Recent evolutionary history predicts population but not ecosystem‐level patterns

Abstract In the face of rapid anthropogenic environmental change, it is increasingly important to understand how ecological and evolutionary interactions affect the persistence of natural populations. Augmented gene flow has emerged as a potentially effective management strategy to counteract negative consequences of genetic drift and inbreeding depression in small and isolated populations. However, questions remain about the long‐term impacts of augmented gene flow and whether changes in individual and population fitness are reflected in ecosystem structure, potentiating eco‐evolutionary feedbacks. In this study, we used Trinidadian guppies (Poecilia reticulata) in experimental outdoor mesocosms to assess how populations with different recent evolutionary histories responded to a scenario of severe population size reduction followed by expansion in a high‐quality environment. We also investigated how variation in evolutionary history of the focal species affected ecosystem dynamics. We found that evolutionary history (i.e., gene flow vs. no gene flow) consistently predicted variation in individual growth. In addition, gene flow led to faster population growth in populations from one of the two drainages, but did not have measurable impacts on the ecosystem variables we measured: zooplankton density, algal growth, and decomposition rates. Our results suggest that benefits of gene flow may be long‐term and environment‐dependent. Although small in replication and duration, our study highlights the importance of eco‐evolutionary interactions in determining population persistence and sets the stage for future work in this area.

the other hand, some populations and taxa persist in spite of small population sizes. For instance, the Apennine brown bear population in Italy has survived for many generations despite being small, inbred, and having startlingly low levels of genetic variation (Benazzo et al., 2017). Given contradicting examples like these, more research is needed to determine the eco-evolutionary predictors of population dynamics and extinction risk.
Gene flow is a nonselective evolutionary force that can mitigate small population problems by reversing the effects of inbreeding depression and genetic drift. By increasing genetic variation, gene flow can also facilitate a faster response to natural selection (Swindell & Bouzat, 2006). Gene flow manipulations are an emerging management strategy in which a small number of immigrants are translocated into a (typically) small, inbred population in an effort to restore connectivity among recently isolated populations and ultimately increase population fitness (Aitken & Whitlock, 2013;Tallmon, Luikart, & Waples, 2004;Whiteley, Fitzpatrick, Funk, & Tallmon, 2015). Genetic rescue, an increase in population growth caused by gene flow, has been shown to produce large and consistent benefits to population fitness (Frankham, 2015;Whiteley et al., 2015). Several iconic conservation success stories resulting from genetic rescue have been documented in species such as Florida panthers (Johnson, Chappell, et al., 2010), greater prairie chickens (Westemeier et al., 1998), and bighorn sheep (Hogg, Forbes, Steele, & Luikart, 2006). Although examples like these suggest that assisted gene flow is an effective management tool, it has not been widely used due to concerns of outbreeding depression, genetic swamping, and disease introduction (Edmands, 2007;Frankham, 2015).
While increasing evidence suggests that gene flow into genetically depauperate populations in the wild can boost genetic variation and thereby increase individual and population fitness (Fitzpatrick et al., 2016;Robinson et al., 2017;Weeks et al., 2017), little is known about how recent episodes of gene flow versus genetic drift affect demography under environmental change. If the environment is poor, gene flow, even if beneficial, may not manifest in higher population sizes. For example, immigration of a single outbred wolf into the isolated Isle Royale population led to an initial increase in heterozygosity without subsequent increases in population size, an outcome thought to be due to poor environmental conditions (Adams, Vucetich, Hedrick, Peterson, & Vucetich, 2011). Conversely, it has not been empirically tested whether populations that have experienced recent gene flow are better able to take advantage of good quality environments than those experiencing high genetic drift.
Understanding how recent evolutionary history determines persistence and adaptive potential under changing environments is urgently important given current levels of habitat destruction, climate change, and anthropogenic disturbance facing natural populations today (Chevin, Lande, & Mace, 2010). It is essential to understand how manipulating evolutionary forces affects population demography and persistence and specifically how human-assisted gene flow may impact the surrounding environment.
Genetic rescue is inherently eco-evolutionary because it involves manipulating an evolutionary force, gene flow, to affect the dynamics of populations. However, it is unknown the extent to which evolutionary manipulations scale up to affect ecosystem dynamics. There is reason to believe that there would be an effect on ecosystem dynamics since genetic rescue often causes dramatic increases in abundance of the "rescued" population (Fitzpatrick et al., 2016). For example, trophic cascades could be affected if increased abundance of a top predator results in higher depletion rates of its primary prey species. Additionally, populations adapted to dissimilar environments may use resources differently and cause significant ecosystem structure alterations in short time periods (Bassar et al., 2010). If immigration from an adaptively divergent population occurs, the ecological role of the recipient population may be altered.
Finally, levels of intraspecific genetic variation can impact the ecological role of populations (Hughes, Inouye, Johnson, Underwood, & Vellend, 2008). For example, different levels of population genetic variation in tall goldenrod, Solidago altissima, were shown to predict arthropod diversity and community structure (Crutsinger et al., 2006). Linking changes in recent evolutionary history to changes in population genetic variation to changes in the ecosystem has not been done experimentally. Understanding such links has important consequences for biodiversity conservation under rapid environmental change.
Trinidadian guppies are a model system frequently used in eco-evolutionary research due to short generation times (3-4 per year; Magurran, 2005;Reznick & Endler, 1982), rapid adaptation in new environments (Magurran, 2005), and predictable phenotypic traits like offspring size, coloration, and life histories based on the environment (Ghalambor, McKay, Carroll, & Reznick, 2007). Streams that drain the Northern Range Mountains in Trinidad have varying predator assemblages, imposing selection pressures that have resulted in the repeated evolution of divergent guppy life histories.
Guppies from high predation (HP) environments, found in lower-elevation stream reaches, have faster maturation rates and more, smaller offspring than guppies found in low predation (LP) environments at higher elevations (Reznick & Endler, 1982). Headwater populations of guppies also typically have lower genetic variation and smaller population sizes due to founder effects caused by geographical barriers such as waterfalls (Barson, Cable, & Oosterhout, 2009).
Previous work shows that inbred guppies often have lower fitness than outbred guppies due to higher susceptibility to parasite infection (Smallbone, Oosterhout, & Cable, 2016), less coloration, fewer mating behaviors (Van Oosterhout et al., 2003), and fewer offspring produced (Johnson, Onorato, et al., 2010). Studies documenting fitness effects of gene flow manipulation in wild populations showed that immigration dramatically increased genetic diversity and population sizes, without the loss of locally adapted traits (Fitzpatrick et al., 2016;Fitzpatrick, Gerberich, Kronenberger, Angeloni, & Funk, 2015).
This study builds on a previous experiment that tested fitness effects of gene flow in replicated populations housed in glass aquaria in a laboratory environment (Kronenberger et al., 2018). individual and population growth rates, we expected resources to be depleted more quickly in those environments. We also predicted that populations with higher genetic variation due to recent gene flow might utilize a more diverse set of resources than populations with lower genetic variation. This is the first study we know of to test ecosystem-level consequences of gene flow manipulation. Results of this experiment enhance understanding of eco-evolutionary interactions by shedding light on how differences in nonadaptive processes, such as gene flow and genetic drift, can influence demography and ecosystem dynamics in changing environments.

| Mesocosm setup
Our experiment took place at Michigan State University's W. K.
Kellogg Biological Station Experimental Pond Facility from June 2017 to October 2017. We acid-washed and filled 24 cattle tanks (1,000 L) with well water and added sand substrate. To create high resource pond communities, we collected a total of 12.5 L of zooplankton-rich water using a plankton tow net from four nearby ponds and inoculated each tank with 500 ml of this mixture to provide a prey resource for the guppies. We also added 15 g of a common aquatic plant species, hornwort (Ceratophyllum demersum), to each tank to provide cover for young fish from cannibalistic adults.
All tanks received a nutrient mixture to increase primary productivity; 10.4 g nitrogen (NaNO 3 ) and 0.33 g phosphorus (NaH 2 PO 4 ) was dissolved in 50 ml of water and added to each tank (Rudolf & Rasmussen, 2013). We added six (4.5 cm 2 ) ceramic tiles to each tank to measure the accumulation of benthic algae. To compare decomposition rates between treatments, 10 × 15 cm mesh bags were filled with 7 g of air-dried mixed leaf litter collected from the W.K. Kellogg Bird Sanctuary and submerged in each tank.

| Experimental design
The guppies used in our experiment initially were collected from LP tributaries of the Quare and Marianne rivers in Trinidad and Tobago, as described in Kronenberger et al. (2018). In that experiment, populations were established in 38 L glass aquaria with 16 guppies of a 1:1 sex ratio. Following a six-month period without Previous studies show low mortality and high recapture success using these methods (Reznick, Butler, Rodd, & Ross, 1996).

| Data collection
We compiled the following data from each tank on a biweekly schedule from late July 2017 to early October 2017 (10 weeks total). Using a depth-integrated tube sampler, we collected 4 L of water from each tank and filtered it through 75 µm mesh to determine zooplankton density. The samples were immediately stored in ethanol and later visualized under a dissecting microscope to calculate copepod and cladoceran density (individuals/L). We extracted one tile of the six total tiles from each tank every other week and weighed the oven-dried algae present to calculate benthic algae accumulation. To quantify primary productivity, pelagic algae were collected by filtering 50-100 ml of water from each tank and storing algal samples on folded papers in foil at −20℃. These samples were later analyzed with a UV/VIS spectrophotometer to quantify the amount of chlorophyll-a (μg/L) in each. We calculated decomposition rates by comparing the dry leaf litter mass (g) before and after the addition of mesh bags containing leaf litter. To ensure that water conditions were held constant between tanks throughout the course of our experiment, every other week we measured dissolved oxygen, temperature, and pH three times over a 24-hr time period in all tanks. All guppies ≥ 14 mm were censused biweekly. New recruits (≥14 mm) were given unique elastomer marks, and all individuals were processed as described above to census populations and determine individual growth.

| Individual growth
We were interested in whether differences in recent evolutionary history led to differences in the rate of growth among individuals in a high resource environment. We only compared growth (measured as standard length) among founding females to avoid confounding juvenile with adult growth and female with male growth, the latter because female guppies have indeterminate growth throughout their lifetimes whereas male growth ceases at maturity. We estimated standard lengths at each sampling occasion using a generalized linear mixed model implemented with "proc glimmix" in SAS version 9.4 (SAS Institute). Residuals were not normally distributed, so we assumed a lognormal distribution as it was superior following a comparison of the model under all available distributions using Akaike Information Criterion corrected for small sample size (AIC C ; Burnham & Anderson, 2002). Fixed effects included the three-way interaction between population (Quare and Marianne), treatment (no gene flow, HP gene flow, and LP gene flow), and sampling occasion (week), in addition to all twoway interactions and main effects. To account for nonindependence resulting from measuring the same individuals over time, we included sampling occasion as repeated random effect with a heterogeneous autoregressive covariance structure, in which the correlation between sampling occasions decreases exponentially as the interval between them increases. This covariance structure fits our expectation of how sampling occasions are related and was best supported according to AIC C comparisons.

| Population growth
To assess whether there were changes in population fitness arising from recent evolutionary history, we used census population sizes to estimate population sizes over time and finite population growth rates of adult fish. Population sizes were estimated using F I G U R E 1 Guppies in the present study were sampled from subset of gene flow treatments from Kronenberger et al. (2018) (a) where recipient populations received either no augmentation (None), gene flow from a high predation (HP) guppy population, or gene flow from a low predation (LP) guppy population. Populations were censused bimonthly for a total of 32 months with a 6-month pregene flow period, 12-month gene flow augmentation period, and 14-month postgene flow period (b). Guppies at month 32 were used to seed mesocosm populations in the present study Month a generalized linear mixed model with a Poisson distribution due to non-normally distributed residuals, the typical distribution for nonoverdispersed count data. As in the individual growth model, fixed effects included the three-way interaction between population, treatment, and sampling occasion in addition to all two-way interactions and main effects. We modeled sampling occasion as a repeated random effect with an autoregressive covariance structure as above, although in this case, the covariance structure was not heterogeneous due to convergence issues. Finite population growth rates were calculated as the total number of individuals in each tank at the end of the experiment divided by 4 (the number of founding individuals) and estimated using multiple linear regression.
Fixed effects included the population and treatment main effects and their interaction. Both models (population size and population growth rate) were run using "proc glimmix" in SAS version 9.4.

| Ecosystem effects
We used linear mixed models to test our predictions about how recent evolutionary history impacts environmental variables (decomposition rates, algal growth, and zooplankton density). For all ecosystem measurements, data from the last sampling period were analyzed and compared among treatments to elucidate the greatest differences between treatments. Models included treatment type (no gene flow, HP gene flow, LP gene flow, and control) as a fixed effect and tank location as a random effect. Each treatment was tested against the null model that included only the random effect.
We also tested whether populations (Quare and Marianne) determined ecosystem structure and used linear models for each environmental variable. We used simple linear regressions to correlate adult fish abundance and total fish biomass with each environmental variable. All models were run using the "lme4" package in R version 3.5.1 (Bates & Maechler, 2009).

| Population growth
Population sizes in individual tanks were highly variable but generally increased over time (Table 1;

| Ecosystem effects
Linear mixed models and linear models used to test the effects of recent evolutionary history on ecosystem variables revealed no effect of gene flow treatment (HP, LP, none, and control) for either population on decomposition rates, benthic algae accumulation, chlorophyll-α level in pelagic algae, or copepod/cladoceran zooplankton density. We did find a significant and positive correlation between total fish biomass and copepod density (R 2 = 2.084, F18 = 4.739, p = .04305). However, this relationship was heavily influenced by a single outlier and then became insignificant when the outlier was removed (R 2 = 0.04126, F17 = 0.7316, p = .4043).

| D ISCUSS I ON
Our study tested how populations with different recent evolutionary histories responded to a scenario of extreme population size reduction followed by a period of high environment quality. We uncovered fitness benefits in the form of individual growth from previous gene flow that only became apparent in high resource environments. We also investigated how evolutionary histories affected ecosystem dynamics. Our results suggest that recent evolutionary histories affect individual and sometimes population growth rates, but do not scale up to induce measurable effects on the ecosystem.
Additional research is needed to fully parse out the relationship between evolutionary history and ecosystem dynamics.

| Effects of evolutionary history on individual growth
When guppies were exposed to novel high resource environments, we found faster individual growth in females from populations that  (Reznick, Bryant, Reznick, & Bryant, 2007). In our study, if faster growth in individuals was caused by immigrants' traits, we would expect the HP gene flow treatments to grow faster than the LP treatments due to their fast life history (i.e., faster growth, more offspring, and earlier maturation times).
However, we found individual growth were higher in all gene flow treatments, regardless of the environment from which the source population originated, suggesting that the observed growth advantage was not solely related to specific traits conferred by immigrants with a fast life history.
Another possible explanation is that individuals from populations with recent gene flow was able to grow faster on the same level of resources due to acquiring and metabolizing resources more effectively. A recent study found that brook trout with recent gene flow exhibited larger body sizes than brook trout that had not received gene flow while living in the same environments. This was possibly caused by overdominance due to heterosis, or increased fitness in hybrid offspring compared to parents, in somatic juvenile growth (Robinson et al., 2017). However, overdominance is unlikely to be the cause of treatment differences in our experiment because multiple generations of admixture in gene flow treatments had occurred prior to our study (Kronenberger et al., 2018). In Kronenberger et al. (2018), the authors genotyped every adult and reconstructed pedigrees to confirm the presence of hybridization. They found that Quare populations produced more hybrids than Marianne populations, but at least some admixture was observed in all gene flow treatments used as source populations for this study. Since treatments were left untampered with for over a year, it is unlikely that heterosis would have occurred in our experiment. TA B L E 1 The results of type 3 tests of fixed effects from our individual growth (standard length) and population growth (population size and finite population growth rate) models. Standard length and population size were modeled as repeated measures, generalized linear mixed models, and population growth rate was modeled as a multiple linear regression (see text for more details). For each model, we report numerator and denominator degrees of freedom and F and p values F I G U R E 2 Estimated marginal means and standard errors of standard length, measured from founding females at the start of the experiment and on each subsequent sampling occasion. Colors correspond to evolutionary history treatments: red, no gene flow (none); blue, gene flow from a high predation source (HP); purple, gene flow from a low predation source (LP). Significant differences (p < .05) are indicated with asterisks; in all cases, HP and LP treatments were similar to one another but significantly different than the none treatment Another possibility is that higher overall genetic variation facilitated a growth benefit to individuals from the gene flow treatments. Genetic variation has been shown in many previous studies to have a positive effect on fitness (Reed, Frankham, Reed, & Frankham, 2016). For example, genetic diversity was positively correlated with mean lifetime expectancy in Chalkhill blue butterflies (Vandewoestijne, Schtickzelle, & Baguette, 2008), honey bee colonies with higher genetic diversity were able to out-forage and gain weight faster than colonies with lower genetic diversity (Mattila & Seeley, 2007), and in blue mussels it was shown that individuals with higher heterozygosity also had higher average growth rates (Koehn & Gaffney, 1984). In the latter study, the authors proposed that faster growth rates in more heterozygous mussels were possible due to a greater availability of energy from the decreased costs of metabolism, specifically oxygen consumption rates. In our study, it is possible that fish with increased genetic variation due to gene flow were able to more efficiently utilize the high resources in outdoor mesocosm environments.
Regardless of the mechanism, the finding of significantly higher

| Effects of evolutionary history on population growth rates
Although population growth rates did not significantly differ among gene flow treatments, there was a trend for higher population growth rates in Quare gene flow treatments, but not Marianne. Similarly, we found that gene flow treatments achieved a significantly higher mean population size than no gene flow treatments in the Quare, but not Marianne. This discrepancy between the two drainages matches results from Kronenberger et al. (2018), namely, that gene flow was only consistently beneficial to recipient populations from Quare. As discussed in Kronenberger

| Effect of evolutionary history on ecosystem variables
In this study, the ecosystem variables we measured (zooplankton density, pelagic algae growth, benthic algae growth, and decomposition rates) did not significantly differ among any treatments (see Supporting Information). Fish density also did not appear to influence ecosystem variables although total fish biomass did have a significant but counterintuitive effect on copepod density where greater fish biomass was positively associated with copepod density. This could be due to an unmeasured indirect effect of guppies feeding on copepod predators such as macroinvertebrates. However, given that this result was dependent on a single outlier, we do not have confidence that it represents a meaningful relationship.
A possible explanation for the lack of differentiation among treatments is that algal and zooplankton communities were able to regenerate at rates fast enough to offset depletion by guppies. A previous study found that the population abundances of many zooplankton species were unaffected by fish predation and that only larger zooplankton species were susceptible to predation (Jack & Thorp, 2002). In our study, we quantified zooplankton density by order (i.e., cladoceran or copepod) and it is possible that a finer resolution of species composition data could have revealed differences among treatments. Also, previous studies have found that algal depletion by zooplankton predation can be offset by the nutrient release due to predation as algal communities use the nutrients to quickly regenerate (Sterner, 1986). However, these explanations do not explain the lack of differentiation of decomposition rates among treatments.
Another factor that could have contributed to these results is that macroinvertebrate species composition and abundance were not measured in tanks. This unmeasured component of the trophic system in our cattle tanks may have played an important role. For example, macroinvertebrates could have been an alternate food source for guppies or affected ecosystem variables and explained individual growth differences among treatments without changes in the resource levels we measured. The lack of differences in ecosystem measurements could also be explained by the relatively short time frame of our experiment or the large size of the cattle tanks with low numbers of fish, making the environments too large for such low fish densities to accrue any measurable changes on ecosystem factors. Our experiment consisted of relatively low replication with only four mesocosms per treatment and four fish per starting population. Given that ecosystem measurements are often highly stochastic, low replication also likely contributed to a lack of power to detect measurable ecosystem differences between treatments.
Interestingly, even though the ecosystem variables we measured were not different among treatments, individual guppy growth and population sizes did differ. This suggests a possible intrinsic mechanism that facilitated individuals from populations that previously received gene flow to grow faster on the same quality and quantity of resources. As discussed briefly above, this could be due to an inverse relationship between heterozygosity and the energetic costs of metabolism, possibly mediated by oxygen consumption. A relationship between heterozygosity and population growth rate has received mixed empirical support in other systems where a positive correlation has been found in American oysters (Koehn & Gaffney, 1984) and some populations of tiger salamanders (Pierce & Mitton, 1982), but not in European plaice (McAndrew, Ward, & Beardmore, 1986) or pink salmon (Beacham & Withler, 1985).

| Environment-dependent benefits of gene flow
Previous gene flow augmentation experiments have focused on fitness benefits that arise in deliberately challenging environments (Hufbauer et al., 2015) or potentially stressful laboratory environments (Kronenberger et al., 2018). While these results make important contributions to our growing understanding of how gene flow can rescue populations, it is also critical to understand how augmented populations will perform in "good" environments or periods of low environmental stress and change. In our study, we

| Caveats, future directions, and broader implications
At first glance, it may seem biologically unrealistic to test eco-evolutionary effects of a tropical fish in temperate ponds. However, guppies are a highly successful invasive species throughout the world-they encounter and thrive in a wide range of novel environments. Due to human introductions (accidental and intentional), guppies have established in sugarcane fields in Australia (Lindholm et al., 2005), artificially heated streams near power plants in Germany (Jourdan et al., 2014), and streams fed by hot springs in Japan (Shoji, Yokoyama, & Kawata, 2007). Although our study was carried out in lentic mesocosms in Michigan, as opposed to their native lotic Trinidadian streams, we think our results can be considered in the contexts of both invasive populations and rapid global change. For example, the lack of observable community effects in our study (in novel and high resource environments), compared with previous findings of effects of population history at the community level in native environments beckons for future work exploring eco-evolutionary dynamics in native versus novel environments. Testing this question using fish densities that are more comparable to natural densities than the low densities in our experiment would be advisable.
Our study sets the stage for additional follow-up research that will elucidate many remaining uncertainties. For example, future experiments that start with larger initial population sizes might better represent the genetic and phenotypic variation within each evolutionary history treatment and reduce the potentially confounding role of founder's effect. We imposed a strong founder's effect by seeding tanks with only four individuals (two males and two females) and this small sample size may not have represented the phenotypic and genetic variation present in the Kronenberger et al. (2018) treatments. To better characterize, the effect of recent evolutionary history on the environment, starting with larger fish densities or smaller mesocosms, may also create higher resource depletion rates and facilitate testing for significant differences in ecosystem variables.
Future studies would also benefit by accounting for the composition and density of macroinvertebrates to understand how they affect other ecosystem variables and whether they serve as an alternate food source. Lastly, longer-term monitoring of mesocosm populations and environments than what was possible here will provide a better opportunity to reveal treatment differences. Ecosystem-level differences may become more detectable as resources are depleted over a longer study period.
As habitat fragmentation increases and native populations become smaller and more isolated, gene flow augmentation is a potentially powerful option to reverse the effects of inbreeding depression and genetic drift (Frankham et al., 2017). Previous work has shown that genetic rescue generally provides an immediate fitness increase to small populations (Whiteley et al., 2015;Frankham, 2016). In this study, we complement studies investigating the initial fitness effects of gene flow augmentation by considering longer-term impacts of recent evolutionary history on individuals and populations in a benign environment. In addition, we investigate the impact of evolutionary history on ecosystem structure. The scenario we studied, in which populations become exposed to new conditions (first a severe bottleneck and then a novel, high resource environment), is highly plausible for contemporary populations-including ones that are not of conservation concern. For example, invasive species often experience a strong bottleneck when few individuals first colonize a new habitat and then are exposed to favorable conditions with many possible niches to exploit and few natural predators (Tsutsui, Suarez, Holway, & Case, 2002). Understanding recent evolutionary history, including nonselective forces like gene flow and genetic drift, may help to predict invasive potential. Additional research is needed in this area to fully understand the eco-evolutionary and conservation implications of how augmented gene flow versus sustained genetic drift and inbreeding affects populations and ecosystems.

DATA AVA I L A B I L I T Y S TAT E M E N T
The datasets generated and analyzed in this study are available at https ://doi.org/10.5061/dryad.fqz61 2jp8.