INFLUENCE OF THE INDIRECT EFFECTS OF GUPPIES ON LIFE-HISTORY EVOLUTION IN RIVULUS HARTII

Transcription

1 ORIGINAL ARTICLE doi: /j x INFLUENCE OF THE INDIRECT EFFECTS OF GUPPIES ON LIFE-HISTORY EVOLUTION IN RIVULUS HARTII Matthew R. Walsh 1,2,3 and David N. Reznick 1,4 1 Department of Biology, University of California, Riverside, California Department of Ecology and Evolutionary Biology, Yale University, New Haven, Connecticut matthew.walsh@yale.edu 4 david.reznick@ucr.edu Received August 4, 2009 Accepted November 19, 2009 Early theories of life-history evolution predict that increased predation on young/small individuals selects for delayed maturation and decreased reproductive effort, but such theory only considers changes in mortality. Predators reduce prey abundance and increase food to survivors. Theory that incorporates such indirect effects yields different predictions. Trinidadian killifish, Rivulus hartii, inhabit communities with and without guppies. Guppies prey on young Rivulus and Rivulus densities decline and growth rates increase when guppies are present. Prior work showed that Rivulus phenotypes from communities with guppies matured earlier and had higher fecundity, consistent with theories that incorporate indirect effects. Here we examined the genetic basis of these differences by rearing 2nd generation, laboratory-born Rivulus from sites with and without guppies under two food levels that match natural differences in growth. Many locality food interactions were significant, often reversing the relationship between communities. Such interactions imply that there are fitness trade-offs associated with adaptation to high or low resource environments. On high food, Rivulus from localities with guppies matured earlier, produced many small eggs, and exhibited increased reproductive investment; these differences reversed on low food. Our results suggest that indirect effects mold Rivulus evolution and thereby highlight connections between community processes and evolutionary change. KEY WORDS: Adaptive plasticity, killifish, nonlethal effects, predation, resource availability. Early theories of life-history evolution consider how changes in age- or size-specific mortality imposed by extrinsic factors, such as predators, influence the trajectory of evolution (Gadgil and Bossert 1970; Law 1979; Michod 1979; Charlesworth 1980; Taylor and Gabriel 1992; Ernande et al. 2004). This theory predicts that increased rates of predation on adult age/size-classes favor the evolution of earlier maturation and increased reproductive effort because such mortality reduces the costs associated with reproduction (i.e., reduced survival and/or reproductive life span) and earlier reproduction is therefore advantageous (Law 1979). For similar reasons, age- and size-specific theory predicts that predation on very small or young individuals selects for delayed maturation and decreased reproductive effort. However, such theory models the effects of mortality alone, independently of ecological factors that may also shape local adaptation. Indirect effects of predation are commonplace and their ecological consequences are well characterized in a diversity of ecosystems (reviewed: Lima 1998; Schmitz et al. 2000; Werner and Peacor 2003; Schmitz et al. 2004; Miner et al. 2005; Preisser et al. 2005). For example, predators often reduce the density of prey, which increases food availability and thereby the growth rates of survivors (Wootton 1994). These changes in population density and prey food availability may influence the optimal life-history strategies of prey. More importantly, when indirect effects are incorporated into evolutionary theory, they can sometimes dominate the direct effect of predators on prey mortality rates and yield predictions 1583 C 2010 The Author(s). Journal compilation C 2010 The Society for the Study of Evolution. Evolution 64-6:

2 M. R. WALSH AND D. N. REZNICK that are the opposite of what is expected if predators only affect mortality rates (Gadgil and Bossert 1970; Stearns and Koella 1986; Kozlowski and Uchmanski 1987; Kozlowski and Wiegert 1987; Abrams and Rowe 1996; Gardmark and Dieckmann 2006). Yet, the empirical study of life-history evolution has almost exclusively used predictions from mortality-only theory as a guide for empirical research. This is despite evidence of disconnects between such theory and trait divergence in nature (Reznick et al. 1996, 2004), as well as selection experiments demonstrating that food and density clearly shape how organisms evolve (Falconer and Latyszewski 1952; Hillesheim and Stearns 1991; Mueller et al. 1991; Hillesheim and Stearns 1992). Thus, in spite of all that we know about the ecological consequences of the indirect effects of predators in natural ecosystems, we know little of their evolutionary consequences (see Walsh and Reznick 2008). A killifish, Rivulus hartii, is found in streams on the island of Trinidad in communities that differ in predation and competition intensity (Gilliam et al. 1993; Fraser et al. 1995, 1999, 2006; Walsh and Reznick 2008). Two such localities are: (1) Rivulus/guppy sites with Rivulus and guppies (Poecilia reticulata), and (2) Rivulus-only sites where Rivulus are the only fish present. These communities are often located in the same stretches of river within tens of meters of one another, and thus do not differ in physical habitat or environmental variables (Walsh and Reznick 2009). Yet, each community is discrete due to the presence of a barrier waterfall that truncates the distribution of guppies but not Rivulus. More importantly, evidence indicates that guppies impact the ecology of Rivulus and thereby the opportunity for selection on life-history evolution. Rivulus are generally 2 3 more abundant in sites that lack guppies (Gilliam et al. 1993; Walsh 2009). The lower population densities of Rivulus in Rivulus/guppy sites are not accompanied by a matching gain in guppy abundance (Gilliam et al. 1993), so adding guppies to the system results in a net decline of the biomass of fish (Rivulus plus guppies) per unit area. Because there are no differences in adult (>30 mm) survival rates between these localities (apparent survival per 12 day: Rivulus/guppy = 0.79 ± vs. Rivulus-only = 0.75 ± 0.042; Walsh 2009), the decline in Rivulus density in Rivulus/guppy sites appears to be mediated by interactions between small size-classes of Rivulus and guppies. Such interactions may occur because Rivulus overlap in size with guppies only as juveniles; guppies rarely exceed 32 mm (Rodd and Reznick 1997), yet Rivulus mature near that size and can grow to 100 mm. In fact, D. F. Fraser and B. A. Lamphere (pers. comm.) have recently shown that guppies strongly reduce the rates of egg to juvenile survival in Rivulus by preying upon very small/young individuals. Life-history theory that models the consequences of age-/size-specific mortality alone, predicts that such predation will cause the evolution of delayed maturation and decreased reproductive effort (Gadgil and Bossert 1970; Law 1979; Charlesworth 1980). Figure 1. Growth in nature versus laboratory. Comparisons in the magnitude of growth differences between Rivulus/guppy and Rivulus-only localities in nature [Left bar = (mean growth in Rivulus/guppy site)/(mean growth in Rivulus-only site)] and food levels that approximate such differences in the lab [Right bar = (mean growth on high food ration)/(mean growth on low food ration)]. The field data are represented by two 24-day and one 14-day mark recapture studies. The laboratory data are from sequential measurements of fish from the Guanapo and Quare Rivers from the current study. Data are from fish between 20 and 40 mm in total length after 20 and days of growth in the laboratory and field, respectively. There were no significant differences between field and laboratory data (Kruskal Wallis: H 1,7 = 0.29, P = 0.59). The presence of guppies is also correlated with an acceleration of Rivulus growth rates; adult Rivulus grow 2 3 faster in sites with guppies compared with Rivulus-only sites (see Fig. 1). At the same time, Rivulus from Rivulus/guppy communities also produce eggs at a higher rate (Walsh and Reznick 2009). A transplant experiment, in which we moved Rivulus from a Rivulusonly locality into a Rivulus/guppy site immediately downstream, caused a rapid increase in the growth rate of the transplanted fish, so that their rates of growth matched those of the residents (Walsh 2009). Such results argue that the naturally occurring variation in growth between sites with and without guppies is mediated by an environmental factor that allows Rivulus from sites with guppies to grow faster and also invest more in reproduction than Rivulus from Rivulus-only localities. Physiological or behavioral responses to guppies (see Morrison 1999; Peacor 2002; McPeek 2004; Schmitz et al. 2004) appear to be an unlikely explanation for the (2 3 ) increase in Rivulus growth in Rivulus/guppy sites because addition of guppies to field enclosures caused the suppression of the growth of Rivulus tothesamedegreeasan equal-sized Rivulus (Gilliam et al. 1993; but see Palkovacs et al. 2009). Therefore, the most plausible explanation is that the faster rates of growth in Rivulus/guppy locales are due to the much lower population densities of Rivulus in these sites. As these localities 1584 EVOLUTION JUNE 2010

3 INDIRECT EFFECTS AND LIFE-HISTORY EVOLUTION are separated by only tens of meters, and because they do not differ in any aspect of the physical environment, including water temperature (Walsh and Reznick 2009), we interpret these differences in growth as most likely being caused by Rivulus from Rivulus/guppy habitats having higher levels of food availability that are a hypothesized indirect consequence of guppy predation. A second possibility is that Rivulus from the low-density Rivulus/guppy sites expend less energy in aggressive intraspecific encounters. Such large changes in growth and resources can have potent evolutionary consequences as several models predict that increased food availability can favor the evolution of earlier maturation and increased reproductive effort (Gadgil and Bossert 1970; Stearns and Koella 1986; Abrams and Rowe 1996). We recently compared phenotypic life-history differences between Rivulus/guppy and Rivulus-only localities across five river drainages (Walsh and Reznick 2009). Strong differences were observed; Rivulus from Rivulus/guppy sites exhibited a smaller minimum size at reproduction, increased fecundity, smaller eggs, and a greater investment in reproduction. This pattern of divergence does not comply with the predictions of theory that models the consequences of age- or size-specific mortality alone, without the inclusion of indirect effects (Gadgil and Bossert 1970; Law 1979; Michod 1979; Charlesworth 1980; Taylor and Gabriel 1992). However, based upon the known differences in population density and growth rates between Rivulus/guppyand Rivulus-only environments, such changes are consistent with some of the predictions of theories that consider the indirect effects of increased mortality rates mediated through changes in population density (Gadgil and Bossert 1970; Stearns and Koella 1986; Hutchings 1993; Abrams and Rowe 1996; Gardmark and Dieckmann 2006). Here we evaluate the genetic basis of trait variation between Rivulus/guppy and Rivulus-only communities by comparing lifehistory differences after two generations of laboratory rearing. We assume any differences that remain are genetic in origin. We reared these fish on high and low food rations to evaluate the evolutionary consequences of variation in resource availability. We adopted this approach from prior selection experiments in which investigators selected for higher individual growth rates or higher fitness at high or low food rations (Falconer and Latyszewski 1952; Hillesheim and Stearns 1991, 1992). For example, Falconer and Latyszewski (1952) selected for increased growth rates in mice reared on high and low levels of food availability. They succeeded in selecting for increased growth rate in both environments, then compared the growth rates of the two selected lines on high and low food rations. Mice selected for faster growth on high rations grew faster than the mice selected on low rations, but only when both were reared on high food rations. There were no growth differences on low food rations. This experiment, and others (Hillesheim and Stearns 1991, 1992), demonstrates that organisms can become adapted to specific food levels and, more importantly, examination of interactions between food treatments and fitness provides an effective means to quantify responses to resource-based selection. Therefore, here we rear all fish communities on two food levels that approximate naturally occurring differences in growth between Rivulus/guppy and Rivulus-only sites (see Fig. 1). If potential increases in mortality on young/small individuals are important in the evolution of Rivulus, then we expect Rivulus/guppy sites to exhibit delayed maturation and decreased reproductive allotment (Gadgil and Bossert 1970; Law 1979; Michod 1979; Charlesworth 1980). Conversely, if the indirect effects of increased mortality mediated through increases in food availability mold evolution in Rivulus, then: (1) we predict that Rivulus from Rivulus/guppy sites will mature earlier and allocate more resources toward reproduction when compared with Rivulus-only communities (Gadgil and Bossert 1970; Abrams and Rowe 1996), and (2) we also predict that there will be significant interactions between fish community of origin and food availability. Rivulus from Rivulus/guppy communities should have a significant fitness advantage over individuals from Rivulus-only communities when compared on high food rations, but this advantage should be lost or reversed when the two are reared on low food levels (Stearns and Koella 1986). Materials and Methods Rivulus were collected from Rivulus-only and Rivulus/guppy communities from the Guanapo and Quare Rivers in January In the Guanapo River, the Rivulus/guppy community was sampled from an area of the main stem of the river, north of a barrier that truncates the distribution of large piscivorous species. Only Rivulus, guppies, and the catfish Rhamdia sebae are found in this stretch of river. The role of Rhamdia in the community is unclear, although an evaluation of stomach contents did not reveal any evidence for piscivory by Rhamdia on guppies or Rivulus (Gilliam et al. 1993). More importantly, Rivulus life-history phenotypes from this locality do not differ significantly from comparable Rivulus/guppy sites that lack Rhamdia in the Guanapo watershed (M. R. Walsh, unpubl. data). The Rivulus-only site in the Guanapo River is located in a tributary to the main river above a barrier waterfall that stopped the migration of guppies and Rhamdia. In the Quare River, Rivulus/guppy and Rivulus-only communities were both located in a tributary to the main river above a barrier that eliminated all other species from this tributary. Laboratory stocks were established from 20 to 25 wildcaught males and females from each locality. A wild-caught female was placed in a 9-L aquarium and paired with a male from the same locality and approximately 10 such pairings were made for each population. Eggs were collected from each pairing for 20 days and reared in petri dishes containing methylene blue to inhibit fungal growth. Upon hatching, larvae from each pairing were EVOLUTION JUNE

4 M. R. WALSH AND D. N. REZNICK placed in aquaria at a density of 8 10 larvae per tank and were reared on ad libitum food. At an age of approximately 50 days, the sexes can be differentiated, and consequently, the density of each tank was reduced to two fish per tank (usually one male and one female). These fish remained in these conditions until they attained sexual maturity. The second generation of fish was generated by randomly arranging eight pairings between males and females within each community (see Fig. S1). For each pairing, we mated one female with a male from a different lineage to avoid any known sibling matings. Most pairings were unique, as each of the crosses from the parental generation was represented by one female and one male in the subsequent generation. However, there were two instances (Guanapo Rivulus/guppy, Quare Rivulus/guppy)where we used a female in an additional pairing because we only had males from the remaining crosses. In both cases, these females were randomly chosen. Overall, we designed these crosses to maintain as much genetic diversity that existed in the wild-caught parental stocks (assuming all wild-caught individuals are unrelated) as all of the wild-caught lineages were, for the most part, equally represented in the subsequent generations. Eggs were collected from these pairings for days and the eggs were reared in the same manner as described above. Upon hatching, 8 12 larvae from each pairing were reared in 9-L aquaria and fed ad libitum. At an age of 20 days, eight fish per pairing were randomly selected to enter the life-history assay. Each of these fish were individually placed in 9-L aquaria and reared until maturity. At random, four fish from each pairing were chosen to receive a high level of food availability, whereas the other four received a low ration level (see below). Rivulus grows 2 3 faster in Rivulus/guppy communities compared with Rivulus-only localities (Fig. 1). As a result, the food levels used in this experiment approximated these natural differences in growth. Each day all fish were given quantified portions of liver paste and/or Brine shrimp nauplii that resulted in growth trajectories that matched the differences in growth exhibited by Rivulus between sites with and without guppies in nature. The high food and low food levels sustained growth rates typical of Rivulus/guppy and Rivulus-only environments, respectively. Dependent variables included age and size at maturation, fecundity, egg size, egg development rate, larval size at hatching, and reproductive allotment (defined below). Maturity in both males and females was quantified by mating each individual with a mature conspecific of the opposite sex for a period of h. Males exhibit clear phenotypic changes associated with maturation; white stripes form along the bottom and top of the caudal fin. Developing males were mated as soon as tail coloration was observed. Prior work showed that females mature later than males (Walsh and Reznick 2008). Therefore mating assays for females began approximately days after males began to mature. An individual was classified as mature if a developing embryo(s) was found on the spawning substrate. When an individual was classified as mature, it was briefly removed from the aquaria with a dip-net and measured for total length and wet weight for estimates of size at maturation. When a mating trial failed to produce a fertilized egg, the individual was isolated and remated every 3 days thereafter. These assays began prior to maturation because all fish were mated at least once preceding maturation. After maturation, females were mated daily for 20 days to quantify fecundity. A random sample of five of these eggs were preserved in formalin and subsequently weighed for estimates of egg size. The remaining eggs were placed in petri dishes to measure rates of egg development and subsequent larval size at hatching. All petri dishes were monitored several times a day for new hatchlings and upon hatching all larvae were preserved in formalin and subsequently weighed for estimates of offspring size. The average number of days between egg laying and egg hatching was used as our measure of egg development rate. An estimate of the per-day allocation toward reproduction was subsequently calculated as: [(mean per day egg production mean egg size)/size of maturity in female] 100. Water temperature was maintained near 25 C throughout all phases of the experiment. STATISTICAL DESIGN AND ANALYSIS Dependent variables were analyzed using general linear models with river entered as a blocking factor and fish community and food level entered as fixed effects. Because the focal rivers do not differ significantly in physical habitat or environmental variables such as water temperature, dissolved oxygen, ph, and salinity (Walsh and Reznick 2009), it is reasonable to assume no river treatment interactions. A preliminary analysis also showed that none of the river treatment interactions are significant. Consequently, the blocking factor was tested over the error mean square (Newman et al. 1997). We first analyzed male and female life-history traits with a multivariate analysis of variance and subsequently analyzed each dependent variable separately (Scheiner 1993). The presence of normality was evaluated with a Kolmogorov Smirnov test, whereas homogeneity of variances was examined with Levene s test. The data for female age at maturation were log transformed to improve normality. Female size at maturation was included as a covariate in the analysis of fecundity, egg size, egg development rate, and size at hatching, whereas the initial size of each fish upon entering the life-history assay was included as a covariate in the analysis of age and size at maturity for both males and females. MISSING VALUES Some fish died during the course of the experiment due to aggressive conspecifics that were placed in the tank for the purposes of 1586 EVOLUTION JUNE 2010

5 INDIRECT EFFECTS AND LIFE-HISTORY EVOLUTION mating or suicidal jumps from the aquaria. Individuals that died prior to maturation did not produce data that could be included in any analyses. Also, two females died after 1 2 days of egg production (one-quare Rivulus/guppy, one-guanapo Rivulus-only). These individuals were not included in the analyses for fecundity and egg size (and egg development rate and size at hatching). Finally, nine females did not produce eggs that ultimately hatched. Such fish did not yield any data on egg development rate and size at hatching. These missing values were not biased toward one fish community because all were represented (Guanapo Rivulus/guppy [1], Guanapo Rivulus-only [4], Quare Rivulus/guppy [2], Quare Rivulus-only [2]). FIELD VERSUS LABORATORY DATA To evaluate the extent to which our laboratory conditions (i.e., food levels) were a fair representation of nature, we compared trait values between the life-history data collected in the present study and field data collected from preserved wild-caught fish. In a recent study (Walsh and Reznick 2009), wild-caught fish were collected from both Rivulus/guppy and Rivulus-only sites across five river drainages (Arima, Guanapo, Aripo, Quare, Turure) and preserved in formalin. Life-history traits that were quantified from these samples, and that can also be compared with data from the present study, include egg size and the minimum size class in which females are reproductively active. This latter trait is measured from preserved samples by sorting all females into 2-mm size-classes and determining the minimum size-class in which the majority of females contained at least one mature oocyte. Because this criterion generally forms a clear division between fish that did and did not contain mature oocytes, it can be used as a proxy for size at maturation. We specifically compared: (1) Rivulus/guppy populations in the laboratory reared on high food versus the field data from these same populations, and (2) Rivulus-only populations in the laboratory reared on low food versus the field estimates from these populations. We made these comparisons because the high and low food levels are meant to approximate growth in a Rivulus/guppy and Rivulus-only environment, respectively. Because there is only one value for minimum size at maturation per population, we compared the trait values from the lab populations with field data from all the five river drainages to increase statistical power. All comparisons were analyzed with a nonparametric Kruskal Wallis test. If the laboratory conditions are an adequate representation of nature, then we do not expect to find significant differences between the field and lab for either trait. Results FIELD VERSUS LABORATORY DATA We first compared differences in trait values between the results of the present study with data collected from wild-caught preserved fish. There were no significant differences in egg size between 2nd generation born Rivulus/guppy fish reared on high food and wildcaught preserved Rivulus/guppy fish from localities in the Quare and Guanapo Rivers (Means ± 1SE: Laboratory = 2.8 ± 0.15 mg, Field = 3.1 ± 0.1 mg, H 1,69 = 2.47, P = 0.12). Nor were there significant egg size differences between Rivulus-only localities reared on low food and preserved Rivulus-only fish from the Guanapo and Quare Rivers (Means: Laboratory = 3.2 ± 0.09 mg, Field = 3.5 ± 0.09 mg, H 1,43 = 2.96, P = 0.09). The minimum sizes at first reproduction were also similar between the field and laboratory data. No significant differences were found between the field and laboratory in Rivulus/guppy (Means: Laboratory = 35.1 ± 0.9 mm, Field = 33 ± 0.53 mm, H 1,7 = 2.58, P = 0.11) or Rivulus-only localities (Means: Laboratory = ± 2.56 mm, Field = 36 ± 1.48 mm, H 1,7 = 0.27, P = 0.6). Thus, the trait values produced by the laboratory experiments, at least for egg size and minimum size at reproduction, approximate the values observed in nature. MULTIVARIATE ANALYSES OF VARIANCE The results of the multivariate analyses of variance revealed significant effects of river (Female: F 7,63 = 4.05, P = 0.001; Male: F 2,141 = 4.8, P = 0.01), fish community (Female: F 7,63 = 4.02, P = 0.001; Male: F 2,141 = 7.8, P < 0.001), food level (Female: F 7,63 = 42.5, P < 0.001; Male: F 2,141 = 227.4, P < 0.001), as well as significant fish community food-level interactions for male and female traits (Female: F 7,63 = 2.69, P = 0.016; Male: F 2,141 = 11.7, P < 0.001). We therefore subsequently evaluated these trends with univariate analyses (Scheiner 1993). STATISTICAL INTERACTIONS In both river drainages, the expression of life-history traits depended upon food level as several statistical interactions between fish community and food level, including male and female age and size at maturation and reproductive allotment, were significant (Table 1). In addition, the fish community food-level interaction was marginally nonsignificant (0.05 < P < 0.1) for fecundity and egg development rate. On high food levels, Rivulus from Rivulus/guppy environments matured earlier and at a smaller size (Table 2; Fig. 2A D), produced more eggs that were smaller and required longer to develop, and invested more heavily in reproduction (Fig. 2E H). On low food levels, these trends either disappeared (egg development) or were reversed (age and size at maturity, fecundity, reproductive allotment). Thus, Rivulusonly fish actually matured earlier and at a smaller size, produced more eggs, and invested more heavily in reproduction than Rivulus/guppy fish when compared at low food levels (Fig. 2). FISH COMMUNITY EFFECTS There were significant differences between Rivulus/guppy and Rivulus-only communities for egg development rate, and a EVOLUTION JUNE

6 M. R. WALSH AND D. N. REZNICK Table 1. Analyses for life-history traits. All traits were analyzed using general linear models with river entered as a blocking factor and fish community and food level entered as fixed effects. When covariates were nonsignificant (P>0.05), they were removed from the model and the data were reanalyzed without them. df Male Male Female Female No. of Egg Egg- RA Size age at size at age at size at eggs/day size development at maturity maturity maturity maturity rate hatching F F F F F F F F F Covariate: Initial size NS NS Female size NS NS Blocking factor: River NS NS 0.12 NS NS Main effects: Fish community NS 0.25 NS 1.26 NS 0.88 NS 2.23 NS NS 0.11 NS Ration NS 1.31 NS Fish Ration NS 0.12 NS NS Error MS (df) (142) (141) (82) (83) (81) 10 8 (80) (72) (81) 10 8 (71) RA, reproductive allotment; NS, nonsignificant. P>0.05,+0.1>P>0.05, P<0.05, P<0.01, P< marginally nonsignificant divergence in male age at maturation (Table 1; Fig. 2). Eggs from Rivulus/guppy females required, on average, 2 days longer to develop (a 13% increase). FOOD EFFECTS There were highly significant effects of food availability on age and size at maturation in males and females (Tables 1 and 2; Fig. 2). Reductions in food resulted in delayed maturation in males and females by an average of 8 and 14 days, respectively. Also, Rivulus matured at smaller sizes on low food rations; males and females were 22% and 26% lighter, respectively, on low food (Table 2). Lower food resulted in the production of fewer eggs and an increase in egg size in all populations (Tables 1 and 2; Fig. 2). Fecundity declined by 27%, whereas egg weight increased by 12% in response to reductions in resource availability. Size at hatching responded in a consistent manner to less food, as there was a 15% increase in larval weight on low food (Tables 1 and 2). RIVER EFFECTS There was a small but significant difference in male size at maturation between rivers (Tables 1 and 2). Males from the Quare River were 7% smaller in wet weight at maturation. Several reproductive characteristics also differed between rivers (Table 1). Female Rivulus from the Quare River produced significantly fewer eggs that required a longer development time and yielded smaller hatchlings. Females from the Quare River also had lower reproductive allotments than their counterparts from the Guanapo River (Table 2). Differences in fecundity and reproductive allotment were strong as Rivulus from the Quare River exhibited 24% and 25% decreases in fecundity and reproductive allotment, respectively, compared with females from the Guanapo River. Discussion Our results demonstrate a strong association between growth and resource availability in nature, which are most likely indirect effects of guppy predation, and the pattern of life-history evolution in Rivulus. This is because the expression of life-history traits in Rivulus/guppy and Rivulus-only communities depended strongly upon controlled levels of food availability that mimic naturally occurring differences in growth (Fig. 2). Rivulus from Rivulus/guppy environments matured earlier and at a smaller size in both males and females. They also produced many small eggs that required longer time to develop, and allocated more toward reproduction (Fig. 2). However, these changes, which parallels the phenotypic trait divergence revealed in a related study (Walsh and Reznick 2009), was observed only on a food level that approximated growth in a Rivulus/guppy environment. When food was reduced to sustain a growth rate typical of a Rivulus-only locality, there was either no difference in the life histories of Rivulus from these two communities or the opposite pattern of trait variation wasobserved. Themagnitudeofthesefood locality interactions were similar in each independent river system (Table 2) and based upon the known differences in population density and growth rates between Rivulus/guppyand Rivulus-only sites (Gilliam et al. 1993; Walsh 2009), these significant fish community food-level interactions indicate that Rivulus are adapted to alternative environments that differ in resource availability (Stearns and Koella 1986; Schluter 2000, p ). Rivulus from Rivulus/guppy 1588 EVOLUTION JUNE 2010

7 INDIRECT EFFECTS AND LIFE-HISTORY EVOLUTION Table 2. Least squares means (SE) for each community on high and low food. Second generation born Rivulus from two sites with guppies (Guanapo and Quare Rivulus/guppy) and two sites that lack guppies (Guanapo and Quare Rivulus-only) were reared on two levels of food availability that mimic the naturally occurring differences in growth between these sites. The values displayed for egg weight and size at hatching are wet-weight measurements. HF, high food; LF, low food. Guanapo Rivulus/guppy Guanapo Rivulus-only Quare Rivulus/guppy Quare Rivulus-only Trait HF LF HF LF HF LF HF LF Male age at maturity (day) 58.1 (1.51) 74.5 (1.62) 63.5 (2.0) 64.9 (2.1) 58.5 (1.6) 72 (1.5) 60.2 (1.5) 65.1 (1.5) Male size at maturity (g) 0.24 (0.01) 0.22 (0.01) 0.28 (0.01) 0.19 (0.01) 0.22 (0.01) 0.19 (0.01) 0.25 (0.01) 0.18 (0.01) Female age at maturity (day) (3.4) (3.3) (2.56) 115 (2.6) (2.5) 115 (3.0) 102 (2.6) (3.1) Female size at maturity (g) 0.56 (0.04) 0.49 (0.03) 0.67 (0.03) 0.45 (0.03) 0.61 (0.03) 0.48 (0.03) 0.67 (0.03) 0.44 (0.03) No. of eggs per day 1.31 (0.13) 0.86 (0.14) 1.29 (0.12) 1.02 (0.11) 1.07 (0.11) 0.57 (0.12) 0.88 (0.12) 0.86 (0.14) Egg size (mg) 2.76 (0.1) 3.19 (0.11) 2.9 (0.09) 3.4 (0.09) 2.7 (0.09) 3.43 (0.09) 2.86 (0.1) 3.34 (0.1) Egg development rate (day) (1.1) 15.3 (1.2) (0.9) 14.7 (1.0) (0.89) (1.11) (0.99) 17.2 (1.16) Size at hatching (mg) 2.0 (0.1) 2.32 (0.11) 2.0 (0.08) 2.44 (0.09) 2.26 (0.09) 2.61 (0.1) 2.3 (0.09) 2.64 (0.1) Reproductive allotment (%) 0.53 (0.09) 0.52 (0.09) 0.54 (0.07) 0.71 (0.07) 0.44 (0.07) 0.34 (0.08) 0.37 (0.07) 0.55 (0.08) communities appear better able to exploit higher levels of food but are less efficient at converting resources into offspring under low food conditions. In contrast, Rivulus-only fish appear well adapted to environments characterized by high population densities, slow growth, low resources, and high levels of intraspecific competition (see also Walsh and Reznick 2008). This logic is based upon: (1) many reciprocal transplant experiments arguing that the presence of fitness trade-offs between divergent environments is an effective indicator of local adaptation (reviewed: Schluter 2000, p ), and (2) the presence of comparable interactions between growing environment and fitness of selected populations in experiments in which Drosophila were reared under contrasting food or density treatments (Mueller and Ayala 1981; Bierbaum et al. 1989; Hillesheim and Stearns 1991; Mueller et al. 1991; Hillesheim and Stearns 1992). Strong connections exist between life-history theories that consider the ecological factors that accompany the indirect effects of changes in mortality rates and the observed trait differences between Rivulus/guppy and Rivulus-only communities. Several models of life-history evolution predict that increased resource availability or faster rates of growth can favor the evolution of earlier maturation (Gadgil and Bossert 1970; Roff 1992; Stearns 1992) and increased investment in reproduction (Gadgil and Bossert 1970; Stearns 1992; but see Kozlowski and Uchmanski 1987; Kozlowski and Wiegert 1987). Such predictions are clearly consistent with the life-history differences observed on high food levels. Two additional frameworks may also be relevant to the results of this study. First, Gardmark and Dieckmann (2006) examined the influence of size-specific mortality on the evolution of size at maturation. They found that a smaller size at maturation can evolve when mortality is targeted at small individuals, mortality decreases with size, fecundity increases with size, and density regulation is manifested during immature size-classes. Such characteristics likely apply to Rivulus. Second, Abrams and Rowe (1996) evaluated interactions between the direct effects of predator-induced mortality and the indirect effects of this mortality mediated through increases in food availability on the evolution of age at maturation. Although this model yields a variety of predictions that depend strongly upon mortality rate functions, it predicts that evolutionary responses to large indirect effects on prey food supply can oppose the direct effect of predators. These latter models also consider the impacts density and food availability and yield predictions that are generally consistent with the life-history differences observed in the present study. Thus, the patterns of divergence in nature and contrasting responses to reduced food in the laboratory collectively argue that the indirect effects of guppies (Reznick et al. 2001), influence life-history evolution in Rivulus. The contrasting patterns of trait variation observed between Rivulus/guppy and Rivulus-only communities likely have EVOLUTION JUNE

8 M. R. WALSH AND D. N. REZNICK A B C D E F G H Figure 2. Differences in life-history traits between fish communities. (A) Male age at maturation, (B) Female age at maturation, (C) Male size at maturation, (D) Female size at maturation, (E) Fecundity, (F) Egg size, (G) Egg development rate, and (H) Reproductive allotment. In each graph, closed circles, Rivulus/guppy; open circles, Rivulus-only. HF, high food; LF, low food; RG, Rivulus/guppy; RO, Rivulus-only. Error bars =±1SE. Significant interactions between fish community and food level were observed for male and female age and size at maturation and reproductive allotment (P < 0.05) EVOLUTION JUNE 2010

9 INDIRECT EFFECTS AND LIFE-HISTORY EVOLUTION important fitness consequences. On high food, Rivulus from Rivulus/guppy environments matured earlier and had higher rates of egg production than Rivulus-only fish, whereas the opposite trends were observed on low food (Fig. 2). This combination of traits will yield higher rates of population growth, and each fish community thus likely exhibits enhanced fitness when reared on a ration that approximates its rate of growth in nature. However, age at maturation plus fecundity does not constitute a complete measure of fitness. The increased duration of egg development observed in Rivulus/guppy localities on a high food ration may reduce some of the fitness benefits associated with earlier maturation and higher fecundity. Furthermore, variation in egg size may contribute to fitness. Rivulus/guppy females produced proportionally smaller eggs on high food rations. This is a potential fitness cost because smaller eggs can reduce offspring quality and survival in fish (Trippel 1995). Yet, several studies have demonstrated that there is little or no fitness loss associated with a smaller offspring size if food is abundant, as is the case in Rivulus/guppy habitats (Hutchings 1991; Einum and Fleming 1999; Bashey 2008). Interestingly, all of the Rivulus localities also produced larger eggs on reduced food (Fig. 2). This parallels changes in offspring size in guppies when food is reduced (Reznick and Yang 1993; Bashey 2006), and because larger egg and offspring sizes typically have higher fitness in more competitive environments (Berven and Chadra 1988; Tessier and Consolatti 1989; Hutchings 1991; Winn and Miller 1995; Einum and Fleming 1999; Bashey 2008), such plastic responses to low food are likely adaptive. Some important similarities and differences exist between a previous experiment in this system and the results of the present study. Walsh and Reznick (2008) compared life-history differences between Rivulus-only sites and localities in which Rivulus co-occur with numerous large predators (hereafter high-predation sites). The direct and indirect effects of predators are evident in high-predation sites; Rivulus suffer higher adult mortality rates in high-predation sites and the presence of predators is correlated with lower Rivulus population density and faster individual growth rates (Gilliam et al. 1993; Fraser et al. 1999; D. F. Fraser and J. F. Gilliam unpubl. data). Unlike the current study, this earlier experiment revealed strong direct effects of predators on the evolution of life histories. Rivulus evolved in a manner consistent with the predictions of theory that modeled the effects of increased adult mortality rates independently of any indirect effects; Rivulus from high-predation sites matured earlier and allocated more toward reproduction (Gadgil and Bossert 1970; Law 1979; Michod 1979; Charlesworth 1980). Similar to the current results, Walsh and Reznick (2008) also provided evidence for an important role for resource availability on evolutionary change; many fish community food-level interactions were significant and the differences between fish communities were reduced or disappeared under low food conditions. Yet, this role for indirect effects was only apparent in the locality-by-food interactions. Thus, the comparisons between Rivulus/guppy and Rivulus-only communities were more heavily dominated by indirect effects. This is because the patterns of divergence in life histories and many significant fish community food-level interactions both indicate a role for the indirect effects of predators on evolutionary change. Most importantly, the results of these two studies together argue that indirect effects of predators, manifested as reduced Rivulus population density and increased growth rates, are a pervasive and potent agent of selection for Rivulus. GROWTH RATES, RESOURCE AVAILABILITY, AND LIFE-HISTORY EVOLUTION Even though several life-history models predict that higher resource availability can favor the evolution of earlier maturation and increased reproductive effort (Gadgil and Bossert 1970; Stearns and Koella 1986; Stearns 1992; Hutchings 1993; Abrams and Rowe 1996), such predictions are not universal. Work by Kozlowski and colleagues predict that increased food favors the evolution of an opposite suite of life-history traits (Kozlowski and Uchmanski 1987; Kozlowski and Wiegert 1987; see also Abrams and Rowe 1996). In the papers by Kozlowski and colleagues, the key trade-off is between growth and reproduction, and a decrease in growth makes it less rewarding to grow rather than mature and reproduce. These contradictory predictions in response to faster rates of growth have been known for some time and are due to the use of different measures of fitness (r vs. R o ) and different assumed relationships between growth, fecundity, and survival (see Roff 1992; Stearns 1992; Reznick et al. 2002). The results of the present study indicate that higher resource availability facilitates the evolution of earlier maturation and higher reproductive effort. Yet, tests of theories that consider these indirect effects of changes in mortality rates are rare and more are clearly warranted. Future theoretical and empirical work needs to more fully consider the relative importance of differential mortality and related changes in population density and resource availability on subsequent evolutionary change. Past theoretical frameworks suggest that the impacts of mortality and resources can interact in a complicated fashion to produce a variety of evolutionary outcomes (see Charlesworth 1980; Abrams and Rowe 1996; Gardmark and Dieckmann 2006). For example, in Kozlwoski and Uchmanski s (1987) model, a uniform increase in mortality, which mortalityonly theories predict should not cause any life-history evolution (Law 1979; Charlesworth 1980), instead favors the evolution of earlier maturation and higher reproductive effort because the impacts of per capita resource availability are also represented (see also Charlesworth 1980; Gardmark and Dieckmann 2006). Given that ecological interactions are often size specific (Werner and Gilliam 1984), further theoretical and empirical consideration EVOLUTION JUNE

10 M. R. WALSH AND D. N. REZNICK of the evolutionary consequences of variation in resources and mortality across multiple life-history stages may be especially fruitful. CONCLUSIONS Our results provide evidence that Rivulus are adapted to differences in resource availability that are a hypothesized indirect effect of guppy predation on small size-classes of Rivulus. In agreement with predictions of several frameworks of life-history evolution (Gadgil and Bossert 1970; Stearns and Koella 1986; Abrams and Rowe 1996), our results indicate that faster rates of growth, which likely reflect increased availability of resources to sustain growth and reproduction, facilitate the evolution of earlier maturation at a smaller size, increased fecundity, smaller eggs, and higher reproductive allotment. Because the predictions of models that only consider the impacts of age-/size-specific mortality are the basis for much research, these results therefore call for a greater consideration of life-history theories that incorporate additional, ubiquitous features of the biotic environment when evaluating evolutionary change in natural populations. ACKNOWLEDGMENTS We thank M. Schrader, C. Oufiero, and D. Fraser for assistance with field work. D. Roff and L. Nunney provided helpful comments. This work was supported by an Explorers Club Exploration Grant, a National Science Foundation Doctoral Dissertation Improvement Grant DEB , and National Science Foundation Grants DEB and EF to DNR. LITERATURE CITED Abrams, P., and L. Rowe The effects of predation on the age and size of maturity of prey. Evolution 50: Bashey, F Cross-generational environmental effects and the evolution of offspring size in the Trinidadian guppy Poecilia reticulata. Evolution 60: Competition as a selective mechanism for larger offspring size in guppies. Oikos 117: Berven, K. A., and B. G. Chadra The relationship among egg size, density and food level on larval development in the wood frog (Rana sylvatica). Oecologia 1988: Bierbaum, T. J., L. D. Mueller, and F. J. Ayala Density-dependent evolution of life history characteristics in Drosophila melanogaster. Evolution 43: Charlesworth, B Evolution in age-structured populations. Cambridge Univ. Press, Cambridge, U.K. Einum, S., and I. A. Fleming Maternal effects of egg size in brown trout in (Salmo trutta): norms of reaction to environmental quality. Proc. R. Soc. Lond. B 266: Ernande, B., U. Dieckmann, and M. Heino Adaptive changes in harvested populations: plasticity and evolution of age and size at maturation. Proc. R. Soc. Lond. B 271: Falconer, D. S., and M. Latyszewski The environment in relation to selection for size in mice. J. Genet. 51: Fraser, D. F., J. F. Gilliam, and T. Yip-Hoi Predation as an agent of population fragmentation in a tropical watershed. Ecology 76: Fraser, D. F., J. F. Gilliam, M. P. MacGowan, C. M. Arcaro, and P. H. Guillozet Habitat quality in a hostile river corridor. Ecology 80: Fraser, D. F., J. F. Gilliam, B. W. Albanese, and S. B. Snider Effects of temporal patterning of predation threat on movement of a stream fish: evaluating an intermediate threat hypothesis. Environ. Biol. Fish. 76: Gadgil, M., and P. W. Bossert Life history consequences of natural selection. Am. Nat. 104:1 24. Gardmark, A., and U. Dieckmann Disparate maturation adaptations to size dependent mortality. Proc. R. Soc. Lond. B 273: Gilliam, J. F., D. F. Fraser, and M. Alkins-Koo Structure of a tropical stream fish community: a role for biotic interactions. Ecology 74: Hillesheim, E., and S. C. Stearns The responses of Drosophila melanogaster to artificial selection on body weight and its phenotypic plasticity in two larval food environments. Evolution 45: Correlated responses in life-history traits to artificial selection for body weight in Drosophila melanogaster. Evolution 46: Hutchings, J. A Fitness consequences of variation in egg size and food abundance in brook trout Salvenlinus fontinalis. Evolution 45: Adaptive life histories effected by age-specific survival and growth rate. Ecology 74: Kozlowski, J., and J. Uchmanski Optimal individual growth and reproduction in perennial species with indeterminate growth. Evol. Ecol. 1: Kozlowski, J., and R. G. Wiegert Optimal age and size at maturity in annuals and perennials with determinate growth. Evol. Ecol. 1: Law, R Optimal life histories under age-specific predation. Am. Nat. 114: Lima, S. L Stress and decision making under the risk of predation: recent developments from behavioral, reproductive, and ecological perspectives. Adv. Study Behav. 27: McPeek, M. A The growth rate/predation risk trade-off: so what is the mechanism? Am. Nat. 163: E88 E111. Michod, R. E Evolution of life histories in response to age-specific mortality factors. Am. Nat. 113: Miner, B. G., S. E. Sultan, S. G. Morgan, D. K. Padilla, and R. A. Relyea Ecological consequences of phenotypic plasticity. Trends Ecol. Evol. 20: Morrison, L. W Indirect effects of phorid fly parasitoids on the mechanisms of interspecific competition among ants. Oecologia 121: Mueller, L. D., and F. J. Ayala Trade-Off between r-selection and K-Selection in Drosophila Populations. Proc. Natl. Acad. Sci. USA 78: Mueller, L. D., P. Guo, and F. J. Ayala Density-dependent natural selection and trade-offs in life history traits. Science 253: Newman, J. A., J. Bergelson, and A. Grafen Blocking factors and hypothesis tests in ecology: is your statistics text wrong? Ecology 78: Palkovacs, E. P., M. C. Marshall, B. A. Lamphere, B. R. Lynch, D. J. Weese, D. F. Fraser, D. N. Reznick, C. M. Pringle, and M. T. Kinnison Experimental evaluation of evolution and coevolution as agents of ecosystem change in Trinidadian streams. Phil. Trans. R. Soc. B 364: Peacor, S. D Positive effects of predators on prey growth rate through induced modifications of prey behavior. Ecol. Lett. 5: EVOLUTION JUNE 2010