| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Population Biology |
Department of Biology, Indiana University, 1001 East Third Street, Bloomington, Indiana 47405 USA
Received for publication August 4, 2006. Accepted for publication January 15, 2007.
ABSTRACT
The genetic architecture of the total phenotype may substantially constrain or enhance the evolution of floral color within populations in response to multiple selection pressures. Using Claytonia virginica I previously identified opposing selection on floral color generated through herbivores and pathogens. Here I ask whether the evolution of floral color in this system is constrained or unconstrained by its phenotypic integration with floral and vegetative traits. Morphological, physiological, and pollen traits were measured on over 400 plants in the field and greenhouse, and these data were used to test whether floral-color morphs differed with respect to other traits and whether the among-trait correlation structure differed across the color morphs. The color morphs varied with respect to most measured traits; however, the pattern of variation was not consistent among them, and there was little evidence of trade-offs with floral color. A common principal components analysis revealed that the pattern of phenotypic integration substantially differed among the color morphs. Combined, these results suggest that floral-color evolution may proceed relatively unconstrained by associations with other traits in this system. The absence of a strong constraint in combination with known fluctuating selective pressures may help to explain observed within- and among-population color variation in this species.
Key Words: Claytonia virginica • common principal components • floral-color variation • phenotypic integration • polymorphism
Several factors can influence the evolution of floral traits including the form and strength of multiple selection pressures, the underlying genetic variation for those floral traits that selection can act on, and genetic correlations between floral and other phenotypic traits (Falconer and Mackay, 1996
; Roff, 1997
). For example, if a population with red and white floral-color morphs undergoes strong and consistent pollinator-mediated selection for increased redness, the prediction that red morphs would increase in frequency over generational time might not hold true for a number of reasons. If phenotypic variation for floral color is the result of environmental influences (e.g., soil pH) rather than allelic differences among individuals, the effects of pollinator-mediated selection in one generation would not be transmitted to subsequent generations because the target of selection is not heritable. Likewise, if floral redness trades off with other fitness-related traits such as pollen viability and this association has a genetic basis, white-flowered morphs might be maintained in the population because evolution toward redder phenotypes would be constrained by the fitness costs of reduced pollen viability. This study utilizes a system that has a distinct floral-color polymorphism with known opposing selective pressures on floral color to ask if the evolution of color in this system might be constrained by trade-offs between color and other fitness-related traits.
Patterns of phenotypic variation within natural population may take several forms. For example, continuous traits (e.g., plant height) are often distributed in a Gaussian-like fashion with an intermediate, distinct peak, whereas discrete traits (e.g., male or female sex organs) often yield patterns of variation with multiple distinct peaks and no overlap among intervals. This second type of phenotypic variation is regarded as being true polymorphism and suggests a relatively simple genetic basis for trait inheritance (e.g., X vs. Y chromosomes). Phenotypic variation may also be manifest as some intermediate between these two if the distribution contains multiple distinct peaks and some overlapping distribution of trait values among them. Floral-color variation in the morning glory (Ipomoea purpurea, Convovulaceae) is a good example of this latter kind of phenotypic variation that is neither completely continuous, nor discrete with only a few classes. In southeastern USA populations, the hue of individual flowers can be characterized as primarily blue, pink, or white with varying intensities (Ennos and Clegg, 1983
). For example, blue-flowered individuals may be intensely blue, dark blue, light blue, or white with blue rays; in addition, there are also at least five pink floral morphs present in nature (Ennos and Clegg, 1983
).
Floral-color variation in spring beauty (Claytonia virginica, Portulacaceae) is similar to that of morning glory in having four floral-color variants in natural populations, ranging from completely white to intensely red, with some overlapping distribution among them (Frey, 2003
, 2004
). The varying degrees of whiteness are produced by the presence of quercetin and kaempferol glycosides, which are cream-colored pigments and co-pigments with anthocyanidins in flowers (Harborne, 1976
, 1988
). In C. virginica, these pigments occur systemically throughout the plant body (Doyle, 1981
, 1983
). Quercetin and kaempferol glycosides have been identified as deterrents to herbivores and fungi in a variety of systems (Bate-Smith and Swain, 1962
; Harborne, 1991
); however, in spring beauty they only deter herbivory (Frey, 2004
). In morning glory, a pleiotropic relationship between pigmentation and herbivore defense has been suggested as a mechanism by which floral-color variation may, in part, be maintained (Simms and Bucher, 1996
; Fineblum and Rausher, 1997
). Likewise, in spring beauty, associations between floral color and herbivore and pathogen defense provide a mechanism by which fluctuating selection generated through natural enemies could, in part, maintain within-population floral-color variation (Frey, 2004
).
Although slug herbivores and fungal pathogens have been identified as opposing, indirect agents of selection on floral color in C. virginica, without knowledge of the underlying pattern of covariation among phenotypic traits, we still are uncertainas to the kind of evolutionary response this selection may generate. The genetic architecture of the total phenotype (floral, vegetative, physiological, and reproductive traits) may substantially constrain or enhance the evolution of floral color within populations in response to multiple selection pressures (e.g., Loeschcke, 1987; Houle, 1991
). For example, several studies have shown that both quercetin and kaempferol glycosides are required for the production of functional pollen (Ylstra et al., 1992
; Taylor, 1995
; Vogt and Taylor, 1995
; Xu et al., 1997
); therefore, if a positive relationship exists between in vivo quercetin and kaempferol concentrations and pollen fecundity (i.e., white-flowered variants produce better quality pollen than red-flowered variants), this correlation could substantially constrain the evolution of redder-flowered morphs. Likewise, some studies have suggested that both of these flavonols may function as a filter against UV-B radiation, but reduce photosystem II photochemical efficiency (Middleton and Teramura, 1993
; Skaltsa et al., 1994
); if this trade-off exists in C. virginica, it could constrain the evolution of whiter-flowered morphs. In addition, pollinator preferences for particular flower sizes or numbers may affect the evolution of floral color, if color is associated with the expression of these traits (Frey, 2004
).
This study is a first step toward understanding how the total phenotype is integrated among the color morphs of C. virginica and how these relationships among traits may constrain or enhance floral-color evolution under previously identified selective pressures. To assess the extent to which phenotypic correlations might constrain an evolutionary response to these selective pressures, I measured a variety of morphological, physiological, and pollen characteristics on over 400 plants grown in the field and greenhouse in two separate seasons. These data were used to (1) ask how floral-color variation was associated with variation of other traits, (2) determine the underlying pattern of phenotypic covariation among the measured traits, and (3) ask whether the phenotypic architecture (among-trait correlation structure) differed among the floral-color variants. If opposing selection can freely make associations between floral color and fitness in this system and help maintain floral-color variation, the results of this study should show a loose association between variation in floral-color and variation in other traits, and weak associations among phenotypic traits with substantial differences in the correlation structure among the color variants. However, if floral color trades off with other fitness-related traits and the architecture underlying the expression of phenotypic traits is consistent among the color morphs, then the evolution of floral color in light of opposing selection could be substantially constrained and potentially lead to the loss of color polymorphism.
MATERIALS AND METHODS
Study system
Claytonia virginica is a spring-flowering, perennial woodland herb distributed throughout northeastern North America. Physiologically independent inflorescences arise from a single corm and are subtended by two opposite, cauline leaves (Jónsdóttir and Watson, 1997
; Morgan, 1998
; Whigham and Chapa, 1999
). In southern Indiana, plants typically emerge from the soil from late-January to early-February, bear all flowering buds upon emergence, and have fully expanded leaves prior to first-flowering (F. Frey, personal observation). Flowers open acropetally, one to two at a time over a period of several weeks from mid-March through mid-April. The flowers are regular with two "sepals," five "petals," five antipetalous stamens, and have a three-carpellate gynoecium with a total of six ovules. Morphologically, the flowers consist of two bracts ("sepals") and a petaloid calyx ("petals"); however, to avoid confusion, this paper retains the common terminology (Zomlefer, 1994
). The predominantly white, bowl-shaped corollas have varying degrees of red shading localized in the veins (described in detail later) that correlate with anther and pollen color (Frey, 2003
). The staminate phase lasts for a single day, and the pistillate phase may last for up to 6 days if the flowers remain unpollinated (Doyle, 1981
). The flowers are self-compatible, but not self-pollinating (Schemske, 1977
; Motten, 1986
). In southern Indiana, the primary pollinator is a solitary bee, Andrena erigeniae (Hymenoptera: Andrenidae).
Study population
There are over 50 recognized cytotypes in the C. virginica aneuploid complex ranging from 2n = 12 to 2n = ca. 192, which fall into one of four flavonoid races that differ with respect to geographical distribution, cytotype, leaf characteristics, and the accumulation of kaempferol and quercetin 3-O-glycosides (Rothwell, 1959
; Rothwell and Klump, 1965
; Lewis et al., 1967
; Lewis and Semple, 1977
; Doyle, 1981
, 1983
). This study was performed using plants from a large woodlot on the Indiana University Bayles Road property in the spring of 2001 and 2002. All plants in this population were of race III and 2n = 16. In 2001, a total of 230 plants were identified as recently emerged perennials and marked by placing a four-inch roofing nail with a painted number next to the base of the emerging inflorescence. In addition, these plants were mapped onto a coordinate-grid system. Ninety of these seedlings were immediately transplanted to a greenhouse in 10-cm pots containing a 50:50 mix of sterilized soil and Metro Mix (Scotts Co., Marysville, Ohio, USA). Twenty of these original 230 were used in a separate study, and the remaining 120 were left undisturbed. In 2002, 250 seedlings were located in a different area of the same woodlot, marked with twenty-centimeter tree spikes, and mapped. A total of 210 plants in 2001 and 250 plants in 2002 were used in this study.
Floral-color variation
As a first step toward asking whether floral-color evolution might be constrained by associations with other traits, I assessed the nature of color variation in this population. Upon first flowering, all 460 plants were assigned a "color class" by sight using a color chart (Frey, 2003
; 1 = white, 2 = light pink, 3 = mauve, 4 = crimson), and quantitative floral-color data were measured with a portable reflectance spectrometer. The quantitative floral-color data and the methods used to collect them are reported elsewhere (Frey, 2004
); however, these results are summarized in a different fashion here to illustrate the discrete nature of color variation in this population. Briefly, spectral data were taken from multiple flowers on multiple plants at different times during the flowering season using an S2000 miniature fiber optic spectrometer with a PX-2 pulsed xenon lamp (Ocean Optics, Dunedin, Florida, USA). Because spectral profiles were characteristically m-shaped with peaks at 450 nm and 650 nm and a valley at 550 nm, a "spectral shape parameter" was calculated as [(%reflectance450 + %reflectance650)/2] – %reflectance550 to allow simple statistical comparisons. Spectral data were measured for 70 plants in the greenhouse in 2001 and 230 plants from the field in 2002. Spectral estimates of floral color among flowers on an individual plant were highly repeatable throughout the flowering season (Frey, 2003
). To assess the validity of these four floral-color classes, I performed a one-way ANOVA on the combined spectral data from each year and used a post-hoc Tukey comparison procedure to test for differences among the classes. There was good support for the presence of four floral-color variants (Fig. 1). Several lines of evidence, including a common-garden experiment and a reciprocal-transplant experiment, suggest that this floral-color variation is due to genetic, rather than environmental, variation (Frey, 2003
, 2004
).
|
). In C. virginica, each inflorescence is subtended by two leaves, which are typically similar in length (Frey, 2003
). At first flowering, the length and mid-width of a single, haphazardly chosen leaf subtending each inflorescence was measured with a set of digital calipers. The length and width of a haphazardly chosen petal on the first and second flower to open on each inflorescence were also measured with a set of digital calipers. Previous work has shown that both within-flower measures of petal size and among-flower measures within an inflorescence are highly correlated (Frey, 2003
). In both field seasons, several plants identified at emergence were later unusable because of deer trampling, falling branches, or some other unknown cause of injury. As a result, of the 460 plants identified for use in this study, inflorescence and flowering bud data were taken from a total of 403 plants, leaf measures were taken from a total of 395 plants, and floral measures were taken from a total of 402 plants. In seven instances, leaves were torn from plants upon first flowering allowing floral measures, but not leaf measures. Leaf and petal area was estimated as an oval (0.25 x 
x length x width). Among-inflorescence leaf measures were averaged for each individual (1–10 estimates per plant), and among-inflorescence flower measures were averaged for each individual (two estimates per inflorescence). The evening before the second and third flower opened on the primary inflorescence (the first to flower on plants with multiple inflorescences), individual flowers were covered with a small pollination bag constructed from bridal veil to prevent insect visitation. Upon opening, one haphazardly selected, freshly dehisced anther from each flower was placed in a sterile, 25-mL scintillation vial. The vials, with lightly screwed caps, were placed in the laboratory to dry naturally. When fully dry, 20 mL of a 1.5% saline solution was added to each vial, which was vigorously shaken for 10 s to disengage pollen from the anther wall, vortexed for 10 s to equally distribute pollen grains throughout the solution, and allowed to rest for 5 s before three estimates of the number of pollen grains in each sample were taken with an Elzone 280 particle counter (Micromeretics, Inc., Norcross, Georgia, USA). The three estimates were averaged to obtain a sample estimate, and the sample estimates from the second and third flower were averaged to obtain a whole plant estimate. Pollen production estimates were successfully taken for 361 plants.
In addition, pollen viability estimates were taken from the second and third flower to open on the primary inflorescence in 2001. The contents of a second haphazardly chosen, freshly dehisced anther from each flower were separately emptied onto a glass slide with a set of fine forceps. Two drops of Alexander's stain (Alexander, 1969
, 1980
) were added to the slide, which was then covered with a coverslip that was lightly ringed with petroleum jelly to create a semipermanent slide. Following activation by passing a flame under the slide four times, pollen viability was assessed under a compound microscope with a 40x objective. Pollen viability estimates from the two flowers were averaged to provide an overall estimate for each individual. Pollen viability data were taken from a total of 148 plants.
Maximal photosynthetic rate and stomatal conductance were measured on the plants transplanted to the greenhouse in 2001. At the third flower stage, a haphazardly selected cauline leaf on the primary inflorescence was chosen and analyzed with a LI-6400 portable gas exchange system (LI-COR, Lincoln, Nebraska, USA) with the following settings: flow = 500, temp = 25°C, quantum flux = 1500 mmol photons·m–2·s–1. A preliminary study showed that these settings maximized individual photosynthetic rate (F. Frey, unpublished data). While each leaf was in the chamber, the edges of the leaf were marked with a fine felt-tipped pen, and the leaf area inside the chamber was traced onto a sheet of graph following measurement. The amount of leaf material inside the chamber during the photosynthetic measurements was estimated by comparing the mass of the graph paper cutout to a previously calculated relationship between graph paper area and graph paper mass (F. Frey, unpublished data). These data were entered into the LI-6400 so that the measures of gas exchange were made with the actual area of the leaf in the chamber. Measures of maximal photosynthetic rate and stomatal conductance were taken from a total of 71 plants.
In C. virginica, photosynthate acquired in a particular season does not contribute to phenotypic trait expression or reproductive success in that season; instead, these resources are stored in belowground corms and used as the sole source of energy in the following season (Whigham and Chapa, 1999
). Upon senescence, corms from these plants were carefully unearthed, cleaned of all soil debris, and weighed. Seven plants had hollow corms, and these data were not included in subsequent analyses (64 total corms).
Statistical analyses
To determine whether the color variants were associated with the particular floral and/or vegetative phenotypes measured, a one-way MANOVA was used to calculate Wilk's 
. A series of separate one-way ANOVAs, with post-hoc Tukey analyses to identify homogeneous subsets, were used to investigate the nature of among-morph variation for each of these phenotypic traits alone. Kendall's coefficient of concordance was calculated to determine the consistency with which the measured traits varied among the color morphs. Differences in maximal photosynthetic rate and stomatal conductance were assessed using an ANCOVA with the leaf area inside the chamber during measurement as a covariate, and a one-way ANOVA was used to determine whether the color variants differed with respect to final corm biomass.
Phenotypic correlations between the five traits measured on all plants in 2001 and 2002 were separately calculated for each color variant, and a sequential Bonferroni correction was used to assess statistical significance (Rice, 1989
). For each phenotypic trait combination, the weighted sum of squares of the z values corresponding to the correlation coefficients (X2) was calculated and compared to a 
2 distribution to determine if there was significant heterogeneity among the four floral-color variants (Sokal and Rohlf, 1981
).
The common principal components method (Flury, 1988
) was used to test whether the structure of covariation between the five phenotypic traits was maintained among the four color morphs in greater depth. In addition to being simply equal or unequal, phenotypic covariation matrices can share complex relationships such as the orientation of particular principal components (Flury, 1988
). Both the "jump-up" and "step-up" approaches of Phillips and Arnold (1999)
were used to infer the degree of multivariate similarity among the color morphs. The jump-up approach utilizes log-likelihood ratio tests to compare a model that assumes complete heterogeneity among covariance matrices (i.e., unrelated) against progressively similar models (i.e., different levels of shared principal components, proportionality, equality) until a statistically significant deviation is encountered. The step-up approach differs from the jump-up approach because it compares hypotheses concerning partial principal component similarity to the next lower model in the hierarchy, rather than the model that assumes complete heterogeneity. In other words, the step-up procedure tests the likelihood that particular model is valid (e.g., two shared principal components) against the next lower model (e.g., one shared principal component), rather jumping all the way down the hierarchy to test the model against complete unrelatedness. Because of unit differences (i.e., inflorescence counts, bud counts, area measures in square millimeters, pollen counts), traits were first standardized to units of within-trait standard deviations (zero mean, unit variance) prior to calculating the phenotypic variance–covariance matrices. Therefore, the four matrices that were analyzed simultaneously under the null hypothesis of no association were actually phenotypic correlation matrices. There is no theoretical consideration that prevents the application of this technique to correlation matrices (Phillips and Arnold, 1999
; Volis et al., 2004
). Both analyses were performed using the program CPC (Phillips, 1998
). With the exception of the CPC analyses, all statistics were performed using SPSS for Mac OSX (SPSS, Chicago, Illinois, USA).
RESULTS
Morphological traits
Intermediately colored plants had substantially larger flowers than either of the two extreme floral color variants (Table 1). In addition, there were moderately significant differences among the color classes with respect to total floral bud number and leaf area (Table 1). For both of these traits, there was a trend for the reddest morph having fewer/smaller values than the other three variants. There were no clear differences among the color morphs with respect to inflorescence number or pollen production per anther. When considered simultaneously, these five traits significantly discriminated the four floral-color variants (MANOVA: Wilk's = 0.880, F15, 956 = 3.031, P < 0.001). However, there was little consistency with respect to how these morphological traits varied among the color variants (Fig. 2; Kendall coefficient of concordance: W = 0.020, df = 4, P > 0.95). Not surprisingly, inflorescence number and total bud number varied in a similar fashion among the color variants as illustrated by the clustering of lines associated with these traits in Fig. 2. However, the patterns of variation for leaf area, petal area, and pollen production per anther were quite distinct.
|
|
2 = 11.30; P < 0.01) and bud number–pollen production per anther (2 = 10.28; P < 0.01). Weak differences among the color morphs were detected for three trait combinations: inflorescence number–pollen production per anther (2 = 9.36; P < 0.05), petal area–leaf area (2 = 7.50; P < 0.06), and bud number–inflorescence number (2 = 7.42; P < 0.06). Just as there was little consistency with respect to how phenotypic traits varied among the color morphs (Fig. 2), there was no clear pattern with respect to which color morphs were driving the observed heterogeneity in the aforementioned correlations. For example, the phenotypic correlation between inflorescence number and pollen production is near zero for all classes except color class four, which shows a negative correlation between these two traits prior to Bonferroni correction (Table 2). In contrast, color class two has a near zero correlation between petal and leaf area, whereas the other color classes have a positive correlation between these two traits prior to a Bonferroni correction (Table 2).
|
|
|
DISCUSSION
Although the number of loci and alleles responsible for the observed pattern of floral-color variation in this system remain undiscovered, independent lines of evidence show that there is indeed a genetic basis for floral color in this system (Frey, 2003
, 2004
). This study uses patterns of phenotypic variation among the color morphs, as well as inferences about the phenotypic architecture drawn from trait correlations, to qualitatively infer patterns of genetic covariation. There is good support for this approach in the literature (e.g., Cheverud, 1988
; Roff, 1996
; Waitt and Levin, 1998
; Armbruster et al.,
1999; Herrera et al., 2002
; Volis et al., 2004
), but it is important to note that these patterns do not necessarily reflect, and likely do not reflect, the exact nature of the genetic structure underlying observed phenotypic associations (Willis et al., 1991
). Therefore, the results of this study should be interpreted with some caution. Nevertheless, these data are particularly useful for understanding the maintenance of floral-color variation in C. virginica in two respects. First, the patterns of phenotypic variation among color variants described here provide a baseline for understanding how floral color may be linked to other traits. Second, the patterns of phenotypic covariation revealed provide insight into the form of multivariate trait evolution that could take place in light of opposing selection on floral color via herbivores and pathogens and of stabilizing selection on flower size via pollinators identified in previous work (Frey, 2004
).
This work suggests three interesting associations between floral color and other traits: plants with intermediately colored flowers had the largest flowers (Table 1) and highest pollen viability (Table 4), and leaf size increased with floral whiteness (Table 1), whereas photosynthetic rate per unit leaf area did not (Table 4). Intermediately sized flowers were found on individuals of all color variants; plants near the smallest end of the distribution of intermediate color classes and plants near the largest end of the distribution of extreme color classes had a comparable size. Therefore, stabilizing selection on flower size would serve to preserve floral-color classes over time. The pollen viability and photosynthetic rate data are different from what was expected (increased whiteness was expected to be associated with higher pollen viability and lower rates of photosynthesis), but provide interesting lines of future study with respect to the maintenance of floral-color variation. For example, if the intermediately colored morphs are heterozygous at loci associated with floral color, their increased siring success relative to the extreme floral-color morphs through greater pollen viability may recreate white- and red- flowering morphs over generational time. In addition, because estimates of photosynthesis per unit leaf area were relatively constant, and there was a strong trend for increased leaf area with increased whiteness, white-flowering plants may have higher net photosynthate production than red-flowering plants, which would affect survivorship and trait expression in the next growing season (Frey, 2003
).
Many of the observed correlations between traits were not significantly different from zero, even before a Bonferroni correction (Table 2). This raises the question of whether this study had sufficient statistical power to detect phenotypic correlations within each of the floral-color groups. A post-hoc power analysis (Phillips, 1998
) showed that this study had approximately 90% power to detect phenotypic correlations as small as approximately 0.45 with significance levels adjusted by the Bonferroni method (initial 
= 0.05), and approximately 80% power to detect correlations as small as approximately 0.25 without Bonferroni correction. Although this study did not have the power to detect whether the bulk of the observed phenotypic correlations were significantly different from zero, these results do suggest that associations among most phenotypic traits are relatively weak and that the differences among the color variants with respect to single correlation comparisons are real.
Consistent with other work in this system (Morgan, 1998
) and another species (Worley and Barrett, 2000
), there was no observed trade-off between flower size and flower number; instead, the phenotypic correlation between flower number and petal area was significantly positive (Table 2), and there was no evidence that the color variants differed with respect to the strength of this association. The lack of a flower number-size trade-off in this system could be due to resource uncertainty or perenniality (Morgan, 1998
), as well as possible environmental effects masking genetically based allocation trade-offs.
There was significant heterogeneity among the four floral-color variants with respect to the multivariate structure of the phenotypic covariance matrices composed of inflorescence number, flowering bud number, measures of petal and leaf size, and pollen production; only the component with the largest eigenvalue was shared among all color variants (Table 3). These results are consistent with univariate comparisons of elements among the floral-color variants, which suggest heterogeneity in several cases. There are at least two ways in which these results could be manifest. First, random environmental effects on the phenotype may have swamped out shared elements of the phenotypic architecture among the color variants, making it appear as though they do not share an underlying pattern of phenotypic integration. Second, the underlying genetic relationships among inflorescence number, flower number, leaf area, petal area, and pollen production may be generally weak. The lack of consistency among color variants with respect to the multivariate phenotypic architecture, as well as the lack of consistency with respect to which floral-color morph diverges from the group with respect to particular trait combinations, suggests that substantial genetic variation for most phenotypic traits exists and that phenotypes may be loosely integrated. Taken together, these data suggest that the floral-color morphs in C. virginica may have different genetic architectures and thus potentially follow different evolutionary courses in light of known indirect, opposing selection pressures on floral color.
FOOTNOTES
1 The author thanks I. Anderson, J. Busch, B. Frey, and K. Frey for assistance in the field, and E. Brodie, III, J. Gastony, and C. Lively for their thoughts and discussion. The quality of this manuscript was substantially improved by constructive comments from W. S. Armbruster, L. Delph, C. Fenster, and an anonymous reviewer. This work was supported by an internal grant to F.M.F. from the Department of Biology at Indiana University. 
2 Author for correspondence (e-mail: ffrey{at}mail.colgate.edu
; present address: Department of Biology, Colgate University, 13 Oak Drive, Hamilton, New York 13346 USA
LITERATURE CITED
Alexander M. P.. 1969. Differential staining of aborted and non-aborted pollen. Stain Technology 44: 117-122.[ISI][Medline]
Alexander M. P.. 1980. A versatile stain for pollen, fungi, yeast, and bacteria. Stain Technology 55: 13-18.[ISI][Medline]
Armbruster W. S. Di Stilio V. S. Tuxill J. D. Flores T. C. and J. L. Velásquez Runk. 1999. Covariance and decoupling of floral and vegetative traits in nine neotropical plants: a re-evaluation of Berg's correlation-pleiades concept. American Journal of Botany 86: 39-55.
Bate-Smith E. C. Swain T.. 1962. Flavonoid compounds. In M. Florkin, H. S. Mason [eds.], Comparative biochemistry: a comprehensive treatise 705-809 Academic Press, New York, New York, USA.
Cheverud J.. 1988. A comparison of genetic and phenotypic correlations. Evolution 42: 958-968.
Doyle J. J.. 1981. Biosystematic studies on the Claytonia virginica aneuploid complex Ph.D. dissertation, Indiana University, Bloomington, Indiana, USA.
Doyle J. J.. 1983. Flavonoid races of Claytonia virginica (Portulacaceae). American Journal of Botany 70: 1086-1091.
Ennos R. A. Clegg M. T.. 1983. Floral color variation in the morning glory, Ipomoea purpurea. Journal of Heredity 74: 247-250.
Falconer D. S. Mackay T. F. C.. 1996. Introduction to quantitative genetics, 4th ed Longman, Essex, UK.
Fineblum W. L. Rausher M. D.. 1997. Do floral pigmentation genes also influence resistance to enemies? The W locus in Ipomoea purpurea. Ecology 78: 1646-1654.
Flury B.. 1988. Common principal components and related multivariate models Wiley, New York, New York, USA.
Frey F. M.. 2003. Floral evolution: how pollinators, herbivores, and pathogens maintain floral-color variation in natural populations of Claytonia virginica Ph.D. dissertation, Indiana University, Bloomington, Indiana, USA.
Frey F. M.. 2004. Opposing natural selection from herbivores and pathogens may maintain floral-color variation in Claytonia virginica (Portulacaceae). Evolution 58: 2426-2437.[CrossRef][ISI][Medline]
Harborne J. B.. 1976. Functions of flavonoids in plants. In T. W. Goodwin [ed.], Chemistry and biochemistry of plant pigments 736-779 Academic Press, London, UK.
Harborne J. B.. 1988. Introduction to ecological biochemistry Academic Press, London, UK.
Harborne J. B.. 1991. Flavonoid pigments. In G. A. Rosenthal, M. R. Berenbaum [eds.], Herbivores: their interactions with secondary plant metabolites 389-429 Academic Press, San Diego, California, USA.
Herrera C. M. Cerdá X. García M. B. Guitián J. Medrano M. Rey P. J. Sánchez-Lafuente A. M.. 2002. Floral integration, phenotypic covariance structure and pollinator variation in bumblebee-pollinated Helleborus foetidus. Journal of Evolutionary Biology 15: 108-121.
Houle D.. 1991. Genetic covariance of fitness correlates: what genetic correlations are made of and why it matters. Evolution 45: 630-648.
Jónsdóttir I. S. Watson M. A.. 1997. Extensive physiological integration: an adaptive trait in resource-poor environments?. In H. de Kroon, J. van Groenendael [eds.], The ecology and evolution of clonal plants 109-136 Backhuys, Leiden, Netherlands.
Lewis W. H. Oliver R. L. Suda Y.. 1967. Cytogeography of Claytonia virginica and its allies. Annals of the Missouri Botanical Garden 54: 153-171.
Lewis W. H. Semple J. C.. 1977. Geography of Claytonia virginica cytotypes. American Journal of Botany 64: 1078-1082.
Middleton E. Teramura A. H.. 1993. The role of flavonol glycosides and carotenoids in protecting soybean from ultraviolet-B damage. Plant Physiology 103: 741-752.[Abstract]
Morgan M. T.. 1998. Female fertility per flower and trade-offs between size and number in Claytonia virginica (Portulacaceae). American Journal of Botany 85: 1231-1236.
Motten A. F.. 1986. Pollination ecology of the spring wildflower community of a temperate deciduous forest. Ecological Monographs 56: 21-42.
Phillips P. C.. 1998. Designing experiments to maximize the power of detecting correlations. Evolution 52: 251-255.
Phillips P. C. Arnold S. J.. 1999. Hierarchical comparison of genetic variance-covariance matrices. I. Using the Flury hierarchy. Evolution 53: 1506-1515.
Rice W. R.. 1989. Analyzing tables of statistical tests. Evolution 43: 223-225.
Roff D. A.. 1996. The evolution of genetic correlations: an analysis of patterns. Evolution 50: 1392-1403.
Roff D. A.. 1997. Evolutionary quantitative genetics Chapman and Hall, New York, New York, USA.
Rothwell N. V.. 1959. Aneuploidy in Claytonia virginica. American Journal of Botany 46: 353-360.
Rothwell N. V. Klump J. G.. 1965. Chromosome numbers in populations of Claytonia virginica from the New York metropolitan area. American Journal of Botany 52: 403-407.
Schemske D. W.. 1977. Flowering phenology and seed set in Claytonia virginica (Portulacaceae). Bulletin of the Torrey Botanical Club 104: 254-263.
Simms E. L. Bucher M. A.. 1996. Pleiotropic effects of flower color intensity on herbivore performance on Ipomoea purpurea. Evolution 50: 957-963.
Skaltsa H. Verykokidou E. Harvala C. Karabourniutis G. Manetas Y.. 1994. UV-B protective potential and flavonoid content of leaf hairs of Quercus ilex. Phytochemistry 37: 987-990.
Sokal R. R. Rohlf F. J.. 1981. Biometry: the principles and practice of statistics in biological research, 2nd ed W. H. Freeman, New York, New York, USA.
Taylor L. P.. 1995. Flavonols: effects on fertility and fecundity. Crop Science 35: 1521-1526.
Vogt T. Taylor L. P.. 1995. Flavonol 3-O-glycosyltransferases associated with Petunia pollen produce gametophyte-specific flavonol diglycosides. Plant Physiology 108: 903-911.[Abstract]
Volis S. Verhoeven K. J. F. Mendlinger S. Ward D.. 2004. Phenotypic selection and regulation of reproduction in different environments in wild barley. Journal of Evolutionary Biology 17: 1121-1131.[CrossRef][ISI][Medline]
Waitt D. E. Levin D. A.. 1998. Genetic and phenotypic correlations in plants: a botanical test of Cheverud's conjecture. Heredity 80: 310-319.
Whigham D. F. Chapa A. S.. 1999. Timing and intensity of herbivory: its influence on the performance of clonal woodland herbs. Plant Species Biology 14: 29-37.
Willis J. H. Coyne J. A. Kirkpatrick M.. 1991. Can one predict the evolution of quantitative characters without genetics?. Evolution 45: 441-444.
Worley A. C. Barrett S. C. H.. 2000. Evolution of floral display in Eichhornia paniculata (Pontederiaceae): direct and correlated responses to selection on flower size and number. Evolution 54: 1533-1545.[ISI][Medline]
Xu P. Vogt T. Taylor L. P.. 1997. Uptake and metabolism of flavonols during in-vitro germination of Petunia hybrids (L.) pollen. Planta 202: 257-265.
Ylstra B. Touraev R. M. Moreno B. Stoger E. van Tunen A. J. Vicente O. Mol J. N. M. Heberle-Bors E.. 1992. Flavonols stimulate development, germination, and tube growth of tobacco pollen. Plant Physiology 100: 902-907.
Zomlefer W. B.. 1994. Guide to flowering plant families University of North Carolina Press, Chapel Hill, North Carolina, USA.
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
