Discussion
Here, we provide the first quantitative synthesis of plasticity rates (λ ) across taxa and apply a novel method to obtain standardized and thus comparable measures of this plasticity parameter. For our focal trait, acclimation of temperature tolerance to temperature change, the shape of the plasticity responses was well described by an exponential decay function (i.e. with the rate λE ). In other words, the absolute rate of change in temperature tolerance when an individual is shifted to a new temperature is proportional to the deviance from the phenotype when completely acclimated to that temperature. We thus validate an assumption that has previously been made in theory describing the evolution of phenotypic plasticity (Lande 2014). In contrast, superior fit by piecewise linear regression was primarily observed in experiments with poor temporal resolution, thus demonstrating the importance of measuring phenotypes at multiple time points in the new environment to determine the optimal model to use when estimating the shape of the plasticity response.
Variation in estimated λE among species was considerable. To put these values into perspective they can be translated into half-times, or how long it takes for the initial deviation from the fully acclimated phenotype to be reduced by 50% after being shifted to a new environment, which is given as ln(2)/λE . The mean observedλE of 0.0347 h-1 corresponds to a half-time of 20.0 h, whereas the minimum and maximum species-specificλE , estimated to 0.0009 and 0.1892 h-1, correspond to half-times of 770.2 and 3.7 h, respectively (based on data where SE λE< 0.01). Using model estimates (Table 2), half-times at 20 °C when the final slope of the decay function is zero are 12.0, 16.9, 27.7, 36.5 and 38.5 h for amphibians, reptiles, insects, crustaceans and fishes, respectively. Thus, our analyses demonstrate considerable systematic variation in rates of phenotypic plasticity among these taxa. This begs the question of why such variation has evolved. One might speculate that due to the higher heat capacity of water compared to air, fishes (all species) and crustaceans (5 out of 7 species in our data) which inhabit aquatic environments experience higher temperature stability than taxa inhabiting terrestrial environments. Furthermore, aquatic habitats are often thermally stratified (lentic habitats such as lakes and oceans) or provide cold-water plumes (rivers), which allows for behavioural thermoregulation under periods of stressful temperatures (Kurylyk et al., 2015; Freitas et al., 2016; Harrison et al., 2016). This may reduce the strength of selection on rapid plasticity compared to in the terrestrial environment occupied by reptiles and insects (all species in our data). Although the amphibians also inhabit aquatic environments (particularly during juvenile life stages), their utilization of thermal refugia in deep or fast flowing water is likely limited. Unfortunately, habitat use is confounded with phylogeny in this data set, preventing direct analysis of the effect of habitat on rate of plasticity in thermal tolerance. Thus, although we do demonstrate evolutionary divergence in plasticity rates among these taxa, it remains an open question as to whether this pattern results from evolutionary adaptation to environmental conditions. However, this question could be addressed in future work that targets populations or species that experience known and contrasting patterns of environmental variability.
We observed a positive relationship between acclimation temperature and rate of plasticity in temperature tolerance. This pattern may be explained by the general relationship that exists between developmental rate and body temperature in ectotherms, which is driven by the positive effect of temperature on biochemical reactions and metabolic rate (Brown et al., 2004). It may also explain the observation that within a species, acclimation to high temperature is achieved faster than acclimation to low temperature (Burton et al., 2020). A relationship between metabolic rate and the rate of plasticity was previously hypothesized and addressed by Rohr et al. (2018), but in a less direct manner. Specifically, Rohr et al. (2018) argued that the effect of metabolic rate on rates of thermal plasticity in ectotherms should be evident as a negative relationship between body size and plasticity rate, because smaller organisms tend to have a higher mass-specific metabolic rate than larger ones. They did not however, calculate rates of plasticity from experiments that were explicitly designed to do so. Rather, they used data from experiments that measure the phenotype at only two time points (z0 andz in our terminology), and from this inferred how the bias in acclimation capacity caused by insufficient acclimation time was influenced by body size. Based on their results it was concluded that rates of plasticity appeared to be higher for smaller organisms. Using a more direct approach we failed to find support for a general relationship between body size and rate of thermal plasticity. Yet, our observation that the rate of plasticity in temperature tolerance is positively related to acclimation temperature suggests a role for metabolic rate in causing some of the variation in plasticity rate across experiments.
Given the general patterns in rate of plasticity observed here, further efforts in studying this plasticity parameter may be fruitful and provide a better foundation for understanding how plasticity evolves in response to environmental variation. From an empirical perspective, including a temporal-dimension in experiments that study plasticity may be included without large costs. In this respect, we make two recommendations. First, a proper choice of model (linear vs. exponential decay) for estimating lambda requires multiple measurements of the phenotype as it responds to the new environment. Our analyses indicate that five or more measurements may be required to adequately establish the shape of the plasticity response (Fig. S6). Superficially, this requirement might appear to substantially increase the workload of such studies in comparison to studies that only estimate the capacity for plastic phenotypic change in a trait. However, once the model that best describes the shape of the plastic response to the new environment is established, a single measurement zt after timet (which must be prior to achievement of full acclimation) in addition to those typically measured (z0 in non-acclimated individuals and z after the full acclimation response has been obtained) is sufficient to accurately estimate Dt , which in turn can be used to calculate the rate of plasticity (λE = ln(Dt )/t for exponential decay orλL = (1-Dt )/t for linear decay). Thus, the workload in such experiments can be greatly reduced by performing a pilot experiment with sufficient temporal resolution (in terms of measurement time points) that provides a precise description of the shape of plasticity response to the new environment before performing more replicates at a lower temporal resolution to obtain the desired estimates of λ . It should be noted thatλE and λL are not directly comparable, because the initial approach towards the fully adjusted phenotype is more rapid under exponential decay. Thus, the relative support for these two types of plasticity responses should be reported. As a second recommendation, experimenters should strive to ensure that complete acclimation to the new environment is achieved prior to measuring z . Our analyses show that failing to do so can, and does, lead to bias in estimation ofλE (Fig. S2, S7). Ideally this is achieved by rearing individuals in all the alternative environments for the whole duration of the experiment (i.e. both prior to and after some of the individuals are transferred into new environments). As pointed out by Burton et al. (2022), this has rarely been done in studies of rates of plasticity. Rather, the majority of studies first acclimate the animals to a single initial environment before shifting them to a new environment and then performing repeated measures of phenotype in this new environment for what typically appears to be a pre-determined (and potentially insufficient) duration.
Natural next steps in research on evolution of plasticity would be to test for links between environmental variation and the evolution of rates of plasticity, and to provide theoretical models that address the co-evolution of plasticity rates and capacity (see Introduction). Although this is beyond the scope of the current paper, our work provides both methodology and novel insights that should stimulate future work along these lines. We also re-emphasize a point made previously (Burton et al., 2022) - that selection on the rate of plasticity might be stronger than selection on the capacity for plasticity. Evolutionary theory posits a central role for phenotypic plasticity in mitigating the fitness impact of environmental variation, but that possessing the potential for such a response is associated with a fitness cost in stable environments (Lande, 2009). Fitness costs of plasticity can be categorized into costs of maintenance and costs of production. Costs of maintenance represent the investment of resources into maintaining the machinery required for detecting and responding to a change in the environment and will be paid at a constant rate independent of environmental conditions (Auld et al., 2010). In contrast, production costs are only paid when the plastic response is triggered and are compensated by the fitness benefits associated with changing the phenotype. If one assumes that the capacity for plasticity can be increased by operating the ‘machinery’ required to change a trait for a longer duration, this will increase production costs but not maintenance costs. Populations living in less variable environments may therefore pay a small price for maintaining their capacity for plasticity (as shown by Van Buskirk & Steiner, 2009), and adaptation of this parameter of plasticity to levels of environmental fluctuations may therefore be relatively modest in magnitude. In contrast, increasing the rate of change in the same trait would require increasing the size or output of that ‘machinery’, with corresponding increases in maintenance costs. Populations living in less variable environments should therefore experience strong selection against maintaining rapid plasticity due to higher maintenance costs, and adaptative evolution across populations may then be expected to be more pronounced for the rate of plasticity. This line of reasoning is also consistent with theoretical results showing that maintenance costs shape the evolution of plasticity to a greater extent than production costs (Sultan & Spencer, 2002). Given these considerations, and the results presented in the current study, it seems prudent to address the hypothesis that adaptation to environmental variation may be more pronounced in terms of rates of plasticity rather than capacity of plasticity. By providing clear evidence that rates of plasticity have diverged among ectotherm classes we show that it is a trait that evolves, and that increased empirical and theoretical focus on the rate parameter is likely to provide a way forward for a more comprehensive understanding of phenotypic plasticity.