Adaptive Radiations: from Field to Genomic Studies

Home , Aquilegia flavescens, Aquilegia longissima, Aquilegia pubescens, Aurone, Leucoanthocyanidin

Adaptive radiations: From field to genomic studies

Scott A. Hodges and Nathan J. Derieg1

Department of Ecology, Evolution, and Marine Biology, University of California, Santa Barbara, CA 93106 Adaptive radiations were central to Darwin’s formation of his phenotype–environment correlation, (iii) trait utility, and (iv) theory of natural selection, and today they are still the centerpiece rapid speciation. Monophyly and rapid speciation for many of for many studies of adaptation and speciation. Here, we review the the classic examples of adaptive radiation have been established advantages of adaptive radiations, especially recent ones, for by using molecular techniques [e.g., cichlids (4), Galapagos detecting evolutionary trends and the genetic dissection of adap- finches (5, 6), and Hawaiian silverswords (7)]. Ecological and tive traits. We focus on Aquilegia as a primary example of these manipulative experiments are used to identify and test pheno- advantages and highlight progress in understanding the genetic type–environmental correlations and trait utility. Ultimately, basis of flower color. Phylogenetic analysis of Aquilegia indicates such studies have pointed to the link between divergent natural that flower color transitions proceed by changes in the types of selection and reproductive isolation and, thus, speciation (3). anthocyanin pigments produced or their complete loss. Biochem- Studies of adaptive radiations have exploded during the last 20 ical, crossing, and gene expression studies have provided a wealth years. In a search of the ISI Web of Science with ‘‘adaptive of information about the genetic basis of these transitions in radiation’’ (limited to the subject area of evolutionary biology) Aquilegia. To obtain both enzymatic and regulatory candidate we found 80 articles published in 2008 compared with only 1 in genes for the entire flavonoid pathway, which produces antho- 1990. Furthermore, citations of these articles have grown expo- cyanins, we used a combination of sequence searches of the nentially, from 34 citations in 1990 to Ͼ3,000 in 2008. The Aquilegia Gene Index, phylogenetic analyses, and the isolation of growing interest in adaptive radiations may be ascribed to several novel sequences by using degenerate PCR and RACE. In total we factors. First, molecular phylogenetics has allowed the identifi- identified 34 genes that are likely involved in the flavonoid cation of monophyly and rapid speciation in taxa outside of the pathway. A number of these genes appear to be single copy in traditionally-recognized examples from isolated islands and Aquilegia and thus variation in their expression may have been key lakes (e.g., ref. 8) and thus has expanded the repertoire of study for floral color evolution. Future studies will be able to use these systems. Second, there has been a resurgence in the study of sequences along with next-generation sequencing technologies to speciation (9). Speciation research has often focused on the follow expression and sequence variation at the population level. development of genetic incompatibilities (e.g., refs. 10 and 11), The genetic dissection of other adaptive traits in Aquilegia should whereas studies of adaptive radiations have focused attention on also be possible soon as genomic resources such as whole-genome the role of divergent natural selection to alternate environments sequencing become available. as a primary cause of reproductive isolation (3, 12–15). Thus, studies of adaptive radiations focus on both the evolution of anthocyanins ͉ evolutionary trend ͉ flower color ͉ hybrid zone adaptations and reproductive isolation. Understanding the processes of adaptation and speciation re- daptive radiations have played, and continue to play, a quires considering taxa that have not yet fully attained reproductive Acentral role in the development of evolutionary theory. isolation. Fundamentally, the ability to make hybrids allows the After his exploration of the Galapagos Islands, Darwin recog- dissection of the genetic basis of traits and permits tests of how nized that variants of species were confined to specific islands. individual genetic elements could affect individual fitness. For In particular, Darwin (1) noted that mockingbirds he collected example, Bradshaw and Schemske (16) made near-isogenic lines of on different islands represented 3 distinct forms and that all of Mimulus cardinalis and Mimulus lewisii, each containing the alter- the forms were similar to mockingbirds from the mainland of nate allele from the other species for a flower color locus. They South America. He further noted that the different islands were found that pollinator visitation patterns radically changed and that, within sight of one another, all quite similar in their habitats, and in the proper ecological setting, an adaptive shift in pollinator apparently geologically young. Based on these observations, preference could possibly occur with a single mutation (16). In Darwin (2) began to suspect that species gradually became addition, studying speciation and adaptation early in the process has modified. Thus, began his search for a process that could account significant advantages because subsequent genetic changes can for such observations, leading to his formulation of the theory of obscure the actual causal mutations (ref. 17; see also ref. 82). For natural selection. example, loss-of-function mutations that confer an adaptive advan- Key to Darwin’s suspicion of common ancestry was the close tage may be followed by additional mutations that would, on their proximity of environmentally-similar islands comprising the own, cause loss of function but were not involved with the evolution Galapagos archipelago. In fact, many classic examples of adap- of the trait itself (18). Thus, recent adaptive radiations are especially tive radiations involve islands or lakes; notable examples include fruitful resources for dissecting the genetic basis of adaptations and Darwin’s finches of the Galapagos, honeycreeper birds and speciation. silversword plants of Hawaii, and cichlid fish of lakes Malawi and Adaptive radiations have also played a major role in identi- Victoria in Africa. The power of these examples for inferring the fying evolutionary trends. This area of study is important be- force of natural selection in evolution stems from the marked diversity in form and function among a group of clearly closely- related taxa. The isolation of islands and lakes makes plausible This paper results from the Arthur M. Sackler Colloquium of the National Academy of Sciences, ‘‘In the Light of Evolution III: Two Centuries of Darwin,’’ held January 16–17, 2009, the inference that the taxa are descended from a single ancestor, at the Arnold and Mabel Beckman Center of the National Academies of Sciences and and the link between variation in form and function makes Engineering in Irvine, CA. The complete program and audio files of most presentations are credible the notion that natural selection and evolution have available on the NAS web site at www.nasonline.org/Sackler࿝Darwin. played key roles. Author contributions: S.A.H. and N.J.D. designed research, performed research, analyzed A great deal of research effort has gone into substantiating the data, and wrote the paper. assumptions behind examples of adaptive radiation. In an influ- The authors declare no conflict of interest. ential book, Schluter (3) laid out the definition of adaptive This article is a PNAS Direct Submission. radiation as having 4 features: (i) common ancestry, (ii)a 1To whom correspondence should be addressed. E-mail: [email protected].

www.pnas.org͞cgi͞doi͞10.1073͞pnas.0901594106 PNAS ͉ June 16, 2009 ͉ vol. 106 ͉ suppl. 1 ͉ 9947–9954 Downloaded by guest on September 29, 2021 cause trends imply predictable patterns in evolutionary history. pollinators will avoid visiting flowers whose nectar reward they There has been, for example, much debate about whether body cannot reach. Recently these alternative hypotheses were explic- size and complexity increase through time (19–23). One way to itly tested with a species-level phylogeny of the North American test for repeatable trends has been to test whether the evolution adaptive radiation of Aquilegia (38). Transitions between major of particular traits is associated with species diversity. For classes of pollinators were found to be significantly directional, example, clades of flowering plants that have independently consisting of bee to hummingbird and hummingbird to hawk- evolved floral nectar spurs show a statistically significant trend moth but no reversals and no bee to hawkmoth transitions. for increased species diversity compared with nonspurred clades Concomitant with these shifts were increases in spur length. In (24–26). This finding has led to the nectar spurs being consid- addition, models of the tempo of evolution strongly indicated the ered a ‘‘key innovation’’ with the implication that they mecha- concentration of spur-length evolution at times of speciation. nistically increase the likelihood of speciation. Thus, as with the broad correlation between the evolution of Another way to test for trends during evolutionary history is spurs themselves and species diversity, this study suggests that to test explicitly for directionality of trait evolution on a phy- spur length is intimately linked with the speciation process, at logeny. Critical to such analyses is the reconstruction of ancestral least in North American species of Aquilegia (38, 39). Function- states (27, 28). Here, maximum-likelihood methods have been ally, matching of spur length to tongue length has been shown to used to statistically distinguish between a 1-rate model (in which be adaptive because Aquilegia flowers with artificially shortened both directions of evolution have the same rate) and a 2-rate spurs have less pollen removed from their anthers, and presum- model (29). However, the ability to test between such models ably deposited on their stigmas, because of the body of the depends on the number of shifts observed (27). Because adaptive pollinator being held further away from the flower (40). radiations often are composed of a large number of species and The pollinator shifts found in Aquilegia entail transitions in have entailed multiple shifts in characters, they are prime foci for other traits besides spur length. The orientation of flowers at tests of directional trends (e.g., ref. 27). anthesis can be either pendent or upright (41). All hummingbird- As the development, use, and availability of genomic tools for pollinated species of Aquilegia have pendent flowers, which is nonmodel organisms has increased (30), the ability to determine common in other hummingbird-pollinated species as well (42). the genetic basis of adaptations and speciation is becoming All 5 inferred shifts from hummingbird to hawkmoth pollination possible for an increasing number of taxa. Long-standing ques- entail a transition from fully pendent flowers to upright flowers. tions about whether particular kinds of genes (e.g., regulatory This shift is adaptive, because experimental manipulation of versus structural; see refs. 31–34) and/or particular types of upright Aquilegia pubescens flowers to the pendent orientation mutations such as substitutions, duplications, or transposable reduced hawkmoth visitation by an order of magnitude (40). One elements are responsible for adaptation and speciation can now bee-pollinated species, Aquilegia jonesii, has upright flowers; be addressed. Adaptive radiations are especially amenable to however, the entire plant is only 2–10 cm tall and the flowers are such studies for a number of reasons. First, because adaptive held just above the foliage (43). Given these constraints, upright radiations have been studied for some time, particular trait flowers offer the best access to the spurs and pollen by bee values have often been substantiated as adaptations. Second, pollinators. recent adaptive radiations often consist of taxa that can be Perhaps one of the most visually-striking features of the North hybridized, thus making possible the genetic dissection of traits. American adaptive radiation of Aquilegia is the diversity of floral Third, adaptive radiations may entail a diversity of adaptive traits color, resulting from multiple, independent shifts (Figs. 1 and 2). or the repeated evolution of the same traits in different lineages Using the same Aquilegia phylogeny as for pollinator transitions, providing multiple comparisons within and between traits in a Whittall et al. (44) reconstructed the ancestral states for the single system. Finally, because recent and rapid adaptive radi- presence (blue or red) or absence (yellow or white) of floral ations necessarily consist of closely-related taxa, the develop- anthocyanins and inferred a significant trend of 7 independent ment of genomic tools for 1 species will likely provide tools for losses and no gains (Fig. 2). These shifts in color appear to be analysis across the entire group (30). Thus, adaptive radiations largely adaptive because all 5 inferred shifts to hawkmoth offer the possibility of determining general molecular trends pollination are coincident with loss of anthocyanin production across traits and whether convergence at the phenotypic level (44), pale flowers set more seed when hawkmoth are present involves convergence at the molecular level. (45), and in manipulative studies hawkmoths preferentially visit Here, we illustrate the use of adaptive radiations to under- pale flowers (41). To consider trends in color shifts when stand the processes of adaptation and speciation by reviewing anthocyanin production is maintained, we inferred the color studies of the columbine genus, Aquilegia. We first outline the produced at ancestral nodes in the Aquilegia phylogeny by using broad evolutionary trends pointing to specific traits as being the inferred pollination syndrome as a guide (38). Given that important in this adaptive radiation and briefly review the hummingbird-pollinated flowers in general tend to be red (42) studies explicitly testing the function of these traits. We then and all but 1 hummingbird-pollinated species of Aquilegia have highlight how genomic studies are aiding in our understanding of red flowers (Aquilegia flavescens has yellow flowers), we inferred the genetic basis of adaptation and speciation, emphasizing ancestral nodes reconstructed as hummingbird-pollinated an- flower color as a primary model. We end with a look toward how thocyanin producers to be red-flowered (Fig. 2). Similarly, given the genetic dissection of more complex traits will be possible in that bee-pollinated flowers tend to be blue and all but 1 the near future. bee-pollinated species of Aquilegia have blue flowers (Aquilegia laramiensis has white flowers) we inferred ancestral nodes Evolutionary Trends in Aquilegia reconstructed as bee-pollinated anthocyanin producers to be In a classic prediction of an evolutionary trend, Darwin (35) blue-flowered (Fig. 2). These reconstructions suggest that there hypothesized that the spurs of flowers and the tongues of have been 2 shifts from blue to red flowers and 2 shifts from red pollinators could undergo a coevolutionary ‘‘race’’ and become to blue. Thus, color shifts in Aquilegia, where anthocyanin increasingly long. However, an alternative hypothesis posited production is maintained, do not exhibit an evolutionary trend. that spurs evolve to fit the already-established tongue length of pollinators, and that spur length increases only when a shift to a Flower Color and the Genetics of Adaptation new and longer-tongued pollinator occurs (36, 37). This second The significant trends in pollinator evolution in Aquilegia, to- hypothesis also predicts a directional evolutionary trend: shifts gether with field experimentation on the functional relevance of to shorter spurs may be less likely because shorter-tongued specific traits, make variation in spur length, flower orientation,

9948 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0901594106 Hodges and Derieg Downloaded by guest on September 29, 2021 Fig. 1. Photographs of Aquilegia flowers. (A) A. coerulea.(B) A. scopulorum, which can be polymorphic for blue (Left) and white (Right) flowers. (C) A. longissima.(D) A. saximontana.(E) A. flavescens.(F) A. pubescens.(G) A. formosa. (H) Natural hybrids between A. formosa and A. pubescens. Photos by N. Derieg (A, B, D, E, and G) and S. Hodges (C, F, and H).

and flower color obvious targets for understanding the genetic basis opposite directions would necessitate the gain of enzymatic of adaptive traits and traits affecting reproductive isolation. How- function. Furthermore, after a loss-of-function transition, addi- ever, focusing on the genetics of flower color evolution is also tional loss-of-function mutations may accumulate, making re- advantageous for practical reasons. Since Mendel’s first experi- versals even less likely (18). In the case of blue to red shifts, ments, flower color has been a primary system for dissecting the substrate specificity for the product of F3H may evolve, making genetics of a biochemical pathway because the phenotype is readily reversals less likely as well (48). observable and relatively simple. Many genes in the anthocyanin Changes in flower color could also occur because of altered biosynthetic pathway (ABP) and their regulators have been de- metabolic flux in side branches of the flavonoid pathway. At each scribed, thus facilitating efforts to identify the likely targets of of the intermediate steps in the ABP side branches lead to the selection. Reds and blues of most flowers result from the accumu- production of other important compounds such as flavones, lation of anthocyanin pigments in the vacuoles of floral tissue cells flavonols, and proanthocyanidins (tannins) (Fig. 3). In fact, the (Fig. 1 A, D, and G). When anthocyanins are absent, flowers are entire flavonoid pathway is likely to be more complicated than yellow or white, largely depending on, respectively, the presence or that depicted in Fig. 3. For instance, in Arabidopsis recent absence of carotenoid pigments (Fig. 1 C and E). Anthocyanin profiling of both flavonoid production and gene expression in biosynthetic genes are a central core to the larger flavonoid flavonoid mutants resulted in the identification of 15 new biosynthetic pathway and are expressed in multiple tissues other compounds and the functional characterization of 2 new genes than flowers (46). The production of all anthocyanins involves 6 in the pathway (49). Changes in the abundance or activity of enzymes [in functional order chalcone synthase (CHS), chalcone side-branch enzymes could result in flux away from or toward flavone isomerase (CHI), flavanone-3-hydroxylase (F3H), dihy- anthocyanin production and affect changes in flower color. For droflavonol 4-reductase (DFR), anthocyanidin synthase (ANS), instance, deep-pink flowers of Nicotiana tobacum can be con- and UDP flavonoid glucosyltransferse (UFGT); Fig. 3]. There are verted to white by overexpressing a side-branch enzyme [antho- 3 basic types of anthocyanins: pelargonidins (orange/red), cyanin- cyanidin reductase (ANR)] (50). Similarly, conversion of white dins (blue/magenta/red), and delphinidins (blue/purple). The pro- flowers to those producing anthocyanins has been achieved in duction of pelargonidins requires just the core enzymes but the Petunia by silencing the side-branch enzyme flavonol synthase production of cyanidins and delphinidins depends on 2 enzymes (FLS) and activating DFR (51). Given that the side-branch [flavonoid 3Ј-hydroxylase (F3ЈH) and flavonoid 3Ј5Ј-hydroxylase pathways result in compounds that are important for a number (F3Ј5ЈH)] that add 1 or 2 hydroxyl groups, respectively, on the of other plant functions such as UV protection and herbivore ␤-ring of the product of F3H (46, 47). Subsequently, DFR, ANS, resistance (52, 53), selection for these compounds could result in and UFGT act to produce anthocyanins, which are then trans- pleiotropic changes in flower color (54). However, changes in ported into the vacuole where they accumulate and produce visual flower color caused by altered biosynthetic flux requires in- colors (Fig. 3). creases in expression levels or enzyme activity and are thus Rausher (47) has described general trends in flower color gain-of-function mutations, which, although certainly possible, evolution: shifts are generally blue to red or from producing are likely rarer than loss-of-function mutations (55). anthocyanin to not, although exceptions do occur. Considering A few studies have begun to determine the genetic basis of the biochemical pathway for anthocyanins, Rausher pointed out natural flower color variation. In Petunia axillaris, white flowers that these trends likely arise because mutations causing loss of can be made pink by the introduction, either through introgres- function are more likely than those causing gain of function. Loss sion or transgenetics, of a functional copy of AN2, the R2R3-myb of function of any enzyme in the core ABP would cause a loss transcription factor that controls expression of genes late in the of anthocyanin production and loss of function of F3ЈHor core ABP (Fig. 3). Sequence analysis of AN2 alleles from a range F3Ј5ЈH would cause a shift from blue to red. Transitions in the of collections indicated 5 independent loss-of-function muta-

Hodges and Derieg PNAS ͉ June 16, 2009 ͉ vol. 106 ͉ suppl. 1 ͉ 9949 Downloaded by guest on September 29, 2021 Aquilegia as a Model for Studying Floral Color Evolution Aquilegia has been used for nearly 50 years to study the evolution and genetics of flower color. Prazmo (58, 59) made extensive crossing studies among purple-, white-, red-, and yellow- flowered species of Aquilegia and followed segregation ratios in F2,F3, and backcross populations. She found that flower color variation could largely be ascribed to 4 independently assorting factors (59): Y, which regulates the existence of yellow chromo- plasts; C and R, which are needed for the formation of anthocyanins; and F, which modifies red anthocyanins into bluish- violet pigments. In crosses between either Aquilegia chrysantha or Aquilegia longissima (both of which lack floral anthocyanins) and species with floral anthocyanins, Prazmo (58, 59) found that a single gene (R) could account for anthocyanin production. Whittall et al. (44) found that multiple genes late in the core anthocyanin pathway, especially DFR and ANS, were down- regulated in A. chrysantha, A. longissima, and Aquilegia pineto- rum, which together form a clade of taxa lacking floral anthocyanins (Fig. 2). These results suggest that Prazmo’s factor R is a transregulator of genes late in the ABP. In crosses with the white form of Aquilegia flabellata, Prazmo also found that a second gene (C) could account for anthocyanin production. Whittall et al. (44) found that only CHS was down-regulated in white A. flabellata. Thus, Prazmo’s factor C is likely a mutation in either a transregulator or the cis-regulatory region of CHS. Significantly, F1 progeny from crosses between white A. flabellata and yellow A. longissima or A. chrysantha produce anthocyanins, confirming independent mutations blocking anthocyanin production (59, 60). Finally, Prazmo’s factor F is a mutation that most likely affects either the expression or the function of F3Ј5ЈH (Fig. 3), because blue/purple-flowered species of Aquilegia primarily produce delphinidins (Fig. 2; ref. 60). Fig. 2. Phylogeny of the North American species of Aquilegia (see ref. 38). Colors at the branch tips indicate flower color. Taxa may be fixed for a flower Biochemical analyses of anthocyanins found in Aquilegia color (solid circles) or polymorphic (multiple colors). Colors at nodes indicate species have also been conducted (44, 60, 61). These studies the most parsimonious reconstruction of color, with the likelihood of produc- suggest that losses of floral anthocyanins are likely caused by ing (filled symbols) or not producing (open symbols) anthocyanins indicated changes in expression patterns of the core enzymatic genes of the by pie diagrams (44). To the right of taxon names are 3 boxes indicating, from ABP or changes causing substrate flux away from anthocyanin left to right, the production (filled symbols) or absence (open symbols) of production. In an extensive analysis of flavonoids and other delphinidins (purple), cyanidins (blue), and pelargonidins (red) based on ref. phenolic compounds, Taylor and Campbell (61) found that 60. Arrows indicate down-regulation of genes late in the core anthocyanin species that lack floral anthocyanins (A. pubescens, A. longissima, pathway in flowers of that species compared with regulation in the anthocyanin-producing species A. formosa, A. canadensis, and A. coerulea (44). Lines A. flavescens) all produce anthocyanins in other tissues. Thus, on the right indicate species that form natural hybrids. genes in the ABP must be functional and, if expressed, would result in flowers with anthocyanins. Similarly, Whittall et al. (44) found that both flavones and flavonols are produced in the tions, suggesting that loss of color arose multiple times (56). flowers of species that lack anthocyanins, confirming that the However, most of these alleles also displayed strong down- genes expressed early in the ABP are functional. regulation and thus it is possible that the loss-of-function alleles In an analysis of floral anthocyanins, Taylor (60) found that arose subsequent to the down-regulation of AN2 expression. In species with blue/purple flowers produced either just delphinidins or a combination of delphinidins and cyanidins, whereas Mimulus aurantiacus, segregation analysis of an F2 population for both color and sequence variants of the core ABP genes, red-flowered species produced both cyanidins and pelargoni- along with analysis of their expression and functional capabili- dins. From North America, he included 2 red-flowered species (Aquilegia formosa and Aquilegia canadensis) and 2 blue- ties, indicated that a locus acting in trans accounts for 45% of flowered species (Aquilegia brevistyla and Aquilegia coerulea). color variation (57). These 2 studies provide the strongest According to our reconstruction of ancestral flower colors (Fig. evidence that changes in transregulators are responsible for 2), these species represent the descendants of 1 shift from blue shifts involving changes in the production of anthocyanins as a to red (A. canadensis from the common ancestor with A. whole. Shifts from blue to red flowers have been studied in brevistyla) and 1 shift from red to blue (the A. coerulea clade from Ј Ipomoea (18). Two mutations, 1 causing a loss of F3 H expres- the common ancestor with the A. formosa clade). The first case sion and 1 causing a change in the specificity of DFR for the likely resulted from loss of function or down-regulation of Ј substrate produced by F3H rather than that produced by F3 H, F3Ј5ЈH and a concomitant increase in flux down the cyanidin are sufficient for this color transition. However, which of these and pelargonidin portions of the ABP. This finding suggests that mutations occurred first and thus produced the initial color shift DFR in the ancestor of A. canadensis was a substrate generalist, is unclear. In these examples, the primary focus has been on the unlike some taxa (e.g., Nicotiana) where DFR is specialized for core ABP genes and their regulators. We know of no study that the products of F3ЈH and F3Ј5ЈH and cannot act on the product also has considered how changes in flux from side branches of of F3H to produce red pelargonidins (62). The second case the core ABP may have affected anthocyanin production in appears to be an example of a reversal, with the recovery of natural systems. F3Ј5ЈH activity and delphinidin production. Interestingly,

9950 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0901594106 Hodges and Derieg Downloaded by guest on September 29, 2021 Fig. 3. A generalized flavonoid biosynthetic pathway. In the nucleus, 3 transregulators (a WD40, a bHLH, and a Myb) coordinately affect the expression of multiple genes of the core pathway by binding to their cis-regulatory elements. Enzymes are indicated in the cytosol with black boxes, and the number of candidate genes identified in Aquilegia is indicated. Biochemical intermediates are indicated with light shading in the cytosol, with the substrate for each enzyme to its left and the product to its right. Biochemicals leading to specific anthocyanins are indicated as pelargonidins (pink–red), cyanidins (light blue–blue), and delphinidins (light purple–purple). Lines from specific enzymes to their products in the vacuole indicate side-branch pathways. Enzymes are: CHS, CHI, UFGT, anthocyanin GST (GST), F3ЈH, F3Ј5ЈH, aurone synthase (AUS), isoflavone synthase (IFS), flavone synthase (FNS), FLS, leucoanthocyanidin reductase (LAR), and ANR.

F3Ј5ЈH is likely expressed and functional in rare individuals of conducted neighbor-joining analysis. We then assigned genes to A. pubescens with blue-lavender flowers (63), suggesting that a functional groups based on their phylogenetic clustering. For functional copy of this gene has been maintained in this species instance, F3H, FLS, and ANS all belong to a large gene family as well. of 2-oxoglutarate-dependent dioxygenases (2-ODDs). However, across multiple species, genes with the same enzymatic function Identifying Candidate Genes for the Flavonoid Pathway form monophyletic clades in phylogenetic analyses of 2-ODDs in Aquilegia (68) (Fig. 4A). Inclusion of Aquilegia sequences in the alignment As we have illustrated, a variety of genes could be involved with of 2-ODDs resulted in a tree with 1 Aquilegia sequence in each evolutionary shifts in flower color in Aquilegia. Flower color of the F3H and ANS clades and 2 sequences in the FLS clade could be affected by the enzymatic function/activity of genes in (Fig. 4A). We used a similar approach to identify F3ЈH and either the core or side branches of the flavonoid biochemical F3Ј5ЈH homologs using the sequences from Bogs et al. (66) and pathway or by mutations affecting gene expression either those from Nakatsuka et al. (67) to identify homologs of UF3GT through cis or trans regulation. Multiple copies of genes could and UF5GT (Fig. 4 B and C). Support for the identification of also be involved. Although some researchers have predicted that UF3GT and UF5GT homologs in Aquilegia but not one for changes are more likely to occur at transregulators (64, 65), UF7GT is the fact that Taylor (60) found only 3- and 5-glucosides because this would allow the structural genes to be used in other and 3,5-diglucoside classes of anthocyanins in Aquilegia. tissues, no study to date has considered the expression and Using the Aquilegia gene index we identified candidates for function of the entire flavonoid pathway (Fig. 3). To initiate such nearly all major genes in the flavonoid pathway (Table 1 and Fig. a study, we sought to identify candidate genes for the flavonoid 4 A–C). However, 1 key gene, the R2R3-myb transcription factor pathway in Aquilegia. Using genes of known function (primarily that controls the expression of multiple genes in the ABP in from Arabidopsis) we conducted tblastn searches of the Aquilegia other systems (56), appeared to be absent from the gene index. gene index (http://compbio.dfci.harvard.edu/tgi/cgi-bin/tgi/ When we searched the index with AN2, the myb transcription gimain.pl?gudb ϭ aquilegia). This database is derived from EST factor that controls anthocyanin production in Petunia flowers sequencing from a broad range of tissues and developmental (56), the best hit, TC30372, did not result in a reciprocal best hit stages of greenhouse hybrids between A. formosa and A. pubes- to any known myb transcription factor controlling flower color. cens, including floral tissue producing anthocyanins. The current Given that changes in expression are likely causes of many of the gene index contains 13,556 tentative consensus (TC) sequences shifts in flower color in Aquilegia, identifying the transregulators and 7,278 singleton ESTs and, thus, likely represents a large of the pathway is particularly important. Using RT-PCR and 3Ј fraction of the expressed genes in Aquilegia. From these initial and 5Ј RACE we identified 2 novel myb sequences, AqMyb12 searches, we used the strongest Aquilegia hits in tblastx searches and AqMyb17 (GenBank accession nos. FJ908090 and of The Arabidopsis Information Resource and National Center FJ908091) expressed in sepals of A. formosa. After alignment for Biotechnology Information, and those that resulted in best with the inferred amino acid sequences of other myb proteins hits to the original characterized gene were retained as potential (69) we found that AqMyb12 and AqMyb17 clustered with other flavonoid pathway genes. known regulators of floral anthocyanins (e.g., Antirrhinum Several genes in the flavonoid pathway belong to large mul- ROSEA1, ROSEA2, and VENOSA and Petunia AN2; Fig. 4D) tigene families, making the criterion of reciprocal best hit for suggesting that these 2 genes are strong candidates as key identifying homologous genes suspect. However, phylogenetic regulators of the ABP in Aquilegia. analysis of a number of these multigene families has identified Our analysis revealed 27 candidates for enzymatic genes in the clades that contain proteins with common function (66–68). flavonoid pathway and 7 candidates for transcription factors Thus, after we identified candidates for these genes through regulating the core anthocyanin pathway. Interestingly, 4 genes tblastx searches, we then included their inferred protein se- of the core ABP (CHI, F3H, DFR, and ANS) appear to exist as quences in ClustalW alignments of genes from these studies and single copies. Loss of enzymatic function of any of these genes

Hodges and Derieg PNAS ͉ June 16, 2009 ͉ vol. 106 ͉ suppl. 1 ͉ 9951 Downloaded by guest on September 29, 2021 Fig. 4. Gene trees of loci with known function and candidate loci for similar function from Aquilegia.(A) Flavonoid 2-ODDs, including F3H, FLS 1, FLS, and ANS. (B) Flavonoid hydroxylases (F3ЈH and F3Ј5ЈH). (C) Glycosyltransferases (3GT, 5GT, and various others). (D) Myb transcription factors (Mybs affecting the regulation of genes in the ﬂavonoid pathway). Within each tree, the outer lines indicate groups of genes that carry out the same enzymatic function. Aquilegia sequences are indicated by Aq followed by the TC number from the Aquilegia Gene Index or AqMYBcl12 and AqMYBcl17.

would cause the loss of anthocyanin production. However, if among species of Aquilegia (65), which makes molecular re- these genes are truly single copy then such mutations would also sources developed in 1 species very likely to be transferable to eliminate the products of the side-branch pathways dependent others. For example, a reverse genetic approach, virus-induced on each enzyme and these losses would occur in all tissues (Fig. gene silencing (VIGS), was developed in A. vulgaris, a European 3). These potentially-detrimental pleiotropic effects likely favor species but using sequence information of ANS from North mutations that only change the expression of these genes in floral American species (A. formosa and A. pubescens) (71). Silencing tissues. In fact, down-regulation of DFR and ANS has been the ANS gene in A. vulgaris converted the normally deep-purple correlated with loss of anthocyanin production in most white/ flowers to white (71), confirming that a single copy of ANS is yellow-flowered species of Aquilegia, suggesting that mutations expressed in floral tissue. Furthermore, the VIGS technique in a common transregulator are responsible for these color shifts itself is applicable across species of Aquilegia (71). (44). Loss of expression of these 2 enzymes would enable the Natural populations of Aquilegia that vary in flower color also continued production of most other products of the flavonoid will offer distinct advantages for uncovering the genetic basis of pathway such as flavones and flavonols in floral tissue and may flower color. The production of controlled crosses in the labo- be particularly favored (44). We also found single copies of F3ЈH ratory can be labor intensive and does not necessarily break up and F3Ј5ЈH and thus changes in expression patterns of these genes linked even at relatively-great physical distances. Thus, a genes is most likely responsible for shifts between blue and red single QTL may harbor many genes that could potentially affect flowers. As described above, loss of enzymatic function in these the trait of interest. For instance, we have identified 34 genes that genes in red-flowered species would make future transitions to may play a role in the core ABP and its side-branch pathways blue flowers especially unlikely. Because red to blue transitions (Table 1), thus implicating an average of nearly 5 such genes per have apparently occurred in Aquilegia (Fig. 2) we would predict chromosome (there are 7 pairs of chromosomes in Aquilegia). Of that F3ЈH and F3Ј5ЈH have likely retained enzymatic function course, there may be even more genes involved in the flavonoid but lost expression in red-flowered Aquilegia species. pathway as the Aquilegia gene index likely does not contain all paralogs of the ABP genes and additional side-branch enzymes Molecular Dissection of Flower Color Variation in Aquilegia are possible (49). Thus, it is likely that any 1 QTL region may In addition to identified candidates for nearly all major genes in harbor multiple genes in the ABP. However, natural populations the flavonoid pathway, 2 features of Aquilegia make it a partic- polymorphic for flower color are likely to have had long histories ularly powerful system for identifying the molecular basis of of recombination and low linkage disequilibria. In such popu- flower color evolution. First is the ability to cross virtually any lations, it should be possible to follow sequence variation at all 2 species (59, 70). Segregating populations allow for quantitative of the genes in the ABP and identify specific genes that correlate trait loci (QTL) analysis and the subsequent ability to identify with (and actually influence) flower color. whether genes in the ABP cosegregate with QTL. In an initial Aquilegia offers many examples of pronounced variation in analysis of flower color in A. formosa and A. pubescens a single flower color. For example, populations of A. coerulea (45, 72) QTL was identified (41) and it should now be possible to and Aquilegia scopulorum are often polymorphic for this trait determine which, if any, of the genes in the flavonoid pathway (Fig. 1B). In addition, many natural hybrids exist between several cosegregate with flower color. The ability to cross species also Aquilegia species. Hybrid populations of A. formosa (red; Fig. allows introgression and the creation of near-isogenic lines as 1G) and A. pubescens (primarily white; Fig. 1F) have been described above. Second is the high degree of sequence similarity studied for many years and produce a broad range of floral

9952 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0901594106 Hodges and Derieg Downloaded by guest on September 29, 2021 Table 1. Candidate genes from Aquilegia for the ﬂavonoid quencing provides a scaffold for mapping variation and identi- pathway and its regulation fying the candidate genes that underlie QTL. The Aquilegia ϫ Gene Characterized gene Aquilegia candidate genome is currently in production for 8 sequencing (at the Joint Genome Institute), and as sequencing costs decline, similar CHS AT5G13930 TC22565 genomic resources should become available for many additional TC32786 natural systems. Second, next-generation sequencing techniques TC22820 open the prospects for sequencing whole transcriptomes from TC26722, DR947012 population samples (73, 74). As we have shown here, even CHI AT3G55120 TC22751 without a whole-genome sequence, characterization of ESTs can F3H AT3G51240 TC21211 provide the necessary scaffolds for analyzing the short reads F3’H AT5G07990 DT768941 these technologies produce. Thus, if such population level anal- F3’5’H TC29760 ysis becomes a reality, then it will be possible to quickly obtain DFR AT5G42800 TC21880, TC30546 both expression and sequence data for genes underlying flower ANS AT4G22880 TC31785 color (75) in a large number of populations. UF3GT AT1G05530 TC21486 Genomic resources will also greatly aid in our ability to dissect UF5GT AT5G17050 TC20545 traits about which we currently understand little in terms of TC33100 underlying biochemistry and genetics. Although traits such as R2R3-MYB EF423868 TC30372 petal spur length and flower orientation have strong affects on AqMYB12 pollinator visitation and resulting pollen transfer (76), our AqMYB17 knowledge of how these traits are genetically influenced and bHLH AT5G41315 TC30032 biochemically expressed is meager. Genes affecting cell size or AB232775 DR940663 cell number may cause spur-length differences between species. WD40 AT5G24520 TC21922, DT757614, DR953251 Variation in flower orientation among species is the result of AT1G12910 TC20110 heterochrony as all flowers of Aquilegia go through a similar FNS AB193314 TC26594, TC29308 pattern of orientation during development (41). Early develop- FLS AT5G08640 TC20335 mental stages of buds are upright; they then become pendent and TC25614 ultimately become upright again. Differences between species LAR DQ129686 TC32378 arise because of differences when anthesis occurs in this se- ANR AT1G61720 quence (41). Little is known about the types of genes that may GST AT5G17220 TC22760 affect such floral differences and thus the genetic dissection of TC24814 these traits is more challenging. However, as noted above, the TC29438 existence of hybrid zones between species that differ in these DR949706 traits offers the possibility that genomic techniques, such as TC25908 association mapping, will allow the identification of the genes TC26513 underlying traits such as these. Transporters AT2G34660 TC27360 AY609318 TC23071 Conclusions AT3G59030 TC33309 Adaptive radiations continue to offer a rich resource for under- AY348872 TC30722 standing the process of evolution. As we have outlined here, they The accession numbers are given (from The Arabidopsis Information Re- are ideal for identifying general evolutionary trends and often source and the National Center for Biotechnology Information) of character- consist of the repeated evolution of the same traits, allowing tests ized genes used to search the Aquilegia Gene Index. Functional characteriza- of whether evolution follows predictable trajectories. As we tions, other than for Arabidopsis genes, can be found in refs. 78–81. When Ͼ1 move forward in an era where genomic resources will be Aquilegia sequence is noted, they differ because of the presence/absence of increasingly available, adaptive radiations offer additional ad- large sequence blocks, suggesting they represent either alternative splice vantages. Because recent rapid radiations have resulted in variants or incompletely processed RNA. closely-related taxa with distinct adaptive features, the genomes are likely to be remarkably similar overall, thus leaving relatively clear signals of which genes are likely involved with speciation phenotypes (Fig. 1H). In addition, we have identified numerous and adaptation (e.g., ref. 77). Furthermore, the development of hybrid populations between taxa with different flower colors genomic tools for 1 species will likely be transferable to other (Fig. 2). All such populations offer research opportunities to taxa in an adaptive radiation (30). Exploration of phenotypes in increase the precision of genotype/phenotype correlations, and the field and their genetic basis provides a powerful approach for perhaps even allow definitive verification of specific genes describing evolutionary processes that have shaped biodiversity. underlying color variation in Aquilegia. For the reasons outlined above, adaptive radiations maximize our ability to detect patterns and test both long-standing and The Future of Genetic Analysis of Adaptations emerging hypotheses about the nature of adaptive evolutionary Despite identification of the likely players in flavonoid pathway change. function and evolution in Aquilegia (Table 1), following the expression and sequence variation in all of these genes in natural ACKNOWLEDGMENTS. We thank John Avise and Francisco Ayala for the opportunity to discuss our work at the Sackler Colloquium and John Avise, or laboratory populations is daunting. With advances in DNA Dolph Schluter, and an anonymous referee for providing comments that sequencing, such analyses may nonetheless soon become possi- greatly improved the manuscript. This work was supported by National Sci- ble in Aquilegia and other natural systems. First, genome se- ence Foundation Grant EF-0412727.

1. Darwin C (1963) Darwin’s Ornithological Notes, Bulletin of the British Museum of 3. Schluter D (2000) The Ecology of Adaptive Radiation (Oxford Univ Press, Oxford). Natural History Series, ed Barlow N (British Museum of Natural History, London), Vol 4. Meyer A, Kocher TD, Basasibwaki P, Wilson AC (1990) Monophyletic origin of lake victoria 2, No 7, pp 201–278. cichlid ﬁshes suggested by mitochondrial-DNA sequences. Nature 347:550–553. 2. Darwin C (1876) Autobiography; reprinted (1958) Autobiography: With Original 5. Petren K, Grant BR, Grant PR (1999) A phylogeny of Darwin’s ﬁnches based on Omissions Restored (Cambridge Univ Press, Cambridge, UK). microsatellite DNA length variation. Proc R Soc London Ser B 266:321–329.

Hodges and Derieg PNAS ͉ June 16, 2009 ͉ vol. 106 ͉ suppl. 1 ͉ 9953 Downloaded by guest on September 29, 2021 6. Sato A, et al. (1999) Phylogeny of Darwin’s finches as revealed by mtDNA sequences. 47. Rausher MD (2008) Evolutionary transitions in floral color. Int J Plant Sci 169:7–21. Proc Natl Acad Sci USA 96:5101–5106. 48. Zufall RA, Rausher MD (2003) The genetic basis of a flower color polymorphism in the 7. Baldwin BG, Sanderson MJ (1998) Age and rate of diversification of the Hawaiian common morning glory (Ipomoea purpurea). J Hered 94:442–448. silversword alliance (Compositae). Proc Natl Acad Sci USA 95:9402–9406. 49. Yonekura-Sakakibara K, et al. (2008) Comprehensive flavonol profiling and transcrip- 8. Hodges SA, Arnold ML (1994) Columbines: A geographically widespread species flock. tome coexpression analysis leading to decoding gene-metabolite correlations in Ara- Proc Natl Acad Sci USA 91:5129–5132. bidopsis. Plant Cell 20:2160–2176. 9. Coyne JA, Orr H (2004) Speciation (Sinauer, Sunderland, MA). 50. Xie DY, Sharma SB, Paiva NL, Ferreira D, Dixon RA (2003) Role of anthocyanidin 10. Mihola O, Trachtulec Z, Vlcek C, Schimenti JC, Forejt J (2009) A mouse speciation gene reductase, encoded by BANYULS in plant flavonoid biosynthesis. Science 299:396–399. encodes a meiotic histone H3 methyltransferase. Science 323:373–375. 51. Davies KM, et al. (2003) Enhancing anthocyanin production by altering competition for 11. Phadnis N, Orr HA (2009) A single gene causes both male sterility and segregation substrate between flavonol synthase and dihydroflavonol 4-reductase. Euphytica distortion in Drosophila hybrids. Science 323:376–379. 131:259–268. 12. Schluter D (1996) Ecological causes of adaptive radiation. Am Nat 148:S40–S64. 52. Winkel-Shirley B (2002) Biosynthesis of flavonoids and effects of stress. Curr Opin Plant 13. Rundle HD, Nagel L, Boughman JW, Schluter D (2000) Natural selection and parallel Biol 5:218–223. speciation in sympatric sticklebacks. Science 287:306–308. 53. Treutter D (2006) Significance of flavonoids in plant resistance: A review. Environ Chem 14. Via S (2001) Sympatric speciation in animals: The ugly duckling grows up. Trends Ecol Lett 4:147–157. Evol 16:381–390. 54. Strauss SY, Whittall JB (2006) Nonpollinator agenets of selection on floral traits. 15. Schluter D (2001) Ecology and the origin of species. Trends Ecol Evol 16:372–380. Ecology and Evolution of Flowers, eds Harder L, Barrett SCH (Oxford Univ Press, 16. Bradshaw HD, Schemske DW (2003) Allele substitution at a flower color locus produces Oxford), pp 120–138. a pollinator shift in monkeyflowers. Nature 426:176–178. 55. Orr HA (2005) The probability of parallel evolution. Evolution 59:216–220. 17. Via S, West J (2008) The genetic mosaic suggests a new role for hitchhiking in ecological 56. Hoballah ME, et al. (2007) Single gene-mediated shift in pollinator attraction in speciation. Mol Ecol 17:4334–4345. Petunia. Plant Cell 19:779–790. 18. Zufall RA, Rausher MD (2004) Genetic changes associated with floral adaptation 57. Streisfeld MA, Rausher MD (2009) Altered trans-regulatory control of gene expression restrict future evolutionary potential. Nature 428:847–850. in multiple anthocyanin genes contributes to adaptive flower color evolution in 19. Jablonski D (1997) Body-size evolution in Cretaceous molluscs and the status of Cope’s Mimulus aurantiacus. Mol Biol Evol 26:433–444. rule. Nature 385:250–252. 58. Prazmo W (1961) Cytogenetic studies on the genus Aquilegia II. Crosses between 20. Damuth J (1993) Copes rule, the island rule, and the scaling of mammalian population Aquilegia ecalcarata Maxim and Aquilegia chrysantha Gray. Acta Soc Bot Pol 30:57–77. density. Nature 365:748–750. 59. Prazmo W (1965) Cytogenetic studies in the genus Aquilegia. III. Inheritance of the 21. Van Valkenburgh B, Wang XM, Damuth J (2004) Cope’s rule, hypercarnivory, and traits distinguishing different complexes in the genus Aquilegia. Acta Soc Bot Pol extinction in North American canids. Science 306:101–104. 34:403–437. 22. Hibbett DS, Binder M (2002) Evolution of complex fruiting-body morphologies in 60. Taylor RJ (1984) Floral anthocyanins of Aquilegia and their relationship to distribution homobasidiomycetes. Proc R Soc London Ser B 269:1963–1969. and pollination biology of the species. Bull Torrey Bot Club 111:462–468. 23. Hibbett DS (2004) Trends in morphological evolution in homobasidiomycetes inferred 61. Taylor RJ, Campbell D (1969) Biochemical systematics and phylogenetic interpretations using maximum likelihood: A comparison of binary and multistate approaches. Syst in genus Aquilegia. Evolution 23:153–162. Biol 53:889–903. 62. Nakatsuka T, Abe Y, Kakizaki Y, Yamamura S, Nishihara M (2007) Production of 24. Hodges SA, Arnold ML (1995) Spurring plant diversification: Are floral nectar spurs a red-flowered plants by genetic engineering of multiple flavonoid biosynthetic genes. key innovation? Proc R Soc London Ser B 262:343–348. Plant Cell Rep 26:1951–1959. 25. Hodges SA (1997) A rapid adaptive radiation via a key innovation in Aquilegia. 63. Chase VC, Raven PH (1975) Evolutionary and ecological relationships between Aqui- Molecular Evolution and Adaptive Radiation, eds Givinish T, Sytsma K (Cambridge Univ legia formosa and Aquilegia pubescens (Ranunculaceae), 2 perennial plants. Evolution Press, Cambridge, UK), pp 391–405. 29:474–486. 26. Kay KM, et al. (2006) Floral characters and species diversification. Ecology and Evolu- 64. Clegg MT, Durbin ML (2000) Flower color variation: A model for the experimental study tion of Flowers, eds Harder L, Barrett SCH (Oxford Univ Press, Oxford), pp 311–325. of evolution. Proc Natl Acad Sci USA 97:7016–7023. 27. Mooers AO, Schluter D (1999) Reconstructing ancestor states with maximum likeli- 65. Whittall JB, Medina-Marino A, Zimmer EA, Hodges SA (2006) Generating single-copy hood: Support for one- and two-rate models. Syst Biol 48:623–633. nuclear gene data for a recent adaptive radiation. Mol Phylogenet Evol 39:124–134. 28. Schluter D, Price T, Mooers AO, Ludwig D (1997) Likelihood of ancestor states in 66. Bogs J, Ebadi A, McDavid D, Robinson SP (2006) Identification of the flavonoid adaptive radiation. Evolution 51:1699–1711. hydroxylases from grapevine and their regulation during fruit development. Plant 29. Pagel M, Meade A, Barker D (2004) Bayesian estimation of ancestral character states on Physiol 140:279–291. phylogenies. Syst Biol 53:673–684. 67. Nakatsuka T, Sato K, Takahashi H, Yamamura S, Nishihara M (2008) Cloning and 30. Abzhanov A, et al. (2008) Are we there yet? Tracking the development of new model characterization of the UDP-glucose:anthocyanin 5-O-glucosyltransferase gene from systems. Trends Genet 24:353–360. blue-flowered gentian. J Exp Bot 59:1241–1252. 31. Hoekstra HE, Coyne JA (2007) The locus of evolution: Evo devo and the genetics of 68. Gebhardt Y, et al. (2005) Molecular evolution of flavonoid dioxygenases in the family adaptation. Evolution 61:995–1016. Apiaceae. Phytochemistry 66:1273–1284. 32. Barrier M, Robichaux RH, Purugganan MD (2001) Accelerated regulatory gene evolu- 69. Nakatsuka T, et al. (2008) Identification and characterization of R2R3-MYB and bHLH tion in an adaptive radiation. Proc Natl Acad Sci USA 98:10208–10213. transcription factors regulating anthocyanin biosynthesis in gentian flowers. Plant Cell 33. Britten RJ, Davidson EH (1969) Gene regulation for higher cells: A theory. Science Physiol 49:1818–1829. 165:349–357. 70. Taylor RJ (1967) Interspecific hybridization and its evolutionary significance in genus 34. King MC, Wilson AC (1975) Evolution at 2 levels in humans and chimpanzees. Science Aquilegia. Brittonia 19:374–390. 188:107–116. 71. Gould B, Kramer EM (2007) Virus-induced gene silencing as a tool for functional 35. Darwin C (1862) The Various Contrivances by Which Orchids Are Fetilized by Insects analyses in the emerging model plant Aquilegia (columbine, Ranunculaceae). Plant (John Murray, London). Methods 6:1–12. 36. Wallace A (1867) Creation by law. Q J Sci 4:471–488. 72. Miller RB (1978) Pollination ecology of Aquilegia elegantula and Aquilegia caerulea 37. Wasserthal LT (1997) The pollinators of the Malagasy star orchids Angraecum sesqui- (Ranunculaceae) in Colorado. Am J Bot 65:406–414. pedale, A-sororium and A-compactum and the evolution of extremely long spurs by 73. Shendure J, Ji H (2008) Next-generation DNA sequencing. Nat Biotechnol 26:1135–1145. pollinator shift. Bot Acta 110:343–359. 74. Kahvejian A, Quackenbush J, Thompson JF (2008) What would you do if you could 38. Whittall JB, Hodges SA (2007) Pollinator shifts drive increasingly long nectar spurs in sequence everything? Nat Biotechnol 26:1125–1133. columbine flowers. Nature 447:706–712. 75. Gilad Y, Rifkin SA, Pritchard JK (2008) Revealing the architecture of gene regulation: 39. Hodges SA, Whittall JB (2008) One-sided evolution or two? A reply to Ennos. Heredity The promise of eQTL studies. Trends Genet 24:408–415. 100:541–542. 76. Hodges SA, Fulton M, Yang JY, Whittall JB (2004) Verne Grant and evolutionary studies 40. Fulton M, Hodges SA (1999) Floral isolation between Aquilegia formosa and Aquilegia of Aquilegia. New Phytol 161:113–120. pubescens. Proc R Soc London Ser B 266:2247–2252. 77. Turner TL, Hahn MW, Nuzhdin SV (2005) Genomic islands of speciation in Anopheles 41. Hodges SA, Whittall JB, Fulton M, Yang JY (2002) Genetics of floral traits influencing gambiae. Plos Biol 3:1572–1578. reproductive isolation between Aquilegia formosa and Aquilegia pubescens. Am Nat 78. Morita Y, Saitoh M, Hoshino A, Nitasaka E, Iida S (2006) Isolation of cDNAs for 159:S51–S60. R2R3-MYB, bHLH, and WDR transcriptional regulators and identification of c and ca 42. Cronk Q, Ojeda I (2008) Bird-pollinated flowers in an evolutionary and molecular mutations conferring white flowers in the Japanese morning glory. Plant Cell Physiol context. J Exp Bot 59:715–727. 47:457–470. 43. Munz P (1946) Aquilegia: The cultivated and wild columbines. Gentes Herbarum 79. Nakatsuka T, Nishihara M, Mishiba K, Yamamura S (2005) Temporal expression of 7:1–150. flavonoid biosynthesis-related genes regulates flower pigmentation in gentian plants. 44. Whittall JB, Voelckel C, Kliebenstein DJ, Hodges SA (2006) Convergence, constraint, Plant Sci 168:1309–1318. and the role of gene expression during adaptive radiation: Floral anthocyanins in 80. Goodman CD, Casati P, Walbot V (2004) A multidrug resistance-associated protein Aquilegia. Mol Ecol 15:4645–4657. involved in anthocyanin transport in Zea mays. Plant Cell 16:1812–1826. 45. Miller RB (1981) Hawkmoths and the geographic patterns of floral variation in Aqui- 81. Mathews H, et al. (2003) Activation tagging in tomato identifies a transcriptional regulator legia caerulea. Evolution 35:763–774. of anthocyanin biosynthesis, modification, and transport. Plant Cell 15:1689–1703. 46. Grotewold E (2006) The genetics and biochemistry of floral pigments. Annu Rev Plant 82. Via S (2009) Natural selection in action during speciation. Proc Natl Acad Sci USA Biol 57:761–780. 106(Suppl):9939–9946.

9954 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0901594106 Hodges and Derieg Downloaded by guest on September 29, 2021