Author for correspondence: Scott A. Hodges Tel: +1 (805) 893 7813 Fax: +1 (805) 893 4724 Email: firstname.lastname@example.org
One of Verne Grant's lasting contributions to plant evolutionary biology has been the recognition that differences between plants in floral characters can have a dramatic impact on both pollinator visitation and pollen transfer and thus affect reproductive isolation between nascent plant species (collectively, floral isolation). Here we review some of the concepts and findings from Grant's work on floral isolation, particularly with respect to the genus Aquilegia (Ranunculaceae). It has now been over 50 yr since Grant first published on the role of floral isolation on reproductive isolation and speciation in Aquilegia and we compare and contrast his findings with our own work on this genus. We find that the data largely support Grant's findings and that Aquilegia will continue to offer great opportunities to learn about the processes of adaptation and speciation.
Verne Grant's impact on plant evolutionary biology is hard to overstate. His work has provided a wealth of ideas and data on a wide range of topics. In particular Grant illuminated the roles of hybridization and polyploidy in speciation and the nature of species themselves in his landmark book Plant Speciation (Grant, 1971, 1981). This work has inspired and informed a great number of students of plant speciation, it certainly has for our own work. Grant may be most widely known for his influential studies on the systematics, genetics and pollination biology of the Polemoniaceae. In addition, he frequently used the genus Aquilegia to illustrate a variety of patterns and processes, such as species relationships (Grant, 1952, 1963, 1971, 1981, 1993b), the evolution of floral isolation (Grant, 1952, 1971, 1976, 1981, 1992, 1993b, 1994a; Grant & Temeles, 1992), the evolution of hummingbird pollination (Grant, 1993b, 1994b), and patterns of hybridization (Grant, 1952, 1976, 1992, 1993a, 1994a).
The purpose of this review is to consider Grant's ideas about floral isolation, the nature of species, reproductive isolation and hybridization in the genus Aquilegia in light of recent advances. We begin by reviewing Grant's earliest work where he outlines his views on the role of floral isolation on reproductive isolation and general patterns of plant diversification. We then consider Grant's specific example of floral isolation as it applies to hybridizing species of Aquilegia. Finally, we consider how Aquilegia may further contribute to our understanding of the genetic basis for adaptations and speciation. We conclude that much progress has been made in understanding the evolution of plant reproductive isolation, that many of Grant's conclusions still hold today, and that his visionary insights continue to point us towards new and exciting studies.
Floral isolation and speciation
One of Verne Grant's enduring contributions to plant evolutionary biology is the concept of floral isolation and its impact on plant speciation. In fact, Grant began his research career with a seminal paper addressing how pollination mechanisms may act as a reproductive isolating barrier among plants (Grant, 1949). He argued that plant species could be interfertile yet remain reproductively isolated by either mechanical isolation (i.e. pollinators visit the flowers of two species but fail to transfer pollen) or by ethological isolation (i.e. pollinator constancy, where pollinators visit one species’ flowers at the exclusion of visits to another species’ flowers). Even at the beginning of his career Grant drew from a broad array of sources to forge his argument citing relevant information from comparative biology, natural history observations, and experimental approaches to show how pollinator behavior and pollen placement mechanisms can cause floral isolation. Of particular interest is his early use of comparative approaches where he argued that either form of floral isolation must be caused by differences between species in floral characters and thus, if reproductive isolation between species has evolved through floral isolation, sister species should differ in floral characters. Grant then tested this hypothesis by classifying plants by pollination system and then determining whether taxonomists had used floral characters to differentiate closely related species (Fig. 1). Grant found that plants pollinated by animals that were likely to show floral constancy (i.e. birds, bees and long-tongued flies) were much more likely to be distinguished by floral characters than plants with pollination systems unlikely to have floral isolation (Fig. 1) and concluded that floral isolation was an important form of reproductive isolation for many angiosperms (Grant, 1949).
In addition, Grant linked the evolution of floral isolation with rates of speciation (Grant, 1949). He argued that groups especially prone to the evolution of floral isolation might also experience higher rates of speciation. To test this idea he compared the species diversity of genera where floral isolation was unlikely to be important in speciation (i.e. genera with generalist pollination) to the species diversity of genera where floral isolation was likely to be important in speciation (i.e. genera pollinated by bees). He found that on average genera pollinated by bees had higher diversity (5.9 species/genus) than genera with unspecialized pollination systems (3.4 species/genus) thus supporting a link between specialized pollination and rates of speciation (Grant, 1949). While Grant's early comparative analyses lacked an explicit phylogenetic foundation, as is commonplace today, similar patterns have been found across the angiosperms using more modern methods (e.g. Dodd et al., 1999).
We have also conducted comparetive studies of species diversity, focusing on how floral morphology may be linked to speciation through its direct impact on floral isolation. Similar to Grant's thesis, we have investigated how the emergence of a novel floral feature – namely the floral spur – has influenced species diversification by increasing the opportunity for floral isolation. Floral spurs are tubular outgrowths of floral organs that contain a reward (usually nectar) for pollinators. A difference between species in the shape, color and orientation of these spurs likely cause mechanical isolation, pollinator constancy, or both (see below). Once nectar spurs evolve they may provide a new focal point for selection to act, thus resulting in reproductive isolation and speciation. To test this hypothesis we compared the species diversity of multiple pairs of sister taxa that differ by the presence or absence of nectar spurs (Hodges & Arnold, 1995; Hodges, 1997a,b). We found a striking pattern whereby there were more species in the spurred clades compared to their nonspurred sister clade in all instances but one (Fig. 2a). These data are among the strongest examples linking an innovative morphological feature with increased rates of species diversification (de Queiroz, 1998). Thus, similar to Grant's predictions, we have found strong evidence linking characters that would promote floral isolation with diversification.
Interestingly, the data in Fig. 2 also suggest that the impact of evolving a new mechanism for floral isolation such as floral spurs may have a rapid impact on species diversification that diminishes over time. The discrepancy in species diversity of pairs of clades peaks in groups with relatively few species in the nonspurred clade (Fig. 2b) and then the difference in diversity declines as the number of species in the nonspurred clade increases. Only in the group with the largest number of nonspurred species is the general pattern reversed (Fig. 2a,b). Assuming that the total number of species in the spurred and nonspurred clade roughly correlates with the time since their common ancestor, then these data suggest that diversification rapidly increases after the evolution of spurs and then subsequently declines. Clearly when spurs first evolve there must be equality in the diversity of the two clades (i.e. 1 : 1). If floral spurs promote an initial and rapid diversification, the discrepancy between clades should increase rapidly (i.e. when there are few nonspurred species). Subsequently, at least two processes could reduce the difference in diversity of the two clades. First, ecological factors, such as habitat, or the number of possible pollinator–plant interactions may place an upper limit the number of possible spurred species thus limiting the clade's diversification rate. Nonspurred clade species may be less constrained at this point in time and continue to diversify thus minimizing the discrepancy in diversity between the two clades. A second possibility is that the nonspurred lineage may evolve ‘nonspur’ innovations causing it to diversify rapidly. This again would cause a decrease in the diversity of the two clades.
Semispecies and the genus Aquilegia
Grant also considered the evolution of Aquilegia as a whole and used it to illustrate the concepts of semispecies and syngameon (Grant, 1971, 1981). Here he defined semispecies as an intermediate stage in speciation, more distinct than geographical races but less so than noninterbreeding sympatric species. The key feature of semispecies in Grant's view is that they are interconnected by gene flow but at a much-reduced level compared to geographical races or populations (Grant, 1963). In fact he used Aquilegia formosa and A. pubescens as an example of semispecies since they often form hybrid populations when they come into contact (see below, Grant, 1971; p. 51). He further referred to the entire genus as an example of a syngameon since all the species or semispecies may be linked by hybridization between some of its members (Grant, 1971; p. 54).
Grant's views of semispecies are clearly reflected in his taxonomic description of Aquilegia. He divided the genus into five groups of species by their floral morphology (Grant, 1952): A. ecalcarata, the only spurless member of the genus; the Vulgaris group with nodding blue or purple flowers, hooked short spurs and long petal blades; the Alpina group which is similar to the Vulgaris group except having straight rather than hooked spurs; the Canadensis group with red and yellow flowers, short, straight and stout spurs and short petal blades; and the Coerulea group with erect yellow or blue flowers, long spurs and long petal blades (Grant, 1952). Although it was not explicitly stated, Grant clearly viewed these lineages as monophyletic when he refers to them as semispecies and super-species (Grant, 1993b, 1994a). Implicit in this view is that each major mode of pollination (i.e. bee, bird and hawkmoth) evolved once because pollinators are the driving force behind differentiation of these groups (Table 1). In addition, using floristic associations, Grant suggests that Aquilegia evolved in Eurasia (Grant, 1952, 1993b) and subsequently spread to North America during the mid-Pliocene (i.e. approximately 3.5 million yr ago; Grant, 1994b), a move that fostered the emergence of hummingbird pollination (i.e. the Canadensis group), as well as hawkmoth pollination (i.e. the Coerulea group). Again, this scenario suggests that each major mode of pollination evolved only once.
Table 1. Summary of species groups as defined by Grant (1952) and records of pollinator visitation for specific species. For each group, the general floral morphology is given along with the species with published reports of pollinator visitation
Molecular phylogenetic studies support many of Grant's views about the evolutionary history of Aquilegia but are equivocal about others. The genus apparently diversified rapidly after the evolution of floral spurs resulting in extremely low levels of DNA sequence divergence among species (Hodges & Arnold, 1994a; Hodges, 1997a,b). These data also support the contention that Aquilegia evolved in Eurasia and subsequently spread to North America. However, ITS-based molecular clock estimates for the emergence of hummingbird pollination suggest an origin that is more recent than the mid-Pliocene. For four hummingbird-pollinated species, there has been an average of 1.75 nucleotide changes in the ITS since the divergence from the Old World taxa, or 0.39% changes per nucleotide (Hodges & Arnold, 1994a, 1995). If this group evolved 3.5 million yr ago, this would give a rate of 0.11% per million yr, a value that is considerably lower than the average estimate of 1.4% per million yr for ITS evolution in plant species calibrated by fossil evidence (Zang et al., 2001). Either Aquilegia has an ITS divergence rate that is an order of magnitude slower than most species, or hummingbird pollination evolved much more recently. Given that there is no particular reason to expect a lowered mutation rate in Aquilegia we suspect that the later scenario is more likely.
Grant's views appear to conflict with the hypothesis that floral isolation has caused the increased diversification rate of Aquilegia because we might expect many transitions in mode of pollination during the evolution of the genus rather than each mode evolving only once. While species pairs that have major differences in pollination (e.g. hummingbird/bee, hummingbird/hawkmoth) are dramatic examples, floral isolation can be achieved between species sharing the same type, or even the same species, of pollinator (Grant, 1994c). Thus, lack of major shifts in pollination mode does not necessarily negate the role of floral isolation in diversification. In fact, bees are the primary pollinator of several of the spurred lineages that display increased levels of diversity (e.g. Delphinium and Aconitum; Hodges, 1997a,b). Similarly, bees are likely the primary pollinator of Old World Aquilegia species (Table 1), which are quite diverse (47 of the approximately 70 described species in the genus). However, among the New World Aquilegia species, strong evidence for the role of floral isolation in diversification would be to find that transitions among major pollinator modes have been a key feature of their history (i.e. multiple origins of each pollinator mode) because these are obvious candidates for effective floral isolation. If Grant's view of monophyletic lineages defined by pollinator modes is correct on the other hand, floral isolation could still be important, just less obvious using only phylogenetic evidence. Detailed field observations to test if floral isolation is occurring between species with similar pollinators would be useful future studies, especially between Old World, bee-pollinated species.
Factors other than floral isolation may have also played a role in the diversification of Aquilegia. Variation in habitat associations (e.g. alpine/forest/desert spring, mesic/dry) is as great among species in Grant's recognized groups as among species in different groups (Grant, 1952; Hodges & Arnold, 1994a). If Grant's hypothesis of relationships is correct then shifts in habitat may have been a creative force in the diversification of Aquilegia, perhaps only secondary to pollination mode. Transitions to different habitats are unlikely to be the sole factor in the rapid radiation of Aquilegia, however. A very closely related clade (i.e. Isopyrum+Enemion; Hodges & Arnold, 1994a; Ro & McPheron, 1997) lacks nectar spurs and presumably floral isolation, but has a very similar distribution in the Old and New World. Thus, this clade has likely had similar opportunity to diversify via habitat differentiation as Aquilegia yet it contains only 19 described species compared with the 70 in Aquilegia, despite its older age. Again, this suggests that the innovation of nectar spurs in Aquilegia has fostered diversification.
Unfortunately, because of the low levels of sequence divergence for the DNA regions examined thus far, we have no way to confidently assess whether the groups defined by Grant are monophyletic and therefore whether each pollination mode evolved once or multiple times. Clearly a detailed species-level phylogeny will be critical in evaluating these different hypotheses. Resolution of species relationships in columbines will undoubtedly require the analysis of multiple nuclear regions such as introns, genes that have evolved especially rapidly (Swanson & Vacquier, 2002), and/or distance methods based on a large number of genetic markers (e.g. AFLP; Beardsley et al., 2003).
Hybridization in Aquilegia
Following his initial work on floral isolation (Grant, 1949), Grant sought out systems in nature where floral isolation could be shown to be affective. He chose to study two species of Aquilegia, A. formosa and A. pubescens at the urging of Jens Clausen (Grant, 1952). The genus was of particular interest because it included many species with large differences in morphology yet it was also known for having minimal sterility among the species (Gregory, 1941; Grant, 1952). Since that time more detailed crossing programs have revealed that while viable F1 seed can be produced between virtually any pair of species in the genus, the ease of producing these hybrids and the fertility of the F1 varies depending on the cross (Prazmo, 1965b; Taylor, 1967). Generally, crosses between Old and New World species are more difficult and show lower pollen fertility of F1 plants than crosses within these groups (Prazmo, 1965b; Taylor, 1967). In nature however, when two species are found in close proximity (such as A. formosa and A. pubescens), they are likely to exhibit few genetic barriers to mating.
Grant (1952) considered many aspects of the biology of A. formosa and A. pubescens in order to identify the cause(s) of reproductive isolation that maintained their distinctiveness. He noted that there were large differences in habitat with A. formosa generally occurring at lower elevations in moist areas while A. pubescens occurred at higher elevations in drier soils (Grant, 1952). He also noted that there were differences in flowering time coinciding with these habitat differences. But it was his observations of pollinators that led him to emphasize that the differences in floral morphology between the two species caused floral isolation. He observed hummingbirds and hawkmoths visiting and pollinating populations of A. formosa and A. pubescens, respectively, and in a hybrid zone these pollinators held this same general pattern (Grant, 1952, 1994a). In addition, the hybrids found in this zone reinforced Grant's view. Rather than all types of recombined floral morphologies (indicative of random mating) he primarily found hybrids that could be classified as backcrosses to one or the other species that he attributed to assortative mating among floral types because of pollinator behavior.
Grant's view was challenged when Chase & Raven (1975) observed hummingbirds and hawkmoths visit both A. formosa and A. pubescens. Their conclusion was that floral isolation did not operate and that selection due to habitat requirements was the primary isolating barrier between these species. In response, Grant (1976) agreed – as he had in 1952 – that habitat plays an important role in isolation, but he questioned Chase & Raven's (1975) conclusions about the role of pollinators, noting that they had failed to examine the frequency of pollinator visits to each species and whether floral constancy was occurring. We evaluated this conflict by observing multiple natural populations of the two species simultaneously and quantifying visitation patterns (Fulton & Hodges, 1999). As Grant predicted, we found that there were striking differences in visitation between the two species with A. formosa visited primarily by bees and hummingbirds and A. pubescens visited primarily by hawkmoths and bees (Fulton & Hodges, 1999). Because these differences in visitation could be a result of either floral differences or habitat differences (these are confounded in natural populations) we also observed visitation to arrays of flowers of the two species in both habitats and found the same patterns (Fulton & Hodges, 1999). Thus, differential visitation to A. formosa and A. pubescens is a result of floral differences and thus floral isolation promotes reproductive isolation between them.
While we have documented substantial differences between A. formosa and A. pubescens in the types of visitors to their flowers, there is some visitation by each pollinator group to both species, particularly by bees (Chase & Raven, 1975; Fulton & Hodges, 1999). In our experiments with arrays of both species of flowers, 34 of 39 individual bees (87%) showed floral constancy to one or the other species (Fulton & Hodges, 1999). Grant (1952) made similar observations and suggested that errant bees were responsible for initiating hybrid zones. One aspect of pollination biology that has not been investigated, but would be especially useful for understanding floral isolation between these two species, is to quantify the effectiveness of each group of pollinator. Our observations of bees suggest that they visit the flowers primarily for pollen collection and therefore they may avoid flowers that are not in male-phase (the species are protandrous). Thus, bees may not be as effective pollinators as hummingbirds and hawkmoths that may not discriminate between flowers with different sexual phases. In addition, our studies have primarily been conducted within pure species stands and we do not know yet the effectiveness of floral isolation in hybrid zones. In order to test this, we are currently conducting a study to quantify mating patterns in a hybrid zone using paternity analysis. This work will determine if floral isolation is likely important in reducing introgression across natural hybrid zones by causing assortative mating due to floral morphology, color and/or flower orientation.
In order to understand which floral characters influence floral isolation between these species, we have conducted a series of manipulative experiments to isolate the effects of specific characters (Fig. 3; Fulton & Hodges, 1999; Hodges et al., 2002). We found that hawkmoths strongly discriminate against pendant flowers (the A. formosa phenotype) of A. pubescens (Fig. 3a). Similarly, hawkmoths favor A. formosa flowers that have upright flowers (Fig. 3d). By contrast, when we artificially shortened the spurs of A. pubescens to mimic the spur length of A. formosa flowers, hawkmoths did not discriminate (Fig. 3b). But, spur length strongly influenced pollen removal with longer spurs having more pollen removed than short-spurred flowers (Fig. 3c). Finally, we showed that floral color likely strongly influences visitation because hawkmoths discriminate between upright A. formosa and A. pubescens (Fig. 3d). Thus, each of these characters influences floral isolation either by affecting pollinator visitation (flower orientation, flower color) or by affecting pollen removal (spur length).
While our data strongly support Grant's conclusion that floral isolation is an important aspect of reproductive isolation between A. formosa and A. pubescens, we agree with both Grant (1952, 1976) and Chase & Raven (1975) that other factors, particularly habitat differences play a role in maintaining the distinction between these species. This is not really surprising, since it has been long understood that reproductive isolation is rarely caused by a single factor (Mayr, 1947). In hybrid zones of A. formosa and A. pubescens, both floral morphology and species-specific genetic markers are closely associated with habitat suggesting that selection acts against hybrids (Hodges & Arnold, 1994b; Yang & Hodges unpublished). Other factors such as floral phenology (Grant, 1952) and pollen competition (Carney et al., 1996) may also play a role. While floral isolation alone fails to provide complete reproductive isolation between these two species, its effect is likely magnified when combined with other factors. The cumulative impact of these processes is what allows species like A. formosa and A. pubescens to maintain their distinctiveness and prevent their amalgamation (Grant, 1994a; Fulton & Hodges, 1999). Potential studies that would be extremely useful in clarifying these issues include quantifying the relative roles of factors affecting reproductive isolation (Ramsey et al., 2003), particularly in hybrid zones where isolation is incomplete.
Future studies – the genetics of adaptation and speciation
Relatively little is known about the genetic basis of species differences (Orr, 2001). However as the methods of genomic analysis become available for nonmodel systems, evolutionary biologist will have the opportunity to dissect the genetic basis of many traits. The features of Aquilegia that drew Grant to study species in the genus over 50 yr ago also make it an ideal system to investigate the genetics of adaptation and speciation today. In order to understand the genetics of speciation it is important to focus attention on species that have recently diverged so that differences between them are more likely to have arisen during speciation rather than after divergence was complete. In addition, for studies of the genetics of either speciation or adaptation, fertile hybrids are necessary between divergent species or lines for the traits of interest. In both of these respects, Aquilegia offers many opportunities. First, species of Aquilegia have large differences in floral morphology that are adaptations for pollination by different animals (Fig. 3, Table 1). Furthermore, species differ to a large degree in their habitat requirements such as the differences described for A. formosa and A. pubescens. There are also many other differences among species in habitat and form ranging from some species with drastically reduced leaves forming alpine ‘cushion’ plants (e.g. A. jonesii) to others with much larger growth forms restricted to desert canyons and springs (e.g. A. longissima). Because these species are intercompatible the differences between them are amenable to genetic analysis such as quantitative trait locus (QTL) studies (Hodges et al., 2002) and the low sequence divergence among species suggest that speciation occurred quite recently (see above).
Numerous historical genetic studies have been conducted on species of Aquilegia which suggest that some characters separating species will be controlled, at least in part, by genes of large affect while other characters will be controlled by many genes of small affect. In particular, the studies by Prazmo (1960, 1961, 1965a) considered the genetics of both floral and vegetative traits. Prazmo made crosses between species of Aquilegia and measured segregation patterns in F2 and backcross populations. She found that some traits followed simple Mendelian ratios and could be accounted for by few genes of large effect while many other traits were likely polygenic. Traits that followed Mendelian ratios were the presence/absence of nectar spurs, whether spurs were straight or curved, whether flowers were erect or pendent, and the basic color of flowers. These traits are exactly the traits that are likely to have a strong affect on pollinator visitation (Fig. 3). Similarly, QTL of large affect for traits involved with floral isolation have been found between species of Mimulus (Bradshaw et al., 1998). This suggests that rapid evolution of floral isolation may be a common feature of many plants. Other traits studied by Prazmo showed continuous variation and she considered these to be polygenic in nature (Prazmo, 1965a). These traits concerned the size and shape of inflorescences, leaves, flower organs and seeds as well as pollen and seed fertility and flowering time (Prazmo, 1965a). The studies of Prazmo were primarily based on crosses between the spurless A. ecalcarata and a number of other species. Thus, her findings may not reflect the genetic basis of differences at the time of speciation. However these studies provide a rich source of predictions for the finer-scale QTL studies possible today using recently diverged taxa. Furthermore, given the high degree of intercompatibility among species of Aquilegia, a very wide range of genetic studies are possible for a diversity of traits differentiating species including habitat preference and growth form as well as the floral features emphasized here.
In the future, evolutionary biologists would like to identify the genes underlying adaptive traits. Such knowledge would allow us to answer questions such as whether particular classes of genes drive adaptive change (e.g. structural or regulatory), whether particular types of mutations (e.g. point, small insertion/deletions, or transposons) are responsible for allelic differences and whether single mutations are the basis of QTL of large effect. To reach this level of understanding it will be necessary to clone genes underlying adaptive differences. Cloning these genes will likely require detailed genetic and physical maps and such studies will be easier in diploid organisms with small genomes such as Aquilegia. All species of Aquilegia are diploid (n = 7) and the genome is one of the smallest among Angiosperms (1C of approximately 410 Mbp, Hodges unpub data). Genetic maps are being developed (Hodges et al., 2002) and a bacterial artificial chromosome library for physical mapping is currently being constructed. Another advantage to utilizing Aquilegia for genomic studies is the rapid radiation of the genus. Because species have evolved so quickly (Hodges & Arnold, 1994a; Hodges, 1997a,b), sequence variation among species is low and therefore detailed knowledge of one or two species will likely be transferable to the entire genus. For example, primers designed to amplify microsatellite loci or single gene regions in one species usually amplify single loci in species throughout the genus (Yang et al., unpublished).
We hope we have made clear the vital role Verne Grant has played in the development of Aquilegia as an experimental system. It has been over 50 yr since he began his pioneering work on this genus and we have no doubt that this work will continue to guide and challenge students of evolutionary biology.
We thank DS Bush, D Kaska, T. Shanahan-Bricker, B Counterman and the many undergraduates who have worked in our laboratory over the years. We especially thank two reviewers who substantially improved the manuscript and, of course, Verne Grant for leading the way. Our studies of Aquilegia have been funded by Grants from the White Mountain Research Station, UCSB and NSF (DEB-9726272 and DEB-0129130).