Plant sexual reproduction during climate change: gene function in natura studied by ecological and evolutionary systems biology.

BACKGROUND: It is essential to understand and predict the effects of changing environments on plants. This review focuses on the sexual reproduction of plants, as previous studies have suggested that this trait is particularly vulnerable to climate change, and because a number of ecologically and evolutionarily relevant genes have been identified. SCOPE: It is proposed that studying gene functions in naturally fluctuating conditions, or gene functions in natura, is important to predict responses to changing environments. First, we discuss flowering time, an extensively studied example of phenotypic plasticity. The quantitative approaches of ecological and evolutionary systems biology have been used to analyse the expression of a key flowering gene, FLC, of Arabidopsis halleri in naturally fluctuating environments. Modelling showed that FLC acts as a quantitative tracer of the temperature over the preceding 6 weeks. The predictions of this model were verified experimentally, confirming its applicability to future climate changes. Second, the evolution of self-compatibility as exemplifying an evolutionary response is discussed. Evolutionary genomic and functional analyses have indicated that A. thaliana became self-compatible via a loss-of-function mutation in the male specificity gene, SCR/SP11. Self-compatibility evolved during glacial-interglacial cycles, suggesting its association with mate limitation during migration. Although the evolution of self-compatibility may confer short-term advantages, it is predicted to increase the risk of extinction in the long term because loss-of-function mutations are virtually irreversible.
CONCLUSIONS: Recent studies of FLC and SCR have identified gene functions in natura that are unlikely to be found in laboratory experiments. The significance of epigenetic changes and the study of non-model species with next-generation DNA sequencers is also discussed.

INTRODUCTION: INTEGRATING EVOLUTION, ECOLOGY AND MOLECULAR BIOLOGY

The effects of changing environments on organisms are one of the foci of contemporary science (Intergovernmental Panel on Climate Change, 2007), and it is essential to understand and predict them. The negative effects of climate change on biodiversity and food production have been sources of concern (Intergovernmental Panel on Climate Change, 2002; Hedhly ). Such effects may depend largely on the responses of plants in terms of sexual reproduction, because plant reproductive success determines the levels of resources that support both biodiversity and the food supply, and because plant sexual reproduction is considered to be particularly vulnerable to the effects of global warming (Hedhly ; Eckert ). Hedhly proposed that plant sexual reproduction under temperature stress could be altered both by phenotypic plasticity (non-genetic responses) and evolution (genetic responses). Phenotypic plasticity is defined as the capacity of a single genotype to produce a series of phenotypes in response to environmental changes. Evolutionary (or microevolutionary) responses imply changes in allele frequencies in populations over generations. Although many studies have focused on the effects of past and future climate changes on the ranges and abundance of species through migration, much less is known about phenotypic and evolutionary responses (Davis ; Gienapp et al., 2008; Chown ).

Molecular genetic studies in laboratory conditions may not be adequate to study responses to changing environments, because laboratory environments can differ from natural habitats, characterized by large, stochastic fluctuations. The typical laboratory environment of A. thaliana includes adequate water, a constant warm temperature, and a lack of natural herbivores, which are not features of most of its natural habitats (Hoffmann, 2002). As a consequence, when mutants and natural accessions of A. thaliana are grown under field conditions, their observed phenotypic plasticity and fitness often differ from those observed under laboratory conditions (Richards ; Wilczek ). Furthermore, mapping the loci of quantitative traits under natural conditions has identified a number of new loci, suggesting that mutations with no phenotypic effects in the laboratory could play specific roles in natural environments (Weinig ). In short, laboratory conditions provide a useful system, but this system may neglect the complexity of its interactions with its naturally fluctuating abiotic and biotic ecosystems. On the other hand, traditional quantitative genetic models may not have sufficient power to distinguish phenotypic plasticity and evolution because their assumptions, including heritability and genetic correlations, are often violated (Gienapp ). Gienapp proposed that the use of genomic approaches to identify ecologically important variations would benefit the study of the biological responses to changing climates.

To understand and predict the biological responses to changing environments, we propose the importance of studying gene functions in natura, or gene functions in natural ecosystems (Fig. 1), analogous to the expression of ‘immunology in natura’ (Quintana-Murci ) in contrast to in vivo (in the living organism) and in vitro (in the test tube). It is essential to analyse gene functions or phenotypes at multiple levels in natural ecosystems, because genes in natura affect the interactions of an organism with individuals in a population of the same species, with other species in the community, or with fluctuating environments or climates (Fig. 1).

Fig. 1.

Potential gene functions in natura or at multiple levels in ecosystems.

To study gene function in natura, it is obviously essential to integrate molecular biology, ecology and evolutionary biology. An established interdisciplinary field is evolutionary genomics (or evolutionary and ecological functional genomics) (Feder and Mitchell-Olds, 2003). Ecologically and evolutionarily relevant genes have been identified and studied using genomics to address evolutionary and ecological questions, including selective advantage in an ecological context and evolutionary timing during historic climatic fluctuations (Shimizu, 2002; Shimizu and Purugganan, 2005; Mitchell-Olds and Schmitt, 2006) (see the study of the evolution of self-compatibility explained below). Moreover, in this short review, we discuss that a broader integration incorporating systems biology and evolutionary genomics provides a powerful framework for predicting biological responses, in particular, plastic responses. Systems biology aims for biological understanding at the system level, and is characterized by large-scale quantitative data, modelling, networks and prediction (Kitano, 2002). So far, most systems biological research has been conducted in interdisciplinary collaborations, by applying mathematical and physical methods to molecular data. However, it should be emphasized that ecologists, evolutionary biologists and geneticists have been the pioneers of mathematical methods in biology (Begon ; Freeman and Herron, 2007; Benfey and Mitchell-Olds, 2008). To integrate the quantitative tools of ecology and evolutionary biology with those of molecular biology would establish a discipline of systems biology within an ecological and evolutionary context, called ‘ecological and evolutionary systems biology’ (Richards ). When combined with molecular biology, ecological data, such as long-term meteorological data, would provide useful tools to aid construction of a predictive model of gene functions in natura. A major aim of ecological and evolutionary systems biology would be to ‘predict’ both future and past responses of organisms to changing environments within the ecological and evolutionary contexts. We note that in climate studies, the terms ‘projection’ and ‘forecasting’ are preferred to ‘prediction’ because they imply a high degree of uncertainty. Here, the word ‘prediction’ is used in accordance with the custom of systems biology, but the uncertainty implied should not be forgotten.

To study the responses to changing environments, a model species like A. thaliana has an advantage because of the availability of a large amount of genetic and genomic data, but a single species cannot be adequate to study various ecologically and evolutionarily important traits. Close relatives of model species provide opportunities to investigate a wide range of ecological and evolutionary phenomena while exploiting the advantages a model species provides (Mitchell-Olds, 2001). This review focuses on a model plant, A. thaliana, and one of its closest relatives, A. halleri. Arabidopsis halleri can be stably transformed using an Agrobacterium-mediated technique (Hanikenne ), and has been used in studies of diverse ecological and evolutionary topics, including self-incompatibility (Bechsgaard ; Castric ; Tsuchimatsu ), perennial growth habit (Aikawa ), heavy metal tolerance (Roosens ; Hanikenne ), defence against herbivores (Shimizu, 2002; Kawagoe and Kudoh, 2010), polyploidization (Shimizu-Inatsugi ; Schmickl ), phylogeography, and population structure (Van Rossum ; Meyer ; Heidel ).

Because the evolution, ecology and molecular biology of plant sexual reproduction have been extensively studied, sexual reproduction provides an ideal platform for interdisciplinary studies. To illustrate gene functions in natura and their potential application for the prediction of biological responses to changing climates, two aspects of our recent research are discussed: (1) as an example of phenotypic plasticity, a systems biological approach to construct a predictive model of the gene expression level of a key flowering gene FLC (FLOWERING LOCUS C) in naturally fluctuating environments; (2) the evolution of self-compatibility as an example of an evolutionary response. Genomic data has shown that self-compatibility in A. thaliana and other species evolved during recent glacial–interglacial cycles. It could be predicted that the evolution of self-compatibility may provide a short-term adaptation but entail a long-term risk of extinction.

PREDICTION OF PLASTIC RESPONSES BY MODELLING GENE EXPRESSION IN NATURAL ENVIRONMENTS

The flowering times of plants are highly plastic in response to diverse environmental factors, including temperature and day length. Recent climate change has shifted the flowering time of many plant species in various ecosystems, even though the flowering of plants in specific seasons is critical for plant reproductive success (Parmesan and Yohe, 2003; Korner and Basler, 2010; Wilczek ; Kobayashi and Shimizu, 2011). The molecular basis of flowering time control has been extensively studied in A. thaliana (Amasino, 2010; Fornara ). Recently, efforts to integrate ecological and molecular functional studies have been made to understand the function of flowering-time control in natural complex environments.

As one of the earliest attempts at an ecological and evolutionary systems biological approach, Wilczek measured the flowering times in a series of A. thaliana strains, including mutants of the major flowering-time genes, in large-scale field experiments. The data were analysed using a photothermal model, which has long been used for the study of phenology (see the supplementary online material by Wilczek ). The model assumes that plants flower when a threshold number of photothermal units accumulate, with input from temperature and day length. By integrating the information derived from the flowering mutants with the molecular genetic network controlling flowering, the model can predict flowering time with a high degree of accuracy. Based on this model, Wilczek also predicted future flowering times within a scenario of global warming.

Using another approach from systems biology, the expression of a key flowering-time gene in the vernalization pathway, FLOWERING LOCUS C (FLC), was monitored in natural habitats (Aikawa ). Because it has been noted that a critical problem in predictions that are based on mechanistic phenology models is the lack of any direct measurement of the internal state of the plant (Chuine ), these gene expression levels as representing such internal states were analysed. The question addressed was: what signals do plants receive that induce flowering in naturally fluctuating temperature regimes? Although it is well known that exposure to constant low temperatures for several weeks (vernalization) induces flowering in many plants, including A. thaliana (Michaels and Amasino, 1999; Sheldon ; Amasino, 2010), such stable conditions rarely exist in unpredictably fluctuating natural environments. For example, cold days could be followed by warm temperatures during autumn or early winter, but flowering in winter in response to such short temperature trends would be deleterious (Stinchcombe ). Therefore, the system of flowering-time control must detect the long-term trends in temperatures, even within the natural fluctuations of the environment, to ensure that they flower at the appropriate time.

FLC encodes a MADS-box transcription factor that functions as a repressor of the floral transition, and its expression is down-regulated by vernalization through epigenetic histone modifications (Michaels and Amasino, 1999; Sheldon ; Bastow ; Sung and Amasino, 2004) (http://www.arabidopsis.org/). Arabidopsis halleri subsp. gemmifera was used because its perennial habit allows the continuous observation of individuals for 2 years. Its FLC homologue (AhgFLC) repressed flowering when overexpressed in A. thaliana, indicating that its function as a floral repressor is conserved in A. halleri (Aikawa ). Tissues were collected from six individuals of a population of A. halleri subsp. gemmifera in central Japan, every week for 2 years, even during drought, flood and snow, and the expression levels of FLC were quantified with real-time PCR (Fig. 2). We note the critical advantage of using sessile organisms in this type of study, as it is generally feasible to repeatedly locate individual plants and to harvest their tissues for molecular studies. Seasonal changes in the AhgFLC expression levels occurred slowly, reflecting the temperature trend of approx. 6 preceding weeks. A time series analysis was conducted to construct a quantitative model of the time course of the expression of this gene. The chilling accumulation model, which has been used in phenological research in ecology (Chuine ), was used to analyse expression levels with the hourly ground temperature recorded using a data logger. The analysis showed that up to 83 % of the variation across 576 expression data points was explained solely by the temperature over the preceding 6 weeks (Aikawa ). The predictions of the model were tested further by exposing plants to controlled laboratory conditions, and the predictions of the model accorded well with the expression levels of AhgFLC in these artificial transplantation experiments.

Fig. 2.

Phenology of Arabidopsis halleri subsp. gemmifera and the relative expression levels of AhgFLC.

From the viewpoint of molecular genetics, FLC is described as encoding a floral repressor, the expression of which is down-regulated by long exposure to low temperatures. If the environments in the natural habitat are taken into consideration, the FLC expression level can be considered a quantitative tracer of the temperature over the preceding 6 weeks (Fig. 1). Using the terminology of electronics or systems biology, the function of FLC in natura is a low-pass filter to extract a long-term trend in temperature by neglecting short-term fluctuations (Bennett ; Aikawa ).

The accuracy of the models developed with the two approaches described above (Wilczek , 2010; Aikawa ) indicates that the molecular basis of flowering time regulation can contribute to the prediction of plant responses to changing climates. To understand fully the functions of flowering-time control genes in natura, processes at the population and community levels as well as other environmental signals should be also incorporated (Fig. 1). For example, selection for early flowering to avoid floral herbivory has been detected in the same population of A. halleri in which AhgFLC expression was studied (Kawagoe and Kudoh, 2010). In terms of the interaction of a plant with other individuals within the population, outcrossing species, like A. halleri, should flower synchronously with conspecifics to maximize the potential for reproductive success (Satake, 2004). A future question to be addressed is how interactions at multiple levels affect the plastic and evolutionary responses of the flowering-time genes to climate changes.

LEARNING FROM THE PAST FOR EVOLUTIONARY PREDICTION: SELF-COMPATIBILITY OF ARABIDOPSIS THALIANA

Another challenge is to predict evolutionary responses to global climate change by learning from the evolution that has occurred during past climate changes. If all evolutionary adaptations were only required on a time scale of millions of years, migration would be the main response to climate change. However, Davis emphasized that adaptive microevolution can occur much more rapidly than this in response to the climatic changes induced by glacial cycles, with periods of 100 000 years, combined with variations on millennial, centennial and even decadal time scales, and further evidence of the rapid evolutionary dynamics affecting ecological traits has recently been reported (Schoener, 2011). At the Evolution 2010 meeting in Oregon in June 2010, a symposium on ‘Moving towards a science of evolutionary prediction’ suggested that the next challenge will be to improve the accuracy of evolutionary prediction (http://www.evolutionsociety.org/SSE2010), although the prediction of evolution is still difficult. Therefore, we propose to investigate relatively simple cases, focusing on loss-of-function mutations.

Here, the evolutionary loss of self-incompatibility is discussed, because it is considered one of the most frequent evolutionary shifts in angiosperms (Stebbins, 1974; Charlesworth, 2006), and because loss-of-function mutations have been shown to underlie this loss (Tsuchimatsu ). First, the molecular basis of self-incompatibility in the Brassicaceae and recent studies of the evolution of self-compatibility in A. thaliana (Shimizu ; Tsuchimatsu ) will be summarized briefly. Then the estimated time of the evolution of self-compatibility will be described using molecular data which suggest its origins during recent glacial–interglacial cycles. Based on these points, how molecular data, focusing on gene function in natura, can contribute to evolutionary prediction is discussed, and the suggestion is made that the evolution of self-compatibility entails the long-term risk of extinction, despite the short-term adaptation it has afforded.

Molecular basis of self-incompatibility and self-compatibility

Many Brassicaceae species, including A. halleri, A. lyrata and Brassica species, have a sporophytic self-incompatible system to avoid self-fertilization. The incompatibility response is based on the female specificity gene S-RECEPTOR KINASE (SRK) and the male specificity gene S-LOCUS CYSTEINE-RICH PROTEIN (SCR; also denoted S-LOCUS PROTEIN 11, SP11), and both are encoded at the S-locus (Takayama and Isogai, 2005). SRK encodes a transmembrane serine/threonine receptor kinase expressed in the stigma, whereas SCR encodes a small cysteine-rich protein ligand present on the pollen surface (Suzuki ; Schopfer ; Takasaki ; Takayama ). The specificity of self-recognition is conferred by S-haplogroups (S-haplotypes or S-alleles), because the SCR and SRK proteins derived from the same S-haplogroups bind one another and inhibit the growth of the pollen tubes (Takayama ; Kachroo ). Tens of S-haplogroups with high sequence divergence are maintained within populations by frequency-dependent selection, because they confer a minority advantage (Castric ; Llaurens ).

Arabidopsis thaliana is self-compatible and a predominant selfer (selfing rate of 97 % or higher) (Abbott and Gomes, 1989; Platt ). The genetic basis of self-compatibility has been extensively studied, and many mutations and deletions at the S-locus have been reported (Kusaba ; Nasrallah ; Tang ; Sherman-Broyles ; Liu ; Shimizu ; Mable, 2008; Boggs ). However, many authors have pointed out that it is difficult to identify the primary inactivating mutation because other genes experience secondary decay once the self-incompatibility function is lost (Igic ; Busch and Schoen, 2008; Boggs ). The S-locus at the population level has been sequenced and genotyped (Shimizu ; Tsuchimatsu ), and a disruptive 213-bp inversion in the male SCR-A gene of haplogroup A or its deletion derivative found in 95 % of European A. thaliana accessions. This pattern contrasts with the population structure, which is thought to have been shaped by glacial–interglacial cycles, suggesting that the inversion mutation conferring self-compatibility spread by natural selection. In contrast to SCR, many accessions express the full-length SRK-A gene. These accessions of A. thaliana showed a self-incompatibility response at the stigma when the pollen grains of A. halleri of haplogroup A were crossed. This evolution has also been successfully reversed experimentally, i.e. when the 213-bp inversion of the SCR gene was restored, the transgenic plants became self-incompatible. These results show that many accessions of A. thaliana are still self-incompatible in terms of the female function, and that a loss-of-function mutation in the male specificity gene was responsible for the evolution of self-compatibility.

The evolution of self-compatibility during glacial cycles

The timing of evolution of selfing or self-compatibility in A. thaliana has been studied by many researchers, and is somewhat contentious. On the one hand, an origin of the selfing estimated to be approximately one million years before present was suggested by a simulation of genome-wide linkage disequilibrium (Tang ), which provides ample time for the evolution of ‘the selfing syndrome’ (small petals/flowers and a reduction in the number of pollen grains) (Barrett, 2002). Conversely, Bechsgaard estimated the time of the evolution of self-compatibility using a molecular evolutionary approach, and developed a method based on the dN : dS ratio (also called the ‘Ka : Ks ratio’, the ratio of the rate of non-synonymous substitutions per site to the rate of synonymous substitutions per site). The dN : dS ratios of functional genes are much <1 because of the functional constraints on the encoded proteins. However, once a gene becomes a pseudogene, the non-synonymous positions evolve neutrally, so the dN : dS ratio will gradually approach a value of 1. Interestingly, the dN : dS ratio of SRK in the A. thaliana lineage shows no pseudogenization signature. The calculations of Bechsgaard et al. showed that A. thaliana has been self-incompatible for at least 91.7 % of the time since it diverged from A. lyrata and A. halleri (Bechsgaard ). To translate this value into years, it is necessary to estimate the time at which A. thaliana diverged from the other Arabidopsis species, which is currently under discussion. If a commonly used divergence time of five million years (Koch ) is assumed, the loss of self-incompatibility is estimated to have occurred more recently than 0·413 million years ago (mya) (Bechsgaard ). Recently, Beilstein noted that the estimate by Koch was based on a misunderstanding of the fossil literature (Beilstein ). More recent reports have proposed a much earlier split at 13·0 mya, with a 95 % confidence interval of 8·0–17·9 mya, according to a calibration based on multiple fossil records by Beilstein ; or 8·7 ± 1·0 mya or 17·9 ± 4·8 mya based on mutation accumulation lines, proposed by Ossowski . Using these values, self-compatibility originated 0–0·64 mya (assuming a minimum divergence estimate of 7·7 mya) or 0–1·88 mya (assuming a maximum divergence estimate of 22·7 mya). The time ranges are well within the Quaternary, which is characterized by 41-ky and 100-ky cycles of glacial–interglacial periods. Therefore, these estimates indicate that the evolution of self-compatibility probably occurred during a glacial–interglacial cycle, providing an example of an evolutionary response to climate change.

The present study supports the recent origin of self-compatibility proposed by the molecular evolutionary analysis in two respects (Tsuchimatsu ). First, all the genes necessary for the self-recognition and rejection response, except SCR, still retain functional alleles in many accessions of A. thaliana. If self-incompatibility were lost a long time ago, the other genetic components would have been degraded, although pleiotropy could have slowed this process (Tantikanjana ). Second, the pattern of polymorphisms at the S-locus has deviated from the genome-wide pattern, which is thought to have been shaped by range expansions from multiple refugia during glacial–interglacial cycles (Sharbel ; Nordborg ; Schmid ; Beck ; Francois ).

In addition to the molecular functional and evolutionary genomic studies of A. thaliana, phylogenetic studies of the family Solanaceae have shown that self-compatible species are short lived because of their higher extinction rates (Goldberg ), and thus supports the hypothesis for recent origin of self-compatibility. Population genetic studies have also suggested a recent origin of self-compatibility in Capsella rubella, A. kamchatica, and North American A. lyrata (Shimizu ; Sugisaka and Kudoh, 2008; Shimizu-Inatsugi ; Foxe , 2010; Guo ; Hoebe ). Taken together, the evolution of self-compatibility during recent glacial–interglacial cycles appears to be a broad, general pattern. It should be noted that time estimates based on population genetics also depend on molecular clock calibration, as described above, and also require a number of assumptions to be made about the population demography. A recent large-scale polymorphism study in A. thaliana showed a continuous gradient of variations along eastern–western Europe (Platt ), suggesting that methods are still required to estimate time and to test neutrality within a continuous population structure.

The recent origins of self-compatibility also suggest that the ‘selfing syndrome’ (Barrett, 2002) evolves rapidly. Although the data presented above demonstrates that the complete loss of self-compatibility occurred recently, this does not necessarily mean that the ancestral A. thaliana was highly outcrossing, with strong self-incompatibility. In many wild species, partial self-compatibility results in mixed mating systems (Goodwillie ). Interestingly, it has been observed that self-incompatibility is weakened in the later stages of flower development in many A. thaliana accessions but not in other accessions (Tsuchimatsu ). Such partial self-compatibility would provide opportunities for reproductive assurance by selfing if no outcrossing pollen is available. It remains to be clarified whether leaky self-incompatibility is the ancestral state of A. thaliana, or represents secondary decay after the evolution of self-compatibility.

Using gene function in natura for evolutionary prediction

Laboratory studies have demonstrated that SCR encodes a protein ligand of SRK. At the population level, functional SCR and SRK genes enforce outcrossing (Fig. 1). Moreover, the studies described above suggested the loss-of-function mutation of SCR contributed to adaptation in a changing climate. Charles Darwin (1876) proposed a hypothesis for reproductive assurance, which states that selfing can be advantageous when mates or pollinators are scarce. The evolution of self-compatibility in many species including A. thaliana appears to have occurred during recent glacial–interglacial cycles. Migration, such as that occurring during range expansion after glacial periods, would be accompanied by a paucity of mates (Baker, 1955). Thus, the origin of self-compatibility glacial–interglacial cycles, together with the spread of the self-compatible mutation at the S-locus, suggests that the loss-of-function mutation of SCR could have been responsible for mating system shift thus allowing reproductive assurance during glacial–interglacial cycles.

Furthermore, the loss-of-function mutation of SCR in A. thaliana illustrates three ways in which integrated studies using molecular data contribute to our understanding and ability to predict evolutionary and ecological phenomena. First, the independent origins of self-compatibility were revealed in the studies of the self-compatibility mutations. The self-compatibility of accessions with haplogroup B, distributed on offshore African islands, evolved independently, because they do not contain the 213-bp inversion in SCR-A (Tang ; Shimizu ; Boggs ). It is also possible that individuals with haplogroup C have yet another independent self-compatible mutation. All natural plants of A. thaliana reported so far are self-compatible, i.e. the self-compatibility phenotype is fixed in A. thaliana, so it is not possible to identify the parallel evolution of self-compatibility from the self-compatible phenotype alone. A study at the species level suggested that the evolution of self-compatibility is one of the most frequent evolutionary transitions in angiosperms (Stebbins, 1974), but it must be even more frequent than has been previously thought because a single species could include parallel transitions. A growing number of studies have shown the parallel evolution of various phenotypes in many species, particularly those caused by loss-of-function mutations (Hoekstra and Coyne, 2007; Shimizu and Purugganan, 2005). Therefore, evolutionary studies based on phenotypic frequencies could overlook the complexity of their evolution. Second, because of the prevalence of loss-of-function mutations, the direction of evolution is often asymmetric or even unidirectional, which is consistent with Dollo's law that states character loss is irreversible (Zufall and Rausher, 2004). Although the evolutionary transition from self-incompatibility to self-compatibility has occurred independently many times, evolution in the opposite direction is considered to be extremely rare (Stebbins, 1974; Igic ; Goldberg ). Evolutionary ecological models that assume symmetric and small-effect mutations have shown that predominant selfing is a stable state among the mating systems (Lande and Schemske, 1985), suggesting that evolution towards predominant selfing is asymmetrically preferred by selection. When we focus on mutation rather than on selection, a simpler explanation is possible, because loss-of-function mutations (nonsense mutations, splicing mutations, frame-shift mutations, etc.) are expected to occur much more frequently than back mutations. Although the transgenic experiments successfully reversed the 213-bp mutation in SCR (Tsuchimatsu ), the chance of this exact mutation occurring under natural conditions is extremely small. Moreover, many self-incompatibility haplogroups are required to maintain efficient outcrossing. Whereas >30 S-haplogroups have been identified in both self-incompatible A. lyrata and A. halleri (Castric ), only three haplogroups are reported in A. thaliana (Shimizu ). The evolution of new haplogroups would require multiple mutations to allow the co-evolution of the male and female specificity genes, although little is known about the molecular basis of their evolution (Newbigin and Uyenoyama, 2005). We would also like to note that rapid reverse evolution may be possible if the alleles are still segregating (Kitano ). Third, future evolvability would be restricted by the nature of the loss-of-function mutations. Once self-incompatibility is lost, it is unlikely to be regained, because the evolution of a new self-incompatibility system is thought to have occurred only several times in the history of the angiosperms (de Nettancourt, 2000). Furthermore, although empirical data are limited, it is generally considered that predominantly selfing species may not respond to environmental changes because their genetic diversity is reduced (Takebayashi and Morrell, 2001; Goldberg ).

What can be predicted about the evolution of self-compatibility from the integrated studies of molecular biology, ecology and evolution? For example, climate models indicate that global warming will have a major impact on alpine flora, including the fragmentation of habitat, because alpine environments will be subjected to dramatic and rapid environmental changes (Parmesan, 2006; Intergovernmental Panel on Climate Change, 2007). When habitats are fragmented, the evolution of predominant selfing by the loss of self-incompatibility may assure the reproduction of plants by conferring a short-term advantage. However, it is reported that self-compatible lineages suffer from extinction (Goldberg ). Thus, it is predicted that these plants will not regain their self-incompatibility in the short term, even if the environment is restored, and will therefore be subject to higher extinction rates in the longer term. In short, rapid adaptive evolution may confer short-term advantages, but entail long-term extinctions.

The mutation of SCR was relatively easy to analyse because its loss of function has a major effect on self-compatibility. However, ecologically and evolutionarily relevant traits may be affected by a large number of small-effect mutations. Analysing these mutations to identify general patterns will require a quantitative and network approach, using genome-wide surveys of wild organisms.

CONCLUSIONS AND PERSPECTIVES

Gene functions in natura

To predict biological responses to changing environments, it is important to analyse gene functions or phenotypes in naturally fluctuating environments, which are often very different from laboratory conditions (Fig. 1). Studies of FLC and SCR have examined gene functions in natura, demonstrating that genes (or alleles) function at multiple levels in ecosystems in the context of various biotic and abiotic interactions (Fig. 1). In the laboratory, FLC functions as a floral suppressor, unless the plant is exposed to low temperatures for several weeks. However, an expression study in naturally fluctuating environments revealed that FLC acts as a tracer of temperature trends in the preceding several weeks (Aikawa ). Such an extended view of gene function is essential for the prediction of plant responses to changing climates. Similarly, we have described that fixation of a non-functional SCR allele could have changed the mating system in plant populations, thereby contributing to reproductive assurance during migrations within glacial cycles. In addition, the interaction with pollinators may be also important for the evolution of self-compatibility. Theoretical analyses have predicted that when selfing is favoured, the frequencies of the mutations that disable the male component of self-incompatibility will increase more than the frequencies of the mutations that disable the female components (Uyenoyama ; Busch and Schoen, 2008). This is because male mutations are propagated through both pollen and seed, whereas female mutations are only propagated through seed. Crossing with A. halleri and transgenic data have shown that an SCR mutation is primarily responsible for self-compatibility according to this prediction. Thus, mutations in the male specificity components suggest that pollinator insects might have assisted the spread of the SCR allele (Tsuchimatsu ) (Fig. 1). In short, genes can affect the interactions of an organism with individuals in a population of the same species, with other species in the community, and with fluctuating environments or climates (Fig. 1). Phenotypes within ecosystems are often called ‘extended phenotypes’ (Whitham ).

The limitations of laboratory experiments are evident in a classic example of the human glucose-6-phosphate dehydrogenase gene (G6PD) (Freeman and Herron, 2007). The G6PD gene is usually considered a ‘housekeeping’ gene, encoding an enzyme in the pentose phosphate pathway. However, deficiency alleles are frequently found in tropical human populations, and these confer resistance to severe malaria. It is reasonable to infer that a function of a deficient G6PD allele is to confer malaria resistance, although it is implausible that a laboratory experiment alone would identify this function in the interaction between humans and other species in an ecosystem.

We would like to add that plasticity, evolution and migration are not independent. Different genotypes show different phenotypic plasticity, so plasticity itself must evolve (Gienapp ). Indeed, the association between the plastic response of flowering and FLC polymorphisms was detected in A. thaliana (Caicedo ). Migration can cause mate limitation and the evolution of self-compatibility (Tsuchimatsu ), and then self-compatibility can, in turn, facilitate migration. Thus, it is important to integrate ecological and evolutionary information to understand the responses of a species to changing climate.

Epigenetics and next-generation DNA sequencing

Epigenetic changes may have played a significant role in evolution and ecology (Kalisz and Purugganan, 2004; Bossdorf ), and more specifically, in organisms' responses to changing climates, but little is currently known. Several studies have started to show heritable epigenetic variations within and among species (Fujimoto ; Johannes ). The transcriptional activity of FLC is mediated by epigenetic histone modifications on the FLC chromatin. During vernalization, the methylation at histone H3 Lys9 and histone H3 Lys27 (H3K27) on the FLC chromatin increases, resulting in the repression of FLC transcription (Bastow ; Sung and Amasino, 2004). In the annual species A. thaliana, the repressed status of FLC is stable, even when the plants are returned to warm temperatures. In contrast, in the perennial species Arabis alpina, the repressed status of PEP1 (FLC homologue) is transient (Wang ). The H3K27 methylation levels on the PEP1 chromatin increase in Arabis alpina during vernalization, corresponding to the expression pattern of PEP1, but they then decrease when the plants experience warmer temperatures. These data suggest that epigenetic regulation underlies the differences between the annual and perennial growth forms. Interestingly, yeasts detect long-term trends of nutrient availability through a slow chemical reaction in a metabolic network (Bennett ), and it is conceivable that the slow epigenetic regulation of FLC underlies the plants' detection of seasonal trends.

In addition to the effect of prolonged low temperature in flowering time regulation discussed above, A. thaliana shows plasticity of flowering time in response to temperatures in the range of non-stress conditions (16–27 °C), called ambient temperatures. It is reported that an increase in growth temperature causes early flowering (Westerman and Lawrence, 1970; Blázquez et al., 2003; Lempe ). Recently, nucleosomes containing histone variant H2A.Z have been found to mediate the thermosensory response (Kumar and Wigge, 2010). H2A.Z-containing nucleosomes wrapped DNA more tightly, and then transcription of one of the key flowering activators, FLOWERING LOCUS T, was changed. These data imply the link between epigenetic regulation through nucleosome assembly and the plasticity of temperature-dependent flowering time. It would be one of challenges to study the significance of epigenetic changes in response to global warming.

The dominance relationship of SCR/SP11 haplogroups is also regulated epigenetically by the DNA methylation induced by a small RNA (Shiba ; Tarutani ). The sporophytic self-incompatibility system in the family Brassicaceae is characterized by dominance relationships between haplogroups, in which heterozygotes often show only one of their two S-specificities (Hatakeyama ). The most recessive alleles exhibit higher frequencies in outcrossing populations (33 % in a population of A. lyrata; Mable ), whereas the dominant alleles show much lower frequencies. Haplogroups in the same dominance class also tend to cluster on phylogenetic trees (Prigoda ). It is noteworthy that haplogroups A, B and C of A. thaliana (Shimizu ) belong to clades of dominant classes. In particular, haplogroup A (AhS04 of A. halleri), from which the self-compatible mutation spread in A. thaliana (Tsuchimatsu ), has been shown experimentally to be a dominant allele in A. halleri (Llaurens ). If a loss-of-function mutant of SCR still retains its dominance relationship and functions as a dominant self-compatible mutation by repressing the expression of another specificity, this allele is expected to spread more rapidly than a recessive self-compatible mutation. Interestingly, studies of Brassica have shown that the S-locus harbours a separate gene (Smi) responsible for the dominance effect, which encodes a small RNA that represses the expression of allelic SCR (Shiba ; Tarutani ). Therefore, the loss-of-function mutation in SCR may yield a dominant self-compatibility allele, which would contribute to the rapid spread of self-compatibility. Furthermore, studies of Brassica have also suggested that self-compatible alleles that are dominant can confer self-compatibility on polyploid species (Okamoto ). It remains to be demonstrated whether Smi is functional in Arabidopsis.

Such recent advances in epigenetic studies in the laboratory indicate that both epigenetic and genetic regulation must be considered together to understand phenotypic plasticity and evolution. To quantify the contribution of epigenetic states in natural environments is another future challenge.

In this review, we have explained that utilizing A. halleri, a close relative of a model plant A. thaliana, enabled the study of perennial habit and self-incompatibility, which is not found in A. thaliana. A future challenge for ecological and evolutionary systems biology will be to apply the methods developed for model species to other species (Karrenberg and Widmer, 2008), in particular to keystone species, which affect many other organisms in an ecosystem. Next-generation DNA sequencers, initially developed for medical purposes, are now revolutionizing ecological and evolutionary biology (Benfey and Mitchell-Olds, 2008; Kobayashi and Shimizu, 2011). These sequencers will provide genome-wide data on ‘non-model’ species, including keystone plant and animal species, even when little genomic information is already available. For example, sequencing cDNA allows genome-wide expression patterns to be investigated, and higher resolution can be achieved by combining this technology with customized microarrays (Toth ; Bellin ; Leakey ). These techniques will facilitate the study of more non-model species, to establish general patterns and to predict their phenotypic plasticity and evolutionary responses to coming climate changes.