Plasticity and genetic adaptation mediate amphibian and reptile responses to climate change

1Climate change is forcing species to pursue their climatic niche as it shifts across landscapes (Chen et al. ). Some species will not track their optimal climates because they disperse poorly or they cannot disperse across natural or human barriers (Loarie et al.; Schloss et al.; Urban et al. ). These species face the threat of high extinction rates (Thomas et al.; Malcolm et al.; Schloss et al. ). However, phenotypic plasticity and genetic adaptation could allow these species to persist in their current locations despite climate change. Here, we review studies that examine evidence for plastic or genetic changes in amphibians and reptiles in response to climate variation.

2Amphibians and reptiles might be particularly sensitive to climate change. Both taxa have already experienced extensive declines and extinctions worldwide (Gibbons et al.; Houlahan et al.; Blaustein and Kiesecker; Sinervo et al. ). Both groups have been called ‘canaries in the coal mine’ based on their supposed sensitivity to environmental change (Kerby et al.; Mitchell and Janzen ). A global assessment further suggests that amphibians are declining faster than birds or mammals (Stuart et al. ). Concurrently, Huey et al. call lizards ‘the new amphibians’ because of their elevated extinction risk (Gibbons et al.; Sinervo et al. ). Climate change has contributed to these threats (Houlahan et al.; Carey and Alexander; Stuart et al. ), and correlative climate envelope models predict that climate change will cause extinction in 12–47% of endemic frogs and 11–49% of endemic reptiles (Thomas et al. ). Sinervo et al. predicted that 20% of lizard species will become extinct by 2080, assuming a limited potential for evolutionary responses. Overall, both amphibians and reptiles are declining, and more tenuous analyses link their declines to climate change. Climate change will likely decrease fitness substantially in some populations and thus could generate strong selection on climate-related traits. However, considerations of these threats rarely account for the mitigating effects of phenotypic plasticity and evolutionary adaptations.

3Adaptive evolution and plasticity offer two ways species can sustain fitness despite climate change (Holt; Urban et al. ). Biologists continue to debate the importance of these mechanisms in mediating climate change impacts (Gienapp et al.; Hoffmann and Sgro ). Merilä and Hendry argue in this issue that neither adaptive evolution nor plasticity should be considered the null hypothesis for climate-related phenotypic change. We agree and further suggest that we should condition our predictions about their relative importance based on available empirical knowledge. Toward this end, we evaluate the degree to which amphibians and reptiles respond to climate variation through evolved genetic differences and environment-induced plasticity. We review studies that evaluate evolutionary or plastic response to spatial or temporal variation in climate and perform a meta-analysis on the proportion of traits showing climate-related differences. We also assess if those changes are adaptive and how accurately phenotypic differences map to climate change versus other correlated environmental variables.

4We now briefly outline the current understanding of evolution and plasticity with respect to climate responses in reptiles and amphibians before conducting our literature review and analysis. Plasticity occurs often in amphibians and reptiles – to such a degree that we expect plasticity in almost every trait we measure, ranging from behavior to morphology to life-history traits (Relyea; Booth; Urban ). Phenological shifts are especially common in amphibians (Beebee; Gibbs and Briesch; Todd et al. ), where changes in breeding phenology are two to four times stronger than responses in other taxonomic groups (Parmesan ). For instance, the dwarf salamander Eurycea quadridigitata migrates to breeding ponds 76 days later than it did 25 years ago (Todd et al. ). These phenological changes likely involve plasticity. However, the evolutionary contribution has never been directly measured, making it difficult for one to conclude about its importance. Lizards are expected to respond to climate change by altering sex ratios, habitat choice, and hatchling traits (Booth ). Scientists have argued both for (Gvozdik ) and against (Huey et al. ) the importance of plastic acclimation to warmer temperatures in buffering climate change effects in lizards. Huey et al. suggested that few tropical forest ectotherms demonstrate acclimation. In contrast, Gvozdik argued that plastic acclimation to higher temperatures might be common, but experimental approaches used to estimate plasticity often are conducted over extremely limited time spans or ignore additional sources of plasticity.

5Can relatively long-lived amphibians and reptiles genetically adapt to climate change? One study parameterized a model of evolutionary responses to climate change based on estimates of phenotypic variation in critical thermal maxima and average evolutionary rates in amphibians and reptiles (Skelly et al. ). The authors predicted that critical thermal maxima could evolve 3.2°C in 50 years – fast enough to track climate changes. However, this estimate ignores potential limits to additive genetic variation at extreme temperatures or selection that is so strong that populations become extirpated before adaptation.

6In contrast to amphibians, scientists have argued that reptiles are less likely to evolve temperature-related traits. This more pessimistic view stems from reptiles’ longer generation times relative to rapid climate change, low heritabilities, and genetic trade-offs among related traits (Janzen; Schwanz and Janzen; Sinervo et al. ). Sinervo et al. estimated that recently extirpated lizard populations had faced a standardized selection intensity of 0.34 on average. This strong selection intensity might be sustained under controlled conditions in a laboratory, but might often lead to extirpation in wild populations. However, empirical evidence remains limited in scope. For example, the evidence for low heritability in traits related to lizard persistence comes from a single population evaluated in the laboratory (Sinervo et al. ), and yet we know that heritabilities vary across species, populations, and environments. Also, genetic variation must be high enough to allow adaptation to rapid climate change. Genetic variation exists within and among populations for temperature-dependent sex determination in reptiles, but modeling studies suggest that it might not be sufficient to counter predicted levels of climate change (Morjan; Mitchell and Janzen ). Hence, previous work suggests divergent views on the ability of amphibians and reptiles to adapt in response to climate change, setting the stage for a more synthetic analysis of the literature.

7To fill the gap in our understanding of evolutionary and plastic responses to climate variation, we searched Web of Science for articles using the following search terms: ‘amphibia*/reptile*’ AND (‘climate’ OR ‘global warming’) AND (‘plastic*’ OR ‘evol*’ OR ‘adapt*’ OR ‘selection’ OR ‘reaction norm’ OR ‘genotype by environment’ OR ‘GxE’ OR ‘phenotyp*’). We further searched literature citations of these articles for additional references. We read each article and only considered studies that presented empirical evidence for phenotypic or genotypic differences associated with climate change (e.g., temperature, precipitation). For each study, we recorded the climate change factor, phenotypic response (e.g., laying date, size, survival), and evaluated evidence for evolved or plastic differences for each species or population. We found 72 estimates recorded in 30 studies that estimated climate-related trait variation in 50 amphibian species and 29 estimates recorded in 24 studies conducted on 18 reptile species (Table 1).

8We considered studies that evaluated phenotypic or genotypic differences in response to temporal (allochronic studies) and spatial variation (synchronic studies) in climate. Allochronic studies evaluate phenotypic responses to climate change over time and provide the strongest inferences (Hendry and Kinnison; Merilä and Hendry ). The alternative to the allochronic study is the synchronic one, which documents phenotypic variation in space and requires that space can be substituted for time. The space-for-time substitution offers weaker inferences especially for genetic change. For instance, a population that adapted to climate variation in the past might not adapt to future climate change if genetic variation is lacking. Substituting space for time likely provides more robust predictions for phenotypic plasticity; plasticity that is induced by spatial climate variation is also likely to be induced by climate variation through time.

9We only uncovered one study of allochronic genetic change in response to climate change in amphibians or reptiles and this one did so indirectly (Phillimore et al. ). Another study evaluated genetic studies through time, but not necessarily on the same populations (Gibbs and Karraker ). Two other studies evaluated spatial differences among populations that had experienced divergence in temperature over time and thus represents a more robust case for applying a space-for-time substitution (Skelly and Freidenburg; Freidenburg and Skelly ). Sixty-three percent of amphibian estimates and 38% of reptile estimates were based on variation across space rather than time. To understand if these disparate designs affected overall outcomes, we tested for an effect of study design.

10We performed a meta-analysis on the proportion of traits that contributed significantly to genetic adaptation, plasticity, and genotype-by-environment interactions. We logit-transformed proportions and weighted each study by the inverse expected variance or np/(1–p) where n is the number of traits analyzed in each study and p is the proportion of traits showing a significant difference (Lipsey and Wilson ). We added 0.5 to studies with zero proportions in order to include them (Veichtbauer ). We counted the responses of different species and different populations reported in the same study as separate estimates. However, we included study, species, and populations as random effects to control for their nonindependence in analyses.

11We conducted analyses in R (v. 3.0.0) and used the lmer function to conduct mixed-effects models. The significance of random effects was assessed using a likelihood ratio test (Pinheiro and Bates ). We also tested for effects of study design (allochronic versus synchronic) and used the pvals function in R to calculate significance values.

12We used a word cloud algorithm (http://www.wordle.net) to visualize the frequency with which researchers measured different traits. This algorithm sets word frequencies proportional to font sizes. We arranged traits alphabetically from left to right. We included all traits analyzed across all studies while also consolidating traits into more general categories (e.g., ‘breeding date’ represents peak calling date, first date of arrival, and observation of egg deposition).

13Most studies on amphibians focused on phenological changes in breeding (Table 1; Fig. 1A; 31% of traits analyzed). Research also focused on amphibian life-history traits related to larval growth (14%), survival (9%), and embryonic development (8%). We found fewer measures of direct physiological effects of temperature (Skelly and Freidenburg ), resistance to climate-related diseases (Chatfield and Richards-Zawacki; Forrest and Schlaepfer ), or adult traits other than breeding phenology (Tryjanowski et al.; Reading; Alho et al. ), despite the potential importance of these traits in mediating climate threats (Carey and Alexander; Corn; Pounds et al. ). Reptile studies focused on a wider range of traits including adult and offspring size (19%), sex determination threshold and sex ratios (15%), nesting site choice and depth (9%), incubation period (8%), and survival (8%; Fig. 1B).

14The inverse-variance weighted proportion of plastic responses to climate variation in amphibians was 0.71 (95% confidence interval: 0.60, 0.80; Fig. 2A) based on 69 estimates. The random effects of species, study, and population did not significantly explain additional variation. The unweighted model indicated a similar proportion (= 0.70). Study design (allochronic versus synchronic) did not significantly alter the estimated proportion of plastic responses to climate variation (P = 0.220).

15The inverse-variance weighted proportion of plastic responses to climate variation in reptiles was 0.69 (0.53, 0.81; Fig. 2A), based on 26 estimates. We detected a significant random effect of species ( = 6.2, P = 0.012) but not of study or population. The unweighted model indicated a similar proportion (= 0.70). Whether the study design was allochronic or synchronic did not significantly alter the estimated proportion of plastic responses to climate variation (P = 0.722).

16Many of the collected studies evaluated how amphibians and reptiles adjusted their breeding phenology in response to warmer climates (Beebee; Blaustein et al.; Schwanz and Janzen; Telemeco et al.; Todd et al. ). In general, the phenological responses of amphibian breeding are among the largest observed – estimated to be two to four times stronger than those in other taxonomic groups (Parmesan ). Thus, the plastic response to climate change is well established and strong in amphibians. Generally, these rapid changes in phenology are assumed to occur too quickly to entail evolution. However, the genetic component of this response has never been evaluated in amphibians or reptiles in common garden experiments. One study estimated no additive genetic variance for breeding phenology in painted turtles, but moderate genetic variation emerged following a warm winter (VA = 0.13; evolvability CVA of 0.23; McGaugh et al. ). Another study evaluated variation in temperature-dependent breeding dates along a latitudinal gradient and used observed differences in reaction norms to argue for the evolution of breeding phenology (Phillimore et al. ). Although the differences in reaction norms are compelling, other uncontrolled environmental factors could explain these differences. Applying complicated statistics to observational data cannot replace the simple elegance of the common garden experiment. Some of these phenological changes over longer time periods might involve genetic changes, but we lack clear evidence to date. Introducing a population from another climate region into a common environment and evaluating differences in phenology between introduced and native populations under the same climate cues would indicate if these responses diverge genetically across landscapes. Determining if they evolve in response to climate change is more difficult, requiring researchers to identify candidate genes, parameterize genetic models, or experimentally estimate trait changes through time.

17Temperature-dependent sex determination presents another important case of plasticity with strong links to climate (Janzen; Mitchell and Janzen ). Warmer climates generally bias sex ratios in reptiles toward females. Some populations could face demographic collapse if high temperatures preclude the development of one sex (Janzen ). Yet, multiple studies suggest that reptiles can respond behaviorally through nesting site choice, depth, and breeding time (Doody et al.; Doody; Refsnider and Janzen ). These behaviors could mediate the effect of hotter temperatures on sex ratios. For instance, water dragons (Physignathus lesueurii) maintain similar incubation temperatures despite changing environmental temperatures by digging shallower nests and selecting more open nest sites in colder regions (Doody et al.; Doody ). In contrast, changes in breeding time and nest depth did not buffer an observed 1.6°C warming period for the lizard Bassiana duperreyi. These lizards nested 4 weeks earlier and buried eggs 15 mm deeper, but eggs still warmed 1.5°C, which threatened their normal sex ratios (Telemeco et al. ). Behaviors might moderate changes in sex ratios, but not always to the degree necessary to mitigate climate change.

18Different thermal regimes were associated with plastic changes in life-history traits such as clutch size, development rate, growth rate, and survival in amphibians (Voss; Galloy and Denoel; Orizaola et al.; Niehaus et al. ). Some of these changes were dramatic. For instance, half as many eggs were laid by palmate newts (Lissotriton helveticus) when raised at 22 vs 18°C (Galloy and Denoel ). Experimentally induced shorter hydroperiods altered survival, development, and growth in multiple European frogs and toads (Richter-Boix et al. ). In contrast, simulated drying did not induce plastic changes in boreal chorus frogs (Pseudacris maculata; Amburgey et al. ).

19Results suggest that plasticity will generally buffer some of the effects of climate change. However, not all traits are plastic, and sometimes plasticity might prove insufficient by itself to buffer climate change.

20Our sample size was small and only one study explicitly evaluated genetic changes over time in response to climate change. However, all 17 studies that evaluated genetic differences found support for adaptive genetic differences in at least one trait. The estimated weighted proportion of genetic responses to climate variation in amphibians was 0.65 (95% confidence interval: 0.46, 0.81), and the proportion in reptiles was 0.66 (0.43, 0.83; Fig. 2A). The unweighted and weighted estimates did not differ substantially from each other, and random effects were not significant for amphibians and reptiles.

21In one amphibian example of a genetic response to climate variation, wood frogs (Rana sylvatica) evolved different critical thermal maxima and thermal preferences between warm and cold ponds that had diverged in temperature over 36 years (Skelly and Freidenburg; Freidenburg and Skelly ). In another case, warmer climates led to increasing proportions of dark morphed red-backed salamanders (Plethodon cinereus; Gibbs and Karraker ). Pool frog tadpoles (Rana lessonae) diverged in temperature-dependent developmental and growth rates across habitats that differed in temperature (Orizaola et al. ). For reptiles, snapping turtles (Chelydra serpentina) from different latitudes differed in their temperature-dependent sex determination patterns when raised in the laboratory (Ewert et al. ), and sex ratios differed among families for alligators and two turtle species (Rhen and Lang ). The overall magnitude of these changes was often not very large. For instance, the study on the evolution of critical thermal maxima in wood frog tadpoles found a difference of 0.4°C between populations inhabiting warm and those inhabiting cold ponds (Skelly and Freidenburg ). Similarly, the occurrence of striped red-backed salamander morphotypes decreased by only 6% over the last century (Gibbs and Karraker ).

22We only found six amphibian and three reptile estimates of genotype-by-environment interactions in response to climate variation. The reptile study (Rhen and Lang ) found evidence for genotype-by-environment interactions in two of the three species they examined. For amphibians, we had enough studies to apply a weighted meta-analysis. The estimated proportion of genotype-by-environment responses to climate variation was 0.48 (95% confidence interval: 0.13, 0.85; Fig. 2A). The unweighted estimate was similar (= 0.47), and random effects were not significant. One genotype-by-climate interaction occurred when R. sylvatica tadpoles varied in hatching depending both on if they came from warm or cold ponds and their exposure to cool or warm environmental conditions (Skelly and Freidenburg ). In R. lessonae tadpoles, larval growth rate and time to metamorphosis were accelerated the most in warm conditions for the population originating from the coldest environment, indicating countergradient variation (Conover and Schultz; Orizaola et al. ).

23Other research suggests the potential for genetic variation to fuel the evolution of sex determination thresholds in reptiles in response to warmer climates. A review of reptiles with temperature-dependent sex determination reports heritabilities ranging from 0.26 to 0.82 (McGaugh and Janzen ). Morjan parameterized a model to predict that threshold temperatures might evolve more quickly than nest site choice, but not rapidly enough to counter predicted climate change. Morjan finds even slower rates of evolution of nest site choice owing to low heritability and high maternal contributions. Heritability of nest site choice in painted turtles (Chrysemys picta) ranged from 0.06 to 0.7 (McGaugh et al. ). Additive genetic variance increased in this trait following warm winters (McGaugh et al. ), which demonstrates that extreme future climates might reveal additive genetic variation to fuel adaptation exactly when it is needed most. However, in other cases, limits to additive genetic variation might exist that would prevent evolution in response to more extreme climate variation. More generally, we highlight the many opportunities for researchers to perform novel experiments that test for the genetic basis of climate-related traits in amphibians and reptiles.

24On a cautionary note, few studies controlled for transgenerational effects, which could affect responses to climate change. Transgenerational effects occur when nongenetic contributions originate from previous generations (e.g., maternal effects). In addition, all our evidence for evolution arose from genetic differences measured across populations rather than through time. This space-for-time substitution is criticized as much as it is applied, but persists because of limited data. The problem is that past adaptive divergence among populations need not indicate the potential for future adaptation, especially at the rates required by climate change. Spatial differences might have evolved over centuries, whereas climate change might require evolution over decades. Also, strong natural selection could produce demographic collapse before population-level fitness can recover (Gomulkiewicz and Holt; Sinervo et al. ).

25On a more positive note, the studies on R. sylvatica tadpoles revealed genetic differences between ponds that had undergone known temporal changes in temperature over just a few decades (Skelly and Freidenburg; Freidenburg and Skelly ), suggesting that space can substitute for time when we have information on recent environmental changes across the landscape. At least some evolutionary rates might be sufficient to meet the demands of the changing climate.

26We next assessed if trait differentiation was adaptive or maladaptive based on authors’ conclusions and logic. In the latter case, we assumed a trait was maladaptive if it reduced fecundity or survival. Only a few studies examining genetic variation also provided information that could be used to assess if changes were adaptive. Of those studies, all six amphibian studies and the one reptile study indicated an adaptive origin for genetic changes.

27Climate change factors generally induced adaptive plasticity (Fig. 2B). The weighted proportion of plastic responses to climate variation in amphibians that was considered adaptive was 0.67 (95% confidence interval: 0.51, 0.79). The random effect of species was not significant (P = 0.813), but study and population significantly explained additional variation ( = 26.1, 8.2, respectively; P < 0.005). The unweighted model indicated the same proportion. The proportion of plastic responses to climate variation in reptiles that was considered adaptive was 0.62 (95% confidence intervals: 0.36, 0.82). The random effects of species, study, and population did not significantly explain additional variation. The unweighted model indicated the same proportion. For both amphibians and reptiles, allochronic studies did not show significantly different proportions of adaptive variation when compared against synchronic studies (P > 0.6).

28Although most observed plasticity was assumed to be adaptive, we also detected substantial maladaptive plasticity (Fig. 2B). Hotter temperatures enhanced learning performance in the lizard B. duperreyi (Amiel and Shine ) and accelerated development and increased juvenile body condition in lacertid lizards (Monasterio et al. ). However, higher incubation temperatures produced lower hatching success, slower righting responses, poorer swim performance, and skewed sex ratios in other reptiles (Janzen; Wapstra et al.; Mickelson and Downie; Micheli-Campbell et al.; Bell et al. ). Despite the near-universality of plastic responses to climate variation, some plasticity will ultimately prove maladaptive under altered climates. Understanding the fitness consequences of climate-induced plasticity requires researchers to delineate plastic responses in experiments, create artificial phenotypes independent of genotype (Waddington ), and then track their long-term fitness in experimental or natural situations.

29We also evaluated the degree to which studies evaluated the causal linkage between climate change and each phenotypic response. Although multiple methods exist (Merilä and Hendry ), our studies assessed causality using common sense, phenotype–environment correlations, or experimental manipulations. Climate was suggested to be involved in phenotypic changes in all the traits analyzed (Table 1). However, these conclusions were often based on common sense rather than rigorous tests.

30Given limited information on adaptive responses to climate change, we argue that we need to intensify our efforts to explore how plasticity and genetic adaptation will attenuate the future impacts of climate change. These mitigating factors might not provide a panacea, but we should estimate their contributions and include them in any rigorous assessment of climate change effects on species. Most of the examples we found required substituting space for time and thus constituted weaker inferences. Quantitative genetic models might be used to predict future phenotypes, but they assume that we can accurately estimate the heritabilities of traits and strength of selection in the wild (Endler ), which is often not the case (Hoffmann and Merilä; Charmantier and Garant ). Genetic correlations can constrain evolutionary responses to climate change (Etterson and Shaw ). Moreover, attempts to predict evolution in the wild have frequently failed. These shortcomings are due to (i) poor estimates of heritability or selection in the wild, (ii) the ability of selection to act on nonheritable phenotypic variation, (iii) the fact that traits are genetically correlated, (iv) rapidly shifting environments, or (v) because of a simple lack of statistical power (Merilä et al. ). All of these reasons cast doubt on our ability to extrapolate current evolutionary responses far into the future.

31Other options include using analogous changes in landscapes as natural experiments (Skelly and Freidenburg ) or performing common garden experiments through time, where relevant phenotypic responses are evaluated in the same populations and using the same methods across long enough periods to capture climate change. These rigorous experiments would reveal evolution in real time, highlight if evolution is keeping up with changing climate, and provide insights into the limits of additive genetic variation to fuel evolutionary responses in the near future. Experimental evolution in the wild provides another compelling option. Researchers could design experiments that manipulate climate factors directly and evaluate evolutionary changes in focal populations or transplant populations into selection regimes that mimic future climates and then track their evolution. However, the lack of allochronic studies in amphibians and reptiles is clearly related to the fact that these species often have long generation times that prevent such comparisons being made over practical time spans. Lastly, a candidate genes approach offers many advantages if sufficient genomic and trait-gene mapping is available or can be accomplished.

32Many amphibians and reptiles have already undergone climate-related extinctions or declines (Gibbons et al.; Parmesan and Yohe; Pounds et al.; Sinervo et al. ), and others are approaching the edge of their thermal tolerance (Huey et al. ). Models predict future declines in amphibians and reptiles owing to climate change (Thomas et al.; Malcolm et al. ). However, these correlative models rarely account for evolutionary change. Strong plasticity or high genetic variance might prove sufficient to generate phenotypic changes that match the demands of a new climate for some species. Alternatively, relatively long generations and limited genetic variation could reduce their capacity for adaptive change after plastic reserves have been exhausted (Morjan; Sinervo et al. ).

33Here, we show that both adaptation and plasticity could play important roles in mitigating climate change for amphibians and reptiles, but data is still limited, especially for genetic responses. Our literature review suggests that plasticity might often mediate changes to climate in both taxonomic groups. A long history of research has focused on plasticity in these species, and we show here that changing climates might often induce adaptive plasticity. Although evidence for genetic adaptations to climate was limited to a handful of mostly synchronic studies, all of these studies found evidence for climate adaptation in at least one trait. Similar assessments of other taxonomic groups in this issue provide evidence for widespread plastic and adaptive responses to climate variation and concurrently highlight the same lack of data for evolutionary responses (Franks et al.; Reusch; Schilthuizen and Kellermann ). If future studies confirm these results, then plasticity and adaptive capacity could dampen some of the doom and gloom associated with research on biotic responses to climate change.