Read an Excerpt

MAJOR EVOLUTIONARY TRANSITIONS IN FLOWERING PLANT REPRODUCTION

THE UNIVERSITY OF CHICAGO PRESS

CHAPTER 1

EVOLUTIONARY TRANSITIONS IN FLORAL COLOR

Mark D. Rausher

Department of Biology, Duke University, Durham, North Carolina 27708, U.S.A.

The tremendous diversity in flower color among angiosperms implies that there have been numerous evolutionary transitions in this character. The conventional wisdom is that a large proportion of these transitions reflect adaptation to novel pollinator regimes. By contrast, recent research suggests that many of these transitions may instead have been driven by selection imposed by nonpollinator agents of selection acting on pleiotropic effects of flower color genes. I evaluate the evidence for these alternative hypotheses and find that while there is circumstantial evidence consistent with each hypothesis, there are no definitive examples of flower color evolution conforming to either hypothesis. I also document four macroevolutionary trends in flower color evolution: color transitions rates are often asymmetrical; biases favoring loss of pigmentation or favoring gain of pigmentation are both observed, but bias favoring transition from blue to red flowers seems more common than the reverse bias; transitions from blue to red often involve inactivation of branches of the anthocyanin pathway; and color transitions often involve loss-of-function mutations. Finally, I discuss how these trends may be related to one another.

Keywords: floral evolution, anthocyanin, natural selection, pollination biology, pollinator choice, pleiotropy.

Introduction

Angiosperms exhibit a tremendous diversity of flower colors, with sister species often differing in the intensity, hue, or patterning of the corolla. This diversity implies that there have been numerous evolutionary transitions in flower color. The observation that floral color is often correlated with other floral traits, resulting in the common recognition of "pollination syndromes" (Faegri and van der Pijl 1966; Fenster et al. 2004), suggests that many of these transitions have been adaptive. Moreover, the apparent importance of showy flowers in attracting pollinators has led to the common interpretation that pollinators are the primary selective agents influencing flower color and that transitions to different colors represent adaptation to different suites of pollinators, a proposition I call the "conventional wisdom" (Faegri and van der Pijl 1966; Grant 1993; Fenster et al. 2004).

In the first part of this article, I inquire into the causes of evolutionary transitions in flower color. I first evaluate the evidence supporting the conventional wisdom. I then consider evidence supporting alternative interpretations, including the possibilities that many flower color transitions are nonadaptive or that many reflect natural selection on pleiotropic effects of genetic variants that affect flower color.

The astounding variety of floral hues, intensity, and patterns of pigmentation makes it appear as if there are few constraints on the evolution of flower color. However, constraints become apparent when one examines macroevolutionary trends in floral color transitions. In the second part of this article, I discuss how properties of flower color genes and the biochemical pathways they encode may contribute to establishing these trends.

Pollinators as Selective Agents on Flower Color

The idea that evolutionary change in flower color reflects adaptation to novel pollinators can be traced back at least as far as Darwin (for historical review, see Fenster et al. 2004). The primary evidence supporting this contention is the existence of "pollination syndromes," groups of floral traits that occur together typically in plants pollinated by a particular agent. Examples include (1) bird-pollinated flowers, which are typically red or orange and have elongated floral tubes, reduced floral limbs, exserted stigmas, and copious dilute nectar; (2) bee-pollinated flowers, which are typically blue or purple and have short, wide tubes, wide limbs, inserted stigmas, and small amounts of concentrated nectar; and (3) mothpollinated flowers, which are typically white and fragrant, have long floral tubes, and open at night.

Although the generality of floral syndromes has been questioned (Robertson 1928; Ollerton 1996, 1998; Waser et al. 1996) and, clearly, not all flowers exhibit standard syndromes, there is substantial evidence to indicate that many species conforming to a particular syndrome are in fact pollinated most effectively by the agent associated with that syndrome (Fenster et al. 2004). This evidence indicates that we should take seriously the hypothesis that interactions with pollinators have driven the evolution of flower color in many, if not all, species.

While the existence of pollination syndromes is consistent with this hypothesis, it is also consistent with others. For example, like any evolutionary change, a particular flower color transition may have occurred by genetic drift. Alternatively, it may have been driven by selection on pleiotropic effects of flower color alleles, even if the change is deleterious with respect to pollinator attraction. In either case, if the flower color change subsequently attracts novel pollinators, selection imposed by these pollinators may mold other floral characteristics to produce a standard floral color syndrome. It is therefore possible for transitions between pollination syndromes with different flower colors to occur without direct selection on flower color by pollinators. Consequently, no matter how common pollination syndromes may be, their existence cannot be taken as definitive evidence that pollinator-mediated selection drives the evolution of floral color.

What type of evidence would constitute a definitive demonstration that a floral color change reflects adaptation to novel pollinators? Ideally, it must be shown that (1) the change was caused by selection and (2) the agent causing that selection was pollinators. As will be seen below, there are no species for which both of these conditions have been demonstrated. This lack of compelling evidence does not necessarily indicate, however, that pollinators are unimportant effectors of flower color evolution. Rather, it just as likely reflects the difficulty of assessing both of these criteria simultaneously for any given plant species. Because of this, it is worthwhile considering how much evidence supports the generality of each condition separately. If, for example, investigations seldom detect selection on flower color variants, we may be led to think that genetic drift is more commonly responsible for flower color evolution than is currently believed.

For information about selection on and pollinator responses to flower color variation, I conducted a literature search on the Web of Science (keywords used were "flower color variation," "flower color evolution," "selection on flower color," and "flower color polymorphism"). Studies were included only if they attempted to determine either whether flower color variants differed in some component of fitness or whether pollinators respond differently to different color variants (table 1). Information was obtained on 24 different species that exhibited variation in floral color. For two species (Antirrhinum majus and Ipomoea purpurea), information was available for two different color polymorphisms, and these are listed separately in table 1. Nine of the examples involve color divergence between populations or interspecific hybrid zones in which selection was examined on potentially introgressing color phenotypes (table 1A). The remainder involved within-population color polymorphisms (table 1B). Although this search cannot claim to be exhaustive, the investigations it turned up are likely to be representative of studies that have examined selection on and pollinator discrimination among flower color variants.

Selection on Floral Color Variation

If the results from these investigations are taken at face value, selection on flower color variants is ubiquitous. Of 21 species that have been examined, 18 exhibit evidence of selection on flower color phenotype (table 1). However, a number of biases and limitations associated with many of these studies restrict the degree of confidence that can be placed on this conclusion. The first is possible reporting bias: it is likely that evidence for selection is more likely to be reported than lack of evidence for selection.

A second limitation is that in most of the investigations listed in table 1, it is impossible to differentiate between selection on flower color itself and selection on genetically linked traits. In only two of the investigations listed in table 1 was any attempt made to randomize or otherwise account for the effects of the genetic background (the two different polymorphisms in I. purpurea), and even in these investigations, associations between the focal polymorphism and moderately linked loci were probably not disrupted. The inability to identify unambiguously the target of selection may mean that direct selection on flower color variation is much less common than the sample seems to indicate.

A third possible bias in many of these studies arises from the fact that in most investigations that attempt to quantify fitness experimentally, selection is measured for only some components of fitness. In cases in which fitness differences are found in one or more fitness components, it is unlikely that fitness differences in unmeasured components will just compensate the measured components to produce net neutrality. By contrast, in cases in which no fitness differences were detected, it is very possible that differences would be exhibited in other fitness components and, hence, in net fitness. This type of bias will lead to underestimating the prevalence of selection. Because lack of selection was reported for only three of the species examined, however, the extent of this bias will be minor in the sample reported here. It should be noted that the four studies that infer fitness differences by comparing distributions of color morphs along a transect with distribution of neutral markers ("cline" in column 3 of table 1) do not suffer from this bias because the approach implicitly considers all components of fitness.

Given these limitations, what is to be concluded about the prevalence of selection on floral color variation? The most legitimate conclusion is that the evidence suggests that color variants are usually not selectively neutral but that this has been shown definitively for only a handful of species (those exhibiting flower color clines). Thus, the current data are consistent with the hypothesis that selection on floral color variation is ubiquitous and therefore likely responsible for many, if not most, evolutionary transitions in flower color, but they also do not exclude alternative hypotheses. This issue will likely be settled only by future investigations that distinguish between direct selection on floral phenotypes and selection on linked traits. This will be most easily achieved by introgressing different flower color alleles into the same genetic background (e.g., Bradshaw and Schemske 2003) and measuring fitness on the resulting isogenic lines. To be definitive, however, this approach will require genetic documentation that the lines are truly isogenic, i.e., that they differ at only the flower color locus.

Selection Driven by Pollinators

As described above, the conventional explanation for floral color transitions is that they represent adaptations to exploiting different types of pollinators. While I argued that available evidence suggests that color transitions are often adaptive, there is little definitive evidence that pollinators are driving that adaptation. It is certainly true that in most cases that have been investigated (13 of 15 species in table 1), pollinators discriminate among color phenotypes and usually visit some morphs more frequently than others. However, while such discrimination could cause fitness differences among color morphs, one can imagine situations in which no fitness differences result from discrimination. For example, consider a situation in which individual pollinators specialize on one flower color morph or the other, so that there is little pollinator movement between morphs. Even though there may be an inherent tendency to specialize on one morph, so that more individual pollinators visit that morph, this differential visitation will not cause differences in either male or female fitness of the two morphs, as long as seed production is not pollen limited. Therefore, simply demonstrating that pollinators discriminate among morphs is insufficient for concluding that pollinators impose selection on flower color.

Instead, some sort of experimental demonstration that pollinators cause fitness differences is required, either by manipulating pollinator access or by manipulating floral characters. Of the 15 species in table 1 that show evidence of pollinator discrimination, only three have been investigated in this way. Irwin and Strauss (2005) examined a two-locus color polymorphism in a naturalized population of Raphanus sativus. Using progeny analysis, they estimated the proportions of the four paternal haploid haplotypes in each of two treatments: one in which pollination occurred naturally and one in which maternal plants were pollinated with a mixture of pollen in which the haplotype frequencies were equal to those produced by flowers in the population. Proportions were different between the two treatments, indicating that some aspect of pollinator behavior generated differences in the male component of fitness among the flower color genotypes.

Using a similar approach, Waser and Price (1981) found that rare white-flowered individuals of Delphinium nelsonii, compared to blue-flowered individuals, had reduced seed production when pollination was by natural pollinators. By contrast, when plants were hand pollinated, there was no difference in seed production. Moreover, the observation that pollinators visited white-flowered plants at a lower rate suggests that the difference in seed production is due to greater pollinator limitation in the whites, caused by pollinator discrimination. This study convincingly demonstrates that pollinator preferences cause differences in the female component of fitness between the floral color morphs.

Finally, Melendez-Ackerman and Campbell (1998) examined a hybrid zone between Ipomopsis aggregata and Ipomopsis tenuituba, in which red-flowered I. aggregata had higher seed production and outcross male success than did hybrid genotypes or the white-flowered I. tenuituba. In addition, pollinators (primarily hummingbirds) preferentially visited I. aggregata. The causal link between differential visitation and fitness differences was established by experimentally manipulating floral color to disassociate effects of flower color from other floral characteristics. When I. aggregata flowers were painted either red or white, red flowers received more visits and sired more seeds than did white flowers. In another experiment, when flowers of the two pure species and the hybrids were painted red, differences in both visitation rates and fitness were almost completely eliminated. It should be noted, though, that selection by pollinators has not been demonstrated outside the hybrid zone; it is therefore unclear that the selection documented in this study is responsible for flower color divergence between the two species.

Because evidence for pollinator-mediated selection has been found when sought, it is tempting to infer that, in general, differential pollinator visitation among color morphs will tend to cause fitness differences. With only three species examined, however, one cannot, at this point, place too much confidence in this inference.
(Continues...)

---

Excerpted from MAJOR EVOLUTIONARY TRANSITIONS IN FLOWERING PLANT REPRODUCTION by Spencer C. H. Barrett. Copyright © 2008 The University of Chicago. Excerpted by permission of THE UNIVERSITY OF CHICAGO PRESS.
All rights reserved. No part of this excerpt may be reproduced or reprinted without permission in writing from the publisher.
Excerpts are provided by Dial-A-Book Inc. solely for the personal use of visitors to this web site.

---