Functional studies from agricultural research offer essential insights into the specific mechanisms that can underlie important trait variation. With respect to drought adaptation, functional and physiological analyses in the crop literature already provide considerable information on the morphological and physiological traits that might be of greatest adaptive importance among tomato species, as well as candidate loci for analyses of drought adaptation. Conversely, evolutionary approaches can identify general patterns of trait variation and response, and can vastly expand the repertoire of natural genetic variation with which to examine (and potentially exploit) functionally important traits. Nonetheless, these reciprocal resources have been underutilized by ecological/evolutionary geneticists and crop scientists alike. Here we highlight four areas where advances could be made towards reciprocal understanding within agricultural and evolutionary fields, with respect to drought responses in tomato.
Identifying and exploiting natural variation for crop improvement
Evolutionary analyses of natural trait variation in wild tomatoes promises to identify exploitable genetic variation that might be valuable for future crop improvement, either directly via targeted trait introgression from wild species (Zamir 2001), or indirectly via expanding the range of genetic variation that is available to understand molecular mechanisms of trait variation (e.g. Alonso-Blanco et al. 2009).
The process of domestication and subsequent artificial selection is known to have several genome-wide genetic consequences including the reduction of variation at both trait and DNA sequence levels (e.g. Bai and Lindhout 2007; Van Deynze et al. 2007). Because of the close relationship among all species in the clade, and the vastly reduced genetic variation within the cultivated tomato, wild tomato species have been the source of multiple agriculturally valuable traits for crop improvement, such as disease resistance and vitamin content in fruits (Zamir 2001; Gur and Zamir 2004; Bai and Lindhout 2007; Lippman et al. 2007). Tomato breeders have previously used climatic or environmental factors to guide the choice of wild germplasm to evaluate for crop improvement (Atherton and Rudich, 1986), however GIS-based environmental characterization (see above) offers a comprehensive quantitative approach to identifying populations most likely to exhibit adaptive traits of interest. These approaches can also simultaneously take into account multiple environmental variables, such as native water availability, temperature, and soil traits, thereby potentially minimizing (or maximizing) the influence of traits that affect more than one environmental factor. As such, in close relatives of economically important species, these approaches can provide a more quantitative and precise set of predictions about which sources of wild germplasm might provide the most useful natural genetic variation.
This trait variation is likely to exceed that found in domesticated varieties both quantitatively and qualitatively. In tomato, investigations of natural variation already show that different wild species likely have alternative mechanisms of adaptation to the same abiotic conditions. For example, physiological water stress responses differ between S. pennellii and S. chilense (see above), making both species sources of new and unique traits for possible crop improvement. In addition to simply identifying traits of agronomic interest, this natural genetic variation can also provide the raw material for new insights into the diversity of molecular mechanisms that determine important trait variation. In other crop and model systems (e.g. Alonso-Blanco et al. 2009), the analysis of genetic variation has provided unprecedented information on the molecular mechanisms that determine important adaptive responses, insights that are unavailable when relying solely on domesticated varieties or mutants for experimental material. For example, Nevo and Chen (2010 and references therein) review studies identifying traits related to drought and salinity stress from multiple wild cereal species.
Finally, GIS-based information can also be combined with species distribution modeling (SDM) to provide geographical predictions of species habitat suitability (see Nakazato et al. 2010). These predictions identify geographical locations that are likely (or unlikely) to be suitable habitats for a species of interest (Kozak et al. 2008). These predictions can be used as geographic guides for future collecting efforts for new, potentially valuable wild germplasm. In this way, GIS environmental characterization can provide quantitative tools to evaluate useful germplasm in existing collections, and SDM can provide an assessment of promising areas for future germplasm collections.
Dissecting the genetic basis of adaptation
Evolutionary biologists are still in the incipient stages of uncovering the genetic basis of adaptation in wild organisms. For some exemplar systems, such as pelvic reduction in threespine stickleback fish (e.g. Chan et al. 2010) or flower color (Rausher 2008), the molecular genetic basis of adaptation is partially known. Progress in these systems owes much to the fact that they focus on ecologically relevant phenotypes that have a relatively simple (e.g. loss-of-function) genetic basis. In contrast, the genetic basis of adaptation for complex traits composed of many interacting constituent parts is in its infancy, for technical as well as biological reasons. Trivially, uncovering the genetic basis of complex traits may be practically difficult because the pertinent phenotypes are difficult to measure. Other complications, such as pleiotropy, linkage, and sensitivity to genetic (epistasis) and external (plasticity and G × E) environments, are more biologically meaningful, but nevertheless obscure inference about causal connections between loci, traits, and adaptation. Importantly, it is this uncharted empirical landscape that has been the purview for most theory of adaptation (Fisher 1930; Orr 2005). Bridging the gulf between theory and data will require methods to address complex, ecologically important traits.
Anatomizing complex traits into multiple, easily measured phenotypes, each of which may have a relatively tractable genetic architecture, is a promising approach. However, it may be difficult to know which phenotypes are important, as there are often multiple suites of traits that are functionally similar. In this context, agricultural research can provide valuable information on which traits are likely candidates for adaptive differentiation in nature. For instance, crop research on physiological dehydration avoidance in tomato species allows approaches like QTL mapping to be more precisely targeted to those traits most likely to reflect adaptive differentiation between species. For example, despite clear differences in tolerance between S. pennellii and S. lycopersicum (see above), functional traits enhancing performance during drought have been mapped to a very limited degree among these and other tomato species (Martin et al. 1989; Foolad et al. 2003; Xu et al. 2008). However, using currently available genetic tools, such as recombinant mapping population, it is relatively straightforward to identify QTL that underpin these physiological and functional differences (e.g. Muir and Moyle 2009). These more targeted efforts, informed by physiological studies in agriculture but incorporating wild species variation, make dissecting the genetic basis of drought adaptation highly promising in tomato and similar systems. For example, for numerous species of cultivated cereal grains, there are many QTL mapping studies of drought tolerance traits that capitalize on natural variation (see Chen et al. 2010 and references therein for a current account).
The candidate trait approach also enables targeted examination of interactions between traits (pleiotropy) and between genes (epistasis) based on a priori theoretical and/or mechanistic expectations. For example, for plant drought responses, there is little reason to expect a strong mechanistic connection between nontranspirational seedling growth and adult WUE. Conversely, leaf size and stomatal density are usually negatively correlated within species, as developmental patterning is determined prior to leaf expansion. Understanding this mechanistic connection suggests explicit expectations for constraints on adaptive response. For example, dehydration avoidance entails both reduced transpirational area and decreased stomatal density (as observed in differences between S. pennellii and S. lycopersicum), implying that multiple mutations should be necessary for selection to break this negative trait correlation. Similar expectations can be generated with respect to epistasis. For example, theory predicts that genes in the same pathway, and therefore affecting the same trait, may exhibit antagonistic epistasis (Phillips 2008), leading to an expected excess of such interactions between QTL underlying the same trait, but not between QTL for different traits. Interestingly, in a QTL analysis of ecophysiological traits differentiating S. pennellii and S. lycopersicum (Muir and Moyle 2009), we detected an abundance of antagonistic epistatic interactions between loci influencing SLA and whole-plant drought responses, consistent with these theoretical predictions.
Together, these complex trait and locus interactions will determine the trajectory of and constraints on adaptive evolution in the wild. Assessing their relative importance relies on experimental systems, like tomato, in which mechanistic and functional data can be combined with predictions from evolutionary genetic theory. Of course, the resulting data will not only offer insights into the genetics of adaptation in the wild. In an interesting inversion of Darwin’s original analogous reasoning between natural and artificial selection, understanding how genetic architecture constrains and facilitates of natural selection can equally act as a guide for informing expectations about the effects of artificial selection on trait variation subject to pleiotropy, epistasis, and constraint.
Assessing functional and evolutionary trade-offs
Integrative approaches to studying trait variation in tomato can also be used to address long-standing questions in the ecophysiological and evolutionary literature (e.g. Chapin et al. 1993). For example, trade-offs arising from biophysical constraints are thought to prevent simultaneous adaptation to multiple contrasting environments (Lambers et al. 2008). Drought adaptation is a canonical example because water loss occurs during photosynthesis, ‘constitut[ing] a fundamental trade-off, evidenced by a negative correlation, between fixing carbon and saving water’ (Arntz and Delph 2001). However, evidence that trade-offs play an important role in divergence and coexistence between closely related species or populations is equivocal (Ackerly et al. 2000; Geber and Griffen 2003; Hereford 2009). Genetic studies of drought adaptation also give conflicting results. McKay et al. (2003) found a trade-off (antagonistic pleiotropy) between two drought escape traits: WUE and flowering time. Conversely, adaptation to drought in the wild oat Avena barbata appears to entail no fitness cost in mesic environments (Latta 2009).
An ecophysiological trade-off is evidenced by a negative cross-environment correlation related to the value of a functional trait. For example, if smaller leaves improve growth during drought by reducing water loss, but decrease growth in well-watered conditions by limiting photosynthesis, then one could conclude there is a trade-off involved in changing leaf size. Two criteria are required for inferring a trade-off:
a negative correlation between performance in contrasting (e.g. dry and wet) environments.
provided criterion 1 is met, performance must be correlated with trait values in the expected direction to differentiate trade-offs from spurious genetic differences.
Both of these can be functionally evaluated in tomato using natural variation for drought adaptive traits. For example, seed germination success and seedling growth vary predictably in response to low and high osmotic stress, depending upon the native water conditions experienced by different species. Using germination media in which the osmotic potential can be artificially manipulated, we found that a species from more water-limited natural environments (S. pennellii), continues to show substantial germination under high osmotic stress (Fig. 3). In comparison, species that occur under more abundant water conditions [e.g. S. lycopersicum (tomato) and S. pimpinellifolium] show precipitous declines in germination rates under high osmotic stress (Fig. 3). Interestingly, in low osmotic stress conditions, seedling growth rates are retarded in the ‘drought-adapted’S. pennellii, in comparison to the other two species (data not shown). Together these results are indicative of a negative correlation between seed germination and seedling establishment in low and high osmotic stress conditions, a necessary condition (criterion 1, above) for demonstrating a functional trade-off between adaptation to alternative resource environments. This apparent trade-off might be related to species differences in seed size. The smaller seed of the drought-adapted species gives it a greater surface area to volume ratio, increasing imbibition, but decreasing maternal provisions that can be marshaled during early growth. This and alternative hypotheses can be tested with further experimental analysis of the loci underlying such traits. We are also testing for dehydration avoidance traits in S. pennellii (see above), which are likewise predicted to exhibit trade-offs. The availability of genetic tools, including recombinant lines that allow the identification of QTL, are particularly valuable in this respect.
Figure 3. Seed germination success of different tomato species under low and high osmotic stress. Seeds were germinated on growth media with variable concentrations of PEG-8000 (polyethylene glycol-8000), a neutral polymer that replaces water in the growth media but is unable to be absorbed by plant roots. Osmotic stress increases from left to right with increasing PEG-8000 concentration. Asterisk indicates a significant difference between species (P < 0.05). Result indicate that, at high osmotic stress, dry adapted Solanum pennellii continues to show appreciable levels seed germination, in comparison to S. lycopersicum (tomato) and S. pimpinellifolium.
Download figure to PowerPoint
The classic evolutionary problem of how species cope with constraints and trade-offs reinforces what crop scientists already know about complex traits. Breeding for tolerance to a single stress may decrease yield in benign conditions or under a different stress. The proper balance of stress tolerance, yield, and other traits thus needs to be tuned to specific environmental conditions (Nevo and Chen 2010). Evolutionary studies can reveal new prospects and challenges for understanding these relationships. Over eons, natural selection has struck upon many innovative solutions to work around constraints, providing insights that may not have been gained from first principles. Conversely, consistent and evolutionarily widespread lack of trait variation in some directions may reveal unforeseen and unavoidable constraints.
Identifying repeated patterns and common functional pathways in adaptive transitions
In the future, integrating broad clade-level comparative analyses of abiotic niche evolution with a more mechanistic understanding of the genetic basis of adaptive species differences (e.g. QTL mapping and functional studies) promises to provide substantial insights into the ‘genetic architecture’ of adaptation. It is in this regard that agricultural studies have the greatest potential to contribute conceptually to evolutionary analysis – that is, via the detailed information they provide on the specific genetic mechanisms responsible for phenotypically important trait changes. Given access to this rich mechanistic information, and the ability to extend it into natural evolutionary settings, comparative evolutionary studies can identify general patterns and processes of adaptation (Harvey and Pagel 1991). These in turn can be used to address ongoing conceptual debates within evolutionary biology including, for example, whether adaptive change is enriched for specific kinds of molecular mutation (e.g. regulatory versus structural changes; Hoekstra and Coyne 2007), whether adaptation frequently involves diversification of gene duplicates (e.g. Oakley 2007; Hahn 2009), and the relative importance of G × E for natural trait expression and responses to selection (e.g. Weinig and Schmitt 2004; Sultan 2007).
For example, Kopp (2009) has advocated a meta-model approach for evolutionary genetics that takes advantage of parallel trait changes in a clade to ask whether and why convergent phenotypes are underpinned by similar genetic changes. Repeated adaptation to novel niches is a pervasive form of parallel evolution, and abiotic adaptive diversification is one instance in which multiple evolutionary transitions can be used to identify patterns in either the traits and/or underlying genetic basis of adaptive phenotypic change. For example, in tomato, patterns of molecular evolution in specific pathways contributing to drought adaptation, e.g. the ABA biosynthesis pathway, can be examined to determine the repeatability of specific evolutionary changes in both domesticated and wild lineages.
These analyses can also contribute to other complementary questions about the nature and predictability of genetic changes that underpin adaptation. For example, there is a growing literature on pathway and network evolution (Flowers et al. 2007, 2009; Rausher 2008; Alvarez-Ponce et al. 2009; Ramsay et al. 2009). Based on metabolic control theory (Kacser and Burns 1973) and similar modeling approaches, it is predicted that genetic changes affecting the function of most enzymes will not affect the phenotype, making them effectively invisible to selection. Rather, control may lie in mutations at a few key loci that, because of their position in a pathway (e.g. branch points [Eanes 1999]), can affect fitness. Within tomato, these theoretical considerations can be used to generate and test predictions about where functional genetic changes are most likely to take place within specific networks and pathways, such as the ABA biosynthesis pathway (see above). In addition to better understanding processes of natural adaptation, these analyses can also be useful for identifying which loci are more likely to be fruitful in manipulating such traits to maximize yield outcomes in agricultural settings.
Some of these conceptual questions lie directly at the intersection of evolutionary and agricultural studies (Hancock 2005; Ross-Ibarra et al. 2007; Purugganan and Fuller 2009). For example, the strength and mode of selection on individual loci during adaptation is debated, with implications for the type (regulatory versus coding) and number of mutations involved. One possible model, influenced by the study of domestication, is that strong selection might fix through loci of major effect that themselves have strong pleiotropic effects (e.g. Clark et al. 2006; Stern and Orgogozo 2008) or that drag along allelic changes in loci that are physically linked to the target of selection (e.g. Olsen et al. 2006). It is unclear whether nature selects as monotonically strongly and/or is as forgiving of pleiotropic consequences (e.g. Westerbergh and Doebley 2002). But if natural were like artificial selection, it would suggest a role for large regulatory changes with compensatory mutations thereafter. The ability to generate detailed mechanistic information in crop systems, and extend this to the study of wild species, promises data that are essential for addressing these fundamental questions about the nature, pace, and comparability of selection by humans and in the wild.