ORIGINAL ARTICLE doi:10.1111/j.1558-5646.2010.00972.x PARALLEL EVOLUTION AT MULTIPLE LEVELS IN THE ORIGIN OF HUMMINGBIRD POLLINATED FLOWERS IN IPOMOEA David L. Des Marais1,2,3 and Mark D. Rausher1,4 1Duke University, Department of Biology and University Program in Genetics and Genomics. Box 90338, Durham, North Carolina, 27708 2E-mail: [email protected] 4E-mail: [email protected] Received September 22, 2009 Accepted January 12, 2010 A transition in flower color accompanying a shift in pollinator guilds is a prominent and repeated adaptation in angiosperms. In many cases, shifts to similar pollinators are associated with similar flower-color transitions. The extent to which this parallelism at the phenotypic level results from parallel changes at the biochemical, developmental, and genetic levels, however, remains an open question. There have been few attempts to determine whether parallelism at these lower levels results from mutation bias or fixation bias of different classes of mutation. We address these issues by examining the biochemical, developmental, and genetic changes that have occurred in red-flowering species of the Mina lineage of morning glories (Ipomoea)andcompare these to the changes reported for I. horsfalliae, which has independently evolved red flowers. Using transgenic techniques, we demonstrate that the transition from blue to red flowers in Mina species is due primarily to down-regulation of the enzyme flavonol-3-hydroxylase (F3H) in flowers but not in vegetative tissues, and that this down-regulation is at least partly due to cis-regulatory change in the gene for F3H. These changes are similar to those exhibited by I. horsfalliae, indicating parallelism at the biochemical and developmental levels, and possibly at the genetic level. KEY WORDS: Adaptation, constraint, anthocyanins, parallelism. Parallel evolution is the independent origin of a character result- by different developmental and genetic changes (Wittkopp et al. ing from changes in the same underlying developmental mech- 2003; Yoon and Baum 2004). anisms (Gould 2002, p. 1079). Although apparent parallelism at Parallelism at the developmental and genetic levels can arise the phenotypic level is frequently observed, it is less often under- for two different reasons. First, it may be that a particular type stood whether the similar phenotypes involved also reflect parallel of phenotypic change can only be accomplished by similar or evolution at the biochemical, developmental, and genetic levels. identical mutations in the same gene. In this case, any time the Spectacular examples exist in which parallel phenotypic changes phenotypic change occurs it will be in the same gene and thus are caused by the same nucleotide substitution (Thompson et al. will appear as parallelism at the genetic level, as well as at higher 1993; Bernasconi et al. 1995; Crandall et al. 1999), by different levels. This scenario has been identified as a type of evolutionary changes in the same genes (Schwinn et al. 2006; Kingsley et al. constraint (Wake 1991). Alternatively, different types of muta- 2009), as well as cases in which phenotypic parallelism is achieved tion (e.g., regulatory vs. function for protein-coding genes) in the same gene, or mutations in different genes of the same path- 3Current address: University of Texas at Austin, Section of way, may produce similar phenotypic change. In this situation, Integrative Biology, 1 University Station C0930, Austin, Texas, parallelism will occur only if one of the mutation types has ei- 78712. ther a substantially higher mutation rate (due either to intrinsic C 2010 The Author(s). Journal compilation C 2010 The Society for the Study of Evolution. 2044 Evolution 64-7: 2044–2054 PARALLEL EVOLUTION OF RED FLOWERS IN IPOMOEA differences in substitution rate or to the size of the target DNA Streisfeld and Rausher (2009a) showed that the biochemical sequence) or a substantially higher fixation probability, once it change in pigment production in I. horsfalliae was due primarily has arisen, than the other types. As has been suggested repeatedly to a marked down-regulation of the enzyme flavonol-3-reductase (see Stern and Orgogozo 2008), differential fixation probability (F3H), which blocks the cyanidin branch of the pathway (see may arise if different types of mutation vary substantially in their Fig. 1), forcing flux down the pelargonidin branch. In addition, degree of associated deleterious pleiotropy. they showed that this pigment change was restricted to floral Because it is normally difficult to characterize the spectrum tissue; in vegetative tissue, plants continue to produce cyanidin of mutations that produce a given phenotypic change, there have derivatives. For I. quamoclit, Zufall and Rausher (2004) identified been few attempts to distinguish between these two causes of de- two genetic changes that could be responsible for the color shift in velopmental and genetic parallelism. However, for flower color I. quamoclit, but did not conclusively demonstrate either actually changes, a rough estimation of this spectrum is provided by re- affected pigment production. One change was a down-regulation ports of the underlying genetic changes involved in spontaneous of F3h, whereas the other was an apparent change in substrate mutations that are maintained as horticultural variants (Streis- specificity of the enzyme DFR. In this report, we use enzyme feld and Rausher 2009a). Here, we use this information to assess functional analyses, gene expression assays, and plant transfor- whether parallel change in flower color in two different lineages mation to determine which of these genes actually caused the shift of morning glories (Ipomoea) are caused by parallel biochemical, to red flowers in I. quamoclit and I. coccinea. In addition, we de- developmental, and genetic change, and whether parallelism at termine whether the tissue specificity of pigment change in these these levels is due to constraint in the form of differential fixa- two species is similar to that in I. horsfalliae. Both of these analy- tion probabilities. In this context, we define parallel biochemical ses were designed to evaluate whether the phenotypic parallelism change as similar flower-color change resulting from the same between these two species is due to underlying developmental pigment type being deployed in the flowers. We define parallel de- and genetic parallelism. velopmental change as a similar phenotypic change resulting from similar change affecting the pattern of flux down the branches of the anthocyanin pathway in similar tissues. Finally, we define ge- Methods netic parallelism as similar phenotypic change resulting from a PLANT MATERIALS AND EXTRACTION OF NUCLEIC similar type of mutation (e.g., enzyme functional knock-out vs. ACIDS down-regulation) in the same gene. Plant material was collected fresh from living specimens of I. The evolution of hummingbird pollination has occurred in- ternifolia, I. quamoclit, and I. coccinea growing in our greenhouse dependently at least four times in the genus Ipomoea (Streis- at Duke University in Durham, NC. RNA was extracted from feld and Rausher 2009a). In each case, a change in flower color buds (including petals, stamens and carpels; collectively “floral” from “blue” to red has accompanied this change, as in the clas- elsewhere in the text) and leaf and stem (collectively “vegetative”) sic bird pollination syndrome (Cronk and Ojeda 2008; Thomson tissue using the Spectrum Plant Total RNA extraction kit (Sigma- and Wilson 2008; bluish flower colors range from blue to purple, Aldrich Corp., St, Louis, MO) according to the manufacturer’s as seen by humans; we use the term “blue” to refer to flow- protocol. DNA was extracted from leaf tissue using a modified ers within this range). In addition, this phenotypic parallelism cTAB protocol (Kelly and Willis 1998). is caused by the same biochemical change: the inferred ances- tral blue pigmentation is due to the production of anthocyanins PIGMENT EXTRACTION AND IDENTIFICATION derived from the precursor cyanidin, whereas the derived red pig- Anthocyanins were extracted from fresh floral and vegetative tis- mentation is due to the production of anthocyanins derived from sue by soaking 250 mg of tissue overnight at room temperature in pelargonidin (Zufall and Rausher 2004; Streisfeld and Rausher 2 M HCl and then boiling for 30 min. Following centrifugation, 2009a). We here focus on two of these lineages: one consisting the supernatant was cleaned three times with ethyl acetate, leaving of the single species Ipomoea horsfalliae, and the other consist- anthocyanidins in the aqueous phase. Anthocyanidins were subse- ing of a small clade (Mina) consisting of approximately 16 red- quently extracted into isoamyl alcohol, which was then evaporated flowered, hummingbird-pollinated species (Austin and Huaman under vacuum. Dried purified anthocyanidins were resuspended 1996; Miller et al. 2004). Within the red-flowered Mina clade, in methanol-HCl (99:1 v/v) and stored at −20◦C until processing. we focus on I. quamoclit, which is sister to the rest of the clade, Anthocyanidins were separated via high performance liq- and I. coccinea. As an outgroup, we examine I. ternifolia,which uid chromatography (HPLC) on a Shimadzu 10A system (Shi- is inferred to be sister to the Mina clade (Miller et al. 2004) and madzu Corp., Columbia, MD). Separation in a 30◦C oven was has blue flowers producing cyanidin derivatives, reflecting the achieved with a propanol gradient from 15% to 27.5%