Habitat partitioning during character displacement between the sexes

Abstract

Ecological differences between the sexes are often interpreted as evidence of within-species ecological character displacement (ECD), a hypothesis with almost no direct tests. Here, we experimentally test two predictions that are direct corollaries of ECD between the sexes, in a salamander. First, we find support for the prediction that each sex has a growth rate advantage in the aquatic microhabitat where it is most commonly found. Second, we test the prediction that selection for ECD in the breeding environment may affect partial migration out of this environment. We found that phenotype-dependent migration resulted in a shift in the phenotypic distribution across treatments, with the highest sexual dimorphism occurring among residents at high founding density, suggesting that migration and ECD can both be driven by competition. Our work illustrates how complex patterns of habitat partitioning evolve during ECD between the sexes and suggest ECD and partial migration can interact to effect both ecological dynamics and evolution of sexual dimorphism.

1. Introduction

Resource competition is often invoked to explain patterns of phenotypic and ecological diversity at a variety of scales. In particular, competition is often suspected to drive evolution of divergent phenotypes [1], a process termed ecological character displacement (ECD). Although ECD is most often envisioned as a driver of divergence between genetically independent lineages, resource competition may also lead to ECD between the sexes of the same species [2]. Observations of habitat partitioning between the sexes are often hypothesized to be the consequence of character displacement between the sexes [3]. Unfortunately, most studies present little or no direct evidence of competition's actual role in the evolution of sex differences in morphology and habitat use. Here, we take a different approach, using a semi-aquatic salamander (figure 1a) study system where past work [4,5] provides explicit evidence that resource competition drives morphological ECD between the sexes at both the micro- and macroevolutionary scales. We make two predictions, bearing on habitat partitioning in ecological communities, that are direct corollaries of the hypothesis of ECD.

Figure 1.

Growth rates and dorsal coloration of adult male and female newts in simulated limnetic and benthic environments. (a) Notophthalmus viridescens, (b) points are least-squares means with 95% CIs and (c) means with 95% CIs. Triangles, male and circles, female.

First, we test for the expected performance trade-offs associated with sex differences in habitat use and growth. We predict that each sex should have a growth rate advantage in its most frequented within-pond microhabitat. This expectation of a fitness trade-off across environments is, in fact, an expected outcome of ECD, in general [6], and more broadly of the maintenance of habitat use polymorphism [7].

Second, we ask whether the negative effects of divergent selection (ECD) in one environment may, in part, be mitigated by phenotype-dependent partial migration (a migration polymorphism where only a subset of the population migrates, usually to or from breeding grounds) to another environment. Based on the observed low performance of intermediate phenotypes in the aquatic habitats [4], we predict that intermediate phenotypes will be most likely to migrate. Thus, our study aims to examine the ecological affects, both direct and indirect, of the evolution of within-species ECD in a system where patterns of competition-driven selection have already been characterized [4,5]. We also unexpectedly discovered plasticity in colour between treatments, which we quantify herein.

2. Methods

(a). Study system

Red-spotted newts (Notophthalmus viridescens) are common semi-aquatic salamanders that are partially migratory; some individuals remain resident year round in breeding ponds, while others undergo a change in skin phenotype and migrate out of the pond to overwinter on land [8]. Individual migration decisions are a plastic response to condition and environmental quality [8].

Newts exhibit sexual dimorphism in head shape and within-pond microhabitat use [4]. Males spend more time in open water, and individuals with more male-like head shape have a high proportion of planktonic prey in their diet, while females are more often found in the benthos and individuals with more female-like head shape have a higher proportion of benthic macroinvertebrates in their diet [4]. Experiments have shown that males and females with extreme morphology have a growth rate advantage over intermediate phenotypes [4]. These results are consistent with the hypothesis that ECD between the sexes is occurring due to competition within ponds [4].

(b). Performance trade-off experiment

We examined growth trade-offs for males and females across within-pond microhabitat. We simulated benthic and limnetic microhabitats by manipulating water level, leaf litter presence and presence of macroinvertebrates or zooplankton, in a single-factor, two-level experiment conducted in 18 tanks (378 l) (nine replicates per treatment). We captured, weighed and measured 72 adult aquatic newts and photographed each for later individual identification. We added four individuals (two male and two female) to each tank; a stratified randomization was used to ensure similar phenotypic distributions across treatments. The experiment was allowed to run until the approximate end of the breeding season when the remaining newts were removed and measured. Following an unexpected observation of plastic dorsal hue colour changes across treatments, we photographed individuals to assay colour (electronic supplementary material).

(c). Partial migration experiment

We performed a single-factor, two-level density (10 or 30 individuals per tank) manipulation in an array of 16 larger (2650 l) artificial ponds to assess the effect of competition and phenotype on individual migration decisions. Although density likely affects many interactions, past work indicates that density is a key driver of the strength of resource competition [4]. The experimental set-up was similar to previous experiments conducted in these tanks [4]. Floating bubble wrap mats covered with leaf litter were added to provide terrestrial substrate for migrating individuals. Two hundred and forty aquatic adult newts were measured (snout-vent length, head-depth, jaw length and gape) and added to tanks in a stratified randomization. These are the same traits measured in the same way as in our past work [4]. Replication of the density treatments was unbalanced (12 low, four high) to achieve the same number of individuals across density levels. The experiment began on 2 June 2016 and was ended on 4 November when all remaining resident individuals were removed and measured.

(d). Statistical analysis

For the performance trade-off experiment, we used a linear mixed model with growth rate (ln(final mass) − ln(initial mass); [4]) or hue value as the response, and sex, treatment and their interaction as fixed effects. We used a generalized mixed model with multinomial error and a generalized logit link to test the effect of density, sex and their interaction on the nominal status of individuals (reside, migrate or die) in the migration experiment. We used a multivariate mixed model with trait values as the response vector, and migration status, sex, trait type and their interactions as fixed effects, as well as the relevant interactions with density. We did not include the main effect of density on morphology because individuals were assigned to density treatments based on morphology (i.e. each tank received a similar distribution of phenotypes). This model allowed for random variation in the migration status effect among tanks. Although this mixed model provides a powerful approach, we also used an alternative approach fitting a quadratic generalized mixed model to assess the effect of sexual dimorphism on residency. This model included residency as a binomial response, and individual discriminant function (df) score (from a df analysis on the sexes), its square, density and its interactions. All models included tank as a random effect. Further details are available in the electronic supplementary material.

3. Results

(a). Performance trade-off experiment

As predicted, males had higher growth rates than females in limnetic habitat, while females had higher growth rates than males in benthic tanks (figure 1b). There was no significant main effect of sex (F_1,42.57 = 0.72, p = 0.40), but the effect of habitat depended on sex (sex × habitat interaction, F_1,42.57 = 4.04, p = 0.0507). Newts had higher growth in the limnetic treatment (habitat effect, F_1,17.01 = 76.75, p < 0.0001; figure 1b), likely due to non-equivalence in total resource abundance across simulated environments. We found a significant difference in dorsal hue across treatments (habitat effect, F_1,15.03 = 31.98, p < 0.0001; figure 1c; sex effect, F_1,40.62 = 0.03, p = 0.86; sex × habitat, F_1,40.62 = 0.01, p = 0.94).

(b). Partial migration experiment

Of the 240 newts, 150 migrated, 54 remained resident and 36 were unrecovered (dead). Individuals in high-density tanks were more likely to migrate or die than individuals in low-density tanks (figure 2a; F_1,19.12 = 3.56, p = 0.048; sex effect, F_1,19.12 = 0.25, p = 0.785; sex × density, F_1,19.12 = 0.14, p = 0.869). This treatment effect on migration and survival led to an approximate equilibration of densities across treatments by the conclusion of the experiment (F_1,14 = 0.42, p = 0.53; figure 2b). Partial migration was non-random with respect to phenotype (migration status × trait effect, p = 0.02; table 1); residents were more sexually dimorphic than migrants (sex × migration status p = 0.045; table 1, figure 2c), although interactions with density treatment were non-significant (table 1). Consistent with these results, density changed the shape of the relationship between morphological discriminant function (df) score and the probability of migration (density × df score², F_1,224.3 = 3.06, p = 0.081), with a reversal in sign of the quadratic relationship across density treatments (figure 2d).

Figure 2.

Treatment effects on phenotype-dependent migration. (a) Density had a significant multivariate effect on migration and death. Confidence intervals are 95%. Dashed line represents the null unit odds ratio. (b) These treatment effects led to an approximate equilibration in final density across starting-density treatments (tank vales, lines indicate means). (c) Sexual dimorphism was higher in residents than migrants. Points are the difference in multivariate least-squares means between the sexes; error bars are standard errors of the difference. (d) Relationship between morphology (discriminant function analysis on the sexes) and probability of remaining resident. The density × phenotype² interaction approached statistical significance. Males, blue circles and females, red crosses. (Online version in colour.)

Table 1.

Fixed effect hypothesis tests from multivariate mixed model analysis of morphology in the migration experiment.

effect	num. df	den. df	F-value	p-value
status	1	99.52	8.69	0.004
sex	1	99.52	12.67	0.0006
status × sex	1	99.52	4.12	0.045
status × trait	3	91.78	3.42	0.0206
sex × trait	3	91.78	28.69	<0.0001
status × sex × trait	3	91.78	1.94	0.1285
status × density	2	105.2	2.09	0.1291
status × density × trait	6	122.4	1.11	0.3596
status × sex × density	2	105.2	1.06	0.35
status × sex × density × trait	6	122.4	0.62	0.7136

4. Discussion

Here, we test for performance trade-offs associated with ECD between the sexes in a salamander. Consistent with predictions, we show that males and females suffer a growth rate trade-off across within-pond habitats; each sex has a growth advantage in the habitat in which it is most commonly found. In addition, we found that partial migration was density- and phenotype-dependent. Migration increased at high density, and individuals at the extremes of the phenotypic distribution were more likely to remain resident at both density levels, leading to increased sexual dimorphism among residents versus migrants. Our results demonstrate the direct effects of ECD on performance in one environment, and moreover that migration and ECD can manifest ecological effects that transcend the environment in which selection occurs.

(a). Fitness trade-offs in aquatic habitat

Models of habitat using polymorphism suggest that steep fitness trade-offs may be necessary for the maintenance of polymorphism in habitat preference [7]. This body of theory is focused on the maintenance of genetically independent morphs, yet the congruence between models of competing lineages and models of ECD between the sexes [9,10] suggests that such trade-offs may also be important in the evolution of sexual dimorphisms in habitat preference.

Our observed growth rate differences across within-pond habitats were small on the absolute scale, but the crossing reaction norms for males and females nonetheless indicate a growth trade-off across these environments. Although growth rate is not fitness, there are several reasons to suspect that the observed trade-off could be important even with the incorporation of other fitness components. First, female habitat use is also related to oviposition and avoiding male harassment [11]. Second, males may be able to chemically search for females more efficiently while active in the open water [12]. This suggests the realized fitness trade-offs may be steeper than for growth rate alone, although this speculation can only be tested with data on individual lifetime reproductive success.

Our finding of a plastic response of dorsal colour to microhabitat treatment was unexpected, in that we did not design the experiment with an a priori interest in this effect. Nonetheless, the finding that both males and females in the limnetic treatment exhibited greener coloration than benthics by the conclusion of the experiment suggests that sexual dimorphism in dorsal coloration observed in the wild [13] may reflect a plastic response to sex differences in within-pond habitat use.

(b). Partial migration between terrestrial and aquatic habitat

We found that migration rate out of the aquatic (breeding) environment occurs non-randomly with respect to both individual phenotype and density. These data suggest that individuals at the extremes of the head shape distribution gain a fitness advantage by residing in ponds, compared with migrating, and that this effect is mediated by resource competition. Although sexual dimorphism was highest among residents at high density (figure 2c), this density × migration interaction was not statistically significant, although the density × morphology interaction approached significance in a binomial model. Moreover, these results illustrate not only that ECD has direct effects on individual microhabitat use and performance in the environment in which selection acts (as shown in the first experiment), but also that ECD in one environment can produce more complex patterns of habitat partitioning across distinct environments.

Although migration depended on male and female phenotype, we found no evidence of an overall sex difference in migration propensity in our experiment (see also [9]). Geographical variation in sex-specific migration [8] suggests that Notophthalmus may present a promising model for studies integrating the evolution of sex differences, partial migration and local adaptation.

(c). Character displacement and migration in a complex life cycle

Models of character displacement between the sexes assume a simple life cycle [2]. Our work in an organism with a complex and variable life cycle suggests that character displacement within one environment and partial migration to another may both affect evolution of ecological sexual dimorphism. In the case of Notophthalmus, partial migration of intermediate phenotypes occurred when selection against these phenotypes in the aquatic environment would have been strongest, suggesting that character displacement between the sexes may be moderated by this phenotype-dependent partial migration, consistent with the macroevolutionary correlation of dimorphism and aquatic life cycles [5]. A limitation of our interpretation is that we have no measures of component fitness from the terrestrial environment.

Ecological processes, namely competition, provide a key to causal explanations for the origins of phenotypic diversity, by mediating evolutionary change, population dynamics and the structure of communities. If such processes are general and conserved through time, competition then provides a mechanism that can unite demography, microevolution and macroevolution. Here, we demonstrate patterns of habitat partitioning in a salamander that can be predicted by previously observed patterns of phenotypic selection (microevolution, [4]) and among-species evolutionary change (macroevolution; [5]). This correspondence suggests a common role for intraspecific resource competition in driving both ecological dynamics of migration and evolutionary change between the sexes across disparate time scales.

Ethics

All experiments were conducted with ethics approval from the BioSciences Animal Care Committee at University of Toronto (Protocol no. 20011715). All fieldwork was performed with the approval of the Koffler Scientific Reserve.

Data accessibility

Data have been uploaded in the Dryad Digital Repository: https://dx.doi.org/10.5061/dryad.1h73pd8 [14].

Authors' contributions

S.P.D.L., S.P. and L.R. designed the study, S.P.D.L. and S.P. performed the experiments and analysed the data, S.P.D.L. and L.R. wrote the manuscript, and all authors revised the manuscript, gave final approval and are accountable for the work herein.

Competing interests

We declare we have no competing interests.

Funding

Funding was from NSERC and the Canada Research Chairs program to L.R.