Molecular ecology of global change
- Box 1
Studying evolutionary responses to climate change
Two approaches are available for documenting past evolution. In spatial studies, locations with contrasting abiotic conditions, for example across latitudes or altitudes, are surrogates for expected temporal changes (Davis & Shaw 2001; Davis et al. 2005). With the appropriate caution, inferences on future evolutionary potential of populations and species are possible. Shortcomings are that evolutionary rates cannot be directly estimated, but divergence-time estimates from molecular markers can be used as proxies (Kinnison & Hendry 2001). Moreover, the evolvability of the fundamental range limits of a species cannot be addressed (Kirkpatrick & Barton 1997).
Retrospective temporal studies follow trait distributions in local populations through time, providing direct evidence that evolution has happened and monitor potential environmental variables that may be causal. However, because temporal and spatial studies identify correlations, deciding whether or not the observed environmental variation caused shifts in trait distribution is not trivial. Inferences can be supported by measuring fitness of local and nonlocal genotypes across contrasting habitat types (Bradshaw et al. 2004), through assessments of the selection differential (Etterson & Shaw 2001) or heritability estimates of target traits (Bradshaw & Holzapfel 2001; Etterson & Shaw 2001).
Selection experiments can overcome these problems because they are manipulative. In artificial selection experiments, mating is determined by the experimenter, a procedure similar to animal or plant breeding. Controlled natural selection instead allows populations under experimentally controlled conditions to freely evolve by fertility- and mortality selection (Fuller et al. 2005; Chippindale 2006). The most appropriate test for evolutionary responses at the end of natural selection experiments is a reciprocal assessment of control and evolved lines under both conditions. Responses to selection are detectable as ‘selection–environment’ interactions, analogous to the detection of local adaptation in the field.
Controlled natural selection is the only method that provides a simulation of the future trajectory of populations, projecting rates of sustained evolutionary change as a function of a defined selection regime, but inferences on genetic architecture of traits are not possible. In artificial selection, information on the genetic architecture of the chosen trait such as heritability and correlations with other traits can be obtained. As a major disadvantage, it remains unclear whether the chosen traits contribute to fitness in the wild (Coyne & Lande 1985).
As a special form of artificial selection, antagonistic artificial selection probes trait correlations that may slow down adaptive trait changes (Beldade et al. 2002b). The experimenter tries to select for trait divergence among correlated traits in order to decide whether or not genetic correlations are due to linkage disequilibrium or due to pleiotropy (Fuller et al. 2005). While the former can be broken up by recombination, the latter cannot (Conner 2002).
Because the population is the basic unit of observation, it should also be the unit of replication. All selection experiments need untreated populations to control for changes induced by the experimental procedure (reviewed in Conner 2003; Fuller et al. 2005). A few intermittent generations without imposing selection will prevent carry-over (e.g. maternal) effects.
- Box 2
Predicting evolutionary effects of increased CO2 concentration
An increase of atmospheric CO2 concentration may be immediately advantageous for plant growth as it alleviates the CO2 limitation of photosynthesis, in particular for plant species without carbon concentrating mechanisms (C3-carbon fixation, Ward & Kelly 2004).
Ward et al. (2000) studied the evolutionary response of Arabidopsis thaliana towards lower and higher CO2 concentration compared to present values in a five-generation-selection experiment. Starting from an outcrossed and genetically diverse base population, seed production increased lines selected at 700 p.p.m. under these novel conditions compared to random control individuals, while growth and survival remained unchanged.
Wieneke et al. (2004) found elevated vegetative growth in Sanguisorba minor, a common herb of European grasslands, as a response to selection by increased CO2-concentration. After six years of elevated CO2 under field conditions, corresponding to presumably 1–6 generations, control and treated plants were compared in the greenhouse under both conditions. The results demonstrate enhanced litter production and carbon sequestration as a response to the selection regime.
Because of the low efficiency of Rubisco, carbon concentration mechanisms (CCM), such as CAM (crassulacean acid metababolism) or C4-carbon fixation have evolved to enrich the CO2 at the point of Rubisco fixation. In accordance with evolutionary theory, one would predict that such energetically costly carbon concentration mechanisms are no longer selected for under elevated CO2. This was indeed found after 1000 generations of experimental selection in the green unicellular algae Chlamydomonas reinhardtii. Strains under increased CO2 concentration were less efficient in carbon fixation after 1000 generations of artificial selection when tested under present-day CO2 conditions (Collins & Bell 2004); a finding which was attributed to conditional detrimental mutations compromising costly carbon-concentrating mechanisms that use bicarbonate (HCO3−) in addition to dissolved CO2 as inorganic carbon source.
Thorsten B. H. Reusch, Fax: +492518324668; E-mail: email@example.com
Global environmental change is altering the selection regime for all biota. The key selective factors are altered mean, variance and seasonality of climatic variables and increase in CO2 concentration itself. We review recent studies that document rapid evolution to global climate change at the phenotypic and genetic level, as a response to shifts in these factors. Among the traits that have changed are photoperiod responses, stress tolerance and traits associated with enhanced dispersal. The genetic basis of two traits with a critical role under climate change, stress tolerance and photoperiod behaviour, is beginning to be understood for model organisms, providing a starting point for candidate gene approaches in targeted nonmodel species. Most studies that have documented evolutionary change are correlative, while selection experiments that manipulate relevant variables are rare. The latter are particularly valuable for prediction because they provide insight into heritable change to simulated future conditions. An important gap is that experimental selection regimes have mostly been testing one variable at a time, while synergistic interactions are likely under global change. The expanding toolbox available to molecular ecologists holds great promise for identifying the genetic basis of many more traits relevant to fitness under global change. Such knowledge, in turn, will significantly advance predictions on global change effects because presence and polymorphism of critical genes can be directly assessed. Moreover, knowledge of the genetic architecture of trait correlations will provide the necessary framework for understanding limits to phenotypic evolution; in particular as lack of critical gene polymorphism or entire pathways, metabolic costs of tolerance and linkage or pleiotropy causing negative trait correlations. Synergism among stressor impacts on organismal function may be causally related to conflict among transcriptomic syndromes specific to stressor types. Because adaptation to changing environment is always contingent upon the spatial distribution of genetic variation, high-resolution estimates of gene flow and hybridization should be used to inform predictions of evolutionary rates.
Humans are, in effect, conducting a planet-wide selection experiment by altering our global ecology. This imposed selection is largely a consequence of ongoing climate forcing by greenhouse gas emissions (IPCC 2007a, b). Biological consequences of rapid climate change include range shifts, behavioural changes, altered phenology and local extinctions. These effects have been observed in diverse ecological settings (summarized in Lovejoy & Hannah 2005). For example, a comprehensive meta-analysis revealed that species ranges have shifted poleward 6.1 km per decade on average, and that 62% of taxa exhibit spring advancement for shifts to reproduction and migrant arrival date (Parmesan & Yohe 2003).
Organismal responses to global environmental change are conventionally categorized as either ecological or evolutionary. While it is difficult to tease apart these categories, the former includes phenotypic plasticity and dispersal, while the latter entails genetic change. Thus, populations can avert fitness decline in the face of environmental change in three ways: be plastic, move or evolve (Jackson & Overpeck 2000). It has been suggested that species may be more likely to shift their distributions rather than evolve in situ (Parmesan 2006), primarily because the evolution of key traits is mutation-limited or otherwise constrained (Bradshaw 1991; Ackerly 2003). While palaeontology and biogeography support this view, diverse organisms are known to have responded to natural selection with remarkable alacrity (Grant & Grant 2002; Reznick et al. 1997; reviewed in Kinnison & Hendry 2001; Reznick & Ghalambor 2001).
As molecular ecologists, we take a synthetic view and consider the interaction between ecological and evolutionary responses to global change. Here, we discuss how an understanding of the genetic basis of phenotypes under selection can help us to predict and mitigate what appear to be predominantly negative effects of global change on population viability. The need to understand the genetic basis of phenotypic responses brings to the fore the fundamental challenge facing molecular ecologists. The field is transitioning from focusing on neutral genetic markers to characterizing selectively important genetic polymorphism (Feder & Mitchell-Olds 2003; Purugganan & Gibson 2003; Vasemägi & Primmer 2005). There are now striking examples that identify the genetic causes of variation in ecologically important traits (e.g. Reusch et al. 2001; Beldade et al. 2002a; El-Assal et al. 2004; Colosimo et al. 2005); these studies and others move us towards a more complete understanding of evolutionary change and local adaptation. Both conservation and evolutionary biology stand to benefit if we direct such successful research strategies at traits that are related to fitness under a changing climate.
Below, we summarize recent progress in identifying climate driven, contemporary evolution at the phenotypic and molecular genetic level. In addition, we discuss study designs for retrospective, simulation and predictive studies of global change-induced evolution. We focus on primary consequences of global climate change, i.e. on direct effects driven by alterations of the physical environment. Selection experiments are highlighted as an indispensable tool for predicting evolutionary responses. Next, we review recent examples of the successful application of molecular ecological tools that have revealed important genes that underlie responses to global climate change. For example, researchers have successfully identified natural genetic variation for phenology, stress tolerance, carbon fixation and carbonate production. Finally, we discuss how molecular ecology provides us with an integrative framework to analyse and predict evolution in a changing ecology.
Globally changing selection and evolutionary responses
The new selection regime
The rate and absolute magnitude of climate change is expected to be greater than that inferred at least for the last four million years (Overpeck et al. 2005). Increases in air and water temperatures are the most obvious primary consequences of contemporary global change (IPCC 2007a, b). For species outside the tropics, the regular sequence of specific climatic conditions throughout the year, seasonal variation, is more important than annual mean values (Bradshaw & Holzapfel 2006). Subtle changes of the temperature regime during a particular seasonal stage may have a pronounced biological effect. For example, in a skipper butterfly Atalopedes campestris, the increase in winter minimal temperatures allowed a northern range shift while changes in summer temperatures were unimportant (Crozier 2003).
Not only are mean values and seasonality of many abiotic variables predicted to change but also their extreme values (Schär et al. 2004). Heat waves, droughts and floods will be superimposed onto increased seasonal variability and may impact ecosystems and communities much earlier than gradual environmental changes (Gaines & Denny 1993). As tolerance is often a threshold function of environmental parameters (Feder et al. 2000), extremes can result in mortality and even local extinction (Ehrlich et al. 1980; Cerrano et al. 2000; Hughes et al. 2003; Reusch et al. 2005). Critical threshold values may be close to contemporary ‘average’ conditions. For example in reef-building corals, coral bleaching and associated mass mortality occurs when temperatures are 2 °C above regular summer sea-surface temperature for only 10 days (Gleeson & Strong 1995). Thus, increasing climate extremes may impose ‘hard’ selection upon species, rapidly selecting for more stress-tolerant genotypes, while changes in seasonality impose ‘soft’ selection, mediated by intraspecific competition (Bijlsma & Loeschcke 2005). Paradoxically, organisms living in the thermally buffered aquatic habitat may be particularly vulnerable (Cerrano et al. 2000; Hughes et al. 2003; Reusch et al. 2005). Due to the high thermal capacity of water, there is less microhabitat heterogeneity in lakes and ocean, in contrast to land, and evaporative cooling is not possible in water (Feder & Hofmann 1999).
In contrast to many climatic variables, the concentrations of major greenhouse gases themselves are expected to increase continually; most importantly, the concentration of CO2 in both the atmosphere and the world's oceans. Because the sequestration of inorganic carbon by aquatic and terrestrial plants constitutes an important biological feedback loop that may ameliorate CO2 increases (IPCC 2007a), knowledge of the ability of plant populations to increase or decrease photosynthetic efficiency under future CO2 concentration are critical for climate predictions (Ward & Kelly 2004). While land plants may benefit through enhanced CO2 concentration, dissolved CO2 makes ocean waters more acidic, potentially interfering with the production of biogenic carbonates through unicellular algae that are an important sink for CO2 (Riebesell et al. 2000).
Direct evidence for evolution: temporal studies
Ongoing alterations of the seasonal sequence of temperature, humidity and precipitation patterns often result in nonoptimal timing of important life-history events (phenology). Because some populations do not adapt to altered seasonality, either phenotypically or genetically, the resulting seasonal mismatch can reduce population mean fitness (Both & Visser 2001; Parmesan 2006). In contrast, photoperiod response has proven to be evolvable in certain cases (e.g. Bradshaw & Holzapfel 2001; Reale et al. 2003), although the associated genetic changes are unknown.
One of the most convincing studies for evolution related to photoperiod examined the pitcher-plant mosquito (Wyeomyia smithii). Repeated laboratory measurements revealed that the initiation of larval diapause has shifted toward a significantly shorter critical daylength, as growing seasons have become longer within the past 30 years (Bradshaw & Holzapfel 2001). The critical daylength of the photoperiodic response revealed a high degree of heritability (Bradshaw & Holzapfel 2001). In addition, Bradshaw et al. (2004) measured up to 88% fitness decline of northern photoperiodic phenotypes under the ‘wrong’, nonlocal photoperiodic regime. Another valuable study decomposed alterations in seasonal behaviour into genetic and plastic components in populations of the Yukon red squirrel, Tamiasciurus hudsonicus, in northern Canada: Reale et al. (2003) were able to determine the heritability of temporal changes in nesting date through an analysis of females breeding in several years. The observed shift towards earlier breeding — 18 days over 10 years — was due in the larger part to phenotypic plasticity, in combination with a smaller genetic component.
Because climate zones are rapidly shifted under global warming, selection favours increased dispersal potential, especially at the edge of ranges. Accordingly, morphological changes have been found in a number of insect species that indicate a response to this type of selection. For example in the UK, the frequency of long-winged forms (macropterous morphs) in two species of crickets (Metrioptera roeselii and Conocephalus discolor) increased in newly established populations at the expanding margin, indicating phenotypic evolution of increased dispersal ability (Thomas et al. 2001). Likewise, in the speckled wood butterfly, colonizing subpopulations were larger and had larger thorax muscles and wing aspect compared to central ones (Hill et al. 1999).
Inferences from past evolution: spatial contrasts
Comparisons of populations from different locations with contrasting habitat types (synchronic approaches sensu Kinnison & Hendry 2001) provide indirect evidence for the potential of adaptive evolution in the face of global change (Box 1). Numerous studies have identified heritable population differentiation in traits that are likely to be important under a changing climate (Jump & Penuelas 2005; Millien et al. 2006). In tree species, there is a large heritable variation in bud set and frost hardiness that covaries with latitude and is correlated with fitness under nonlocal conditions (Savolainen et al. 2004). In the herbaceous Solidago sempervirens, flowering time in a common garden differed by as much as 40 days (Goodwin 1944). In a conceptually similar study, Mousseau & Roff (1989) reveal fine-tuned adaptation to a latitudinal cline in season length in the striped ground cricket, Allonemobious fasciatus. Here, onset of diapause is highly heritable and under diverging selection across the cline. The critical variable that is left unknown is how long the populations have been separated; that is, the number of generations required to achieve the observed divergence.
Latitudinal gradients that are apparently smooth at higher spatial resolutions (say, 1 m to 1 km), are better described as habitat mosaics (Helmuth et al. 2005). As a corollary, informative environmental contrasts, relevant in a global change context, can also be found at small spatial scales (100 m to 1 km). In a series of experiments on natural populations of Drosophila melanogaster inhabiting the north- and south-facing slope of ‘Evolution Canyon’ in Israel, Nevo (2001) utilized the natural contrast of temperature and desiccation stress to study rapid evolutionary divergence with respect to microclimate at the level of phenotype, gene expression, molecular evolution and sexual selection (overview in Korol et al. 2006). Here, the proximity, and thus the high likelihood of gene flow between the populations, suggests that strong environmental selection can be met with genetic response.
CO2 vents in geologically active areas of Iceland provide unique opportunities to study evolutionary consequences of variation in atmospheric CO2 concentration. Certain species of plants growing near the springs are distributed across a natural CO2 gradient. Cook et al. (1998) used this ‘natural experiment’ to compare morphological and physiological traits of plants growing in different CO2 microhabitats. Under increased CO2, plants did not sequester more carbon but invested less into the photosynthetic apparatus, including Rubisco synthesis. There was some evidence that plants allocated excess carbon into other metabolic functions, and that this altered allocation had a heritable component.
Patterns of genetic polymorphism
Latitudinal frequency shifts of genetic marker polymorphism provide another line of evidence for climate-mediated selection (Caicedo et al. 2004; Stinchcombe et al. 2004). For example, the variations at two marker loci in Drosophila melanogaster in Australia, one chromosomal inversion and one enzyme polymorphism (alcohol dehydrogenase) are all correlated with latitude. Umina et al. (2005) compared recent patterns of variation to those sampled in 1979 and 1982. Both linear clines changed in intercept but not slope between the 1980s and 2002, corresponding to a displacement of climate zones of 400–800 km (Umina et al. 2005). Similar changes have been reported in D. subobscura in Europe (Rodriguez-Trelles & Rodriguez 1998). In another fruit fly species, D. robusta, temporal changes in clinal polymorphism occurred at several locations in North America independently, providing firm evidence for climatic selection driving allele-frequency change (Levitan & Etges 2005). In a remarkable time series from 1947 to 2003, the clinal distribution of a chromosomal inversion polymorphism linked to thermal adaptation also shifted upwards in an altitudinal transect in D. robusta, in accordance with climate warming (Etges et al. 2006). Shifts in clinal polymorphism are suggestive of climate-driven selection, but direct fitness estimates of segregating genotypes in contrasting environments are needed to demonstrate the adaptive value of the genetic polymorphism (for ADH locus see Alahiotis 1982).
Experimental selection: simulating evolution as a response to global change
The above examples convincingly demonstrate that evolution has occurred in response to global change. However, what we really would like to know is whether and how evolution will play out in the near future, and whether it would be fast enough to keep track with environmental change (Gomulkiewicz & Holt 1995). Selection experiments provide the only means to test and predict the response to projected conditions; for example, when simulating predicted increases in atmospheric or aquatic CO2 concentrations (Collins & Bell 2004; Li et al. 2006; Box 2), or by translocating populations to microclimatic locations representative of future climatic conditions (Etterson & Shaw 2001; Rehfeldt et al. 2002; Box 1).
Assessing selection within a single generation informs us which phenotypes are likely to contribute more than others to the next generation, but such an approach does not yield information on the evolutionary response (Endler 1986; Fuller et al. 2005). How trait values change across generations depends on the pattern of additive genetic variances and covariances of targeted traits, the G-matrix (see also Schluter 1996). However, within-generation estimates of selection may be the only option, for example if organisms are long-lived, and/or when the effects of selection are to be measured in a natural setting.
Translocation experiments are an underused, but very valuable, tool for testing the effects of semicontrolled natural selection. By moving local genotypes to a climate zone that is predicted as a future condition, their evolutionary potential can be directly assessed (Etterson & Shaw 2001; Rehfeldt et al. 2002). In a seminal study, Etterson & Shaw (2001) employed a nested, paternal half-sib design to estimate additive genetic variances and covariances for drought-related traits in the legume Chamaecrista fasciculata. They exposed populations to nonlocal conditions that simulate drier and warmer climates predicted within 40 years. Although genetic variation for traits critical to fitness at drier locations was observed, the direction of selection on trait combinations conferring higher fitness was antagonistic, slowing down adaptive evolution to such an extent that the species would not be able to keep pace with the predicted change.
The ‘gold standard’ for making predictions is controlled selection experiments over several generations (Box 1). Such experiments have revealed substantial shifts in trait distributions in just a few generations (Table 1), including photoperiod responses (Pulido et al. 2001; Van Dijk & Hautekeete 2007), traits related to dispersal (Ogden 1970; Zera 2005), stress tolerance (Cavicchi et al. 1995; Hoffmann et al. 1997; Gilchrist & Huey 1999; Lerman & Feder 2001) and photosynthetic rates under altered CO2 concentration (Ward et al. 2000; Collins & Bell 2004; Wieneke et al. 2004; Box 2).
Table 1. Selection experiments simulating global change. Study type: L, laboratory; F, field; G, greenhouse; Ng, number of generations. Traits and evolutionary response: (+), trait value responds positive; (–), responds negative; (0), no change
| Abiotic stress tolerance |
|L, 150*|| Drosophila melanogaster ||Natural, different mean temperature (18, 25, 28 °C)||Survival in heat shock (+)Threshold temp f acclimation (+)induction of tolerance (0)||Mean temperature correlated with tolerance to heat shock|| Cavicchi et al. (1995)|
|L, 120|| D. melanogaster||Natural, water vapour (truncation selection)||Water retention (+40%), water storage (+30%), dehydration tolerance (0)||Two of three mechanisms responded, but not water-loss rate|| Gibbs et al. (1997)|
|L, 32*|| D. melanogaster||Artificial, knockdown temperature +/–||Knockdown temperature (+ in + lines, – in – lines)||Up-selected lines: variation in knockdown temperature exhausted after 30 generations|| Gilchrist & Huey (1999)|
|L, 26|| D. melanogaster||Artificial, desiccation resistance (survival)||Female survival (+) h2 0.11–0.19||One or two genes of large effect invokedbody size showed correlated response Tolerance to water loss and rate of water loss evolved|| Telonis-Scott et al. (2006)|
|L, 10|| D. melanogaster ||Artificial, desiccation||Desiccation resistance (+)Heat-shock resistance (+)||Correlated responses desiccation/ heat-stress tolerance|| Hoffmann & Parsons (1989)|
|F, 2||Barley Hordeum vulgare||Artificial, drought tolerance||Reproductive investment = grain yield under stress (+) h2 = 0.35-0-67||Rapid evolution of drought tolerance with up to 57% more grain|| Ceccarelli et al. (1998)|
|F, 2–8||Maize Zea mays||Artificial, drought tolerance||Biomass (+) N-accumulation under drought (+)||Drought tolerance reveals correlated response with N uptake and metabolic efficiency|| Bänziger et al. (1999)|
|G, 3||Wild mustard Sinapis arvensis||Natural, water supply||Flowering time (−) seedling growth rate (+)||Evolution of stress avoidance, trait correlations accelerate adaptation|| Stanton et al. (2000)|
| Photoperiodic responses and phenology |
|L, 2||Blackcap warbler Sylvia atricapilla||Artificial, autumn migration date||Migration date (+), population mean +7 daysh2 = 0.34–0.45||Rapid evolution possible, trait mean shifted to 1-week delay high heritability|| Pulido et al. (2001)|
|G, 9||Sea beet; Beta vulgaris ssp. maritima||Artifical, lower critical daylength for flowering||Critical daylength (–) h2 = 0.56||Rapid evolution population mean from 13.5 reduced to 11 h|| Van Dijk & Hautekeete (2007)|
| Adaptation to increased CO2 |
|F, 1–6†|| Sanguisorba minor ||Natural, increased pCO2||Number of leaves (+)Reproductive traits (0)||Enhanced carbon sequestration through litter production|| Wieneke et al. (2004)|
|G, 5||Mouse-ear cress, Arabidopsis thaliana||Natural, increased pCO2||Biomass (+), reproductive traits (0)||Adaptation to increased CO2 may enhance reproductive rate|| Ward et al. (2000)|
|L, 1000|| Chlamydomonas rheinhardtii ||Natural, increased pCO2||Growth rate (0), photosynthesis rate (0)||Conditional deleterious mutations impairing CCM‡|| Collins & Bell (2004)|
| Composite selection regime |
|G, 7|| Brassica juncea ||Natural, increased temperature mean + variability, increased pCO2||3 vegetative and 10 reproductive traits (0), only vegetative biomass (+)||Inbreeding prevented adaptive evolution|| Potvin & Tousignant (1996)|
| Dispersal associated traits |
|L, 25||Cricket Gryllus firmus||Artificial, flight vs. flightless wing morph||Wing length (+), fatty acid metabolism (+)||Correlated responses of several physiological traits such as fat metabolism|| Zera (2005)|
|L, 5||Flour beetle Tribolium castaneum||Migration between culture vessels||Rates of exchange among culture vessels (+)||Rapid evolution for migrational behaviour possible|| Ogden (1970)|
Particularly instructive in the context of climate-driven selection is a large body of selection experiments in Drosophila that addressed the evolvability of abiotic stress tolerance, in particular to temperature and desiccation stress. Although the focus of these studies was not explicitly on global-change-related selection, they complement work conducted in natural environments discussed above. Flies were selected in a natural scheme using constant temperature regimes (Hoffmann & Parsons 1989; Cavicchi et al. 1995), as well as by imposing artificial selection on specific components of the heat-shock tolerance (Hoffmann et al. 1997; Bubliy & Loeschcke 2005; Telonis-Scott et al. 2006). In Drosophila, the evolution of stress responses is often correlated among stressors. Flies (D. melanogaster) selected for desiccation hardiness were also more tolerant to heat stress and vice versa (Hoffmann & Parsons 1989; Bubliy & Loeschcke 2005). In part, this is likely due to physiological overlap of stress response. These results, however, could not be repeated by Telonis-Scott et al. (2006), suggesting that some trait correlations reflect specific genetic architectures of populations (Harshman & Hoffmann 2000).
None of the changes in abiotic variables predicted under global change comes in isolation. Rather, complex combinations of selective factors will be the norm: an increase in sea-surface temperatures will come along with increasing acidification, while on land, drought will be correlated with increasing temperatures, with a high likelihood of synergistic effects observed in organismal responses (Drake et al. 2005). The factorial combination of abiotic variables in controlled selection experiments are needed to decompose the effects of various selective agents. Unfortunately, we are not aware of any such study that has tested environmental effects independently and in combination.
The value of laboratory selection experiments has been questioned (Harshman & Hoffman 2000). The common practice of keeping all other conditions benign while only imposing stressful conditions with respect to one environmental factor, may produce results that do not pertain to more complex selection regimes. For example, Gibbs et al. (1997) found that populations of D. melanogaster enhanced desiccation resistance by taking up higher amounts of water before being subjected to drought conditions, and by reducing water loss during water deprivation. However, in nature, the first option is unavailable under water, thus only rates of water loss evolve (Gibbs 1999), compromising the extrapolation of data to field situations. However, the adaptive value of acclimation to heat stress by heat hardening was demonstrated in D. melanogaster under natural conditions, suggesting that fitness assessments are possible to yield novel insight into the evolution of stress tolerance (Loeschcke & Hoffmann 2007).
The molecular genetic basis of evolution under global change
As suggested above, it is critical to identify the genetic basis of traits important under global-change-induced selection. A good starting point is to consider (i) genes that have been demonstrated to be responsible for traits important under global change; and (ii) how this knowledge will complement work at the level of phenotypes. In the following section we highlight molecular genetic details gleaned from in genetic model systems. These studies will serve as starting points for candidate gene approaches in targeted nonmodel organisms.
Genes underlying stress tolerance
The predicted increase in extreme events under global change will impose episodic stress upon organisms, such as desiccation or heat, favouring tolerant phenotypes (Bijlsma & Loeschcke 2005). Enzymes with a key function for coping with stress are molecular chaperones. These enzymes protect diverse metabolic functions by assuring correct folding of many enzymes during protein biosynthesis, and by rapid degradation of broken proteins. Chaperones are particularly important in ectothermic organisms that lack the variety of homoeostatic mechanisms of endothermic taxa. An important group of chaperones are heat-shock proteins (hsps) that are encoded by genes that belong to several gene families and that are grouped by their molecular weight (for an overview for animals, see Feder & Hofmann 1999; for plants, see Boston et al. 1996). While hsps are structurally conserved, their expression level is under selection and often varies in a pattern consistent with a thermal environment (overview in Sorensen et al. 2003). For example, Osovitz & Hofmann (2005) found lower temperature thresholds for the induction of hsps in northern vs. southern sea-urchin populations. Similar threshold by environment patterns at hsp-loci that are consistent with climate-driven selection have been identified in Drosophila species (Frydenberg et al. 1999; Frydenberg et al. 2003).
Some of the most widely studied genes belong to the hsp70 (70 kDa hsp) family. Here, cis-acting regulatory changes probably underlie differences in expression level (Michalak et al. 2001; Bettencourt et al. 2002). Interestingly, in Drosophila, the rate of evolution in hsp-promoter regions seems to be accelerated because they are prime targets for transposon insertions, with concomitant changes in gene activity (Walser et al. 2006). Heat-shock transcription factors could also be important targets of selection under heat stress. These trans-acting factors, in particular heat-shock factors hsf, deserve closer attention (Feder & Hofmann 1999; Baniwal et al. 2004). The expression of hsf increased in response to experimental selection for heat-stress tolerance in Drosophila (Lerman & Feder 2001), suggesting that this represents a second potential genetic mechanism for altering hsp expression aside from cis-acting regulatory changes.
Genetic polymorphism associated with photoperiodic responses
Life-history events such as hatching, diapause, flowering or migration need to be precisely coupled to seasonality in many animals and plants (e.g. Both & Visser 2001). Under climate warming, existing photoperiod cues in the genetic programmes of local populations are suboptimal, because an identical daylength now corresponds to altered temperature regimes. In order to accurately determine the photoperiod, many species measure daylength using an internal, self-regulatory oscillating system with a period of c. one day (hence circadian). Most molecular genetic knowledge comes from genetic model systems, notably Arabidopsis and Drosophila. In Arabidopsis, past work has focused on genes involved in the regulation of flowering; a trait that is highly variable among populations and correlated with local climate. Four major pathways receive endogenous (circadian clock, hormones) and exogenous (temperature, light) signals. Floral integrator genes summarize all four different input pathways and serve themselves as transcription factors for the development of flowers. Although over 100 genes have been identified to be involved, natural polymorphism has only been identified at a few loci (Koornneef et al. 2004).
The flowering locus C (FLC) integrates signals from the autonomous and vernalization pathway. This MADS-Box transcription factor suppresses flowering by default, leading to a winter annual habit in higher latitudes. FLC interacts synergistically with its upstream repressor, frigida (FRI) (Johanson et al. 2000). In nature, allelic variants of FRI and FLC control flowering time epistatically (Caicedo et al. 2004). Whereas for FRI most allelic variants are null alleles, for FLC several functional allelic variants have been found in natural populations (Caicedo et al. 2004). Natural variation within another regulator pathway, the photoperiod pathway, has been identified in the blue light photoreceptor gene CRY2. Two different variants show different sensitivities to entrainment by sunlight and hence revealed different stimulation for flowering by short days (El-Assal et al. 2004).
The mechanisms of molecular genetics of the circadian regulatory system among flowering plants, in particular crop species, is currently under intense investigation. Other members of the Brassicaceae seem to have conserved much of the regulatory pathways elucidated in A. thaliana. In sea beet (Beta vulgaris, Chenopodiaceae) a quantitative trait locus (QTL) analysis revealed the action of several minor loci (Van Dijk & Hautekeete 2007). In rice (Oryza sativa), a subtropical plant, the flowering regulation networks lack the vernalization pathway, including the genes FRI and FLC. Here, natural variation for flowering time is found at different points of a partly homologous network of genes regulating flower induction. Natural polymorphism has been identified in HD1 (homologous to the Arabidopsis gene constans) and in HD3a (homologous to flowering locus T of Arabidopsis, Izawa et al. 2003). While the mechanistic basis of flowering time is clearly distinct across monocotyledons and dicotyledons, the usefulness of model systems as a template for identifying adaptive genetic variants in closer allies awaits further investigation.
In animals, genes underlying seasonal adaptation and circadian rhythms are best understood in the Drosophila species. A key gene is period that is involved in the auto-catalytic circadian loop. In D. melanogaster sampled across Europe, the clinal polymorphism of the period gene (Costa et al. 1992) is correlated to different capabilities of temperature compensation under local conditions, which are in turn related to the length of a polymorphic repeat element consisting of two paired amino acids, threonine–glycine (Sawyer et al. 1997). A more northern allele was correlated with a period significantly shorter that 24 h. This allele was better able to maintain relative stability of circadian rhythm under variable temperature than a southern allele with a more accurate 24-h period. Such genetic differentiation was also found over distances as small as 400 m at the north- and south-facing slopes of ‘Evolution Canyon’ under widely varying microclimatic conditions (Korol et al. 2006). Homologues of period that also show polymorphism associated with different photoperiodic phenotypes have been identified in other insect species (Takahisa et al. 2002), suggesting that a candidate-gene approach may be a valuable strategy for uncovering the genetic basis of photoperiodic shifts in this order.
The ‘ecogenomic toolbox’: uncovering the genetic basis of adaptation to global change
Current genetic models only cover a restricted range of life histories, metabolic diversity and habitat types. Thus, the genetic/genomic diversity of life is poorly represented (Jackson et al. 2002). Among the underrepresented species are many keystone or ecosystem engineering species such as trees, corals or seagrasses that provide the structural basis of communities. This gap seriously compromises predictions on the fate of communities and ecosystems in the face of global change. Fortunately, this gap is clowing rapidly. Genomic tools (‘ecogenomic’ tools, Ouborg & Vriezen 2007) are rapidly gaining broader application (Jackson et al. 2002; Feder & Mitchell-Olds 2003; Purugganan & Gibson 2003; Vasemägi & Primmer 2005). Such approaches will allow systematic identification of key genes and pathways underlying traits that are critical for population persistence under global change.
QTL approaches. QTL approaches take advantage of crosses between phenotypically divergent lines. QTL describe Mendelian-inherited locations in a genome that explain statistically significant proportions of phenotypic variation in a segregating population (see recent reviews by Erickson et al. 2004; Slate 2005). QTL analyses have lead to the successful identification of genes underlying phenotypic divergence in genetic model species; for example in gene loci affecting flowering time (Johanson et al. 2000; Izawa et al. 2003; Michael et al. 2003; El-Assal et al. 2004) or temperature tolerance (Somorjai et al. 2003; Norry et al. 2004; Morgan & Mackay 2006). An important advance are novel statistical tools that allow trait mapping in unmanipulated, outbred populations, solely based on pedigree information (Slate et al. 2002). The identification of causal gene polymorphism in the vicinity of the marker locus requires additional resources such as large insert genomic libraries, restricting the full utility of QTL approaches to genetic model organisms. Nevertheless, the resources required for QTL analysis and subsequent gene identification will become available increasingly also for nonmodel species (see, for example, Hofmann et al. 2005), given the methodological progress in obtaining large numbers of genetic markers at relatively low cost.
Even if an identification of the genetic polymorphism that is causally responsible for a given trait is not possible, information on the number and mode of action of genes contributing to complex traits is important for predicting the rate and magnitude of sustained evolutionary change (Lande 1981; Barton & Turelli 1987; Erickson et al. 2004). QTL approaches can analyse whether genetic correlations are transient due to gametic phase disequilibrium or caused by tight physical linkage or pleiotropy (Hawthorne & Via 2001). While, in the former case, they can be broken up by selection, the evolutionary response is more constraint in the latter case (Lynch & Walsh 1998).
Genome scans. Efficient marker technology now allows for the development of tens of genetic markers at reasonably low costs (Luikart et al. 2003; Morin et al. 2004; Bensch & Akesson 2005; Bouck & Vision 2007). Genome scans use moderate to large numbers of genetic markers to sample or ‘scan’ the entire genome at the population level. In a bottom-up approach, divergent selection across environments or among contrasting phenotypes can be associated with marker polymorphism due to genetic hitchhiking (Maynard Smith & Haigh 1974). In cross-population studies, marker loci displaying a larger population genetic differentiation than expected under a neutral null model are presumably under positive (divergent) selection (Vitalis et al. 2001; Beaumont 2005). Genome scans have recently been applied to polymorphisms relevant under global change. In beech (Fagus sylvatica), populations were genotyped across an altitudinal gradient used as surrogate for climate change. Among a total of 254 scored amplified fragment length polymorphism (AFLP) fragments, one showed exceptionally high genetic differentiation across different altitudes, suggesting diverging, climate-driven selection (Jump et al. 2006).
A second, more challenging step after a genome scan is to characterize the underlying functional polymorphism linked to marker loci. To this end, detailed genetic information is needed in the vicinity of the target marker locus, limiting the utility of genome scans in nonmodel organisms where no linkage map is available. One new marker type, EST (or gene)-linked microsatellites (Bouck & Vision 2007), may partly alleviate this problem because in this approach, a genetic polymorphism is directly linked to a gene with a putative function.
Using the association of genotypes with contrasting habitats to infer past selection is risky because these patterns are influenced by recent demography and phylogeography (Hewitt 1999). An important advance is therefore to implement study designs that allow for replicated tests of molecular genetic adaptation. For example, in an AFLP-based genome scan in a coastal snail, small-scale genetic differentiation was determined between high- and low-shore intertidal populations, strongly differing in desiccation and thermal stress (Wilding et al. 2001). Reassuringly, these parallel scans revealed a number of identical marker fragments (15) to diverge more than under a population genetic null model. Experimental replication is easily possible when environmental gradients are found on a small spatial scale; for example, altitudinal climatic gradients or gradients in desiccation and exposure associated with tidal elevations (Wilding et al. 2001; Bockelmann et al. 2003; Campbell & Bernatchez 2004).
Transcription profiling. Natural variation in gene regulation is a critical determinant of phenotypic variation (Doebley & Lukens 1998; Ferea et al. 1999; Van Laere et al. 2003). Our ability to quantify gene expression via mRNA abundance has provided an important new target for empirical research. Researchers can now use microarray technology to survey the genome for elements that are up- or down-regulated in specific conditions (Gibson 2002; Ranz & Machado 2006). Microarrays quantify the expression level of between hundreds and thousands of mRNAs that can be drawn, for example, from different populations, or from within populations but under contrasting abiotic conditions (e.g. stress/nonstress).
Transcription profiling has recently been questioned as an effective tool for the discovery of genes that are functionally important and display variable expression (Feder & Walser 2005). There are limits to the inference from this approach which we discuss below. However, we should not be overly pessimistic. For example, Quinn et al. (2006) used expression profiling to search for genes and metabolic networks involved in the calcification of the marine coccolithophorid algae Emiliania huxleyi, that play an important role in the global carbon cycle. The study compared the expression patterns of algae grown under calcifying and noncalcifying conditions. Several genes were identified that play a role in the formation of biogenic calcareous structures, including a novel gamma-class carbonic anhydrase involved in the concentration and delivery of inorganic carbon prior to calcification.
Several studies have measured transcriptomic responses to a number of environmental changes such as to increased CO2 concentration, extreme temperatures, drought or salinity (e.g. Atienza et al. 2004; Rizhsky et al. 2004; Swindell 2006). However, such assessments cannot determine whether gene-expression regulation is adaptive and heritable. Studies that address the evolution of gene transcription by comparing the end product of adaptive divergence, for example heritable expression differences among ecotypes, are still rare (but see Derome & Bernatchez 2006; Li et al. 2006). One example relevant to global change is the seminal study of thermal adaptation in the killifish Fundulus heteroclitus, a small, coastal fish widely distributed along the east coast of North America (Whitehead & Crawford 2006). Individual F. heteroclitus were sampled from five locations differing by as much as 12 °C in annual mean temperature along the latitudinal gradient and were raised in a common garden environment. The variance in transcriptomic responses of 329 genes were analysed to partition within- and among-population components of transcription regulation. Although the majority of significant expression differences were neutral, a substantial fraction of transcripts showed evidence for selection. Thirteen genes were found to be under divergent selection, while purifying selection acted on 24 genes. The most parsimonious explanation is that these differences constitute adaptations of transcriptional regulation to different thermal regimes (Whitehead & Crawford 2006).
Due to the large number of hypotheses being asked, transcription profiling using high-throughput microarrays suffers from the detection of false positive gene candidates, while mRNA abundance may be a poor indicator of final protein abundance due to post-translational modifications (Feder & Walser 2005). Clearly, physiological measurements of protein abundance for selected gene candidates, or even proteomic approaches, are a valuable addition to pure array studies (de Vienne et al. 2001).
Alternatives to full-blown microarray/oligonucleotide-array approaches in nonmodel organisms exist that do not require the expensive development of a microarray. In cDNA–AFLP studies, bulk mRNA samples from different experimental conditions are first transcribed into cDNA (Bensch & Akesson 2005). Transcripts are then tagged and identified after an AFLP separation. Fragments differing in intensity among conditions or populations can be excised from gels and sequenced. Using such an approach, Knight et al. (2006) compared the transcriptomic response of two locally adapted populations of Boechera holboellii (an Arabidopsis relative) that differ in their tolerance to drought, presumably due to local adaptation. The study tested both populations under drought and no-stress conditions and identified 450 gene fragments that were differentially expressed among water supply conditions, several of which may underlie higher water-use efficiency in the xeric population.
Synergism among ecogenomic approaches. Complementary ecogenomic approaches are often necessary to identify the genetic basis of traits (Vasemägi & Primmer 2005). In particular, the combination of ‘top-down’ approaches, such as QTL mapping, with ‘bottom-up’ approaches, such as genome scans and transcription profiling, can be successful. Several recent studies highlight the potential of a combined QTL/genome scan approach (Rogers & Bernatchez 2005). In a study comparing Atlantic salmon populations (Salmo salar) across habitat types, one microsatellite marker locus (Ssa14) that was under divergent selection between populations in a genome scan (Vasemägi et al. 2005) coincided with a QTL associated with thermal tolerance (Somorjai et al. 2003). In a hybrid sunflower inhabiting salt marshes, three QTLs associated with salt tolerance were independently confirmed using a microsatellite genome scan. Strongly reduced variability at the three marker loci was inferred to be the result of strong directional selection that fixed a single allelic variant that conferred salt tolerance (Edelist et al. 2006).
QTL mapping has also been combined with expression profiling to place gene-expression differences among segregating lines on a linkage map (Schadt et al. 2003). In this approach, gene-expression data are treated as molecular ‘phenotypes’, and the question asked changes to where the transcriptional regulation is located. The identification of map locations for such ‘eQTL’ allows inferences on the nature of regulatory changes. When ‘eQTL’ and genomic QTL coincide, then the regulatory polymorphism is close to the target gene, or cis-acting. On the other hand, when the ‘eQTL’ maps to a distant location, then the polymorphism leading to expression changes is trans-acting (Hughes et al. 2006). The relative proportion of both regulatory modes determines the potential speed of regulatory evolution (Hughes et al. 2006), as well as the presence of genetic correlations that are antagonistic (Ranz & Machado 2006).
Future research directions
What is the genetic basis of evolutionary stasis?
Given the many recent examples of rapid evolution (reviewed in Kinnison & Hendry 2001; Reznick & Ghalambor 2001), a central research question is when and why populations fail to evolve (Antonovics 1976; Bradshaw 1991; Both & Visser 2001; Meriläet al. 2001; Blows & Hoffmann 2004; Nussey et al. 2005). For example in populations of the Australian fruit fly Drosophila birchii, a species which is distributed along a humidity gradient from tropical to subtropical latitudes, no response in terms of increasing tolerance of desiccation stress could be measured in selection lines (Hoffmann et al. 2003), putting this species at risk of local extinction under the predicted drier climate. Other experimental reports of stasis include a lack of response in upper (but not lower) temperature tolerance in a fish (Heterandria formosa) after eight generations of artificial selection (Baer & Travis 2000).
Stasis despite directional selection may be due to a general lack of polymorphism of genes underlying traits, due to negative trait correlations caused by pleiotropy or linkage, or due to metabolic costs and associated trade-offs. Molecular ecology approaches hold great promise to disentangle these different causes and hence to explain evolutionary stasis. Insufficient genetic variation for selected traits may be directly assessed once the genetic basis of traits has been identified using approaches outlined in the previous sections. Particularly instructive would be recent examples of selection experiments where no trait evolution could be observed despite strong directional selection (Gilchrist & Huey 1999; Baer & Travis 2000; Van Dijk & Hautekeete 2007), or field studies that monitored selection over time but failed to identify trait shifts (Both & Visser 2001; Nussey et al. 2005).
Whether or not novel mutations or only standing genetic variation contribute to adaptive polymorphism, is also an open question. Within the life-time of a researcher, the mutation limitation of evolution can be best addressed in short-lived organisms, often unicellular, that have short generation times (Lenski & Bennett 1993; Collins & Bell 2004). Regardless of the generation time, in some cases of experimental selection, lines respond seemingly interminably (examples in Falconer & McKay 1998), while in others, variation is exhausted rapidly (Blows & Hoffmann 2004).
Trait correlations as a cause of stasis. Correlations among traits may accelerate (Stanton et al. 2000) or constrain adaptive evolution (Blows & Hoffmann 2004). For example, if natural selection acts antagonistically on positive trait correlations this may pose another important constraint to evolutionary responses (Etterson & Shaw 2001). The molecular genetic basis of negative trait correlations can be a result of linkage disequilibria among two or more loci, or due to antagonistic pleiotropy (Coyne & Lande 1985; Cooper & Lenski 2000; Roff & Fairbairn 2007). Distinguish ing between these proximate mechanisms is imperative for predicting evolutionary trajectories in the face of environmental change (Cooper & Lenski 2000). While linkage can be rapidly broken up by recombination, pleiotropic gene action will remain a constraint to selection for much longer (Conner 2002). Whole-genome transcription profiling revealed that the actual amount of pleiotropy in the genome may be pervasive, with 100–200 genes having a trans-acting regulatory effect upon a total of 1716 downstream genes in yeast (Yvert et al. 2003).
Molecular genetic costs as limits to trait evolution. Stress tolerance is a special form of physiological plasticity that is itself under genetic control and can evolve (Schlichting & Pigliucci 1995; Nussey et al. 2005). The evolution of plasticity is constrained by costs, otherwise all organisms would be indefinitely plastic, i.e. live and perform well in all environments (Angilletta et al. 2003; Sorensen et al. 2003). Functional genetic analysis has revealed important insights into the nature of such costs. For example, Drosophila melanogaster lines selected for increased heat-stress tolerance expressed less, not more, mRNA of hsp70 compared to control lines (Bettencourt et al. 1999; Sorensen et al. 1999). This paradoxical finding possibly reflects long-term costs of hsp70 expression due to shut-down of other enzyme synthesis (Heckathorn et al. 1996; Taji et al. 2004), direct metabolic toxicity (Krebs & Feder 1997) or allocation to hsp-production (see also Clarke 2003).
Trade-offs as constraints for adaptive evolution. Limits to the evolution of physiological tolerance can also be determined by allocation conflicts, or trade-offs, in particular when organisms face multiple stresses, as is the case with global change. Molecular genetic approaches, in particular transcriptomics, are beginning to identify the functional basis of synergism among stressor impacts. It has recently been suggested that organismal trade-offs reflect physiological conflicts resulting from interfering transcriptomic syndromes (Stearns & Magwene 2003). Potential for genomic conflict arises particularly in situations where the up-regulation of one response pathway is detrimental for another physiological function (Roff 2007). This hypothesis is supported by two plant studies that find surprising specificity in gene-expression changes when stressors are applied singly or in combination (Atienza et al. 2004; Rizhsky et al. 2004). In the study by Rizhsky et al. (2004), little overlap in the transcriptomic response of Arabidopsis to drought vs. heat stress was observed, suggesting a collision of tolerance-mediating pathways among different stressors (Rizhsky et al. 2004). Under drought and heat as common stressors, a unique set of 454/24 000 genes was up-regulated in a combined response that was not observed under either stressor alone.
Which genetic responses underlie phenotypic evolution as a response to global change?
Studies of climate-change-induced evolution under simulated and natural conditions rarely integrate phenotypic and genetic evolutionary responses. Experimental work in Drosophila is an exception (Table 2). Here, genetic changes that occurred during the course of artificial selection were also monitored at the protein or genetic level. One focus was on candidate loci that explain variance in temperature tolerance. Responses to selection at the genetic level were documented for the production of hsps (Bettencourt et al. 1999; Sorensen et al. 1999), for allele-frequency shifts in 5′-promoter-region variants of hsp alleles (Bettencourt et al. 2002), and finally in the expression of hsf-transcription factors that in turn regulate hsp expression (Lerman & Feder 2001). Very recently, the effects of artificial selection among heat-stressed and control lines were compared using full-genome transcription profiling (Nielsen et al. 2006; Sorensen et al. 2007). These studies found novel candidate genes other than hsp genes that changed their expression patterns as response to selection and may thus underlie thermotolerance in Drosophila, and possibly, other insect species. It is important to consider whether genetic changes observed during experiments are adaptive at longer time scales, and in nature where the pleiotropic fitness effects and multiple stressors and species interactions are important (see also Harshman & Hoffman 2000).
Table 2. Experimental selection and genetic changes in Drosophila species. Species: Dm, Drosophila melanogaster; Db, D. buzzatti. Ng, number of generations; hsp, heat-shock protein; hsf, heat-shock transcription factor
|Db||7–64||Natural, cycling temps with heat shock vs. constant|| hsp70|| hsp70 expression (–) in larval lines(+) in adult selection lines||larval selection lines suggest costs of hsp70 expression, other mechanism of temperature tolerance invoked|| Sorensen et al. (1999)|
|Dm||> 200||Natural, different mean temperature|| hsp70 expression|| hsp70 expression (–) in larval lines(+) in adult selection lines||costs of hsp70 expression identified|| Bettencourt et al. (1999)|
|Dm||> 200||Natural, different mean temperature|| hsp70, 5′ end allelic variants|| cis-acting hsp70 alleles change in frequency||latitudinal cline in allele frequencies consistent with selection results|| Bettencourt et al. (2002)|
|Dm||> 200||Natural, different mean temperature|| hsf- transcription factor|| hsf-transcription factor activation in heat shock:low temp lines (+)||complex patterns, partly congruent to hsp70 expression|| Lerman & Feder (2001)|
|Dm||11–27||Artificial, different stressors (heat shock, mean temperature, dessication, starvation, cold)||Whole genome expression array (Affymetrix, 13 000 genes)||262 genes differentially expressed among selection regimes, mostly down-regulation compared to control lines||expression patterns partly overlapping among stressorshsp role not confirmed†|| Sorensen et al. (2007)|
|Dm||20*||Artificial, adult sublethal heat shock||Whole-genome- expression array (Affymetrix, 13 000 genes)||no-stress conditions; 108 genes up-regulated, 10 down||novel role of photosensing pathway genes for temperature tolerance†|| Nielsen et al. (2006)|
How is genetic diversity related to adaptive evolution?
Quantifying anonymous genetic variation using molecular markers has been one of the most successful research lines within the field of molecular ecology. The prevailing consensus posits that genetic variation in natural populations is eroded by random loss of alleles and/or through inbreeding when effective population sizes are small, leading to population and individual fitness declines and enhanced extinction risk (Frankham 1995; Spielmann et al. 2004).
There are, however, few studies that link rates and potential for adaptive evolution to ongoing loss of genetic variance due to drift and inbreeding; although theory predicts that rate and potential are diminished (Lynch 1996). In one selection experiment, the failure of the crucifer Brassica juncea to respond to selection of a combination of temperature and increased CO2 concentration was interpreted as result of experimentally induced inbreeding (Potvin & Tousignant 1996). Inbreeding may also be included as an experimental variable (Bijlsma et al. 1997). In the light of ongoing habitat destruction and population fragmentation that will increase genetic erosion (Frankham 1995), experiments are highly warranted that follow adaptive evolution not only as a function of selection but also take into account initial genetic diversity at marker loci.
How does the evolutionary response compare among species?
Comparative molecular ecology provides a framework for understanding and predicting differential evolutionary response. It is now clear that genotypic evolution is repeatable across a variety of taxonomic scales (Wood et al. 2005). This observation suggests that predicting change at the genetic level across taxa may be possible. One testable prediction would be that species or populations which cannot evolve lack polymorphism at key genes or metabolic pathways (see also Blows & Hoffmann 2004). Closely related species that vary in their ability to respond to similar selection pressures are ideal for such comparative studies. For example, while Drosophila melanogaster populations rapidly evolve desiccation tolerance (Hoffmann & Parsons 1989; Gibbs et al. 1997; Telonis-Scott et al. 2006), its congener D. birchii failed to respond despite 30 generations of artificial selection (Hoffmann et al. 2003).
Because rapid evolution often involves altered gene regulation (Doebley & Lukens 1998; Ferea et al. 1999; Van Laere et al. 2003), data on differences in transcriptional response are invaluable for understanding the nature of genetic limits to adaptation. Such an approach was followed in a comparison of two congeneric species that differ widely in salt tolerance; Arabidopsis thaliana and its relative, salt cress Thellungiella halophila. In an example of heterologous microarrays derived from the genomic model, salt cress revealed a number of transcriptomic adaptations when exposed to salt stress, as compared to its nontolerant relative. Several genes associated with induced stress response in Arabidopsis were constitutively expressed in the halophyte (Taji et al. 2004). Species-level, comparative studies may shed light on genes that determine the fundamental niches of species (Kirkpatrick & Barton 1997) and thus will greatly enhance prediction once more genomic data are available (van de Mortel & Aarts 2006).
How does gene flow potentiate or limit adaptive evolution?
A research area where ‘old’ and ‘new’ molecular ecology can unite is the role of gene flow in promoting local adaptive evolution. Traditionally, gene flow has been viewed as an antagonistic process opposing local adaptation by introducing locally maladaptive variants (Slatkin 1987; Kirkpatrick & Barton 1997), but its role may change when the environment changes. In spatially shifting climatic envelopes, gene flow may provide novel alleles from preadapted populations (Davis & Shaw 2001; Shirley & Sibley 2001; Harding & McNamara 2002). Interestingly, such beneficial alleles are expected to spread faster than neutral alleles (van Valen 1982; Morjan & Rieseberg 2004). In widespread species in particular, marginal populations at the warmer distributional border may be a source of adaptive alleles to central parts of the range (Davis & Shaw 2001; Ayre & Hughes 2004; Hewitt & Nichols 2005). Gene flow can be reduced under anthropogenic habitat fragmentation and due to natural geological barriers (Hewitt 1999), highlighting the importance of considering corridors and constraints to population movement (Manel et al. 2003).
Secondary contact of locally adapted subpopulations as a result of gene flow can also generate extreme phenotypes that exceed the variance displayed by parental populations. Such transgressive hybridization may be common among animals and plants and may speed up adaptive evolution; in particular when evolutionary responses are constrained by major gene loci that are antagonistic (Rieseberg et al. 2003). A good example is the work by Lewontin & Birch (1966) who demonstrated that admixture of differentiated genomes triggered range expansion in the Australian fly, Dacus tryoni.
In order to understand the complex interactions among local genetic diversity, genetic divergence at the metapopulation scale and gene flow, the integration of gene-exchange estimates using high-resolution neutral markers, with studies on adaptive polymorphism and evolution, is an important future direction (Harding & McNamara 2002; Hanski & Gaggioti 2004). Researchers can now take advantage of novel methods to detect the directionality of gene flow (Gaggiotti et al. 2002; Manel et al. 2005). In addition, novel techniques can allow to distinguish between historical and ongoing genetic exchange (Wilson & Rannala 2003), while Bayesian estimation of population parameters overcome traditional limitations of statistical estimation (Beaumont & Rannala 2004). In addition, genetic markers can now be developed quickly (Zane et al. 2002; Bouck & Vision 2007), and the incorporation of more loci allows gene-flow estimates of greater precision (Gaggiotti et al. 2002). Increasingly, landscape models of population connectivity go beyond simple distance-based estimates of dispersal; an important step towards predicting the potential spread of favourable genotypes/genes throughout the distributional ranges of populations (Manel et al. 2003). Over a distance of 1800 km, Veliz et al. (2006) have modelled the frequency cline of two allozyme polymorphisms (MPI, mannose phosphate isomerase and GPI, glucose phosphate isomerase) in the acorn barnacle (Semibalanus balanoides) as a function of selection and gene flow. They were able to identify an equilibrium solution that correctly predicted allele frequencies in the face of ample gene flow.
This review is guided by the recognition that global change alters ecological conditions, and thus the selection regime (Bradshaw & McNeilly 1991; Lynch & Lande 1993; Bradshaw & Holzapfel 2006). In light of the palaeoecological and fossil record, ecological responses — in particular range shifts and migratory changes — seem more important than rapid evolution for the persistence of species and biological communities (Hewitt 1999; Jackson & Overpeck 2000; Ackerly 2003; Parmesan 2006; but see Hellberg et al. 2001; Roy & Pandolfi 2005). On the other hand, many physiological traits that are currently under directional selection — such as stress tolerance or photoperiodic response — are not likely to be preserved in the fossil record, but are nevertheless critical for individual fitness and population persistence. Moreover, some traits such as altitudinal range limits reveal a great deal of variability within genera, indicating the potential of niche shifts as a response to climate change (Peat & Fitter 1994).
We thus argue that any ‘either-or’ controversy should be abandoned in favour of a synthetic view that unifies ecology and evolution (Hairston et al. 2005). A central role for synthesis will be to define population-level attributes such as connectedness and effective size, as well as a species’ life-history attributes that are correlated with the relative importance of local adaptive evolution. We also note that all three major ecological responses, namely migration, plasticity and extinction, interact with evolutionary processes (Davis et al. 2005). Migration and dispersal mediate the rapid diffusion of favourable alleles throughout a species’ distributional range, supplying individuals in the expanding, leading edge with genes and alleles necessary to adapt to previous conditions experienced within the species’ central area of distribution (van Valen 1982; Davis & Shaw 2001; Hewitt & Nichols 2005). Dispersal capabilities in animals and plants also evolve rapidly (Ogden 1970; Venable et al. 1995; Cody & Overton 1996; Hill et al. 1999; Thomas et al. 2001; Zera 2005), which forces us to consider evolution as an important component of range expansions. Finally, stasis, the failure of a population to evolve despite selection, is an alternative way of looking at ecological extinctions. As Antonovics (1976) pointed out over 30 years ago, the ecological question of species abundance and distribution, and hence of persistence in the future, is in fact a genetic question: what is setting the limits to natural selection?
With one of the most drastic environmental changes in the earth's history just ahead, a fundamental goal is to predict biotic response to global change, in particular, whether or not critical traits can evolve fast enough to avert local extinction (Lynch & Lande 1993; Gomulkiewicz & Holt 1995). Organisms, like climate systems, are almost overwhelmingly complex, and making predictions about such systems is daunting. Mutant analysis, i.e. the systematic destruction and recovery of genes, has been critical for finding molecular genetic correlates of many physiological and morphological traits (Chippindale 2006). Now we need to know more about natural allelic variants associated with a particular phenotype that influence fitness in nature (Koornneef et al. 2004; Colosimo et al. 2005; Cano et al. 2006). This challenge is in front of us, forcing us to take an organized, informed and rigorous approach. Indeed, the challenge presented by climate change is the ultimate test of evolutionary ecology. The discipline of molecular ecology will certainly play a major role in identifying genetic attributes and key genes that contribute to a causal analysis of evolution and stasis in the face of global change.
Thorsten BH Reusch holds a chair in Evolutionary Ecology at the University of Münster as one of the core groups of the newly founded Institute for Evolution & Biodiversity. He is interested in the role and maintenance of genetic diversity in natural populations, working on diverse aquatic species including sticklebacks, trematodes, seagrasses and pondweeds. Troy E. Wood recently started as a Scientist at the University of Münster with the Plant Evolutionary Ecology group. He studies plant speciation in the genus Ipomopsis, as well as the role of polyploidy in the establishment of distinct lineages in vascular plants.