Article

Rapid morphological change of a top predator with the invasion of a novel prey

  • Nature Ecology & Evolution 2108115 (2017)
  • doi:10.1038/s41559-017-0378-1
  • Download Citation
Received:
Accepted:
Published online:

Abstract

Invasive exotic species are spreading rapidly throughout the planet. These species can have widespread impacts on biodiversity, yet the ability for native species, particularly long-lived vertebrates, to respond rapidly to invasions remains mostly unknown. Here we provide evidence of rapid morphological change in the endangered snail kite (Rostrhamus sociabilis) across its North American range with the invasion of a novel prey, the island apple snail (Pomacea maculata), a much larger congener of the kite’s native prey. In less than one decade since invasion, snail kite bill size and body mass increased substantially. Larger bills should be better suited to extracting meat from the larger snail shells, and we detected strong selection on increased size through juvenile survival. Using pedigree data, we found evidence of both genetic and environmental influences on trait expression and discovered that additive genetic variation in bill size increased with invasion. However, trends in predicted breeding values emphasize that recent morphological changes have been driven primarily by phenotypic plasticity rather than micro-evolutionary change. Our findings suggest that evolutionary change may be imminent and underscore that even long-lived vertebrates can respond quickly to invasive species. Furthermore, these results highlight that phenotypic plasticity may provide a crucial role for predators experiencing rapid environmental change.

Main

Humans are rapidly spreading exotic species throughout the world1. Evidence is mounting that invasive species undergo rapid phenotypic changes as they invade new ecosystems2. However, the ecological and evolutionary consequences for native species, particularly vertebrates, are only beginning to be understood3,4. If adaptive phenotypic plasticity and biological evolution can operate rapidly in native species, particularly species of conservation concern, this could fundamentally alter conservation strategies and models aimed at predicting the effects of human-induced environmental change5,6.

Long-lived species are thought to be vulnerable to rapid environmental change7, in part because of their inability to respond at the same pace as rates of change. However, research in this area is scarce. Indeed, nearly all examples of evolutionary response by native species to invasions come from plants and invertebrates with relatively short generation times (for example, less than one year)8,9. Furthermore, little is known about the mechanisms that underlie mostrapid phenotypic responses that have been documented10. For example, to what extent are the changes due to evolution, and to what extent are they due to phenotypic plasticity? Given that many flagship and umbrella species used to guide conservation decisions are long-lived vertebrate predators, and that the loss of predators can have unappreciated ecological and evolutionary consequences11, understanding the potential for such species to rapidly respond to human-induced change is essential. Nonetheless, empirical evidence that exotic prey can induce plastic and micro-evolutionary responses in native predators has been lacking12, except in the case of a toxic prey13.

We investigated the morphological response of an endangered, long-lived predator, the snail kite (Rostrhamus sociabilis plumbeus; Fig. 1a) to the invasion of an exotic apple snail (Pomacea maculata; formerly known as P. insularum14). The snail kite is a wetland-dependent raptor that feeds almost exclusively on apple snails (Fig. 1b), using its curved beak and long talons to extract snail meat from shells15. Over the past decade, P. maculata—considered one of the world’s worst invaders16—has swept across the range of snail kites in the United States, reaching very high densities (2–100 times that of native snails) where it has invaded17 (Fig. 1c). Snail kites, particularly fledglings, have difficulty handling the exotic P. maculata, because of their larger size compared to the native apple snail, P. paludosa18 (Fig. 1b,d). Nonetheless, kites frequently consume and feed P. maculata to nestlings19, and nesting distributions of kites have closely tracked the invasion17 (Fig. 1c). Recent demographic analyses show that population growth rates of snail kites have increased since the invasion of P. maculata17, while the population had previously severely declined and was near extinction20.

Fig. 1: The endangered snail kite, its novel prey and the invasion across the kite’s geographic range.
Fig. 1

a, Snail kites are dietary specialists that have morphological traits, particularly bill size and shape, that have adapted to extracting apple snails from their shells. b, The exotic apple snail (P. maculata; right) is a novel prey21 for snail kites, because it is much larger than the native congener (P. paludosa; left), leading to implications for foraging and demography17,19. c, The invasion first occurred in Lake Tohopekaliga, where P. maculata had become established by the 2005 breeding season (orange). By 2009, P. maculata had established in several wetlands (red). Snail kite breeding closely tracked the invasion sequence, where bars show changes in the proportion of nests over time with the invasion (pre-invasion 2003, 2003–2004; initial invasion 2005, 2005–2008; post-invasion 2009, 2009–2012). Annual averages are shown, n = 1,778 nests. Test for change in the proportions of nests across regions over time: F 4,45 = 13.1, P < 0.0001. d, Snail kites do feed on the much larger exotic snails. The relative frequencies of snail sizes consumed by snail kites in 2013–2014 (n = 903) are shown, taken from snail shells collected at foraging perches throughout the range.

Here we examined whether this novel prey21 has driven a rapid increase in the bill, body and tarsus size of this long-lived vertebrate predator. We coupled 11 years of extensive demographic data with quantitative genetic analyses on kites spanning 350 km across the breeding range of this endangered bird. We expected an increase in the size of fledgling kites over time since the invasion, because of the increased abundance of large snails and the importance of kite morphology for foraging success22,23,24. We tested whether these traits have undergone directional selection since the invasion, because we expected that larger billed kites may more easily forage on this prey. Using pedigree data, we isolated genetic and environmental variance in morphological traits and contrasted changes in the genetic component of traits (that is, the predicted breeding values) with that of phenotypic changes, to address the extent to which changes were driven by evolution, phenotypic plasticity or both. An evolutionary response would require the presence of both natural selection and additive genetic variation of traits under selection25.

Results

The invasion of this prey started in the kite breeding habitat at a single wetland (Lake Tohopekaliga; Fig. 1c) in late 2004; by 2009, P. maculata was abundant in several wetlands in the region17. Since the invasion, the morphology of fledgling snail kites has changed considerably, with bill length and body mass increasing by approximately 1 s.d. over 8 years (Fig. 2a–f). Trait distributions shifted only after P. maculata had spread to several wetlands used by snail kites (Fig. 2a–c). Overall, mass increased with the invasion of the prey (t = 4.19, d.f. = 1,586, P < 0.001; Supplementary Tables 1, 2), and tarsus and bill length increased, even while accounting for changes in mass and annual trends (tarsus: t1,585 = 2.88, P = 0.004; bill length: t1,584 = 3.02, P < 0.003). Furthermore, bill and tarsus size increased over time, after accounting for changes in mass (year effects, tarsus: t1,585 = 2.20, P = 0.028; bill length: t1,584 = 5.63, P < 0.0001; Supplementary Tables 1, 2). Increases in bill length and body mass may be beneficial given the large size of this prey, if such trait changes confer increased fitness.

Fig. 2: Rapid change in morphology, but not predicted breeding values, of fledgling snail kites with the invasion of P. maculata.
Fig. 2

ac, Probability densities of traits by year from 2003 to 2012. df, Changes based on pre- (2003–2004) and post-invasion (2009–2012) data, where ‘post’ refers to the time after which exotic snails had spread to several sites within the snail kite’s breeding range. gi, Changes in predicted breeding values for each morphological trait. For di, the centre line represents the median, hinges represent 1st and 3rd quartiles, whiskers indicate the 1.5× interquartile range and points are outliers. Panels af show standardized phenotypic traits, such that variation is illustrated in units of s.d. For i, body mass breeding values were scaled (divided by 10) to improve visualization.

Biological invasions have the potential to greatly alter selection on native species, providing a catalyst for rapid evolution26. First-year survival of snail kites varied substantially among cohorts, ranging from 9% (2004) to 62% (2011). On the basis of analyses that estimated selection gradients27,28, we found evidence for directional selection through first-year survival, where larger kites and kites with larger beaks were more likely to survive their first year (Fig. 3, Supplementary Table 3 and Supplementary Fig. 1), but stabilizing selection on tarsus length. Given the importance of first-year survival to the population viability of this critically endangered raptor20, the selection for larger-billed and larger-bodied individuals and the change in these traits over time can help illuminate why recent increases in first-year survival and population growth have occurred17.

Fig. 3: Snail kites with large bills and larger mass were more likely to survive the first year of life since the invasion of P. maculata.
Fig. 3

ac, Results from selection analyses28 are shown for apparent juvenile survival on morphology in the snail kite after the invasion, pooling across sexes and years (2004–2012 cohorts; n = 582). We detected directional selection gradients, β (selection parameters accounting for correlated traits), on bill size and body mass, and stabilizing selection (quadratic gradients, γ) on tarsus length (*P < 0.01, **P < 0.001, based on bootstrapping). Shaded areas represent 95% prediction intervals. Rugs (vertical lines) show raw data for trait distributions of survivors (orange) and non-survivors (grey). For analyses by sex and for selection differentials, see Supplementary Table 3.

We used quantitative genetic analyses with an animal model approach29,30, which has been emphasized as a rigorous means of interpreting potential phenotypic plasticity and evolutionary responses to environmental change in wild populations31. We decomposed phenotypic variance in morphological traits to additive genetic variance and environmental variance, which accounted for nest, wetland and annual environmental variation. Results indicated significant additive genetic variance and heritability (Fig. 4), but relatively low coefficients of additive genetic variation32, CVA (Supplementary Table 4), for each trait. Additionally, increased additive genetic variances and heritabilities of bill length in invaded wetlands (Fig. 4 and Supplementary Table 4) suggested cryptic genetic variation, or genetic variation that is only expressed under novel conditions33, occurred with establishment of P. maculata. In particular, changes were greatest for bill length—the exact trait expected to influence foraging success on exotic prey.

Fig. 4: Variance components of morphology of snail kites in invaded and uninvaded wetlands suggest cryptic genetic variation for morphology.
Fig. 4

ac, Estimates (±s.e.; n = 590) of additive genetic variance (V A), environmental variance (V E) and heritability (h 2; the proportion of total variance of a trait due to additive genetic variance). Significant changes in variance components are denoted by an asterisk, and based on 95% confidence intervals for the change in the posteriors of variance components. See Supplementary Table 4 for further variance decomposition.

The increase in sizes of morphological traits was probably due to phenotypic plasticity based on three lines of evidence. First, we tested for changes in the genetic component of morphological traits using predicted breeding values34, both as linear trends over time and as a function of the three relevant time periods for the invasion that have been described previously17. None of the traits showed evidence for changes in predicted breeding values (P > 0.4; Fig. 2g–i and Supplementary Table 5). Although breeding values were correlated with phenotypic values (r > 0.81), there were no substantial changes in breeding values over time (Fig. 2g–i and Supplementary Fig. 1). Second, growth rates of nestlings appeared highly responsive to food availability, with greater growth rates in invaded versus non-invaded wetlands (Supplementary Fig. 2). Finally, there was a mismatch between the strength of selection, additive genetic variance and observed trait change. Snail kites are long-lived birds and, using previously published demographic data17, we calculated that generation time in snail kites was approximately 5–8 years during our study, such that the observed rapid morphological change occurred over approximately 1–1.4 generations. When applying the breeder’s equation35, the expected rate of trait change from micro-evolutionary responses was on the order of 0.05–0.1 s.d. (or 0.006–0.013 per year), which was clearly lower than observed mean trait change (Fig. 2, Supplementary Fig. 2 and Supplementary Tables 2, 6).

Discussion

We provide clear evidence for rapid morphological change in a long-lived vertebrate predator with the invasion of a novel prey. The potential for an invasive prey to elicit such a rapid response from a long-lived native predator is remarkable36. Bill length not only increased in size, but also increased relative to body size; kites are now larger with especially large beaks. Although there have been some examples of invasive species inducing morphological change in wild populations, the mechanisms for change have remained unknown10,13. By testing for selection and partitioning trait variation into additive genetic and environmental components, we provide insight into the potential roles of phenotypic plasticity and evolutionary response to invasive prey. We found that there has been directional selection on the first-year survival of snail kites, with larger kites and kites with larger bills (controlling for size) surviving better. Interestingly, directional selection and additive genetic variance have not yet led to significant evolutionary changes in morphology in this population. Instead, snail kites appear to be growing large with disproportionately large bills through phenotypic plasticity in response to the new, larger food source. It is no surprise that a large, abundant prey could increase body size, but the bills of these birds are now larger than would be expected relative to body mass.

We found evidence that additive genetic variation, particularly in bill length, increased over invasion time. This finding is important, because additive genetic variation is necessary for evolution, and it is lacking in some threatened and endangered taxa. When additive genetic variation is low, populations may not be able to evolve when conditions change, sometimes leading to extinction37. Genetic variation expressed in response to novel environmental conditions is termed cryptic genetic variation33,38, but evidence of such changes in wild populations has been rare. Cryptic genetic variation provides an unappreciated means by which evolutionary responses can proceed even in populations with previously documented low additive genetic variation38. Cryptic genetic variation, together with phenotypic plasticity, may be have crucial roles allowing native species to respond quickly to invaders39.

The breeder’s equation and breeding values are often used to detect evolutionary change31,40,41,42. By contrasting these methods to observed phenotypic change, we found that phenotypic plasticity, not evolution, is probably responsible for most of the morphological changes in the snail kite at this point in time. It is important to note that the breeder’s equation does not account for some evolutionary processes that may be relevant for morphological change (for example, correlated selection)43, and selection may be operating on other parts of the life cycle (for example, adult survival). However, trends in predicted breeding values suggested that rapid phenotypic change was mostly from plasticity, corroborating predictions from the breeder’s equation. Furthermore, nestling growth rates were greater in wetlands where invasive prey had established, consistent with plastic responses of nestlings to increases in food availability44,45.

Here we are witnessing the initial stages of the response of snail kites to an invasive prey, illustrating effects on genetic variation and the role of phenotypic plasticity in facilitating the shift to a novel prey. The presence of directional selection and additive genetic variation suggests that substantial evolutionary change may occur in the coming years. But there are some limitations to our conclusions. First, we had limited morphological data prior to the invasion of P. maculata. Although standardized demographic monitoring has occurred since 1996, only since 2003 did this work incorporate standardized morphological measurements. In addition, in 2003, the population was alarmingly small and was declining46, such that the statistical power for interpreting variance components of traits prior to invasion was low and we were not able to estimate whether selection has changed with invasion. Second, our animal model approach for estimating variance components relied on a social pedigree. For many songbirds, extra-pair paternity is common47; however, the role of extra-pair paternity in raptors is thought to be less prevalent than in songbirds and is hypothesized to decrease with population density48. If extra-pair paternity occurs in snail kites, then our estimates for additive genetic variance may be conservative49,50. Finally, we did not isolate the alleles responsible for trait variation. Although the quantitative genetic analyses clearly illustrate a large component of additive genetic variance, further work that isolates the alleles responsible for trait variation would be useful to better understand the mechanisms of morphological change.

These results illustrate that long-lived, top predators can rapidly respond to invasive species through phenotypic plasticity. Phenotypic plasticity may prove to be crucial to imperilled species, allowing them to rapidly respond as their environment changes. We found evidence of increased genetic variation and directional selection, suggesting that an evolutionary response is imminent. But the long-term evolutionary and ecological consequences of this invasion remain uncertain. Our results highlight the value of considering phenotypic plasticity and evolution when developing plans for conservation and management of endangered species facing rapid environmental change26, even for long-lived species.

Methods

Study species

The snail kite (R. sociabilis) is an extreme dietary specialist. In the United States, the snail kite is a critically endangered subspecies (R. s. plumbeus), that previously fed almost exclusively on the freshwater apple snail (P. paludosa)15, the only Pomacea species native to the United States6. The size of P. paludosa (hereafter the native snail) rarely exceeds 60 mm in shell length18 (Fig. 1d). Historically, 98.5% of the kite’s diet consisted of native snails ranging from 20–60 mm24. The success of snail kite’s meticulous handling process is probably snail size-dependent and limited by kite morphology, particularly by bill length, toe length and overall body size18,23. Here we focus on the entire breeding range in the United States (Fig. 1c), where snail kites are endangered, although snail kites are also found in South and Central America and the Caribbean15.

P. maculata (formerly P. insularum; hereafter, the exotic snail), is invasive in the United States, and is typically 2–3 times larger and 4–5 times more massive than the native snail18. P. maculata is a ‘novel’ prey21, because it was not resident in this region until recently and its large body size is not observed in native taxa in this region. The larger size of exotic snails can impact kite foraging performance, leading to higher drop rates, longer handling times, difficulties with meat extraction, and the potential for negative daily energy balances for juvenile kites18,19. Nonetheless, exotic snails constitute a growing proportion of the snail kite’s diet22, because exotic snail populations have expanded rapidly throughout Florida16,17, whereas many native snail populations have declined51. Recently fledged snail kites are less adept than older individuals at handling and extracting snails19, with starvation likely to be a leading cause of death for young kites15, such that we expected strong selection pressures on traits that may facilitate foraging on exotic snails. Indeed, bill length and body size of many avian predators are correlated with prey size52 and changes to available food resources can lead to directional selection on adult and fledging-age morphology53,54.

Study area

Morphological and demographic data were collected as part of a long-term monitoring program across the breeding range of the snail kite in the United States46, which, during our study (2003–2014), primarily included three regional freshwater systems in peninsular Florida: the Everglades, Lake Okeechobee and the Kissimmee River Valley (KRV)17, each consisted of multiple constituent wetland sites15. Snail kites in Florida are often considered a single, spatially structured population55, with regular movements among wetlands both within and between years56,57.

P. maculata has been established in Florida since at least 2002, and many wetlands across the state were invaded between 2004 and 200616. Lake Tohopekaliga (in the KRV) was the first wetland extensively used by snail kites that had a major invasion of P. maculata, which spread throughout the lake in late 2004 and early 200518,19 (Fig. 1c). By fall 2008 and the 2009 breeding season, exotic snail populations had invaded Lake Okeechobee and the remaining wetlands that were used by kites in the KRV17. Although exotic snails have not yet colonized substantial interior portions of the Everglades22, they are common in many canals, stormwater treatment areas and other marginal habitats throughout the region16 where kites commonly forage and breed17. Although snail density estimates were not available across much of the kite’s range, existing quantitative data indicate that exotic snail presence in a given wetland is strongly associated with both density and size distributions of the kite’s prey base18,19, which impacts kite foraging behaviour and demography17,19. Given the semi-nomadic tendencies of snail kites, increased movement rates to invaded wetlands17 and the widespread distribution of exotic snail populations, it is likely that kites fledged in uninvaded wetlands encounter, and rely upon, exotic snails at some time during dispersal. Nonetheless, we assessed differences between individuals from invaded versus uninvaded wetlands in all analyses for which there were sufficient data (see below).

Morphological measurements, parental assignment and data selection

Snail kite nests were located and monitored during systematic surveys conducted across the kite’s range throughout the breeding season each year (typically around 6 surveys per site per year)46. When nestlings were near fledging, we recorded morphological measurements, equipped individuals with unique alphanumeric leg bands and obtained feather samples (to ascertain sex using genetics). Protocols were approved and conducted under the University of Florida Institutional Animal Care and Use Committee permit number 201005469.

Between 2003 and 2012, we took standardized morphological measurements: length of exposed culmen (without cere) along the upper mandible from the proximal end of the cere to the bill tip (±0.1 mm; hereafter bill length), tarsus length (±0.1 mm), flattened wing chord from the wrist to the tip of the longest primary (±0.1 cm) and body mass (±1 g). We aged nestlings to within ±1 day58. In snail kites, extra-pair copulation has not been observed15, and both parents incubate, tend young and will chase away intraspecific intruders15, so we assumed that banded adult(s) at the nest were true parents and used this information to establish pedigrees.

To obtain adequate sample sizes while minimizing variation due to differences in age of measurement, we restricted analyses to individuals processed within a one-week age window surrounding the mean fledging age of 29 days15. Use of this criterion resulted in measurements on 590 individuals (304 females, 286 males) between the ages of 26–32 days from 10 annual cohorts (2003–2012). For snail kites in Florida, at 25 days of age, tarsus length has reached adult size and body mass during the nestling period has reached an asymptote (Supplementary Fig. 2). Both adult and fledging-age snail kites display reversed sexual size dimorphism in some traits, particularly body mass, with females being larger than males on average; however, male and female morphologies overlap considerably and both sexes reach asymptotic size at generally the same age (Supplementary Fig. 2).

Tarsus length is a commonly used univariate indicator of overall body size59, and linear measurements of other structural components relevant to foraging on apple snails (for example, toe length) often correlate strongly with tarsus length60. Fledging body mass, too, can provide valuable information on body size61 and energy reserves62. Here we considered both tarsus length and body mass as indicators of body size.

The snail kite’s bill has a central role in the ability to extract snails23. Snail kite bill length is only weakly correlated with the other traits (r < 0.26; Supplementary Table 5). Notably, snail kite bill length does not reach an asymptote prior to fledging, but instead continues with almost linear growth throughout the nestling period63. We controlled for the effects of age on bill length in our 26-to-32-day-old chicks in one of two ways, depending on the modeling framework of a given analysis and whether bill length was included as a response or predictor variable. When including bill length as a response variable (for example, estimating phenotypic variance components), we directly incorporated age as a fixed covariate (see below). In order to assess the effects of bill length on a given response (for example, survival), on the other hand, required a standardized measurement that was comparable across individuals. Assuming a linear relationship with age, we used the slope coefficient from a linear regression model (βage = 0.22; R2 = 0.39, F2,961 = 309.10, P < 0.001) to estimate bill length at 32 days (hereafter age-corrected bill length) based on the measured bill length and age at measurement for each individual.

Trends in phenotypic traits since invasion

We evaluated the effects of the establishment of exotic snails and time since invasion on snail kite morphological traits; however, we also accounted for other factors that have been shown, or are predicted, to influence these traits. We tested for linear year effects (that is, time since invasion) using generalized linear models (GLMs; Gaussian error, identity link). To account for known sources of variation, we included the effects of sex and, as appropriate, age of measurement (see above). We also considered models that included exotic snail presence in the natal wetland (as a categorical factor) to test for the possibility that exotic snails may provide supplemental food resources to kites17, which could potentially have proximate effects on fledgling morphology, particularly if morphology responds plastically to resource availability45,64. In analysing variation in tarsus and bill length, we used standardized body mass (as a covariate) to account for potential variation in nutritional state and correlations between body mass and other size traits.

We performed all analyses in the program R v.3.0.265. We ranked models and evaluated their relative performance using Akaike’s Information Criteria corrected for small sample size (AICc) and AICc weights (w i ). We considered effects of a covariate to be significant when the 95% confidence interval for its parameter estimate did not overlap zero.

Viability selection

Measuring fitness components

To assess selection on fledging-age size traits, we used survival to reproductive age (that is, first-year, or juvenile, survival66) as a component measure of fitness. Annual survival is relatively constant in adult snail kites (≈ 85–90%), whereas juvenile survival is lower and more variable, typically ranging between 15 and 60%66, making this a critical period for viability selection. We used 10 years of band re-sight data (2005–2014) to assess survival of individuals from 9 cohorts (2004–2012). We classified individuals re-sighted during any year subsequent to fledging as survivors, while individuals that were never re-sighted were classified as non-survivors. Although individuals from earlier cohorts may have been available for detection over more years, they were no more likely to be classified as survivors than individuals from later cohorts (see Supplementary Fig. 3).

In other contexts17,66, snail kite survival has been typically analysed in a Cormack–Jolly–Seber (CJS) modelling framework to account for imperfect detection. Annual estimates of survival obtained from CJS methods that incorporated all banded individuals were nearly identical to estimates based on the above classification method (linear regression: β0 = −0.01, β1 = 0.98; R2 = 0.91, F1,7 = 74.29, P < 0.001), which is not surprising, because we have no reason to expect that resighting would be dependent on morphological traits per se. Because coefficients from CJS models are not equivalent to the parameters used in standard equations from quantitative genetic theory, and because selection is based on relative (not absolute) fitness, we used our classification method to provide selection estimates consistent with other similar studies67.

Selection analyses

To quantify the forces of selection acting on each trait, we estimated four common selection parameters from quantitative genetic theory: standardized linear (S) and quadratic (c) selection differentials, and standardized linear (β) and quadratic (γ) selection gradients27. Selection gradients provide estimates of selection on correlated traits, whereas selection differentials ignore such correlations27. We report both selection differentials and gradients, because we found statistically significant, but often weak, correlations between all pairs of traits (r = 0.05–0.26; Supplementary Table 5) and we were interested in separating direct and indirect selection. We expected positive directional selection to be acting on all size traits, but we also examined quadratic parameters to determine whether stabilizing selection might constrain directional change27.

We used recent advances in selection analyses that unify estimation of selection parameters and their visualization through the implementation of generalized additive models (GAMs) on absolute fitness values28. This approach provides quantitative estimates of selection differentials and gradients for non-normal fitness components, such as survival. We fit GAMs for binary survival data, using a logit link function and assuming a binomial error distribution with the ProgramCodeProgramCode LineGroup FixedLinemgcv package in R68. To estimate selection differentials, we fitted univariate GAMs, whereas to estimate selection gradients, we included all three traits into one GAM. We initially considered a tensor product smooth function that allowed non-linear interactions among traits in our selection gradient model, but there was no evidence for such effects (based on model comparisons using AIC), so we only included additive spline effects in the final model. With each generalized GAM function, differential and gradient parameters were estimated based on numerical approximations of first and second partial derivatives of relative fitness, averaged over the distribution of observed phenotype28. To interpret the significance of selection differentials and gradients, we used a bootstrap approach (n = 2,000 samples) implemented in the ProgramCodeProgramCode LineGroup FixedLinegsg package in R28. Because of sexual size dimorphism15, we assessed viability selection on size traits for each sex independently as well as with sexes pooled. For all selection analyses, we centred and scaled covariates within sexes to facilitate pooling data across sexes. We then back-transformed scaled covariates for visualization.

Quantitative genetic analysis

Estimating variance components

A prerequisite for an adaptive evolutionary response in a given trait is that the trait must have sufficient additive genetic variation on which natural selection can act35. To assess whether fledgling-age morphology in snail kites met this criterion, we used an animal model approach29,30 to estimate phenotypic variance components and derive metrics of trait evolvability and heritability of each quantitative trait. Animal models extend the generalized linear mixed model framework by incorporating a relationship matrix (based on known pedigree data), which accounts for levels of relatedness between pairs of individuals and is used to define the variance–covariance structure for additive genetic random effects. These models are robust to unbalanced and missing data, can accommodate fixed effects, and, by taking advantage of all known familial relationships, animal models provide greater statistical power than traditional quantitative genetic approaches29,30.

For each trait, we applied a model that included additive genetic and common environment random effects. We included nest as a random effect to account for non-independence of siblings (that is, common rearing environment), and we included wetland and year as random effects to account for larger-scale common environment effects acting on groups of nests located proximally in space and time (for example, local prey availability). We included sex as a fixed effect in all models. In models for bill length, we also included nestling age as a covariate. After controlling for fixed effects, we partitioned total phenotypic variance (VP) into additive genetic (VA), environmental (VE) and residual (VR) components. We further partitioned environmental variance into components deriving from effects of common nest environment (VEn) and, at a larger-scale, common wetland (VEw) and year (VEy ) environments. Estimates of genetic variation must be standardized to enable comparison across traits. We standardized these estimates in two ways to infer evolvability32 by calculating the coefficient of additive genetic variation (CVA; the square root of additive genetic variation divided by the trait mean; expressed as a percentage) and heritability (the ratio of additive genetic variation to phenotypic variation).

We also estimated variance components using the subsets of individuals fledged in wetlands with and without exotic snails. Exotic snails provide supplemental food resources to breeding kites17, and could potentially influence offspring phenotypes through environmentally mediated effects (for example, larger egg sizes or post-hatching provisioning rates)69 or through genetic effects. Evolutionary theory predicts that directional selection could reduce VA under some circumstances35. However, the invasion of a novel prey is not typical and has the potential to reveal cryptic genetic variation, or genetic variation that is only expressed under novel conditions33, especially shortly after the addition of the novel prey as examined here. Thus, by contrasting VA and VE in invaded and non-invaded wetlands, we tested predictions regarding resource stabilization (that is, a lower VE and lower proportion VE in invaded wetlands) and directional selection reducing VA in invaded wetlands, as well as novel conditions potentially releasing genetic variation (that is, increased VA in invaded wetlands). We did not attempt to estimate variance components before the initial invasion in 2004, because we did not have sufficient sample sizes to do so (at that time, the population size of snail kites had declined to alarmingly low levels). To fit these models, we used Markov chain Monte Carlo analyses with Bayesian inference using the ProgramCodeProgramCode LineGroup FixedLineMCMCglmm package in R. This approach allows proper quantification of all uncertainty in random variables, which can be difficult with frequentist methods30. To do so, we used a burn-in of 10,000 samples, thinning rate of 50, and ran a total of 110,000 samples, resulting in 10,000 thinned samples from the posterior distribution. We used trace plots and the Heidel diagnostic to interpret convergence. We used priors based on observed sampling variances30.

Estimating and predicting evolutionary change

We used two approaches to infer evolutionary change. First, we contrasted observed phenotypic change with that predicted from the breeder’s equation35,42. Second, we quantified changes in predicted breeding values over time34.

The breeder’s equation predicts rates of evolutionary change, R, based on the product of heritability, h2, and selection differentials, s, (R = h2s). We quantified h2 and s as described above (where s is taken from univariate selection analyses). R is a measurement of predicted evolutionary response per generation in s.d. of the trait of interest. To interpret R relative to observed phenotypic change, we estimated generation time for snail kites using demographic parameter estimates of snail kites post-invasion provided in ref. 17 to populate a two-stage (first year and after-first year) matrix model. With this matrix model, we used the ProgramCodeProgramCode LineGroup FixedLinegeneration.time function in the ProgramCodeProgramCode LineGroup FixedLinepopbio package to estimate generation time for snail kites70.

Because the breeder’s equation has several assumptions43, we also quantified predicted breeding values for individuals within our animal model analyses by following the approach outlined in ref. 34. Predicted breeding values can be quantified through best linear unbiased predictors for the additive genetic component of the animal model analysis. We use posterior distributions from the random effects of breeding values to test for changes in predicted breeding values over time with linear models on the posteriors, based on a linear, fixed effect of time (year) and by testing for changes across the three relevant time periods described in ref. 17: 2003–2004 (pre-invasion), 2005–2008 and 2009–2012. In this analysis, year was included as a random effect, minimizing concerns regarding confounding annual environmental variation with that of genetic change40 (note that results did not change when removing year as a random effect). We also tested for changes beyond that expected by genetic drift34, and the results were similar (Supplementary Table 6).

Life Sciences Reporting Summary

Further information on experimental design is available in the Life Sciences Reporting Summary.

Data availability

The snail kites is a listed federally endangered species and the specific location data are sensitive and have not been made available. However, the general data used for these analyses are available at figshare: https://figshare.com/s/e5824db8a8b523caf3b0.

Additional Information

Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

References

  1. 1.

    Vitousek, P. M., D'Antonio, C. M., Loope, L. L., Rejmanek, M. & Westbrooks, R. Introduced species: a significant component of human-caused global change. N. Z. J. Ecol. 21, 1–16 (1997).

  2. 2.

    Maron, J. L., Vila, M., Bommarco, R., Elmendorf, S. & Beardsley, P. Rapid evolution of an invasive plant. Ecol. Monogr. 74, 261–280 (2004).

  3. 3.

    Hanfling, B. & Kollmann, J. An evolutionary perspective of biological invasions. Trends Ecol. Evol. 17, 545–546 (2002).

  4. 4.

    Shine, R. Invasive species as drivers of evolutionary change: cane toads in tropical Australia. Evol. Appl. 5, 107–116 (2012).

  5. 5.

    Sih, A., Ferrari, M. C. O. & Harris, D. J. Evolution and behavioural responses to human-induced rapid environmental change. Evol. Appl. 4, 367–387 (2011).

  6. 6.

    Moran, E. V. & Alexander, J. M. Evolutionary responses to global change: lessons from invasive species. Ecol. Lett. 17, 637–649 (2014).

  7. 7.

    Henle, K., Davies, K. F., Kleyer, M., Margules, C. & Settele, J. Predictors of species sensitivity to fragmentation. Biodivers. Conserv. 13, 207–251 (2004).

  8. 8.

    Carroll, S. P. et al. And the beak shall inherit — evolution in response to invasion. Ecol. Lett. 8, 944–951 (2005).

  9. 9.

    Prentis, P. J., Wilson, J. R. U., Dormontt, E. E., Richardson, D. M. & Lowe, A. J. Adaptive evolution in invasive species. Trends Plant Sci. 13, 288–294 (2008).

  10. 10.

    Langkilde, T. Invasive fire ants alter behavior and morphology of native lizards. Ecology 90, 208–217 (2009).

  11. 11.

    Estes, J. A. et al. Trophic downgrading of planet Earth. Science 333, 301–306 (2011).

  12. 12.

    Berthon, K. How do native species respond to invaders? Mechanistic and trait-based perspectives. Biol. Invasions 17, 2199–2211 (2015).

  13. 13.

    Phillips, B. L. & Shine, R. Adapting to an invasive species: toxic cane toads induce morphological change in Australian snakes. Proc. Natl Acad. Sci. USA 101, 17150–17155 (2004).

  14. 14.

    Hayes, K. A., Cowie, R. H., Thiengo, S. C. & Strong, E. E. Comparing apples with apples: clarifying the identities of two highly invasive neotropical Ampullariidae (Caenogastropoda). Zool. J. Linn. Soc. 166, 723–753 (2012).

  15. 15.

    Reichert, B. E. et al. in The Birds of North America Online (ed. Poole, A.) https://doi.org/10.2173/bna.171 (Cornell Lab of Ornithology, Ithaca, 2015).

  16. 16.

    Rawlings, T. A., Hayes, K. A., Cowie, R. H. & Collins, T. M. The identity, distribution, and impacts of non-native apple snails in the continental United States. BMC Evol. Biol. 7, 97 (2007).

  17. 17.

    Cattau, C. E., Fletcher, R. J. Jr, Reichert, B. E. & Kitchens, W. M. Counteracting effects of a non-native prey on the demography of a native predator culminate in positive population growth. Ecol. Appl. 26, 1952–1968 (2016).

  18. 18.

    Darby, P. C., Mellow, D. J. & Watford, M. L. Food-handling difficulties for snail kites capturing non-native apple snails. Fla. Field Nat. 35, 79–85 (2007).

  19. 19.

    Cattau, C. E., Martin, J. & Kitchens, W. M. Effects of an exotic prey species on a native specialist: example of the snail kite. Biol. Conserv. 143, 513–520 (2010).

  20. 20.

    Martin, J., Kitchens, W. M., Cattau, C. E. & Oli, M. K. Relative importance of natural disturbances and habitat degradation on snail kite population dynamics. Endanger. Species Res. 6, 25–39 (2008).

  21. 21.

    Saul, W.-C. & Jeschke, J. M. Eco-evolutionary experience in novel species interactions. Ecol. Lett. 18, 236–245 (2015).

  22. 22.

    Wilcox, R. & Fletcher, R. J. Jr. Experimental test of preferences for an invasive prey by an endangered predator: implications for conservation. PLoS ONE 11, e0165427 (2016).

  23. 23.

    Snyder, N. F. R. & Snyder, H. A. A comparative study of mollusk predastion by limpkins, Everglade kites, and boat-tailed grackles. Living Bird 8, 177–223 (1969).

  24. 24.

    Sykes, P. W. Jr. The feeding habits of the snail kite in Florida, USA. Colon. Waterbirds 10, 84–92 (1987).

  25. 25.

    Strauss, S. Y., Lau, J. A. & Carroll, S. P. Evolutionary responses of natives to introduced species: what do introductions tell us about natural communities? Ecol. Lett. 9, 354–371 (2006).

  26. 26.

    Stockwell, C. A., Hendry, A. P. & Kinnison, M. T. Contemporary evolution meets conservation biology. Trends Ecol. Evol. 18, 94–101 (2003).

  27. 27.

    Lande, R. & Arnold, S. J. The measurement of selection on correlated characters. Evolution 37, 1210–1226 (1983).

  28. 28.

    Morrissey, M. B. & Sakrejda, K. Unification of regression-based methods for the analysis of natural selection. Evolution 67, 2094–2100 (2013).

  29. 29.

    Kruuk, L. E. B. Estimating genetic parameters in natural populations using the ‘animal model’. Phil. Trans. R. Soc. Lond. B 359, 873–890 (2004).

  30. 30.

    Wilson, A. J. et al. An ecologist’s guide to the animal model. J. Anim. Ecol. 79, 13–26 (2010).

  31. 31.

    Merila, J. & Hendry, A. P. Climate change, adaptation, and phenotypic plasticity: the problem and the evidence. Evol. Appl. 7, 1–14 (2014).

  32. 32.

    Hansen, T. F., Pelabon, C. & Houle, D. Heritability is not evolvability. Evol. Biol. 38, 258–277 (2011).

  33. 33.

    Gibson, G. & Dworkin, I. Uncovering cryptic genetic variation. Nat. Rev. Genet. 5, 681–690 (2004).

  34. 34.

    Hadfield, J. D., Wilson, A. J., Garant, D., Sheldon, B. C. & Kruuk, L. E. B. The misuse of BLUP in ecology and evolution. Am. Nat. 175, 116–125 (2010).

  35. 35.

    Falconer, S. & Mackay, T. F. C. Introduction to Quantitative Genetics 4th edn (Pearson Education, Harlow, 1996).

  36. 36.

    Mooney, H. A. & Cleland, E. E. The evolutionary impact of invasive species. Proc. Natl Acad. Sci. USA 98, 5446–5451 (2001).

  37. 37.

    Lande, R. & Shannon, S. The role of genetic variation in adaptation and population persistence in a changing environment. Evolution 50, 434–437 (1996).

  38. 38.

    Paaby, A. B. & Rockman, M. V. Cryptic genetic variation: evolution’s hidden substrate. Nat. Rev. Genet. 15, 247–258 (2014).

  39. 39.

    West-Eberhard, M. J. Developmental Plasticity and Evolution (Oxford Univ. Press, New York, 2003).

  40. 40.

    Postma, E. Implications of the difference between true and predicted breeding values for the study of natural selection and micro-evolution. J. Evol. Biol. 19, 309–320 (2006).

  41. 41.

    Merila, J., Kruuk, L. E. B. & Sheldon, B. C. Natural selection on the genetical component of variance in body condition in a wild bird population. J. Evol. Biol. 14, 918–929 (2001).

  42. 42.

    Reale, D., McAdam, A. G., Boutin, S. & Berteaux, D. Genetic and plastic responses of a northern mammal to climate change. Proc. R. Soc. Lond. B 270, 591–596 (2003).

  43. 43.

    Morrissey, M. B., Kruuk, L. E. B. & Wilson, A. J. The danger of applying the breeder’s equation in observational studies of natural populations. J. Evol. Biol. 23, 2277–2288 (2010).

  44. 44.

    Beissinger, S. R. & Snyder, N. F. R. Mate desertion in the snail kite. Anim. Behav. 35, 477–487 (1988).

  45. 45.

    Lacombe, D., Bird, D. M. & Hibbard, K. A. Influence of reduced food availability on growth of captive American kestrels. Can. J. Zool. 72, 2084–2089 (1994).

  46. 46.

    Reichert, B. E., Kendall, W. L., Fletcher, R. J. Jr. & Kitchens, W. M. Spatio-temporal variation in age structure and abundance of the endangered snail kite: pooling across regions masks a declining and aging population. PLoS ONE 11, e0162690 (2016).

  47. 47.

    Griffith, S. C., Owens, I. P. F. & Thuman, K. A. Extra pair paternity in birds: a review of interspecific variation and adaptive function. Mol. Ecol. 11, 2195–2212 (2002).

  48. 48.

    Mougeot, F. Breeding density, cuckoldry risk and copulation behaviour during the fertile period in raptors: a comparative analysis. Anim. Behav. 67, 1067–1076 (2004).

  49. 49.

    Charmantier, A. & Reale, D. How do misassigned paternities affect the estimation of heritability in the wild? Mol. Ecol. 14, 2839–2850 (2005).

  50. 50.

    Firth, J. A., Hadfield, J. D., Santure, A. W., Slate, J. & Sheldon, B. C. The influence of nonrandom extra-pair paternity on heritability estimates derived from wild pedigrees. Evolution 69, 1336–1344 (2015).

  51. 51.

    Darby, P. C., Fujisaki, I. & Mellow, D. J. The effects of prey density on capture times and foraging success of course-hunting adult snail kites. Condor 114, 755–763 (2012).

  52. 52.

    Hespenheide, H. A. Ecological inferences from morphological data. Annu. Rev. Ecol. Syst. 4, 213–229 (1973).

  53. 53.

    Grant, P. R. & Grant, B. R. Unpredictable evolution in a 30-year study of Darwin’s finches. Science 296, 707–711 (2002).

  54. 54.

    Tornberg, R., Monkkonen, M. & Pahkala, M. Changes in diet and morphology of Finnish goshawks from 1960s to 1990s. Oecologia 121, 369–376 (1999).

  55. 55.

    Robertson, E. P., Fletcher, R. J. Jr & Austin, J. D. Microsatellite polymorphism in the endangered snail kite reveals a panmictic, low diversity population. Conserv. Genet. https://doi.org/10.1007/s10592-017-1003-1 (2017).

  56. 56.

    Bennetts, R. E. & Kitchens, W. M. Factors influencing movement probabilities of a nomadic food specialist: proximate foraging benefits or ultimate gains from exploration? Oikos 91, 459–467 (2000).

  57. 57.

    Reichert, B. E., Fletcher, R. J. Jr, Cattau, C. E. & Kitchens, W. M. Consistent scaling of population structure across landscapes despite intraspecific variation in movement and connectivity. J. Anim. Ecol. 85, 1563–1573 (2016).

  58. 58.

    Snyder, N. F. R., Beissinger, S. R. & Chandler, R. E. Reproduction and demography of the Florida Everglade (snail) kite. Condor 91, 300–316 (1989).

  59. 59.

    Rising, J. D. & Somers, K. M. The measurement of overall body size in birds. Auk 106, 666–674 (1989).

  60. 60.

    Newton, I. Population Ecology of Raptors (Poyser, London, 1979).

  61. 61.

    Field, D. J., Lynner, C., Brown, C. & Darroch, S. A. F. Skeletal correlates for body mass estimation in modern and fossil flying birds. PLoS ONE 8, e82000 (2013).

  62. 62.

    Price, T., Kirkpatrick, M. & Arnold, S. J. Directional selection and the evolution of breeding date in birds. Science 240, 798–799 (1988).

  63. 63.

    Beissinger, S. R. Mate Desertion and Reproductive Effort in the Snail Kite. PhD thesis, Univ. Michigan (1984).

  64. 64.

    Vannoordwijk, A. J., Vanbalen, J. H. & Scharloo, W. Heritability of body size in a natural population of the great tit (Parus major) and its relation to age and environmental conditions during growth. Genet. Res. 51, 149–162 (1988).

  65. 65.

    R Core Team. R: A Language and Environment for Statistical Computing (R Foundation for Statistical Computing, Vienna, 2016).

  66. 66.

    Martin, J., Kitchens, W. M. & Hines, J. E. Natal location influences movement and survival of a spatially structured population of snail kites. Oecologia 153, 291–301 (2007).

  67. 67.

    Kingsolver, J. G. et al. The strength of phenotypic selection in natural populations. Am. Nat. 157, 245–261 (2001).

  68. 68.

    Wood, S. N. Generalized Additive Models: an Introduction with R (Chapman & Hall and CRC, Boca Raton, 2006).

  69. 69.

    Charmantier, A. & Garant, D. Environmental quality and evolutionary potential: lessons from wild populations. Proc. R. Soc. B 272, 1415–1425 (2005).

  70. 70.

    Stubben, C. & Milligan, B. Estimating and analyzing demographic models using the popbio package in R. J. Stat. Softw. 22, 1–23 (2007).

Download references

Acknowledgements

We thank J. Orrock, B. Reichert, E. Robertson and M. Morrissey for providing comments on earlier versions of this manuscript. This project was funded by USGS’s Greater Everglades Priority Ecosystems Science (GEPES), the US Army Corps of Engineers, and US Fish and Wildlife Service.

Author information

Affiliations

  1. Department of Wildlife Ecology and Conservation, University of Florida, Gainesville, FL, 32611, USA

    • Christopher E. Cattau
    •  & Robert J. Fletcher Jr
  2. Department of Biology, University of Florida, Gainesville, FL, 32611, USA

    • Rebecca T. Kimball
  3. Department of Entomology and Nematology, University of Florida, Gainesville, FL, 32611, USA

    • Christine W. Miller
  4. US Geological Survey, Florida Cooperative Fish and Wildlife Research Unit, University of Florida, Gainesville, FL, 32611, USA

    • Wiley M. Kitchens

Authors

  1. Search for Christopher E. Cattau in:

  2. Search for Robert J. Fletcher Jr in:

  3. Search for Rebecca T. Kimball in:

  4. Search for Christine W. Miller in:

  5. Search for Wiley M. Kitchens in:

Contributions

C.E.C. conducted the study; C.E.C., R.J.F., R.T.K. and C.W.M. analysed data; C.E.C. and R.J.F. wrote the initial manuscript; C.E.C., R.J.F., R.T.K., C.W.M. and W.M.K. edited the manuscript; W.M.K. and R.J.F. secured funding.

Competing interests

The authors declare no competing financial interests.

Corresponding author

Correspondence to Robert J. Fletcher Jr.

Electronic supplementary material

  1. Supplementary Information

    Supplementary Tables 1–6, Supplementary Figures 1–4

  2. Life Sciences Reporting Summary