Report

Selection on Heritable Phenotypic Plasticity in a Wild Bird Population

See allHide authors and affiliations

Science  14 Oct 2005:
Vol. 310, Issue 5746, pp. 304-306
DOI: 10.1126/science.1117004

Abstract

Theoretical and laboratory research suggests that phenotypic plasticity can evolve under selection. However, evidence for its evolutionary potential from the wild is lacking. We present evidence from a Dutch population of great tits (Parus major) for variation in individual plasticity in the timing of reproduction, and we show that this variation is heritable. Selection favoring highly plastic individuals has intensified over a 32-year period. This temporal trend is concurrent with climate change causing a mismatch between the breeding times of the birds and their caterpillar prey. Continued selection on plasticity can act to alleviate this mismatch.

Phenotypic plasticity—defined as the ability of a single genotype to alter its phenotype in response to environmental conditions—is an important mechanism by which populations can respond rapidly to changes in ecological conditions (13). Plasticity in life history traits is ubiquitous in animal populations (1), with traits often varying within the lifetimes of individuals depending on the conditions they experience (4, 5). It is typically conceptualized and measured using reaction norms: linear functions describing the change in a trait across an environmental gradient (3, 6). Laboratory research has shown that genetic variation for plasticity exists (7, 8) and that heritable plasticity can respond to artificial selection (2, 9).

Given that many species are currently experiencing long-term anthropogenically driven environmental change (10, 11), a better understanding of how natural selection acts on plasticity under altered levels of environmental variation in the wild is imperative. Detailed analyses of within-population variation in life history plasticity are rarely undertaken in naturally occurring populations, because such analyses require data from large numbers of individuals breeding repeatedly across their lifetimes. Recent research using mixed-effects linear models has shown that individuals in two wild vertebrate populations vary in their levels of life history plasticity (4, 12). At present, little is known about the consequences of environmental change for the action of natural selection on plasticity and, ultimately, the ability of populations to continue to respond adaptively to environmental variation. Here we present data from a wild bird population showing temporal trends in natural selection on heritable phenotypic plasticity in the timing of reproduction, which are concurrent with changes in climate and the timing of food availability.

After a warm spring, female passerines often breed earlier than they do after a cold spring (13, 14). This is a result of phenotypic plasticity (14, 15), an individual-level response to temperature. Such a response is considered adaptive because it synchronizes the birds' phenology with the temperature-dependent hatching times and growth rates of the caterpillars they rely on to feed their nestlings (16, 17).

A long-term study of great tits (Parus major) in the Hoge Veluwe, one of the Netherlands' largest national parks, has revealed that after recent warming of spring temperatures in the region, the timing of growth of their caterpillar prey has advanced while the phenology of the birds has not (17). As a result, over the past three decades, the laying dates of female great tits have moved closer to the peak in the caterpillar biomass, so that the peak in demand for food for their offspring no longer coincides with the peak in prey availability (14, 17). Selection on the heritable component of great tits' plastic responses to spring temperatures could act to reduce this phenological mismatch (14).

We used information on laying dates for 833 females breeding in more than 1 year between 1973 and 2004 to examine variation among females in their laying date reaction norms. A random-coefficients model of laying dates (18) showed that after a warm spring, on average, females began laying earlier than they did after a cold spring (Table 1). We found significant variation between females in both their estimated laying date at the average spring temperature [likelihood ratio test (LRT) = 226.73, df = 1, P < 0.001] and the magnitude of their response to spring temperature (LRT = 27.07, df = 2, P < 0.001). Females in this population lay early after a warm spring, but the magnitude of this plastic response varies between females (19). There was a significant, positive correlation between elevation and slope (r = 0.40, LRT = 15.41, df = 1, P < 0.001): Females that lay early in the average environment are also the most plastic females.

Table 1.

Linear mixed-effects model of 2195 laying date observations from 833 female great tits that bred in more than 1 year during the period 1973 to 2004. Estimated covariance (female, female × spring temperature) = 1.16 ± 0.31 (SE).

Random effects
Term Variance SE LRT
Year of breeding 9.54 2.55 558.33View inline
Female 8.05 0.76 226.73View inline
Female × spring temperature 1.05 0.31 27.07View inline
Residual 14.97 0.64
Fixed effects
Term Wald statistic df Wald/df
Spring temperature 45.47 1 45.47View inline
Age 116.91 1 116.91View inline
Age × spring temperature 7.26 1 7.26View inline
  • View inline** P < 0.01.

  • View inline*** P < 0.001.

  • Significant genetic variation in a trait must exist for there to be any response to selection (20). We generated predictors for the two components of each female's reaction norm: her laying date in the average environment (elevation), and her change in laying date in response to temperature (plasticity or slope) (3). We used an “animal model” (21) to estimate the genetic component of phenotypic variance in predictors of female elevation and slope (18). We found that significant genetic variation for laying date plasticity exists in the Hoge Veluwe great tit population and that laying date plasticity was significantly heritable [h2 = 0.30 ± 0.14 (SE), z(<>0) = 2.21, P < 0.05 (Fig. 1A)]. Genetic variation and heritability estimates for laying date elevation were relatively high but were not significantly greater than 0 [h2 = 0.24 ± 0.14, z(<>0) = 1.73, P > 0.05 (Fig. 1B)]. However, the genetic correlation between slope and elevation was highly positive and not significantly different from 1 (rA = 0.77 ± 0.18, z(<1) = 1.28, P > 0.05).

    Fig. 1.

    Significant genetic variation exists for laying date plasticity. The bar plots show “animal model” estimates of residual and additive genetic variance (gray) and heritability (white) with SE bars for (A) laying date–spring temperature slope and (B) laying date elevation. The left y axis shows variance component values for the gray bars; the right y axis shows predicted heritabilities for the white bars. Asterisks above bars indicate an estimate that is significantly greater than 0 (*P < 0.05, ***P < 0.001).

    To investigate selection on laying date plasticity across the study period, we measured the relationship between a female's lifetime reproductive success (LRS) and predictors of her laying date elevation and slope (18). There was evidence for directional selection on both reaction norm components, and there was no evidence of stabilizing or correlated selection (table S1). Females that laid earlier in the average environment (low elevation) and responded more strongly to temperature (more negative slope) had significantly more of their offspring recruit into the population as breeding adults (standardized selection gradients for elevation and slope: –0.094 ± 0.039 and –0.085 ± 0.039, respectively).

    Over the study period, selection favoring females that advance their laying dates strongly in response to warm spring temperatures increased (Fig. 2; slope × cohort interaction: F1,804 = 7.22, P < 0.01). It is clear from both the fitted interaction (Fig. 2A) and the data themselves (Fig. 2, B to D) that selection has been strongest in the last two decades of the study, during which time the phenological mismatch with the peak in caterpillar biomass first emerged and then increased (14). The same pattern of changing selection over time is observed on estimates of females' laying date elevation (table S2) (18).

    Fig. 2.

    Selection on plasticity is increasing over time. (A) The fitted model for unstandardized LRS (“fitness”) of female great tits. “Plasticity” is the predictor of a female's laying date–spring temperature slope; the predictors are centered on zero, so negative values represent females that advance laying more strongly than average after a warm spring. The surface plot is constrained by the linear predictions of the model (table S2); the range on the plasticity axis represents the 25% to 75% quartiles of the raw data. (B to D) Plasticity predictor quartiles were estimated across the entire study period; mean LRS values for each quartile (with SE bars) are shown for females first breeding during the periods 1973 to 1982 (B), 1983 to 1992 (C), and 1993 to 2002 (D). Quartile 1 contains the most plastic females.

    It appears that the strong correlation between females' elevation and slope renders these two components of their reaction norms indistinguishable. Selection favored those highly plastic females that also lay early on average. How can we explain the correlation between elevation and slope? Both plasticity and elevation in laying date may be correlated to some unmeasured aspect of individual quality or condition (4, 12, 22). If birds differ in their ability to lay early in the year because of variation in some aspect of individual quality—for example, because of differences in their ability to gather resources—high-quality birds will lay early in warm years and later in cold years and will have steeper slopes and lower elevations, whereas poor-quality birds will usually lay later regardless of temperature and will therefore have shallower slopes. The early/plastic birds would be expected to match their reproductive timing better with the peak in caterpillar biomass, especially as spring temperatures become increasingly warm.

    To substantiate the relationship between the great tits' reaction norms and the mismatch with the peak in food availability, we estimated each female's “lifetime synchrony” (18). Improved lifetime synchrony was associated with increased laying date plasticity (F1,806 = 189.4, P < 0.001) and earlier breeding in the average environment (F1,806 = 582.4, P < 0.001). Furthermore, the synchrony of highly plastic females increased over time relative to less plastic individuals (slope × cohort interaction: F1,804 = 55.8, P < 0.001). The observed changes in selection on plasticity, as the phenological mismatch has increased over time, appear to be driven by the fact that highly plastic females breed in closer synchrony with the peak in food availability and hence have more resources available for provisioning their young.

    Female LRS has decreased across the study period (Fig. 2; cohort main effect: F1,804 = 9.92, P < 0.01). If such a decline persists alongside increased mismatching of phenologies between birds and caterpillars, the population's viability may ultimately be threatened. A phenotypic response to recent selection on laying date reaction norms cannot yet be demonstrated in this population (23), although the presence of additive genetic variance for plasticity means that a response to selection is predicted (20, 21). However, a microevolutionary response to selection on the laying date reaction norms toward lower elevations and stronger plasticity would be expected to result in closer synchrony between the great tits' laying dates and the peak in food availability, and ultimately could alleviate the trophic mismatch.

    We have shown that selection affects life history plasticity and that it can change with prevailing ecological conditions to potentially alter reaction norms in a wild population. This finding has wider implications because climate change has the potential to induce mismatches in the timing of breeding between trophic levels across a wide variety of ecosystems (17, 24, 25). The capacity for evolutionary change in phenological reaction norms shown here represents a potential means for natural selection to alleviate such mismatches and their ultimately negative consequences for population viability and ecosystem function (11, 26). However, it remains to be seen whether microevolutionary change in reaction norm shape can occur fast enough to keep up with the rapid rate of change in ecological conditions.

    Supporting Online Material

    www.sciencemag.org/cgi/content/full/310/5746/304/DC1

    Materials and Methods

    SOM Text

    Tables S1 to S3

    References

    References and Notes

    View Abstract

    Navigate This Article