Key Questions on Phenotypic Plasticity in Eco-Evolutionary Dynamics PDF

Journal of Heredity, 2016, 25–41 doi:10.1093/jhered/esv060 Symposium Arti...

Journal of Heredity, 2016, 25–41 doi:10.1093/jhered/esv060 Symposium Article Advance Access publication August 22, 2015 Symposium Article Key Questions on the Role of Phenotypic Plasticity in Eco-Evolutionary Dynamics Andrew P. Hendry Downloaded from https://academic.oup.com/jhered/article/107/1/25/2622828 by guest on 08 November 2021 From the Redpath Museum & Department of Biology, 859 Sherbrooke St. W., Montreal, Quebec H3A OC4, Canada. Address correspondence to Andrew P. Hendry at the address above, or e-mail: [email protected]. Received February 28, 2015; First decision March 26, 2015; Accepted July 16, 2015. Corresponding Editor: Robin Waples Abstract Ecology and evolution have long been recognized as reciprocally influencing each other, with recent research emphasizing how such interactions can occur even on very short (contemporary) time scales. Given that these interactions are mediated by organismal phenotypes, they can be variously shaped by genetic variation, phenotypic plasticity, or both. I here address 8 key questions relevant to the role of plasticity in eco-evolutionary dynamics. Focusing on empirical evidence, especially from natural populations, I offer the following conclusions. 1) Plasticity is—not surprisingly—sometimes adaptive, sometimes maladaptive, and sometimes neutral. 2) Plasticity has costs and limits but these constraints are highly variable, often weak, and hard to detect. 3) Variable environments favor the evolution of increased trait plasticity, which can then buffer fitness/performance (i.e., tolerance). 4) Plasticity sometimes aids colonization of new environments (Baldwin Effect) and responses to in situ environmental change. However, plastic responses are not always necessary or sufficient in these contexts. 5) Plasticity will sometimes promote and sometimes constrain genetic evolution. 6) Plasticity will sometimes help and sometimes hinder ecological speciation but, at present, empirical tests are limited. 7) Plasticity can show considerable evolutionary change in contemporary time, although the rates of this reaction norm evolution are highly variable among taxa and traits. 8) Plasticity appears to have considerable influences on ecological dynamics at the community and ecosystem levels, although many more studies are needed. In summary, plasticity needs to be an integral part of any conceptual framework and empirical investigation of eco-evolutionary dynamics. Subject areas: Quantitative genetics and Mendelian inheritance Keywords: adaptation, adaptive divergence, community structure, contemporary evolution, ecosystem function, heritability, population dynamics, rapid evolution. Eco-evolutionary dynamics is an emerging research field that con- and Ghalambor 2001; Hendry et al. 2008). In the other direction, siders interactions between ecology and evolution as they play out evolutionary change on those time scales can have important eco- on contemporary time frames. These interactions can take place in logical consequences at the population, community, and ecosystem either direction (Figure 1). In one direction, ecological change can levels (Thompson 1998; Hairston et al. 2005; Fussmann et al. 2007; cause evolutionary change on time frames ranging from only a few Kinnison and Hairston 2007; Pelletier et al. 2009; Schoener 2011). to hundreds of generations (Hendry and Kinnison 1999; Reznick Given that these effects flow in both directions (eco-to-evo and © The American Genetic Association. 2015. All rights reserved. For permissions, please e-mail: [email protected] 25 26 Journal of Heredity, 2016, Vol. 107, No. 1 on genotypes, and so influences evolutionary responses to ecological change and ecological responses to evolutionary change. In short, plasticity needs to be an integral part of any general framework for eco-evolutionary dynamics—and the present article is designed to make steps in that direction. Plasticity has been the focus of considerable interest in recent dec- ades, partly inspired by several key books (Schlichting and Pigliucci 1998; West Eberhard 2003). I obviously cannot cover all of the nuances in a single (admittedly long) article, and so I instead focus on key aspects of plasticity that are necessary for exploring eco-evolu- tionary dynamics. I specifically address 8 key questions that are much discussed in the literature, attempting to answer them by reference to empirical data—especially from natural (as opposed to laboratory) populations. This article is intended as a complement to a previously published companion article on “Key questions in the genetics and Downloaded from https://academic.oup.com/jhered/article/107/1/25/2622828 by guest on 08 November 2021 genomics of eco-evolutionary dynamics” (Hendry 2013). How to Infer Plasticity By far the most common approach to studying phenotypic plasticity is to implement experimental manipulations under otherwise con- trolled conditions. Such experiments yield information on the phe- notype produced by a given genotype under different conditions, a Figure 1. Conceptual diagram outlining the basic elements of eco- relationship called the “reaction norm” (Scheiner 1993; Schlichting evolutionary dynamics. Phenotypic traits in a focal species can influence the and Pigliucci 1998). Often, the goal is to compare plasticity between population dynamics of that species, which can then influence the structure of different groups, such as different time periods or species or popula- the community in which that species is embedded, as well as the functioning tions, which can be accomplished by comparing slopes and eleva- of the overall ecosystem. In addition, phenotypic traits in the focal species tions (or other features when nonlinear) of their reaction norms. can directly (i.e., not through population dynamics) influence community Such experiments work most elegantly for species where inbred structure and ecosystem function. Ecological effects at the population, community, and ecosystems levels can then feedback through plasticity or lines or clones can be generated because a single genotype can be selection to influence phenotypic traits. These phenotypic changes will be examined in multiple environments. However, this approach is not passed on to the next generation to the extent that they are heritable. This possible for many other organisms, where different “genotypes” must figure and caption are the same as in the previously-published companion instead be represented by different full-sibling families (i.e., different paper “Key questions in the genetics and genomics of eco-evolutionary individuals from a given family are split between different condi- dynamics” (Hendry 2013). tions) or by unrelated individuals randomly sampled from the differ- ent groups (usually different populations). The assumption in these evo-to-eco), feedback loops can emerge, and these loops can rein- cases is that substantial genetic differences are not present between force or dampen the ecological or evolutionary changes (Post and the individuals from a given family/population exposed to the dif- Palkovacs 2009). ferent conditions. If this assumption is correct, the resulting family/ Eco-evolutionary dynamics are driven by interactions between population-level reaction norm should be representative of the aver- the environment and organismal phenotypes. It is typically assumed age reaction norm of genotypes from those families/populations. that these phenotypes have a genetic basis, which has often (but not Once average reaction norms are estimated for different groups always) been established for the eco-to-evo pathway. By contrast, a (e.g., clones, populations, or species), a number of outcomes might genetic basis for evo-to-eco effects has been confirmed only rarely emerge—here conceptualized in linear form (the following alterna- (Hendry 2013), with a logical alternative being phenotypic plasticity tives correspond to Panels A–F in Figure 1 of Morbey and Hendry (encompassing developmental plasticity, environmental induction, 2008). First, environmental conditions might not have a plastic acclimation, inducible defenses, maternal effects, and epigenetics). effect on the trait (flat reaction norms) and the groups might not In fact, plasticity is expected to be very important in shaping both differ genetically (identical reaction norms). Second, environmen- phenotypic change in response to ecological change and ecologi- tal conditions might have a plastic effect on the trait (nonflat reac- cal change in response to phenotypic change (see also Miner et al. tion norms), but the groups might not differ genetically (identical 2015). These influences need to be incorporated into the developing reaction norms). Third, environmental conditions might not have framework of eco-evolutionary dynamics for reasons that I briefly a plastic effect (flat reaction norms), but the 2 groups might dif- introduce here and detail later. First, the current level of plasticity fer genetically in trait expression (different elevations). Fourth, in a population typically will have evolved as a result of past selec- environmental conditions might have a plastic effect for which the tion (as opposed to drift), and so plastic changes expressed by indi- groups differ genetically (different slopes), with the genetic and viduals have a genetic basis and are often adaptive. In such cases, plastic influences reinforcing each other (i.e., cogradient: sensu one can think of plastic changes as a contemporary manifestation Conover and Schlutz 1995). Fifth, environmental conditions might of historical genetic change. Second, plasticity can evolve on con- have a plastic effect that differs in direction between the 2 groups temporary time scales, and so phenotypic changes in a population (slopes differ in sign). Sixth, environmental conditions might have a might reflect evolving plasticity. Third, plasticity modifies selection plastic effect for which the groups differ genetically, with the genetic Journal of Heredity, 2016, Vol. 107, No. 1 27 and plastic influences this time opposing each other (i.e., counter- thus leaving plasticity as the default explanation (Merilä and Hendry gradient: sensu Conover and Schlutz 1995). 2014). First, groups that differ phenotypically in nature can be raised A number of considerations attend reaction norms and their under common-garden conditions to see if those differences vanish, estimation (Schlichting and Pigliucci 1998). First, although con- which thus implies plasticity, with a classic example being James cepts are most easily envisioned for only 2 environments, reaction (1983). Second, estimates of selection and genetic variation can be norms can be quantified for any number of environments and for used in the breeder’s equation or the Robertson–Price Identity to continuous environmental variables. Second, although reaction predict the likely contribution of evolution, leaving plasticity as the norms are typically represented as linear functions, they can take explanation for any change not explained thereby (Crozier et al. any shape. Third, reaction norms can be quantified for phenotypes 2011). Third, animal model analyses can measure genetic change of any sort. Traditional organismal phenotypes include behavior, based on breeding values and thus, corresponding to the remainder, physiology, color, morphology, life history, and various fitness met- any plastic contributions (Merilä et al. 2001). Fourth, the Price equa- rics; but reaction norms also can be evaluated for variables such tion can be used for post hoc partitioning of phenotypic changes into as gene expression (Swindell et al. 2007; McCairns and Bernatchez those due to selection versus various forms of plasticity (Ellner et al. 2010) or protein expression (Tomanek 2008; Martínez-Fernández 2011). Each of these approaches can be informative but each is also et al. 2010). In addition, phenotypes can be continuous functions of attended by inferential caveats (details in Merilä and Hendry 2014). Downloaded from https://academic.oup.com/jhered/article/107/1/25/2622828 by guest on 08 November 2021 time, such as growth curves, which have been called “function-val- ued traits” (Kingsolver et al. 2001; Stinchcombe et al. 2012). Fourth, additive genetic (co)variances and heritabilities can be calculated for Evidence From Nature the slopes and elevations of reaction norms (or other parameters I now address key questions surrounding phenotypic plasticity and its for nonlinear functions), just as for trait values in a single environ- role in eco-evolutionary dynamics from the perspective of both causes ment. Moreover, one can evaluate plasticity as the genetic variance of the trait in each environment along with the genetic covariance and consequences. With respect to the causes of plasticity, we need to know how the ecological environment shapes the evolution of plastic- between environments, which also works for function-valued traits ity (eco-to-evo-to-pheno). Thus, I will first revisit several key questions (Via and Lande 1985; Via et al. 1995). Fifth, it is sometimes possible that are commonly considered in the literature. 1) To what extent is to study the specific genes or gene regions underlying some aspects plasticity adaptive (i.e., increases fitness), as opposed to a nonadaptive of plasticity (Gutteling et al. 2007), with prime examples being heat or maladaptive response to (for example) stressful conditions? 2) To shock proteins (Rohner et al. 2013) and DNA methylation patterns what extent does plasticity have limits or costs, which will influence (Herrera and Bazaga 2010). Sixth, reaction norms induced by a par- selection and potential responses to environmental change? 3) What ticular environmental variable likely depend on levels of other envi- environmental and organismal characteristics favor the evolution of ronmental variables (G × E becomes G × E × E), and so results will plasticity? From the perspective of consequence, we need to know be context-dependent. how plasticity shapes the evolutionary dynamics of populations and The above approaches typically involve experimental manipu- their ecological effects (pheno-to-evo-to-eco). Some of the questions lations of the environment, but plasticity also can be studied in here are classic whereas others are rather new. 4) To what extent does an observational approach: individuals/populations in nature can plasticity aid colonization and responses to environmental change—a be followed through time to quantify the relationship between demographic consequence? 5) Does plasticity constrain or promote environmental conditions and trait expression. At the individual genetic evolution and 6) ecological speciation? 7) How fast can plas- level, this approach relies on the same individuals experiencing, ticity evolve, which is particularly germane given the focus of eco-evo- and responding to, different environments at different episodes in lutionary dynamics on short time scales. Finally, I address an emerging their life (Nussey et al. 2007). In long-lived birds or mammals, for question: 8) How might plasticity have community/ecosystem effects? example, the breeding times of an individual across years can be Most of these questions, which I have intentionally framed in a related to temperature in those years, allowing the estimation of manner typical of the literature, ultimately prove hard to answer defin- individual-level reaction norms for breeding time in relation to itively. As a result, my conclusions often resort to a vague “sometimes temperature (Nussey et al. 2005; Charmantier et al. 2008; Husby yes and sometimes no” or “maybe” or “we don’t know.” While this et al. 2010; Porlier et al. 2012). Of course, such analyses focus ambiguity initially might seem unsatisfying, it reflects the empirical on traits showing “labile” plasticity that can be adjusted by an reality and it highlights the need for progress, which is ultimately more individual on an episode-by-episode basis; most obviously various exciting and interesting than a cut-and-dried “yes” or “no” answer aspects of behavior and physiology. The same approach is often that brooks no debate. As noted by one of the reviewers of this article, impossible for developmentally plastic traits that then become ambiguous conclusions and vague answers likely arise when “either “fixed”, such as many (although not all) aspects of morphol- the available data is not sufficient to answer, or the questions have ogy and life history. At the population level, the observational been framed in too vague terms that cannot have a general answer”. approach relates average trait values to average environmental Thus, when I have to be vague or waffle, I will try to distinguish which conditions across years (Phillimore et al. 2010). Although such of these 2 causes is most likely, and then suggest how the ambiguity analyses are extremely common (Parmesan and Yohe 2003), they might be resolved through more research or more refined questions. have severe limitations (Merilä and Hendry 2014). For instance, factors other than plasticity can cause temporal changes, and unmeasured correlated traits or environmental variables might Question 1: To What Extent is Plasticity Adaptive? influence observed trends (see also Schlichting and Pigliucci 1998). Some types of plasticity are clearly adaptive, such as immune Fortunately, theory-motivated analytical improvements are being responses to parasites or behavioral avoidance of predators, developed (Michel et al. 2014). whereas other types of plasticity are clearly not adaptive (Grether Another set of approaches for inferring plasticity seeks to rule 2005; Ghalambor et al. 2007). For instance, resource limitation out (or partition out) genetic contributions to observed differences, can cause developmental problems that generate phenotypes of no 28 Journal of Heredity, 2016, Vol. 107, No. 1 benefit to the organism. Given these alternative possibilities, it is traits) clearly differ dramatically in individual plasticity, its genetic important to not only quantify plasticity but to also evaluate its basis, and its adaptive significance (Husby et al. 2010). Dramatic adaptive significance. One way to do so is through experiments intra-specific variation in these properties is also present on small where plastic responses are induced and changes in fitness are spatial scales, as demonstrated by work on Blue Tits (Cyanistes monitored. For instance, defensive responses to a particular enemy caeruleus) (Porlier et al. 2012). I will return to estimates of selec- often decrease vulnerability to that enemy. As a specific exam- tion on plasticity in Question 2—because they have also been used ple, herbivory on plants decreases following herbivore-induced to infer costs of plasticity. increases in volatile chemicals (Kessler and Baldwin 2001), setose trichome density (Agrawal 1999), and spine length (Milewski et al. Conclusion 1991). Similarly, predation decreases following predator-induced Plasticity is sometimes adaptive, sometimes maladaptive, and increases in body depth in Carassius carassius carp (Brönmark and sometimes neutral. This vague answer reflects the vague way in Miner 1992) and shell thickness in Physa acuta snails (Auld and which this question is typically posed. A much more informative Relyea 2011). Plastic responses of this sort are expected to influ- question would be: What are the conditions under which plas- ence population dynamics. As an example, animals that evolved ticity has the greatest adaptive value? It seems likely that such on islands without predators often lack adaptive antipredator Downloaded from https://academic.oup.com/jhered/article/107/1/25/2622828 by guest on 08 November 2021 conditions occur when traits have different optimal values under behaviors (Cooper et al. 2014), and so suffer major declines when different environmental conditions that the population has rou- a predator is introduced (Sih et al. 2010). tinely experienced in the past, particularly when reliable environ- Even in cases where plasticity is seemingly adaptive, caveats mental cues allow appropriate and timely plastic changes (Padilla and nuances exist. For instance, defenses induced by exposure to and Adolph 1996; Reed et al. 2010). These expectations will be one enemy might be disadvantageous in the presence of a differ- revisited in Question 3. In addition, some of the vagueness of the ent enemy (DeWitt et al. 2000), and induced defenses can be costly answer arises because the adaptive significance of plasticity can in general. (More generally, the adaptive value of plasticity can be be considered in a particular environment (e.g., when exposed to considered in one environment but whether it is adaptive overall a particular predator) or over the entire life time of the organism. requires assessment across multiple environments.) Moreover, the It is critical to make these distinctions in empirical studies and above examples were targeted investigations of specific changes meta-analyses. expected a priori to be adaptive, whereas more diverse results are obtained when traits are chosen more objectively. For example, Caruso et al. (2006) exposed 2 wildflowers (Lobelia cardinalis and Question 2: To What Extent is Plasticity Costly or Lobelia siphilitica) to wet or dry conditions, measured a series of Limited? phenotypic traits related to photosynthesis, and used above-ground Organisms faced with variable environments might evolve geneti- biomass as a surrogate for fitness. The 2 species showed different cally based adaptive divergence or might instead use plasticity to levels of plasticity in different traits, and the consequences ran the mold phenotypes to current conditions (Or both, including adap- gamut from adaptive to maladaptive to neutral. As will be consid- tive divergence in plasticity.). In the absence of constraints, plasticity ered further in the next question, a series of similar studies have been would seem the best of these alternatives because it should be the performed with other organisms and the results are highly variable most immediately responsive to environmental change. Yet adap- with respect to the adaptive significance of plasticity (Van Kleunen tive divergence is common (Schluter 2000; Hereford 2009; Hendry and Fischer 2005; Auld and Relyea 2011). 2013), which suggests that plasticity must have constraints in the Most studies of the adaptive significance of plasticity are con- form of costs or limits (DeWitt et al. 1998; Auld et al. 2010, Murren ducted under controlled experimental conditions, such as common et al. 2015). gardens or mesocosms. Given that these arenas do not include all One suggested method for assessing costs of plasticity is to potential selective forces, the overall adaptive significance of plas- relate fitness in a given environment to trait values in that envi- ticity often remains uncertain. The alternative is to evaluate plas- ronment and to plasticity between environments (Van Tienderen ticity and its consequences in natural populations (Nussey et al. 1991; DeWitt 1998; Scheiner and Berrigan 1998). The data are 2007). This approach is rarely implemented owing to logistical then analyzed in a Lande and Arnold (1983) style multiple regres- constraints, but we do have some informative case studies. I would sion model, where one predictor is the trait value in an environ- especially like to highlight a contrast between 2 studies of individ- ment and the other predictor is the difference in trait value between ual plasticity in populations of great tits, one in the Netherlands environments. The partial regression coefficient for the latter term (Nussey et al. 2005) and one in the United Kingdom (Charmantier provides an estimate of the cost of plasticity while controlling et al. 2008). In each case, plasticity was quantified as the extent for mean trait value. When this coefficient is negative (selection to which individual birds changed in their breeding date between against plasticity), a cost is inferred. When this coefficient is posi- years as a function of changes in temperature, and this individual tive (selection for plasticity), a “cost of canalization” (benefit of plasticity was related to lifetime reproductive success. In the Dutch plasticity) is inferred. Van Buskirk and Steiner (2009) performed population, individuals differed dramatically in plasticity, selec- a meta-analysis of 27 studies reporting 536 separate selection tion favored increased plasticity, and current levels of plasticity estimates. Costs of canalization were found to be as common as were insufficient for fully adaptive responses to climate change costs of plasticity, and both types of costs were relatively weak and (Nussey et al. 2005). Results were opposite in the UK popula- rarely significant (see also Van Kleunen and Fischer 2005; Auld tion: individuals did not differ strongly in plasticity, plasticity et al. 2010) (Figure 2). At face value, these results might be taken was not under selection, and the current levels of plasticity were to mean that costs of plasticity are not strong (see also Auld et al. sufficient for fully adaptive responses (Charmantier et al. 2008). 2010). In reality, however, these analyses test for selection on plas- More recent work has formally compared these 2 studies and, ticity (i.e., Question 1), which will reflect a combination of costs although some conclusions change, the different populations (and and benefits acting on plasticity across the various environments/ Journal of Heredity, 2016, Vol. 107, No. 1 29 Downloaded from https://academic.oup.com/jhered/article/107/1/25/2622828 by guest on 08 November 2021 Figure 2. Distribution of estimates of selection on plasticity (540 estimates from 27 studies), with individually significant estimates (P < 0.05) shown as the dark portions of bars. These estimates are from standardized multiple regression analyses that also include the mean value of the trait. Negative values imply a net cost of plasticity, whereas positive values imply a net benefit of plasticity (or a cost of canalization). These data are from van Buskirk and Steiner (2009) and were provided by J. van Buskirk. conditions/contexts experienced during the selection interval. As example, selection might favor boldness in the presence of poten- a result, a lack of selection against plasticity could reflect a com- tial mates but shyness in the presence of potential predators (Smith bination of strong benefits and strong costs offsetting each other and Blumstein 2007), and yet bold individuals might remain bold during the selection interval. in both contexts as a result of limited moment-to-moment flexibil- Plasticity certainly has limits—both ultimate and proximate. ity. Such syndromes could have important consequences for a vari- In an ultimate sense, some phenotypic changes will be forever ety of ecological and evolutionary processes (Wolf and Weissing impossible through plasticity, just as some phenotypic changes 2012). At present, however, the relative frequency and importance will be forever impossible through evolution. In a proximate of syndromes in causing maladaptive context-dependent behavior sense, the plasticity currently present within a population is often is unknown. Another uncertainty is the extent to which behavioral insufficient for fully adaptive responses to environmental change. syndromes reflect hard limits to behavioral plasticity, as opposed For example, although some birds can plastically match their to adaptive responses to past selection resulting from, for exam- breeding time to appropriate conditions, such as the timing of ple, high costs of excessive plasticity. What is known from meta- peak caterpillar abundance, migratory birds cannot breed before analyses is that 1) the behavioral repeatability of individuals is they arrive. Migratory timing, which is often genetically based, highly variable (Bell et al. 2009)—that is, behaviors are sometimes thus places a limit on what can be achieved through plasticity of very repeatable and sometimes not (Figure 3), and 2) different per- breeding time (Both and Visser 2001; Gill et al. 2014). Further sonality axes (boldness, exploration, aggression) can influence fit- examples of limits to plasticity are legion: for instance, current ness components (reproductive success and survival) in a variety plasticity appears insufficient for responding to climate change in of ways (Smith and Blumstein 2007). British frogs (Phillimore et al. 2010), a number of birds (Nussey Costs and limits of plasticity should be context dependent: for et al. 2005; Gill et al. 2014), and many plants (Willis et al. 2008; example, costs might be strong only when plastic responses are Wolkovich et al. 2012; Van Buskirk et al. 2012). What remains large and environmental conditions are stressful. The first pos- uncertain is just how prevalent and important are these limita- sibility (large responses) was considered by Lind and Johansson tions (Murren et al. 2015). (2009) through a comparison of common frog (Rana temporaria) Another context for considering limits to plasticity is the populations that showed large versus small plastic changes in devel- idea of behavioral syndromes: suites of “correlated behaviors opmental timing in response to simulated pond drying. Costs of expressed either within a given behavioral context (e.g., correla- plasticity were found only in populations that showed the largest tions between foraging behaviors in different habitats) or across plastic responses (see also Merilä et al. 2004). However, it is difficult different contexts (e.g., correlations among feeding, antipredator, to separate selection on plasticity from selection on trait values in mating, aggressive, and dispersal behaviors)” (Sih et al. 2004). The a given environment if the 2 are correlated such that individuals basic idea is that different individuals fall at different positions with the greatest plasticity produce the most extreme traits values along behavioral or “personality” axes, which makes it difficult to (Auld et al. 2010). The second possibility (stressful conditions) was alter behaviors from one context to another (Sih et al. 2004). For considered in the meta-analysis of van Buskirk and Steiner (2009). 30 Journal of Heredity, 2016, Vol. 107, No. 1 Conclusion Plasticity must have costs and limits but these constraints are highly variable, often weak, and hard to detect. Moreover, it has proven difficult for studies to reliably separate limits, costs, and benefits, all of which might interact and be context-dependent (Auld et al. 2010). A great need exists for more studies that partition the fitness con- sequences of plasticity between different aspects of an organism’s life, such as different ages, environments (e.g., different predators, parasites, diets, and competitors), and fitness components (survival, fecundity, mating success). Such studies could more effectively assess the various context-specific benefits and costs and also how these factors combine to determine overall fitness consequences. It seems likely that costs of plasticity will be highest when plastic changes are greatest, when environmental conditions are stressful, and in rarely experienced environmental conditions (because past selection will Downloaded from https://academic.oup.com/jhered/article/107/1/25/2622828 by guest on 08 November 2021 have had less opportunity to reduce costs). In addition, it has been suggested that limits to plasticity are most likely in cases of relaxed selection and variable selection intensities (Murren et al. 2015). Figure 3. Distribution of 659 repeatability estimates from studies of behavior. Question 3: What Environmental and Organismal Repeatability is the variance in behavior among individuals divided by the Characteristics Favor the Evolution of Plasticity? sum of the variance among individuals and the variance across repeat Given that the adaptive benefits and costs/limits of plasticity vary measurements within individuals. For improved presentation, one very low repeatability value (−0.95) is not shown. These data are from Bell et al. (2009). among traits, organisms, and environments, the evolution of plastic- ity should vary at these same levels. For instance, theoretical models have shown that adaptive phenotypic plasticity readily evolves when selective conditions are variable. Some of these models have altered optimal phenotypes/genotypes through time for a single population (Gabriel 2005; Stomp et al. 2008; Svanbäck et al. 2009; Gomez- Mestre and Jovani 2013; Ezard et al. 2014), whereas others have analyzed meta-populations where optimal phenotypes/genotypes vary across space (Levins 1968; Via and Lande 1985; Van Tienderen 1997; Thibert-Plante and Hendry 2011; Scheiner and Holt 2012). These models consistently suggest that greater plasticity is favored when 1) spatial variation is greater, 2) dispersal is higher, 3) tem- poral variation is greater, 4) environmental cues are more reliable, 5) genetic variation for plasticity is higher, and 6) costs/limits of plas- ticity are lower. I now summarize studies testing the first 4 of these predictions, with the final 2 being discussed elsewhere in this article. 1. Several empirical studies have tested the prediction that higher plasticity should evolve when environments are more spatially heterogeneous. Lind and Johansson (2007) examined plasticity in common frog populations from 14 islands off the coast of Sweden. Islands with more spatial variation in pond-drying regimes (some ponds dry quickly and others slowly) were found to have frogs with greater plasticity in developmental timing when exposed Figure 4. Distribution of estimates of selection on plasticity for plants and to simulated drying regimes (water volume changes) (Figure 5). animals in low stress or high stress conditions (redrafted from Van Buskirk and Steiner 2009). Shown are means and standard errors from a mixed Along the same lines, Baythavong (2011) showed that plasticity model controlling for other factors (van Buskirk and Steiner 2009), which is for the plant Erodium cicutarium was higher in environments presumably why they don’t closely match the raw data in Figure 2. Negative with more fine-grained spatial variation. Although a number of values imply a net cost of plasticity, whereas positive values imply a net other such studies further support the above expectation, too few benefit of plasticity (or a cost of canalization). have been conducted to warrant sweeping generalizations. 2. If spatial variation favors the evolution of plasticity (as above), Specifically, costs of plasticity were highest when environmental greater plasticity is expected to evolve under higher dispersal rates, stress was greatest, at least for animals (Figure 4)—although this which increase the spatial variation experienced by a given lineage. result is not universal (Steiner and Van Buskirk 2008). Not surpris- Lind et al. (2011) used the frog system described just above to sug- ingly, then, costs of plasticity depend on properties of organisms, gest (statistical significance was lacking) that phenotypic plasticity traits, and environments. was greater when gene flow (based on DNA microsatellites) was higher among islands with different drying regimes. A potential uncertainty in such analyses is ascertaining whether higher gene Journal of Heredity, 2016, Vol. 107, No. 1 31 high dispersal (planktonic larvae) (Figure 6). The inference is that planktonic species have limited control over the conditions they experience and should therefore evolve higher plasticity, with a recent example being eastern oysters, Crassostrea virginica (Eier- man and Hare 2015). 3. Several studies inform the expectation that populations experi- encing greater temporal variation will evolve greater plasticity. For example, temporal variation in fish predation on Daphnia is present in lakes with fish but not in lakes without fish—and plasticity in kairomone-induced phototactic responses of Daph- nia is correspondingly higher in the former than the latter (De Meester 1996). Interestingly, an opposite result is seen for some traits in Trinidadian guppies, where the population not experienc- ing the predator shows greater plastic responses to predator cues (Torres-Dowdall et al. 2012)—probably as a result of correlated Downloaded from https://academic.oup.com/jhered/article/107/1/25/2622828 by guest on 08 November 2021 responses to selection on the mean phenotype. Returning to sup- portive examples, Gianoli and Gonzalez-Teuber (2005) compared Figure 5. For frogs on recently colonized Swedish islands, within-island heterogeneity of pond drying regimes is correlated (across islands) with 3 populations of the plant Convolvulus chilensis that experience the degree of plasticity frogs show in their development time. Plasticity is dramatically different inter-annual variation in precipitation and “mean development time for the offspring of a female under constant water therefore drought stress. Four traits showed plastic responses level, minus the development time under the artificial pool drying treatment.” to simulated drought conditions in the laboratory and, in each Heterogeneity is the coefficient of variance in pool drying on an island. These case, plasticity was greatest for the population that experienced data are from Lind et al. (2011) and were provided by M. Lind. the greatest temporal variation in nature. However, the adaptive significance of plasticity could be confirmed for only one of the traits: foliar trichome density. 4. Plastic responses should evolve only when an environmental cue provides a reliable and timely indicator of appropriate adaptive phenotypes (Padilla and Adolf 1996; Reed et al. 2010). This topic has been studied extensively in plants that respond to crowding conditions by elongating internodes and accelerating flowering, with the first response helping to escape competition for light and the second response helping to increase reproduction before death. The environmental cues that initiate these responses are overall irradiance and the ratio of red to far red wavelengths, both of which are indicators of vegetation-generated shade. However, the same cues will not reliably indicate local com- petitive conditions in woodland habitats, where shade is mostly determined by larger trees. As expected, populations from non- woodland habitats show greater responses to light cues than do populations from woodland habitats (Morgan and Smith 1979), and reciprocal transplant experiments in Impatiens capensis have confirmed the adaptive significance of these differences (Dono- hue et al. 2000, 2001). Evidence that plasticity is stronger under more predictive conditions has also been reported for animals (Porlier et al. 2012). Conclusion Figure 6. Plasticity in marine invertebrates with low dispersal (nonplanktonic larvae) or high dispersal (planktonic larvae). Shown are means and confidence Multiple lines of evidence support the expectation that greater trait intervals for the magnitude of plasticity (Hedges’ d) based on different plasticity evolves in more variable environments, when environmen- experimental treatments in common garden or reciprocal transplant experiments. tal cues are more reliable, and when costs are lower. This plasticity Shown are calculations based on all data—similar results are obtained in reduced can then buffer performance and fitness across a range of environ- analyses (one study per species). These data are from Hollander (2008). ments (Lynch and Gabriel 1987; Chevin et al. 2010; Lande 2014; Resuch 2014). Yet counter-examples exist, such as the maintenance flow is the cause or the consequence of higher plasticity (Crispo of high plasticity in isolated populations experiencing relatively sta- 2008). The role of dispersal was evaluated more generally in a ble environments (Torres-Dowdall et al. 2012; Wiens et al. 2014) meta-analysis of 258 experiments on plasticity in marine inver- and the failure of generalists to evolve in variable environments tebrates (Hollander 2008). In accordance with the expectation, in some laboratory experimental evolution studies (Condon et al. species with low dispersal (nonplanktonic larvae that had “vivipa- 2014). Thus, while the general expectations are often upheld, numer- rous/ovoviviparous development or direct development from ben- ous exceptions point to the importance of additional interacting fac- thic egg masses”) showed lower plasticity than did species with tors (Angilletta 2009; Condon et al. 2014). 32 Journal of Heredity, 2016, Vol. 107, No. 1 Question 4: To What Extent Does Plasticity Aid flycatchers, Ficedula albicollis (Przybylo et al. 2000). By contrast, Colonization and Responses to Environmental phenotypic plasticity seems insufficient for fully adaptive responses Change? to climate change in other instances (see Question 2). The next step Large environmental shifts should pose problems for populations should be the transition from trait changes to fitness consequences. because existing phenotypes will not be well suited for the new con- The populations referenced in the above paragraph all persisted ditions. In such cases, organisms are expected to shift their pheno- in the face of environmental change, and perhaps adaptive plasticity types in an adaptive direction, which might then make the difference was the reason, although explicit confirmation is not available. A more between persistence versus extirpation. This “phenotypic rescue” can informative analysis, however, would be to consider the role of plas- occur if populations undergo adaptive genetic change (“evolution- ticity in populations showing alternative demographic responses to ary rescue”), if individuals move to more appropriate locations, or climate change. For example, Willis et al. (2008) recorded changes in if individuals manifest adaptive plasticity (“plastic rescue”) (Chevin the flowering time and abundance of plant species over 150 years in et al. 2010; Yamamichi et al. 2011; Barrett and Hendry 2012; “Thoreau’s Woods,” Concord, MA. In this location, the species that Gomez-Mestre and Jovani 2013; Kovach-Orr and Fussmann 2013; were extirpated were those that showed low plasticity in flowering Ezard et al. 2014). In the present question, I focus on how pheno- time in relation to temperature. The implication is that persistence of the remaining species, whose flowering time advanced by an aver- Downloaded from https://academic.oup.com/jhered/article/107/1/25/2622828 by guest on 08 November 2021 typic rescue might be achieved through plasticity in 2 contexts: in situ environmental disturbance (e.g., climate change) and the intro- age of 7 days, was at least partly due to plastic rescue. As another duction of populations to new environments. example, limits to plasticity in pied flycatchers (Ficedula hypoleuca) The basic idea behind plastic rescue is that individuals evaluate have prevented sufficient change in breeding time, which has caused altered conditions and adjust their phenotypes appropriately, which population declines (Both and Visser 2001; Both et al. 2006). It thus might then increase mean population fitness and thereby enhance seems that plasticity will be sufficient for phenotypic rescue in some persistence and colonization of new environments. This phenome- instances, whereas evolutionary changes will be needed in others (see non has been called the “Baldwin Effect” (Simpson 1953; Price et al. also Phillimore et al. 2010). With this recognition, plasticity has been 2003; Ghalambor et al. 2007; Crispo 2007) following its exposi- increasingly incorporated into population viability and evolutionary tion by Baldwin (1896, 1902). Baldwin further suggested that, once rescue models for specific taxa (Baskett et al. 2009; Gienapp et al. adaptive plasticity occurred, genetic change would be expected in the 2012; Vedder et al. 2013). The upshot of these analyses is that plastic- direction of the plastic response. This second step has been termed ity, as long as current cues reliably predict appropriate future pheno- “genetic accommodation” (West Eberhard 2003; Schlichting and types, generally should have a positive effect on population persistence. Wund 2014). Waddington (1953, 1961) argued that the specific type When organisms are introduced into new environments, adaptive of post-plasticity genetic change would be canalization of the trait plasticity might play a key role in colonization, persistence, and inva- such that the new phenotypes would no longer require environmen- siveness (Baker 1965; Richards et al. 2006; Hulme 2007). One way tal induction, a phenomenon he called “genetic assimilation” (West to inform this possibility is to compare levels of plasticity in fitness- Eberhard 2003; Crispo 2007; Schlichting and Wund 2014). Spalding related traits between invasive and noninvasive species. Early quali- (1873) had a similar idea, as described by Price (2008, p. 133). tative reviews for plants yielded inconclusive results, with greater Despite the appeal of these ideas, it has been argued that concrete plasticity found as commonly for noninvasive species as for invasive evidence is lacking (De Jong 2005) and that the opposite sequence species (Bossdorf et al. 2005; Richards et al. 2006). However, a more (evolution first, then plasticity) is also possible (Scheiner and Holt recent quantitative meta-analysis that examined 75 invasive/noninva- 2012). Here I will consider evidence for the first part of the idea: sive plant species pairs came down decisively in favor of greater plas- plasticity aids persistence, colonization, and invasiveness. ticity in invaders (Davidson et al. 2011). A related, but independent, Perhaps the best evidence for the importance of plasticity in line of inquiry asks whether behavioral plasticity promotes invasion responding to environmental change comes from studies of phenolog- success in animals (Wright et al. 2010). Sol and colleagues (2008, ical responses to climate warming. Many organisms have advanced 2012) found that brain size (expected to be correlated with behav- the timing of spring life-history events (e.g., flowering, breeding, ioral flexibility) and foraging innovation (a measure of behavioral migration) as temperatures have increased and winters shortened flexibility) were positively associated with the probability that intro- over the past 50 years (Parmesan and Yohe 2003). It is hard to ascer- duced birds and mammals became invasive. At the same time, how- tain whether these changes are the result of genetic evolution or phe- ever, some species with modest brain sizes become invasive and some notypic plasticity (or both), mainly because the common methods species with large brain sizes don’t (Figure 7): that is, the variance for confirming a genetic basis for phenotypic change (e.g., common- explained isn’t very high. Overall, then, although behavioral plas- garden experiments) are difficult to apply in a temporal (allochronic) ticity in animals (and trait plasticity in plants) might sometimes aid context (Gienapp et al. 2008; Merilä and Hendry 2014). That is, responses to new environments, it certainly isn’t a universal solution. it is hard to take genotypes that live at different times and assess them under the same conditions, with the exception being organ- Conclusion isms with dormant stages (e.g., seeds or resting eggs: De Meester Plasticity sometimes aids colonization of new environments 1996, Boersma et al. 1998; Cousyn et al. 2001) or when the common and responses to in situ environmental change. However, plastic garden environment can be exactly duplicated at different times. responses aren’t always necessary or sufficient in these contexts. In Without disputing the importance of evolution in at least some phe- one sense, this qualified answer reflects data deficiency: very few nological changes (Bradshaw et al. 2006; Merilä and Hendry 2014), studies have examined the contributions of plasticity to population plasticity also must often be important. For instance, the study of the dynamics in the face of environmental change. In addition, no exper- UK population of great tits (Charmantier et al. 2008; Vedder et al. imental studies in nature have assessed the role of plasticity in medi- 2013), suggested that plasticity was entirely sufficient for adaptive ating such challenges, such as by assessing the responses of more or responses of reproductive timing to climate change. Similar argu- less plastic genotypes. (Studies of invasive species often show that ments have been made for other species, including Gotland collared invaders are more plastic but cannot confirm that plasticity was a Journal of Heredity, 2016, Vol. 107, No. 1 33 variation in the trait can be increased in at least some environments, thus aiding trait evolution in those environments. In addition, simula- tions have suggested that plasticity can alter genetic architecture so as to increase the production of adaptive phenotypes (Fierst 2011). Given that nearly everything seems possible in theory (Paenke et al. 2007), I will here focus on empirical observations relevant to the key ideas. The phenomenon of counter-gradient variation, where genetic effects are in the opposite direction to plastic effects, provides a nice demonstration of how maladaptive plasticity can promote, indeed necessitate, compensatory adaptive genetic change (Conover and Schultz 1995; Levins 1968; Conover et al. 2009). A well-known example is growth rate in Atlantic silversides, Menidia menidia. Northern and southern populations of this fish have similar body sizes in nature despite better growing conditions in the south. When raised in a common garden, however, fish from the northern popula- Downloaded from https://academic.oup.com/jhered/article/107/1/25/2622828 by guest on 08 November 2021 tion grow faster and to a larger size than do fish from the southern population. In this case, plastic effects on growth that result from environmental differences have led to the evolution of compensat- ing genetic differences in intrinsic growth rate (Conover and Present 1990; Present and Conover 1992). The contrasting pattern of co- gradient variation, where plastic effects are in the same direction as genetic effects, can imply that plasticity has reduced genetic diver- Figure 7. In birds, invasion potential (probability of establishment following gence (Byars et al. 2007). (Of course, initial plasticity could well have introduction) is positively related to residual brain size (brain size corrected allowed the cogradient genetic variation to evolve—as in genetic for body size). Each point represents a different species and similar results accommodation.) As both patterns are known to exist in nature, are obtained if phylogeny is controlled through independent contrasts. These the important question becomes: how common is each? Although a data are from Sol et al. (2012) and were provided by D. Sol. formal meta-analysis has not been conducted, Conover et al. (2009) summarized more than 60 examples of counter-gradient variation, while finding many fewer examples of cogradient variation. key contributor to the invasion.) However, I expect that the above Counter-gradient variation thus provides a particularly obvious answer will remain qualified even when more data are gathered. situation where plasticity can promote genetic change—because the Instead, a more profitable question might be: Under which condi- plasticity is maladaptive and thus imposes selection for genetic com- tions is plasticity most likely to aid colonization and in situ rescue? pensation. Another such situation occurs when plastic change in one I suggest that the answer is likely to be when 1) the trait is particu- trait necessitates genetic change in other traits: that is, altering one larly important for fitness, 2) the new conditions are similar to those aspect of the phenotype requires compensatory genetic changes in previously experienced by a lineage (plasticity is then more likely other aspects of the phenotype. As a clear example, the introduction to have been shaped by past selection), 3) plasticity can accomplish of a predator (curly-tailed lizards, Leiocephalus carinatus) caused large phenotypic changes, 4) plasticity isn’t very costly, and 5) the a prey species (Anolis sagrei lizards) to plastically shift their habi- traits are behavioral or physiological as these should be the most tat to narrow perches in trees, which imposed selection for shorter malleable traits on short time scales. legs (Losos et al. 2006). Of course, the opposite effect is also pos- sible: behavioral thermoregulation (plasticity) reduces exposure to Question 5: Does Plasticity Promote or Constrain extreme temperatures and thereby reduces selection for physiologi- Genetic Evolution? cal temperature adaptation (Huey and Kingsolver 1993). A number of arguments have been advanced for how plasticity might It is much more difficult to ascertain whether adaptive plasticity promote or constrain adaptive genetic change. On the constrain- in a trait promotes genetic change in the same trait. Some correlative ing side, the main argument is that plasticity shields the genotype support for this idea comes from studies of ecological speciation, from selection, thereby slowing adaptive genetic change (Huey and as will be discussed further in Question 6. However, the best evi- Kingsolver 1993; Linhart and Grant 1996; Ghalambor et al. 2007). dence would come from experiments showing that populations with This argument applies mainly to adaptive plasticity, whereas mala- greater adaptive plasticity in a trait also show faster adaptive evolu- daptive plasticity would be expected to increase selection for adap- tion of the same trait. One such experiment has been performed. tive “genetic compensation” (West Eberhard 2003; Grether 2005; Schaum and Collins (2014) performed a laboratory experimental Ghalambor et al. 2007). On the promoting side, a common argument evolution study using 16 lineages of Ostreococcus (a marine green is that plasticity allows colonization of, and persistence in, extreme algae microbe) that initially differed in CO2-related plasticity for environments (Question 4), which thereby increases selection on that “oxygen evolution rates” (generating oxygen through chemical reac- trait or other traits (West Eberhard 2003; Schlichting and Wund 2014). tion). During 400 generations of rearing under constant or fluctuat- Another suggested positive influence is that plasticity can expose oth- ing CO2 conditions, lineages with higher ancestral plasticity showed erwise cryptic genetic variation (not expressed under normal condi- faster evolution of population growth rates (a measure of fitness). tions) to selection (West Eberhard 2003; Suzuki and Nijhout 2006; This positive relationship between plasticity and evolution was, as Pfennig et al. 2010; Moczek et al. 2011; Gomez-Mestre and Jovani expected, strongest in the treatments with fluctuating CO2. 2013; Schlichting and Wund 2014). Related to this point, increasing Evidence for the converse—plasticity in a trait constrains genetic genetic variance in reaction norm slopes is one way in which genetic divergence in that trait—also could be generated through the 34 Journal of Heredity, 2016, Vol. 107, No. 1 above-suggested experiments. Given their current scarcity, we can start with correlative support from 2 sorts of comparisons—one based on populations and one based on traits. For the first, populations showing greater plasticity in a trait should show lower genetic diver- gence (among populations) in that trait. For the second, traits showing greater plasticity should show lower genetic divergence. Exemplifying a population-based comparison, Misty Lake and Misty Inlet stick- leback (Gasterosteus aculeatus) show strong genetic divergence in a number of adaptive traits, whereas Misty Lake and Misty Outlet stickleback show no genetic divergence but rather plastic differences (Sharpe et al. 2008). Of course, cause and effect is here difficult to establish given that high gene flow between the lake and outlet popu- lations (Roesti et al. 2012) could prevent genetic divergence, leaving plasticity as the only recourse (as opposed to plasticity evolving and then limiting genetic divergence). Exemplifying a trait-based compari- Downloaded from https://academic.oup.com/jhered/article/107/1/25/2622828 by guest on 08 November 2021 son, we have the adaptive responses of fish to low dissolved oxygen. In many species, fish from low-oxygen environments have larger gills so as extract more oxygen, and they also have smaller brains due to the resulting limitations on cranial space. Crispo and Chapman (2010) collected populations of the cichlid fish Pseudocrenilabrus multicolor victoriae from different oxygen environments in nature and raised their offspring under high and low oxygen conditions in the labora- tory. Essentially all of the resulting variation in gill size was plastic, with no apparent genetic differences among populations (Figure 8). By contrast, brain size was less plastic and showed more genetic variation among populations (Figure 8). Similar findings (more plastic traits show lower genetic divergence) emerged in a study of the effects of predators on guppies (Torres-Dowdall et al. 2012). Although the above observations are consistent with the idea that plasticity constrains genetic divergence, causation is hard to establish and, regardless, too few studies have been conducted to invite generalization. Moreover, the effects of plasticity could be— indeed they are often expected to be—transient during the course of evolution, such as in the case of genetic assimilation. Detecting such effects requires the tracking of genetic and plastic contributions dur- ing the course of environmental change or in controlled experiments Figure 8. Population-level reaction norms for gill size (PC1 of measurements standardized to a common body size) and brain size (standardized to a (e.g., Schaum and Collins 2014). common body size) for 6 populations of Pseudocrenilabrus multicolor victoriae raised under low and high oxygen conditions in the laboratory. Conclusion These data are from Crispo and Chapman (2010). Plasticity will sometimes promote and sometimes constrain genetic evolution. In this case, my vague answer mostly reflects a vague ques- formation. The debate has crystallized around 2 opposing schools of tion: that is, the opposing expectations—and everything in between— thought, which I here dichotomize for the sake of argument. (Both should be differentially likely under different conditions. Thus, we here perspectives, and various intermediates, are acknowledged in most clearly need a better question, such as Under which conditions does publications.) The first perspective is an extension of the “Baldwin plasticity promote versus constrain genetic evolution? Some predic- Effect” described in Question 4. It argues that plasticity facilitates tions are that promoting effects will be most likely when 1) plastic- colonization of new environments, or the use of new resources, ity enables colonization/persistence where it would not be otherwise after which phenotypes are exposed to divergent selection that possible, 2) plasticity in one trait (e.g., behavioral flexibility that alters causes adaptive genetic divergence and hence ecological speciation resource use) results in altered selection on other traits, 3) selection is (Skúlason and Smith 1995; Smith and Skúlason 1996; Robinson and on plasticity itself, 4) plasticity exposes otherwise cryptic genetic varia- Parsons 2002; West Eberhard 2003; Pfennig et al. 2010). One branch tion, and 5) plasticity is maladaptive. Although correlative tests of these of this argument specifically emphasizes behavioral flexibility that hypotheses will be useful, particularly informative approaches would results in the use of new resources, which can then enhance specia- be experimental. In particular, more versus less plastic genotypes could tion through a process sometimes called “behavioral drive” (Wyles be introduced into new environments and subsequent adaptive evolu- et al. 1983). The opposing school of thought is an extension of an tion monitored—as Schaum and Collins (2014) did in the laboratory. idea from Question 5, arguing that plasticity shields the genotype from selection and thereby reduces genetic divergence and hampers Question 6: Does Plasticity Help or Hinder speciation (Price et al. 2003; Ghalambor et al. 2007; Crispo 2008; Ecological Speciation? Svanbäck et al. 2009; Thibert-Plante and Hendry 2011). Behavior The previous question focused on variation within species, whereas also could have constraining effect by allowing organisms to main- the present question considers the same issues with respect to species tain their use of a particular resource even as environments change, Journal of Heredity, 2016, Vol. 107, No. 1 35 which would reduce divergent selection (Duckworth 2009). I start environments, or resources. Such imprinting can lead to assortative by summarizing and evaluating 3 empirical observations suggested mating that allows genetic divergence and the evolution of repro- to indicate that plasticity promotes ecological speciation. ductive barriers. In birds, nestlings sometimes imprint on the songs of their fathers (Price 2008), with male offspring later singing—and female offspring later preferring—similar songs. The result can be 1. Plasticity within species, ideally demonstrated in ancestral forms, mating isolation between groups whose songs have diverged for what- is sometimes in the same direction as genetic differences between ever reason (Price 2008). Remarkably, male and female nestlings of species. One clear example is trophic morphology in fishes, brood-parasitic Vidua finches imprint in a similar way on the songs of where “limnetic” versus “benthic” diets cause plastic divergence their host species, which leads to assortative mating between finches in trophic morphology in a direction that parallels genetically parasitizing different hosts (Payne et al. 2000). In insects, larvae some- based divergence between closely-related species (Day et al. times imprint on the plant on which they feed (“conditioning”) and 1994; Robinson and Parsons 2002; Adams and Huntingford then preferentially select those plants during mating and oviposition, 2004; Wund et al. 2008). The common inference therefrom is thus generating mating isolation between groups using different host that ancestral plasticity initiated and promoted the subsequent plants (Funk et al. 2002). In salmonid fishes, juveniles often imprint on genetic divergence. chemical properties of their natal site and then strongly “home” back Downloaded from https://academic.oup.com/jhered/article/107/1/25/2622828 by guest on 08 November 2021 2. Character displacement between species is sometimes facilitated to that site for reproduction, which reduces gene flow between popu- by polyphenism (different, discrete phenotypes emerge when the lations (Hendry et al. 2004). Considering these examples as instances same genotype is exposed to different environments), which can of “positive” imprinting, “negative” imprinting also can reduce gene then sharpen reproductive barriers (Pfennig and Pfennig 2009). flow: for example, exposure to heterospecifics can strengthen prefer- In spadefoot toads, for example, 2 species (Spea bombifrons and ences against them (Price 2008; Delbarco-Trillo et al. 2010). In each S. multiplicata) can develop either herbivorous or carnivorous tad- of these cases, reproductive isolation depends on individuals being poles: but, when reared together, S. multiplicata produces many exposed to different environments, with another example being the fewer carnivores than does S. bombifrons. In addition, S. multipli- commensal bacteria in Drosophila that influence the chemical signals cata from ponds with more S. bombifrons in nature are genetically that drive mating isolation (Sharon et al. 2010). less likely to produce carnivores in a common-garden environment What of the opposing school of thought—that plasticity retards (Pfennig and Murphy 2002). The common inference therefrom is speciation? Empirical support might be provided through evidence that selection in sympatry has enhanced ancestral polyphenism that groups with lower plasticity speciate more often, or that the and thereby exaggerated species divergence. traits determining reproductive isolation between species are not 3. Plasticity might be greater in taxonomic groups that are more especially plastic (especially in the ancestor). Formal tests of these speciose, as a number of studies have shown. Nicolakakis et al. predictions have not been performed; however, many populations (2003) reported that innovation rate, a proxy for behavioral in different environments show strong plastic differences and yet flexibility, is positively related to the number of species within minor—if any—reproductive isolation. Following up on examples bird taxa. Sol et al. (2005) showed that relative brain size, which from Question 5, gene flow is high between populations where phe- is correlated with behavioral flexibility, is positively related notypic divergence has a primarily plastic basis in Pseudocrenilabrus to the number of subspecies in Holarctic passerines. Tebbich from different oxygen environments (Crispo and Chapman 2008) and et al. (2010) pointed out that the bird group that has diversi- between the Misty Lake and Outlet stickleback populations (Roesti fied most in Galápagos, Darwin’s finches, shows very high levels et al. 2012). Despite such putative examples, the idea that plasticity of behavioral flexibility. Pfennig and McGee (2010) used sister hampers speciation has not yet been subject to rigorous testing. group comparisons to show that fish and amphibian lineages that include polyphenic species are more speciose. The common Conclusion inference from such findings is that plasticity generally promotes Plasticity will sometimes help and sometimes hinder ecological spe- diversification, speciation, and adaptive radiation. ciation (see also Duckworth 2009). This vague answer reflects both weak data and a vague question, with a better question mirroring Each of the above arguments is consistent with idea that plasticity those suggested for the above questions: Under what conditions does promotes speciation, yet none of them provides strong evidence. One plasticity help versus hinder ecological speciation? With respect to problem is that no meta-analysis has yet quantified the extent to data, current empirical tests are insufficient to allow general conclu- which plasticity within species is in the same direction as divergence sions as to how often and when each result is most common. Among between species. Moreover, observed plastic effects within species are the improvements suggested above, the most critical would be con- often much smaller than observed differences between species, even trolled experiments that examine progress toward ecological spe- if they are in the same direction (e.g., Losos et al. 2000). Another ciation in lineages that are initially more or less plastic. In general, limitation of the first 2 types of analysis is that a low-plasticity I predict that plasticity is especially likely to have positive effects in “control” comparison is not normally considered: that is, speciation the various manifestations of imprinting, when mating cues depend might have been even more likely/rapid/dramatic if plasticity wasn’t on environmental exposure, and when dispersal occurs after (rather present. Most critically, the identical prediction (plastic and genetic than before) plastic changes occur (Thibert-Plante and Hendry 2011). differences are in the same direction) also emerges from arguments that plasticity constrains divergence. In the third type of analysis, the level of plasticity is often unknown in the ancestral species. As a Question 7: How Fast Does Plasticity Evolve? result, it is difficult to establish whether plasticity was the cause or Many studies dichotomize phenotypic change into that caused by the consequence of high diversification. genetic change versus that caused by plasticity. In reality, both effects One process by which plasticity is particularly likely to promote can occur at the same time and can influence each other—as has been speciation occurs when juveniles imprint on parents, conspecifics, described previously. Moreover, plasticity can evolve and such change 36 Journal of Heredity, 2016, Vol. 107, No. 1 should have important consequences for population dynamics, includ- Fussmann 2013). Very few empirical studies have directly assessed this ing “rescue” (Question 4). It is therefore important to ask how quickly question but a few examples will illustrate the possibilities, starting plasticity can evolve and what factors increase or decrease this rate. with community influences and then moving to ecosystem influences. A first point is that many studies have documented the evolution Many foraging traits of fishes are phenotypically plastic in response of reaction norms on the time scale of decades, with a classic example to diet. For instance, fish fed on zooplankton diets (as opposed to being the phototactic behaviour of Daphnia in response to changing benthic diets) tend to have longer gill rakers and changes in jaw mor- fish predation (De Meester 1996; Boersma et al. 1998; Cousyn et al. phology that increase foraging efficiency on those food items (Day 2001). Many other examples exist—and I will here mention 2. The and McPhail 1996). Because these traits have dramatic influences Asian shade annual plant Polygonum cespitosum colonized North on aquatic prey communities (Harmon et al. 2009; Palkovacs and America in the early 1900s and has recently spread into more open Post 2009), diet-induced trophic plasticity should influence prey com- habitats. In the 10 years following this niche expansion, the plant has munities (Lundsgaard-Hansen et al. 2014). Such effects have not yet evolved increased plasticity in root allocation and physiological traits been demonstrated formally in nature but they would be fascinating in response to open versus shaded conditions (Sultan et al. 2013). The to explore, not the least because they show a strong chance of feed- Asian shore crab Hemigrapsus sanguineus was first reported in North backs. That is, plastic changes in traits that influence foraging success America in 1988 and feeds on native marine mussels Mytilus edulis. on a given food type should reduce the availability of that food type Downloaded from https://academic.oup.com/jhered/article/107/1/25/2622828 by guest on 08 November 2021 At present, mussels in areas where the crab has invaded (southern (and induce its evolution) which should then influence further plas- New England) show inducible shell thickening in response to water- ticity and selection. Another situation for which plasticity is almost borne H. sanguineus cues, whereas mussels in areas where the crab certainly critical for community structure is the relative phenology of has not invaded (northern New England) do not (Freeman and Byers interacting species (Both et al. 2009; Phillimore et al. 2012). 2006). At the same time, however, a number of other studies have Rates of feeding, metabolism, and growth dramatically influence documented instances where plasticity did not evolve even on long biological stoichiometry, “the balance of energy and multiple chemical time scales despite a change in selection pressure. A particularly obvi- elements in living systems”, by altering the consumption and excre- ous example is the retention of antipredator behavior long after the tion of various elements (Elser et al. 2000; Matthews et al. 2011). predator is no longer present (Lahti et al. 2009). These latter cases As a result, plastic changes in these rates could have dramatic effects likely reflect relaxed selection (the trait is not expressed in the absence on the availability and transfer of elements within and between com- of the cue), in which case trait evolution would occur only through the munities and ecosystems. As one example, Schmitz (2013) argued that relatively slow processes of drift and mutation. increasing animal metabolic rates with increasing temperature owing As always, selected examples can only take us so far and any hope to climate change should cause “phenotypically plastic shifts in animal of generality must come from meta-analyses. In one meta-analysis, elemental demand, from nitrogen-rich proteins that support produc- Crispo et al. (2010) analyzed 20 studies that measured plasticity in tion to carbon-rich soluble carbohydrates that support elevated energy 2 or more populations, at least one of which was subject to recent demands.” The resulting change in diets should then have important human disturbance and at least one of which was not. The authors consequences for carbon cycling (Schmitz 2013). As another example, calculated rates of change for plasticity in darwins and haldanes, 2 Dalton and Flecker (2015) showed that the presence of dangerous common metrics of rates of change in phenotypic traits (Hendry and predators (simulated with predator cues in the laboratory) decreased Kinnison 1999). Results showed that disturbed plant populations often N excretion rates of guppies by 39%, which could have important evolved changes in plasticity and that different taxa and traits showed consequences for this limiting nutrient in their stream ecosystems. different responses. Based on a qualitative comparison between Crispo Another likely arena for ecosystem effects of plasticity is for organ- et al. (2010) and Hendry et al. (2008), rates of evolution of plasticity isms that produce chemical resources that are used by many other were qualitatively similar to rates of evolution of mean phenotypes. organisms, such as plants producing CO2 or fixing nitrogen. Such effects seem particularly likely given the great plasticity in these pro- Conclusion cesses depending on environmental conditions, such as ambient levels Plasticity can show considerable evolutionary change on contempo- of CO2 or nitrogen, as well as temperature and humidity. Collins and rary time scales, although the rates of this evolution are highly vari- Gardner (2009) provide a “worked example” of how to calculate the able. These findings confirm theoretical expectations that plasticity potential contribution of plasticity,