating that male-biased populations reduce pelagic prey abundances, while female-biased populations have larger impacts on benthic prey (figure 1; electronic supplementary material, table S1). Note that prey were identified as coarse taxonomic groups, creating the intriguing possibility that more nuanced prey responses may arise within groups, and even within species (i.e. potential (co)evolutionary responses). More broadly, while newts are generalist predators [30], our results suggest that much of their generalism arises not because each individual is a generalist, but rather because males and females occupy somewhat different trait spaces and therefore play different roles in the community [20,36]. Realistic shifts in the sex ratios of generalist predators may then have considerable impacts on prey populations and communities, necessitating a consideration of sexual dimorphism and intraspecific variation more broadly when investigating ecological communities.

Beyond their impact on particular taxa, our study is the first to experimentally demonstrate that sexual dimorphism can have community-level impacts. While past work has suggested that sexual dimorphism can influence the co-occurrence of competitors [19,22], these broad correlative studies have not been able to separate cause and effect (i.e. whether sexual dimorphism is a cause or effect of community composition) or demonstrate directly that sexual dimorphism is ecologically relevant. Conversely, our study makes clear links between sexual dimorphism, the ecology of the species and community composition, ultimately showing that realistic shifts in sex ratios can have very large effects on prey communities (explaining 33% of all variation). Interestingly, while not tested here, such shifts in prey community composition are likely to have important consequences for abiotic fluxes [9,37], creating links between the evolution of the sexes and the functioning of entire ecosystems. In short, our study conclusively demonstrates the broad importance of sexual dimorphism for the assembly of contemporary ecological communities, a pattern that is likely to have consequences for ecosystem multi-functioning.

Here we have presumed that all differences in measured abundances and community composition arise because of changing abundances. However, an interesting possibility is that our results could in part reflect changing micro-habitat use by prey. In particular, benthic (pelagic) prey in female (male) treatments may move to the sides of the mesocosms, avoiding both predation and our sampling [38]. For this process to drive our results would require that many organisms be able to detect differences in predator sex ratio [38], which we view as unlikely. If such a process were occurring, we believe it would be the first demonstration of whole communities responding behaviourally to shifting sex ratios. Further experimentation should aim to conclusively separate likely consumptive effects from the behavioural effects of predator sex ratios on their prey.

An important question arising from our study is whether sexual dimorphism is consequential in natural communities. In the narrow sense, our experiment closely mimicked natural variation in newt sex ratios [25], created mesocosms with realistic newt densities [30] and assembled ecological communities with nearly natural levels of complexity. We therefore argue that the effects of sexual dimorphism in our experiment are realistic and are likely to occur in natural ponds. Somewhat more broadly, Notophthalmus viridescens has diverged into several (sub)species, with the (sub)species investigated here being both the least aquatic and the least dimorphic [39]. As a result, sexual dimorphism in the (sub)species studied here ought to have the most modest impact on prey, both because of the relatively subtle differences between the sexes and because they simply spend less time foraging in aquatic environments [39]. We therefore argue that sexual dimorphism can have important consequences for ecological interactions, but the degree to which differences between the sexes structure ecological communities will depend on the species expressing dimorphism, and more broadly the macroevolution of clades [37,39].

Beyond the specific case of newts impacting prey, we can begin to speculate on the broader importance of sexual dimorphism in structuring ecological communities. First, more impactful species (e.g. keystone species) and those species with more dramatic differences in sex ratios should have larger effects on ecological communities. While newts are consequential predators [26,30], they have a smaller impact on prey than other co-occurring species (e.g. large fish) and exhibit fairly typical variation in sex ratios relative to other tetrapods [21]. In this respect, the impact of changing newt sex ratios on community structure that we observe may not be exceptional. We also might expect large trait differences to confer large ecological effects of sexual dimorphism [14]. Again, newts are by no means exceptionally dimorphic, and striking ecologically relevant sexual dimorphisms are common across a range of taxa [14]. Thus, although our experiments demonstrate explicitly the community-wide effects of ecological sexual dimorphism in a specific study system, these types of community-wide effects may be commonplace across a range of species and communities.

A final and crucial consideration when evaluating the importance of sexual dimorphism in ecological communities is that traits will only impact communities by having an ecological function. In our case, differences in head morphology have functions related to micro-habitat selection and foraging [20], with predictable consequences for prey communities (figure 3). Dimorphism in these ecologically relevant traits arose (at least in part) because male and female newts diverged through ecological character displacement caused by resource competition [20], rather than because of discordant sexual selection that may act to create dimorphisms in ecologically irrelevant traits. Overall, sexual dimorphism may have been particularly important in our system because the dimorphic traits being considered are clearly ecologically relevant [14,20]. A simple prediction would then be that sexual dimorphisms created by ecological mechanisms should have larger downstream consequences for ecological communities than dimorphisms arising from sexual selection. However, a crucial caveat to this prediction is that sexually selected traits can adopt adaptive, naturally selected functions secondarily [40]. So although the evolutionary mechanisms underlying the separate evolution of the sexes will partly determine the impact of sexual dimorphism on ecological processes, community-wide effects of sexual dimorphism may nonetheless manifest even in cases where the divergence between the sexes reflects an outcome of sexual selection.

By considering sexual dimorphism in the context of food webs and communities, our work has demonstrated the potential importance of differences between the sexes for community ecology. Our work also adds to a growing realization that sex-specific effects need to be considered to achieve a fuller understanding of ecological and evolutionary dynamics [14,19]. We have shown that ecological sexual dimorphism in a predator can have meaningful consequences for the impact of a predator on prey populations and communities (figures 1–3). Virtually all sexual species exhibit some degree of sexual dimorphism [12,13], and such intraspecific trait differences can have marked impacts on species interactions [6,11], suggesting that sexual dimorphism will often have broad impacts on ecological communities.

Data accessibility

Additional data available as part of the electronic supplementary material.

Authors' contributions

D.S. and S.D.L. conceived of the study. S.D.L. completed the fieldwork. D.S. sorted samples, conducted analyses and wrote the first draft of the paper. Both authors revised and approved the final version of the manuscript.

Competing interests

We declare we have no competing interests.

Funding

Funding was provided to D.S. by NSERC CGS-D and to S.D.L. by grants from NSERC and the Canada Research Chairs Program to Locke Rowe.