Experimental evaluation of evolution and coevolution as agents of ecosystem change in Trinidadian streams

Abstract

Evolution has been shown to be a critical determinant of ecological processes in some systems, but its importance relative to traditional ecological effects is not well known. In addition, almost nothing is known about the role of coevolution in shaping ecosystem function. Here, we experimentally evaluated the relative effects of species invasion (a traditional ecological effect), evolution and coevolution on ecosystem processes in Trinidadian streams. We manipulated the presence and population-of-origin of two common fish species, the guppy (Poecilia reticulata) and the killifish (Rivulus hartii). We measured epilithic algal biomass and accrual, aquatic invertebrate biomass, and detrital decomposition. Our results show that, for some ecosystem responses, the effects of evolution and coevolution were larger than the effects of species invasion. Guppy evolution in response to alternative predation regimes significantly influenced algal biomass and accrual rates. Guppies from a high-predation site caused an increase in algae relative to guppies from a low-predation site; algae effects were probably shaped by observed divergence in rates of nutrient excretion and algae consumption. Rivulus–guppy coevolution significantly influenced the biomass of aquatic invertebrates. Locally coevolved populations reduced invertebrate biomass relative to non-coevolved populations. These results challenge the general assumption that intraspecific diversity is a less critical determinant of ecosystem function than is interspecific diversity. Given existing evidence for contemporary evolution in these fish species, our findings suggest considerable potential for eco-evolutionary feedbacks to operate as populations adapt to natural or anthropogenic perturbations.

1. Introduction

Recent evidence suggests that evolution may play a critical role in shaping ecological processes over contemporary time scales. This evidence is derived largely from studies of how phenotypic and genetic differences among individuals and populations influence the dynamics of populations and the characteristics of communities and ecosystems. Although ecological changes directly associated with ongoing evolution have rarely been observed in nature (but see Grant & Grant 2006; Pelletier et al. 2007; Kinnison et al. 2008), widespread contemporary evolution of ecologically important traits suggests that evolution could be an important driver of population, community and ecosystem dynamics (Thompson 1998, 1999; Hairston et al. 2005; Kinnison & Hairston 2007).

The majority of studies investigating the community and ecosystem effects of intraspecific diversity involve foundation plant species (i.e. plants that structure a community or ecosystem). Intraspecific variation in plant traits, especially leaf chemistry, has been shown to drive variation in soil processes and arthropod communities (Whitham et al. 2006; Bailey et al. 2009; Johnson et al. 2009). Variation among arthropod communities, in turn, can have implications for trophic interactions involving avian predators (Bailey et al. 2006). Recently, the ecological effects of evolutionary divergence have been extended to include the top-down effects of predators. Fish populations that are locally adapted to exploit certain prey resources selectively remove favoured prey, thereby reshaping community composition (Yonekura et al. 2007; Post et al. 2008; Palkovacs & Post 2009). Just as bottom-up effects can cascade up food webs from plants to arthropods to birds (Bailey et al. 2006), top-down effects can cascade down food webs from fishes to zooplankton to phytoplankton (Post et al. 2008). This bidirectionality suggests that there are numerous mechanisms by which evolution may influence ecosystem processes by altering the nature of food-web interactions.

Although intraspecific diversity has been shown to be important for a variety of population, community and ecosystem processes, the relative importance of evolutionary effects compared with traditional ecological effects remains poorly known (Hairston et al. 2005; Ezard et al. 2009). In addition, prior studies have focused on the ecosystem consequences of diversity within single species, even though food webs present opportunities for evolution in multiple interacting species. As such, almost no empirical attention has been paid to potential ecological effects arising via local coevolutionary processes, even though coevolution is a ubiquitous feature of ecological communities (Thompson 2005) and can be a key driver of the evolution of ecologically important traits (e.g. Brodie et al. 2002; Benkman et al. 2003). Understanding the relative magnitude of ecological, evolutionary and coevolutionary effects in communities and ecosystems is a critical step towards assessing whether evolution deserves broad consideration from ecologists (Johnson & Stinchcombe 2007).

In the experiment described here, we evaluated the relative importance of species invasion (a traditional ecological effect), intraspecific evolution and local coevolution for ecosystem processes in model Trinidadian stream ecosystems. Species invasion changes species-level diversity in ecological communities, thereby providing a recognized interspecific baseline against which to measure the importance of intraspecific evolution and coevolution. We focused on two fish species that are native to Trinidad—the guppy (Poeciliidae: Poecilia reticulata) and the killifish (Rivulidae: Rivulus hartii). To examine the effects of species invasion, we compared experimental ecosystems containing Rivulus with experimental ecosystems containing Rivulus and guppies. To examine the effects of intraspecific evolution, we compared experimental ecosystems containing Rivulus from the same population and guppies originating from either of two evolutionarily divergent populations. To examine the effects of local coevolution, we compared ecosystems where the fish community comprised Rivulus and guppies collected either from the same site (locally coevolved) or from different sites (non-coevolved). For each ecosystem comparison, we calculated effect sizes to evaluate the magnitude of evolutionary and coevolutionary effects relative to the (ecological) effects of guppy invasion.

2. Study system

We based our experiment on naturally occurring stream ecosystems in the Northern Range Mountains of Trinidad. Streams in this region are characterized by series of barrier waterfalls that create discrete community and habitat segments. Downstream segments contain large fish predators (e.g. Crenicichla alta, Hoplias malabaricus), open forest canopies and high-resource conditions (high standing stocks of epilithic algae), whereas upstream segments contain no large piscivorous fishes, closed forest canopies and lower resource conditions (Grether et al. 2001; Reznick et al. 2001). Guppies and Rivulus are common in both downstream and upstream habitats. However, Rivulus also occupy headwater streams that lack guppies (Gilliam et al. 1993). Therefore, three identifiable fish communities exist within most drainages: species-rich high-predation communities (HP); Rivulus–guppy low-predation communities (LP); and Rivulus-only communities (RO). Historically, guppies from HP and LP sites have probably invaded RO sites, setting the stage for evolutionary and coevolutionary effects in stream ecosystems.

There are several reasons to suspect that guppy invasion, guppy evolution and Rivulus–guppy coevolution might alter ecosystem properties. One reason involves the potential influence of guppy invasion and evolution on the body size structure of the fish community. Size structure may have important implications for excretion rates, nutrient dynamics and primary production. For a given biomass of fishes, an assemblage of smaller individuals is expected to produce higher excretion-driven nutrient fluxes than an assemblage of larger individuals, which can serve to stimulate primary production (Vanni & Layne 1997; Vanni 2002; Hall et al. 2007). Surveys of HP, LP and RO sites across four Northern Range drainages (Arima, Aripo, Guanapo and Marianne) indicate that the average size of an adult guppy is approximately one-fifth the average size of an individual Rivulus (0.15 g versus 0.75 g; E. P. Palkovacs 2008, unpublished data). Thus, the invasion of guppies into RO streams changes the body size distribution of the fish community. Similarly, guppy populations show evolutionary divergence in life-history traits that also shape body size distributions. Compared with the guppies from LP environments, guppies from HP environments experience higher mortality rates due to predation, driving the evolution of smaller size (and earlier age) at maturity and greater reproductive investment in the form of more, smaller offspring (Reznick 1982; Reznick & Endler 1982; Reznick & Bryga 1996; Reznick et al. 1996). Experimental introductions of guppies from HP sites transplanted to RO sites demonstrate that such life-history differences evolve on contemporary (ecological) time scales (Reznick & Bryga 1987; Reznick et al. 1990, 1997).

A second reason to suspect that guppy invasion, guppy evolution and Rivulus–guppy coevolution may influence ecosystem properties involves interspecific and intraspecific differences in diets and trophic interactions. Guppies are omnivores, consuming both epilithic algae and aquatic invertebrates (Dussault & Kramer 1981; Grether et al. 2001). By contrast, Rivulus are strict predators, foraging on aquatic and terrestrial invertebrates and small fishes (Gilliam et al. 1993; Fraser et al. 1999). The introduction of omnivores into food webs may decrease the biomass of primary producers by reducing the strength of trophic cascades (Bruno & O'Connor 2005; Duffy et al. 2007; Ho & Pennings 2008). At the intraspecific level, guppy populations show divergence in feeding morphology that could underlie dietary differences. Guppies from HP localities have shallower heads with mouths oriented dorsally, whereas guppies from LP sites have deeper heads and more terminal mouths (Langerhans & DeWitt 2004; Hendry et al. 2006; C. K. Ghalambor, J. A. Walker & D. N. Reznick 2004, unpublished data). Similar morphological differences are evident in pelagic versus benthic forms of many lacustrine fish species (Robinson & Wilson 1994) and may have important consequences for guppy foraging behaviour (Robinson & Wilson 1995). In addition, ecological interactions between guppies and Rivulus, including predation by adult Rivulus on small guppies (primarily males and juveniles) and competition for shared prey resources, may set the stage for local coevolution to influence ecosystem properties.

3. Material and methods

(a) Experimental design and analysis

We assembled experimental mesocosms to test the relative effects of species invasion, evolution and coevolution on stream ecosystems. Our experimental treatments were: (i) RO Rivulus-only, (ii) RO Rivulus+HP guppies, (iii) RO Rivulus+LP guppies, and (iv) sympatric LP Rivulus+LP guppies. We employed two methods of statistical analysis. First, treatments were compared using a priori orthogonal contrasts. Contrast 1 (invasion) tested the effects of a traditional ecological variable, species invasion, on ecosystem properties. The invasion contrast, designed to test the mean effect of guppy invasion into RO sites (ignoring intraspecific differences), compared the mean of the RO treatment (treatment 1) with the mean of the three Rivulus–guppy treatments (treatments 2–4). Contrast 2 (evolution) tested the effects of guppy evolution on ecosystem properties, keeping the Rivulus population-of-origin constant. The evolution contrast compared the mean of the HP guppy treatment (treatment 2) with the mean of the LP guppy treatment (treatment 3). Contrast 3 (coevolution) tested the effects of Rivulus–guppy coevolution on ecosystem properties. The coevolution contrast compared the mean of the non-coevolved Rivulus–guppy treatments (treatments 2 and 3) with the mean of the coevolved Rivulus–guppy treatment (treatment 4). Second, we used ANOVA with post hoc Tukey's honestly significant difference (HSD) tests to examine pairwise differences between treatments. Using pairwise tests, we examined the alternative scenarios by which guppies can invade RO streams—from an HP site (treatment 1 versus treatment 2) or an LP site (treatment 1 versus treatment 3). We also tested for differences between the RO treatment and the Rivulus–guppy coevolved treatment (treatment 1 versus treatment 4) because these ecosystem states represent the eco-evolutionary equilibrium states exhibited in this system. Note that the pairwise comparison for guppy evolution (treatment 2 versus treatment 3) was already performed as one of our orthogonal contrasts. Orthogonal contrasts and pairwise tests were evaluated at α=0.05. Statistical analyses were performed using SPSS v. 11.0.4 for Mac OS X (SPSS Inc., Chicago, IL, USA).

Response variables of primary interest included algal standing stocks at the end of the experiment, algal accrual rates over the duration of the experiment, total invertebrate biomass (and the biomass of predators and grazers) at the end of the experiment and decomposition rates over the duration of the experiment. In order to examine the mechanisms by which evolutionarily divergent guppy and Rivulus populations shaped experimental ecosystems, we estimated size-specific guppy and Rivulus excretion rates and examined the algal content of guppy diets. Experimental methods are described in detail below. Variables were log transformed for analysis to satisfy the assumption of normality. To aid the comparison of the magnitude of ecological, evolutionary and coevolutionary effects for each response variable, we calculated means for all treatments and effect sizes for orthogonal contrasts. Effect sizes were calculated as Cohen's d (Cohen 1988).

(b) Stream mesocosms

We constructed 16 experimental stream mesocosms adjacent to Ramdeen Stream, a second-order tributary of the Arima River on the south slope of Trinidad's Northern Range Mountains. Each mesocosm was 2.6 m long×0.5 m wide and received flow from the adjacent stream, piped through a 1500 l settling tank to remove silt and debris. A mixture of gravel and sand was placed at the bottom of each mesocosm. Water depth was maintained at approximately 5 cm at the inflow and approximately 20 cm at the outflow to simulate natural habitat variation. Flow rates in each mesocosm were adjusted daily, and flow was maintained at approximately 67 l h⁻¹ throughout the duration of the experiment. Vertical shade cloth (80 cm high) was erected around the sides of each channel. Light levels, monitored throughout the experiment using HOBO light loggers (Onset Computer Corp., Bourne, MA, USA), indicated that the light environment in the mesocosms was typical of open canopy conditions. Two hollow bricks (30.5 cm×19.5 cm×10 cm) were placed in each mesocosm to serve as habitat structure for fishes. The tops of all mesocosms remained open to allow terrestrial infall, but large leaf debris was removed daily. The experiment was allowed to run for a total of 28 days, starting on 27 June 2007.

We used two source populations for guppies (HP and LP) and two source populations for Rivulus (LP and RO). The LP site was the same for both species. All fishes were collected from the Guanapo River (HP: 10°41′46″ N, 61°15′43″ W; LP: 10°42′37″ N, 61°16′2″ W; RO: 10°43′17″ N, 61°16′51″ W). Collecting sites were identified based on surveys of fish communities conducted by Gilliam et al. (1993). Ten Rivulus (treatment 1) or five male guppies, five female guppies and five Rivulus (treatment 2–4) were introduced into each mesocosm to maintain equal biomass of fishes across all mesocosms. Fishes were introduced on 27 June 2007. Natural size distributions were approximated, with the exception that Rivulus greater than 55 mm in total length were excluded due to the risk of predation on guppies (which would have altered our treatments). Before being introduced into the experiment, all fishes were individually marked with injected elastomer (Northwest Marine Technology Inc., Shaw Island, WA, USA), photographed and weighed. All fishes were collected at the termination of the experiment (28 July 2007), photographed and weighed. Juvenile guppies born during the experiment were collected from all mesocosms at the end of the experiment. No juvenile Rivulus were found in the mesocosms. Differences between treatments in total fishes (numbers and biomass) and guppy reproductive output (numbers and biomass) were evaluated using ANOVA with Tukey's HSD tests.

We introduced invertebrates into each mesocosm one week before the introduction of fishes. An area of Ramdeen Stream equal to the area of all mesocosms was sampled for invertebrates using a 63 μm sieve, and predators (primarily Odonata) were removed. The remaining live invertebrates were homogenized and added in equal amounts to each mesocosm one week before the introduction of fishes. At the termination of the experiment, invertebrate communities were sampled. Two passes along the length of each mesocosm were performed using a kicknet (15 cm×15 cm, 1 mm mesh). This procedure was followed by a 5 min visual inspection to collect any large, mobile macroinvertebrates that escaped the net. Samples were preserved in 70 per cent ethanol.

We identified invertebrates to the level of family according to Merritt & Cummins (1996) and Maharaj & Alkins-Koo (2007). Length was determined for each specimen. Mass was found using length–mass regressions from Benke et al. (1999). Mass for Calamoceratidae was found using the regression equation for Limnephilidae, according to McNeely et al. (2007). Total invertebrate biomass for each mesocosm was estimated by extrapolating from the area sampled to the total mesocosm area.

We added 11 litterbags (10 cm×10 cm, 1 mm mesh, containing 3–4 g freshly picked, dried Pachystachys coccinea leaves) to each mesocosm at the start of the experiment and removed one, weekly, to estimate decomposition rates. Five bags were retained as handling controls (mean handling loss=0.344 g). Decomposition rate was calculated as the slope of the ln-transformed per cent remaining leaf mass regressed on the incubation period in days since the start of the experiment (processing coefficient; Hauer & Lamberti 2006).

We added 12 unglazed tiles (5 cm×5 cm) to each channel at the start of the experiment. Two tiles were chosen at random and removed each week to measure algal accumulation in terms of chlorophyll a concentrations (chl a). The tiles were placed in plastic bags in a −20° C freezer for at least 24 hours prior to extraction to facilitate algal cell lysing. We added 10 ml of 95 per cent ethanol to each bag and extracted in the dark at room temperature for 24 hours. We measured the fluorescence of the diluted extract using an Aquafluor handheld fluorometer (Turner Designs, Sunnyvale, CA, USA) calibrated with a solid standard. The chl a concentration was calculated using the total tile area and extract volume corrected for dilution. The algal accrual rate was calculated as the slope of the regression of chl a concentration on days since tile introduction.

We measured nutrient concentrations at the outflow of each mesocosm on 27 July 2007. Nitrogen was measured as NH₄ using an automated phenate method and as NO₃ using an automated cadmium reduction method (USEPA 1983). Phosphorus was measured as PO₄ using an automated ascorbic acid method (USEPA 1983). Nutrient analyses were performed at the Odum School of Ecology Analytical Chemistry Laboratory at the University of Georgia. Nutrient concentrations in source water entering the mesocosms were also measured on this date: PO₄=1.7 μgP l⁻¹, NH₄=4.4 μgN l⁻¹ and NO₃=202 μgN l⁻¹.

(c) Excretion rates

We estimated size- and sex-specific excretion rates for guppies and Rivulus using short-term excretion trials. Rivulus excretion trials were performed after dark on 26 July 2007, and guppy excretion trials were performed during daylight hours on 27 July 2007. For Rivulus, three individuals were selected from each mesocosm and placed independently into three different plastic ziplock bags. For guppies, individuals were selected from each mesocosm and grouped into four sex/size classes (1 adult female, 2 adult males, 3–5 subadults and 5–8 juveniles). Guppies smaller than 0.1 g wet mass were categorized as juveniles, and guppies larger than 0.1 g wet mass, but unidentifiable as either male or female, were categorized as subadults. Sex/size class groups were placed independently into four different plastic ziplock plastic bags. Each bag contained 200 ml filtered water from the mesocosm source water. Plastic bags containing fishes were placed into buckets containing stream water to prevent temperature stress. After 15 min, 60 ml of water was removed from each bag, filtered through a Pall AE glass fibre filter (Pall Corp., East Hills, NY, USA) and frozen until analysis.

NH₄ and PO₄ concentrations from each excretion bag were measured as described above. To estimate excretion rates, we calculated the difference between nutrient concentrations in excretion bags after incubation minus the average concentration for two control bags (incubated without fishes). We divided this difference by the number of fishes per bag and multiplied by bag volume to obtain an excretion rate per fish. We calculated mass-specific excretion as the per fish rate divided by the average wet mass of fishes per bag. We then estimated the total excretion flux driven by Rivulus and guppies in each mesocosm by applying mass-specific excretion rate estimates to the empirical sex/size distributions obtained via fish censuses conducted on 28 July 2007. We calculated the contribution of nutrients to each mesocosm via guppy and Rivulus excretion relative to nutrient concentrations in source water (as measured on 27 July 2007 and assuming a constant flow rate of 67 l h⁻¹). Differences in total N- and P-excretion (for Rivulus and guppies) were examined using orthogonal contrasts and pairwise tests, as described above.

(d) Guppy diets

Algal consumption was measured from adult guppies collected from the mesocosms at the termination of the experiment and frozen for preservation. We extracted chl a from guppy diets by making a transverse incision across each abdomen with a scalpel to expose stomach contents and incubating them in 95 per cent ethanol for 24 hours before measuring chl a concentrations as described above. Dietary algal content was expressed in terms of chl a mass per fish wet mass. Differences in dietary algae were examined using two-way ANOVA with Tukey's HSD tests, with treatment and sex included as fixed effects. Sample sizes (females, males) were: treatment 2, n=13, 15; treatment 3, n=12, 15; treatment 4, n=9, 13.

4. Results

(a) Fish biomass and guppy offspring production

Total fish biomass at the end of the experiment did not differ among treatments (ANOVA: F_3,12=1.820, p=0.20; table 1). However, the addition of guppies and the source of guppies did alter the overall size structure of the fish community (i.e. the total number of fishes comprising that biomass differed significantly; ANOVA: F_3,12=62.349, p<0.01; table 1). The number of fishes was the lowest in the RO treatment (Tukey's HSD: treatment 1 versus treatment 2, p<0.01; treatment 1 versus treatment 3, p<0.01; treatment 1 versus treatment 4, p<0.01), the greatest in the treatment containing HP guppies (Tukey's HSD: treatment 2 versus treatment 3, p<0.01; treatment 2 versus treatment 4, p<0.01) and did not differ between the treatments containing LP guppies (Tukey's HSD: treatment 3 versus treatment 4, p=0.99). Guppy source effects resulted from significant differences in reproductive output, in terms of both offspring numbers (ANOVA: F_2,9=13.877, p<0.01; table 1) and offspring biomass (ANOVA: F_2,9=10.653, p<0.01; table 1). Compared with LP guppies, HP guppies had more numerous offspring (Tukey's HSD: treatment 2 versus treatment 3, p<0.01; treatment 2 versus treatment 4, p<0.01) and a greater biomass of offspring (Tukey's HSD: treatment 2 versus treatment 3, p<0.01; treatment 2 versus treatment 4, p=0.02). These findings are consistent with previously documented life-history divergence between guppy populations locally adapted to HP versus LP habitats.

Table 1.

Numbers (n m⁻²) and biomass (g m⁻²) for all fishes and for guppy offspring born in the mesocosms (mean±s.e.), collected at the end of the experiment.

	total fishes		guppy offspring
treatment	(n m−2)	(g m−2)	(n m−2)	(g m−2)
1. RO Rivulus	7.83 (0.18)	9.49 (0.39)	—	—
2. RO Rivulus+HP guppies	53.55 (6.85)	10.51 (0.64)	42.05 (6.49)	2.02 (0.32)
3. RO Rivulus+LP guppies	23.60 (2.63)	8.80 (0.66)	11.76 (2.43)	0.38 (0.14)
4. LP Rivulus+LP guppies	24.45 (2.05)	9.15 (0.48)	12.66 (2.18)	0.47 (0.09)

(b) Algae, invertebrates and decomposition rates

Orthogonal contrasts. Guppy invasion, a traditional ecological factor, did not have a significant effect on algal biomass, algal accrual rates, invertebrate biomass (total, grazers or predators) or decomposition rates. Guppy evolution had a significant impact on algal biomass (ANOVA: F_1,12=6.463, p=0.03; figure 1b) and algal accrual rates (ANOVA: F_1,12=7.989, p=0.02), with HP guppies causing increases in algae compared with LP guppies. Rivulus–guppy coevolution had a significant effect on total invertebrate biomass (ANOVA: F_1,12=6.364, p=0.03; figure 1c), with the coevolved fish community reducing aquatic invertebrate biomass relative to the non-coevolved community. The difference in invertebrate biomass between the coevolved and non-coevolved treatments resulted primarily from a change in the biomass of invertebrate predators (ANOVA: F_1,12=8.162, p=0.02). Predatory invertebrates were not included in the invertebrate introduction stock. Therefore, those present at the end of the experiment colonized the mesocosms during the course of the experiment (adult odonates were frequently observed depositing eggs into the mesocosms). Decomposition rates were not significantly different for any contrast. However, the absence of any differences in decomposition rates may have been due to the small mesh size of the litterbags, which prevented many invertebrate shredders and fishes from accessing the leaf litter. Full results for orthogonal contrasts are provided in appendix A.

Figure 1.

Effects of (a) guppy invasion, (b) guppy evolution and (c) Rivulus–guppy coevolution on aquatic invertebrates and epilithic algae (mean±s.e.) in stream mesocosms.

Pairwise comparisons. Aside from the above significant effect of guppy evolution on algal biomass and accrual, pairwise comparisons were not statistically significant. Neither guppy invasion from an HP site (treatment 1 versus treatment 2) nor guppy invasion from an LP site (treatment 1 versus treatment 3) had a significant effect on algal biomass, algal accrual rates, invertebrate biomass (total, grazers or predators) or decomposition rates. Likewise, we did not find a significant difference between the RO treatment and the coevolved Rivulus–guppy treatment (treatment 1 versus treatment 4) for algal biomass, algal accrual rates, invertebrate biomass (total, grazers or predators) or decomposition rates. However, the pattern of treatment means for invertebrate biomass (total, predators and grazers) was consistent with the findings from the orthogonal contrasts (i.e. the lowest biomass in the coevolved treatment, treatment 4). Treatment means (±s.e.) and results of pairwise comparisons are provided in appendix B.

(c) Excretion rates

Orthogonal contrasts. Fish excretion contributed substantially to nutrient concentrations in the mesocosms, and guppy excretion contributed more nutrients than did Rivulus excretion. Orthogonal contrasts revealed a significant effect of guppy invasion on the total nutrient flux due to fish excretion for NH₄ (ANOVA: F_1,12=17.205, p<0.01) and a marginally significant effect for PO₄ (ANOVA: F_1,12=3.600, p=0.08). Relative to nutrient concentrations in source water, guppy excretion increased total N concentrations (measured as NH₄+NO₃) by an average of 7.5 per cent and P concentrations (measured as PO₄) by an average of 58.5 per cent. Rivulus excretion increased N concentrations by an average of just 0.9 per cent and P concentrations by an average of 6.1 per cent. The effect of guppy evolution on total excretion was significant for NH₄ (ANOVA: F_1,12=5.656, p=0.04; figure 2a), with HP guppies contributing more NH₄ via excretion than LP guppies. The effect of guppy evolution was not significant for PO₄ (ANOVA: F_1,12=1.001, p=0.34; figure 2b). Relative to source water, HP guppy excretion increased N concentrations by an average of 11.4 per cent and P concentrations by an average of 82.9 per cent, whereas LP guppy excretion increased N concentrations by an average of 5.6 per cent and P concentrations by an average of 46.4 per cent. Rivulus–guppy coevolution had a significant effect on NH₄ excretion (ANOVA: F_1,12=5.407, p=0.04), with the non-coevolved fish community contributing more NH₄ to the mesocosms than the coevolved community. Coevolution did not have a significant effect on PO₄ excretion (ANOVA: F_1,12=0.496, p=0.50). For the evolution and coevolution contrasts, differences in total fish excretion were driven overwhelmingly by differences in guppy excretion rates, not by differences in Rivulus excretion rates, which did not differ among treatments for either NH₄ or PO₄.

Figure 2.

Mesocosm-level excretion rates (mean±s.e.) for guppies from high-predation (HP) and low-predation (LP) populations coexisting with Rivulus from the Rivulus-only (RO) site ((a) N-excretion and (b) P-excretion).

Pairwise comparisons. Guppy invasion from an HP site had a significant effect on total NH₄ excretion (Tukey's HSD: treatment 1 versus treatment 2, p<0.01) but not on total PO₄ excretion (Tukey's HSD: treatment 1 versus treatment 2, p=0.14). Guppy invasion from an LP site had a marginally significant effect on total NH₄ excretion (Tukey's HSD: treatment 1 versus treatment 3, p=0.06) but no effect on total PO₄ excretion (Tukey's HSD: treatment 1 versus treatment 3, p=0.57). Somewhat surprisingly, based on overall differences in Rivulus versus guppy excretion rates, we did not find a significant difference between the RO treatment and the coevolved Rivulus–guppy treatment for either NH₄ excretion (Tukey's HSD: treatment 1 versus treatment 4, p=0.23) or PO₄ excretion (Tukey's HSD: treatment 1 versus treatment 4, p=0.64). However, mean values were greater for the coevolved treatment (appendix B).

Differences in fish excretion rates were somewhat related to differences in nutrient concentrations (measured at the outflow of each mesocosm) for NH₄ and PO₄, but not for total N (measured as NH₄+NO₃). The HP guppy treatment (treatment 2) had the highest excretion rates and the highest nutrient concentrations for NH₄ and PO₄, but the lowest concentrations for total N. Nutrient concentrations in the water are the product of both excretion and uptake via algal and heterotrophic microbial activity. Therefore, increased algal growth in the HP guppy treatment may have reduced NO₃ concentrations in the water, despite higher rates of fish excretion.

(d) Guppy diets

Overall, algae consumption by guppies differed significantly among treatments (ANOVA: F_2,71=5.039, p=0.01). The amount of algae consumed was significantly greater for LP guppies than for HP guppies (Tukey's HSD: treatment 2 versus treatment 3, p=0.04; treatment 2 versus treatment 4, p=0.01; figure 3) and significantly greater for males than for females (ANOVA: F_2,71=17.257, p<0.01; figure 3). The interaction between treatment and sex was not significant (ANOVA: F_2,71=1.086, p=0.34).

Figure 3.

Algal content of guppy diets (mean±s.e.) for males and females from high-predation (HP, white bars) and low-predation (LP, grey bars) populations coexisting with Rivulus from the Rivulus-only (RO) site.

(e) Effect sizes

Compared with the effects of guppy invasion, guppy evolution and Rivulus–guppy coevolution had relatively large effects on ecosystem properties (figure 4). For algal accrual rates, guppy evolution had the largest effect size (4.6 times as large as guppy invasion), followed by guppy invasion. For invertebrate biomass, Rivulus–guppy coevolution had the largest effect size (4.7 times as large as guppy invasion) followed by guppy evolution (1.8 times as large as guppy invasion). For decomposition rates, effect sizes were relatively small for all contrasts, although the effects of guppy evolution and Rivulus–guppy coevolution tended to be slightly larger than the effect of guppy invasion. For N- and P-excretion rates, guppy invasion had the largest effect size, followed by guppy evolution and Rivulus–guppy coevolution.

Figure 4.

Effect sizes (Cohen's d) for guppy invasion (white bars), guppy evolution (hatched bars) and Rivulus–guppy coevolution (grey bars) derived from orthogonal contrasts. (Contrasts that were significant at the α=0.05 level are indicated by an asterisk.)

5. Discussion

(a) Effects of invasion, evolution and coevolution

Ecologists have traditionally focused on functional diversity at or above the species level to explain community and ecosystem patterns and processes (Hooper et al. 2005). For this reason, species introductions and invasions have garnered much attention. Invading species introduce new phenotypes into communities, thereby initiating novel ecological interactions and modifying existing ones (White et al. 2006). However, phenotype distributions and ecological interactions can also change as a result of evolutionary and coevolutionary processes including those occurring over contemporary (ecological) time scales (Thompson 1999; Hairston et al. 2005). We examined the effects of guppy evolution and Rivulus–guppy coevolution on model Trinidadian stream ecosystems and compared these effects with the effects of guppy invasion. Our results show that the magnitude of evolutionary and coevolutionary effects can exceed those of traditional ecological effects.

Guppies and Rivulus show substantial interspecific differences in body size, excretion rates and trophic interactions. Owing to body size differences, guppies drove higher nutrient excretion rates than did Rivulus. In addition, guppies consume epilithic algae, whereas Rivulus do not. Due to these interspecific differences, guppy invasion might be expected to have important impacts on ecosystem properties. However, our results show that guppy invasion (the substitution of half the biomass of Rivulus for an equal biomass of guppies) did not drastically alter algal dynamics, invertebrate biomass or decomposition rates (figures 1a and 4). The lack of a significant effect of guppy invasion on algae may be due to the opposing forces of nutrient excretion and algal consumption. Algal standing stocks represent primary production minus consumption. Guppy invasion caused a significant increase in N-excretion and a marginally significant increase in P-excretion (figure 4), which are expected to increase primary production. However, algae consumption by guppies may have offset any increases in primary production, resulting in no net change in algal biomass or accrual rates. We observed a small increase in aquatic invertebrate biomass caused by guppy invasion, but the effect was not significant (figures 1a and 4). This lack of a significant effect may have been due to either substantial invertebrate consumption by guppies and/or compensatory feeding by Rivulus when released (somewhat) from intraspecific competition.

Guppy populations display substantial intraspecific variation in life-history traits and morphology as a result of local adaptation to different predation regimes. HP guppies drove significantly higher rates of N-excretion compared with LP guppies (figure 2), perhaps as a consequence of life-history evolution on fish community body size structure. LP guppies consumed greater quantities of algae compared with HP guppies (figure 3), perhaps due to differences in morphology and foraging behaviour. Owing to these differences, guppy evolution might be expected to have important effects on ecosystem properties. Our results show that guppy evolution did indeed have a significant influence on algal biomass and accrual rates, with HP guppies driving increases in algae relative to LP guppies (figures 1b and 4).

Why did guppy evolution cause significant changes in algal biomass and accrual rates while guppy invasion did not? For guppy invasion, the effects of nutrient excretion and algal consumption appeared to work in opposite directions on algal biomass. However, for guppy evolution, these factors appeared to work in concert. Compared with LP guppies, HP guppies excreted N at higher rates, increasing primary production, and fed on algae at lower rates, decreasing algal consumption. Both effects served to increase algal standing stocks, and the net result was a significant increase in algal biomass and accrual rates for mesocosms containing HP guppies (figures 1b and 4).

Interestingly, guppy populations from both HP and LP localities showed consistent sex differences in algal consumption, with males of both types consuming more algae than females (figure 3). Previous studies have suggested that male and female guppies show different patterns of morphological divergence across predation regimes and other habitat features (Hendry et al. 2006) and show different degrees of phenotypic plasticity when faced with alternative food presentations (Robinson & Wilson 1995). One possible explanation for sex-specific aspects of phenotypic divergence and plasticity in trophic morphology is the link between algae consumption and male guppy colour patterns, which are important as mating cues and respond to the quantity of algae-derived carotenoids in the diet (Grether 2000; Grether et al. 2005).

In streams where Rivulus and guppies coexist in the absence of other fish species, predation and competition may drive coevolution between these species. We found that the coevolved Rivulus–guppy treatment had a lower biomass of aquatic invertebrates than the non-coevolved treatments (figure 1c and 4). The reduction of a shared prey resource in the coevolved treatment suggests that competition between guppies and Rivulus may have selected for niche convergence (Scheffer & van Nes 2006) and enhanced competitive ability (Hairston 1980). If coevolved guppies and Rivulus can more efficiently exploit aquatic invertebrates, this hypothesis could explain why coevolution significantly decreased invertebrate biomass. Behavioural observations conducted in the mesocosms provide some support for this interpretation. Guppies and Rivulus showed less habitat segregation and interspecific avoidance behaviour in the coevolved treatment than in the non-coevolved treatments (B. A. Lamphere 2007, unpublished data). However, further work is needed to test this hypothesized mechanism.

The current paradigm for predicting ecological function is based primarily on overall phenotypic similarity or phylogenetic relatedness—the more dissimilar or distantly related two forms, the more probable they will be functionally distinct (Webb et al. 2002; Hooper et al. 2005). This assumption underlies the use of species as functional units, as overall phenotypic differences found among species are generally larger than differences within species. Based on this assumption, one would predict that the relative effect sizes in our experiment would be: guppy invasion>guppy evolution>Rivulus–guppy coevolution. However, our results did not generally conform to this pattern (figure 4). The effect of guppy evolution was greater than the effect of guppy invasion for algal biomass, algal accrual rates, invertebrate biomass and decomposition rates. The effect of Rivulus–guppy coevolution was greater than the effect of guppy invasion for invertebrate biomass and decomposition rates. These results suggest that phenotypic dissimilarity and phylogenetic relatedness are not always reliable predictors of ecological function. Given that contemporary evolution is prevalent in many species (Kinnison & Hendry 2001), including those examined in this study, our findings also suggest that ongoing evolution may merit greater consideration in both basic and applied ecology.

The results of our experiment show that coevolution may be an important factor shaping ecosystem processes. The existence of coevolutionary effects in ecosystems is not unexpected. Ecological processes are often determined not by the phenotypic traits of one species (although the traits of some species may have inordinately large effects; Post & Palkovacs 2009), but by how traits mediate interactions among many species (Urban & Skelly 2006). What is unexpected about our results is the magnitude of coevolutionary effects. For invertebrate biomass and decomposition rates, the effect of coevolution was larger than the effects of either invasion or evolution (figure 4).

Interestingly, our results suggest that it is the source of invading guppies, rather than the invasion of guppies per se, which determines algal biomass and accrual rates, key ecosystem properties. Relative to the RO condition, HP and LP guppy invasions had opposite effects on algal biomass and accrual rates. Perhaps equally intriguing, Rivulus–guppy coevolution returned algal biomass and accrual rates to levels similar to their pre-guppy states. This finding suggests a note of caution for non-experimental studies of invasion. If we were to have compared algal biomass and accrual rates in natural streams at the eco-evolutionary equilibrium states that occur in nature (RO and coevolved Rivulus–guppy), we may have concluded that the addition of guppies to the ecosystem has no significant effects. This scenario provides a potential example of how evolutionary processes may mask ecological dynamics (Yoshida et al. 2007), an area of research that certainly warrants further investigation.

(b) Eco-evolutionary feedbacks

In this study, we examined the ecological effects of phenotypic divergence—the approach commonly used in community and ecosystem genetics (Whitham et al. 2006, 2008; Johnson & Stinchcombe 2007). We are able to draw inferences about the effects of evolution on ecosystem dynamics because previous research has shown that the phenotypes we examined are heritable and subject to contemporary evolution. However, it remains a frontier to experimentally examine the ecosystem effects of dynamically evolving (and coevolving) populations in the wild. Our results suggest that the ecosystem consequences of species invasions may derive, in part, from post-invasion evolution to new environmental conditions and that coevolution with other community members may contribute to community and ecosystem resilience. However, one potentially critical element that can only be entirely captured using dynamic experiments is the eco-evolutionary feedback (Fussmann et al. 2007; Post & Palkovacs 2009).

Eco-evolutionary feedbacks occur when organisms change the biotic or abiotic conditions of their environment and those changes then influence the direction of evolution (Laland et al. 1999; Palkovacs & Post 2008; Post & Palkovacs 2009). Such feedbacks can change the ecological and evolutionary trajectories of systems, causing them to deviate from expectations based on fixed phenotype experiments (Habets et al. 2006). We did not directly test eco-evolutionary feedbacks in this experiment, but we can use the results to speculate about potential feedback mechanisms. As revealed by previous studies, guppies from HP habitats typically experience greater resource availability due to increased algal standing stocks than do guppies from LP habitats (Grether et al. 2001; Reznick et al. 2001). The availability of algal resources appears to contribute to the evolution of female growth rates (Arendt & Reznick 2005) and can influence the evolution of male guppy colour patterns (Grether 2000; Grether et al. 2005). Increases in algal biomass associated with HP environments have previously been interpreted as extrinsically determined features of the ecosystem—features to which guppies respond evolutionarily. By contrast, our findings suggest that guppy populations can influence algal availability as a by-product of the evolution of life-history traits, body size differences, morphological traits and dietary preferences. The scope for guppies to shape their ecosystem in natural streams is currently unknown. However, the experiment reported here raises the real possibility that previously documented evolutionary responses to increased algal availability may represent (at least in part) eco-evolutionary feedbacks. To examine this possibility, we have recently undertaken a series of dynamic experiments tracking the eco-evolutionary consequences of guppy invasion, guppy evolution and Rivulus–guppy coevolution as they unfold in the wild.

Humans are a long-recognized global driver of species invasions. Therefore, conservation biologists focus much attention on the ecology of species invasions and on developing policies to slow the rate of species introductions. Only relatively recently has it been recognized that humans are also a global driver of intraspecific phenotypic change (Palumbi 2001; Hendry et al. 2008). Eco-evolutionary dynamics—the effects of ecology on contemporary evolution and the reciprocal effects of evolution (and coevolution) on ecological processes—provide a new framework to understand natural systems. However, broader integration of eco-evolutionary dynamics into applied ecology and conservation biology will probably depend upon additional studies similar to ours that provide empirical insights into the relative importance of evolution and coevolution in areas of conservation concern, such as species invasions, over-harvesting, habitat alteration and global climate change.

Appendix A

Table 2.

ANOVA tables for a priori orthogonal contrasts (SS, sum of squares; MS, mean squares; *p<0.05).

source	d.f.	SS	MS	F-ratio	p-value
algal biomass
treatments	3	0.129	0.043	2.131	0.15
C1 (invasion)	1	1.097×10−5	1.097×10−5	0.001	0.98
C2 (evolution)	1	0.129	0.129	6.463	0.03*
C3 (coevolution)	1	5.797×10−5	5.797×10−5	0.003	0.96
residual	12	0.243	0.020
total	15	0.372
algal accrual rate
treatments	3	0.133	0.044	2.768	0.09
C1 (invasion)	1	0.005	0.005	0.329	0.58
C2 (evolution)	1	0.128	0.128	7.989	0.02*
C3 (coevolution)	1	5.058×10−7	5.058×10−7	3.161×10−5	0.99
residual	12	0.192	0.016
total	15	0.325
total invertebrate biomass
treatments	3	0.240	0.080	2.602	0.10
C1 (invasion)	1	0.021	0.021	0.666	0.43
C2 (evolution)	1	0.022	0.022	0.710	0.41
C3 (coevolution)	1	0.197	0.197	6.364	0.03*
residual	12	0.369	0.031
total	15	0.609
predator invertebrate biomass
treatments	3	0.405	0.135	3.001	0.07
C1 (invasion)	1	0.027	0.027	0.599	0.46
C2 (evolution)	1	0.011	0.011	0.237	0.64
C3 (coevolution)	1	0.367	0.367	8.162	0.02*
residual	12	0.540	0.045
total	15	0.945
grazer invertebrate biomass
treatments	3	0.021	0.007	0.025	0.99
C1 (invasion)	1	0.011	0.011	0.038	0.85
C2 (evolution)	1	4.322×10−4	4.322×10−4	0.002	0.97
C3 (coevolution)	1	0.010	0.010	0.036	0.85
residual	12	3.334	0.278
total	15	3.355
decomposition rate
treatments	3	0.007	0.002	0.197	0.90
C1 (invasion)	1	7.500×10−7	7.500×10−7	6.250×10−5	0.99
C2 (evolution)	1	0.002	0.002	0.148	0.71
C3 (coevolution)	1	0.005	0.005	0.429	0.53
residual	12	0.140	0.012
total	15	0.147
N-excretion rate
treatments	3	4.551	1.517	24.943	<0.01*
C1 (invasion)	1	2.770	2.770	17.205	<0.01*
C2 (evolution)	1	0.911	0.911	5.656	0.04*
C3 (coevolution)	1	0.871	0.871	5.407	0.04*
residual	12	1.929	0.161
total	15	6.481
P-excretion rate
treatments	3	0.015	0.005	24.943	<0.01*
C1 (invasion)	1	0.011	0.011	3.600	0.08
C2 (evolution)	1	0.003	0.003	1.001	0.34
C3 (coevolution)	1	0.001	0.001	0.496	0.50
residual	12	0.033	0.003
total	15	0.048

Appendix B

Table 3.

Response variables (mean±s.e.) for all treatments and post hoc pairwise comparisons between treatments from Tukey's HSD tests (*p<0.05; see text for details).

	RO Rivulus; treatment 1	RO Rivulus, HP guppies; treatment 2a	RO Rivulus, LP guppies; treatment 3b	LP Rivulus, LP guppies; treatment 4c
algal biomass (mg chl a m−2)	26.92 (2.53)	38.73 (9.03)	20.96 (1.17)	26.98 (1.17)
algal accruald	0.88 (0.07)	1.36 (0.26)	0.75 (0.11)	0.96 (0.04)
invertebrate biomass (mg m−2)	73.11 (15.50)	90.76 (11.98)	117.82 (19.03)	55.96 (11.20)
predator biomass (mg m−2)	49.85 (9.08)	73.38 (9.22)	98.29 (24.11)	36.26 (9.81)
grazer biomass (mg m−2)	11.47 (8.52)	6.17 (3.32)	9.58 (7.52)	4.31 (1.51)
decomposition ratee	0.14 (0.02)	0.15 (0.02)	0.14 (0.02)	0.15 (0.02)
N-excretion rate (mg N −1)	0.18 (0.01)	1.66* (0.30)	0.99 (0.15)	0.76 (0.21)
P-excretion rate (mg P h−1)	0.01 (0.00)	0.10 (0.04)	0.06 (0.01)	0.06 (0.01)

^aGuppy invasion from HP (Tukey's HSD: treatment 1 versus treatment 2).

^bGuppy invasion from LP (Tukey's HSD: treatment 1 versus treatment 3).

^cEco-evolutionary equilibrium states (Tukey's HSD: treatment 1 versus treatment 4).

^dSlope (chl a concentration on days since introduction).

^eSlope (ln per cent remaining leaf mass on days since introduction).