SEARCH

SEARCH BY CITATION

Keywords:

  • constraints;
  • distyly;
  • floral integration;
  • flower morphology;
  • heterostyly;
  • phenotypic evolution;
  • phenotypic matrix;
  • tristyly

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information
  • Although the6 magnitude and pattern of correlation among floral traits (phenotypic integration) is usually conceived as an adaptation for successful pollination and reproduction, studies on the evolution of plant reproductive systems have generally focused on one or a few characters. If evolutionary transitions between reproductive systems involve morphological floral adjustments, changes in the magnitude and pattern of phenotypic integration of floral traits may be expected.
  • In this study, we focused on the evolutionary dynamics of a complex adaptive trait, the extent of reciprocity (reciprocal placement) among sexual organs in a heterostylous species, and explored the associated changes in phenotypic floral integration during the transition from tristyly to distyly. The extent of reciprocity and both the magnitude and pattern of floral integration were characterized in 12 populations of Oxalis alpina representing the tristyly–distyly gradient.
  • Although the extent of reciprocity increased along the tristyly–distyly transition, the flower size diminished. These adjustments did not affect the magnitude, but did affect the pattern, of floral integration.
  • Changes in the pattern of floral integration suggested that allometric, functional and pleiotropic relationships among floral traits were affected during this evolutionary transition.

Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Flower diversification in animal-pollinated species is thought to result from the selective pressures exerted by pollinators (Herrera et al., 2006; Johnson, 2006). Flowers are thus envisioned as mechanical devices optimized through the action of natural selection to maximize pollen transfer between potential mates (Darwin, 1877; Stebbins, 1970). In the long term, this selective process is expected to result in a precise relationship between the flowers of animal-pollinated species and pollinator morphology. Because flowers are modular entities composed of several relatively independent organs (perianth, androecium and gynoecium), the reciprocal adjustment between flowers and pollinators should involve the evolution of particular allometric relationships among floral organs, that is, the combination of trait values maximizing the rates of pollen delivery and receipt (Nilsson, 1988; Alexandersson & Johnson, 2002). Natural selection is expected to favour the evolution of floral integration, in other words, high phenotypic covariances among traits associated with pollinator attraction and reward, pollen delivery and pollen receipt (Berg, 1960; Stebbins, 1970; Conner & Via, 1993; Armbruster et al., 1999; Herrera, 2001; Herrera et al., 2002; Armbruster et al., 2004; Badyaev, 2004; Anderson & Busch, 2006; but see Ordano et al., 2008).

The floral phenotype can be represented as a square matrix of phenotypic variances and covariances (Chernoff & Magwene, 1999; Steppan et al., 2002; Smith & Rausher, 2008). This matrix allows the calculation of both the magnitude and pattern of floral integration, the first describing the strength with which floral traits correlate with each other, and the second referring to the overall structure of the phenotypic correlations among traits (Ordano et al., 2008). Depending on its magnitude and pattern, floral integration may favour or constrain the response to the selective pressures exerted by pollinators or other agents (Campbell, 2009). Evolutionary changes in the floral phenotype may be constrained if the trajectory favoured by natural selection differs from the direction of the axis of maximum variation of the flower (i.e. when the direction of the eigenvector associated with the largest eigenvalue of the variance–covariance matrix of the flower differs from the direction of the vector of selection gradients) (Kirkpatrick & Lofsvold, 1992; Schluter, 1996; Merilä & Björklund, 2004). The higher the discrepancy between these two axes, the stronger the restrictions on adaptive evolution (Arnold, 1992; Schlichting & Pigliucci, 1998). Evolutionary transitions between reproductive systems involving adaptive changes in the functioning and morphology of flowers offer an ideal opportunity to examine the evolution of phenotypic integration.

The remarkable evolutionary lability exhibited by heterostylous reproductive systems (Ornduff, 1972; Weller, 1992; Domínguez et al. 1997) makes them interesting study subjects for addressing the evolution of floral integration. Literature on heterostyly shows several examples describing evolutionary transitions from a polymorphism with three floral morphs (tristyly) to one with only two floral morphs (distyly) (Ornduff, 1972; Charlesworth, 1979; Weller, 1986, 1992; Eckert & Barrett, 1994; Barrett et al., 2004; Weller et al., 2007). Heterostyly is characterized by the reciprocal positioning of anthers and stigmas between floral morphs (Lloyd & Webb, 1992; see Fig. 1a,d), and by the presence, in many species, of a heteromorphic incompatibility system preventing seed production after self- and same-morph pollinations. Theoretically, efficient pollen transfer in heterostylous species requires a precise relationship between flower and pollinator morphology, because pollen from different stamen levels adheres to specific parts of the pollinator’s body (Ganders, 1979; Barrett & Glover, 1985). Therefore, the precise location of sexual organs (anthers and stigmas) within and between flowers (i.e. the reciprocal placement of anthers and stigmas from different floral morphs; hereafter reciprocity) is the primary mechanism maintaining the evolutionary stability of heterostylous systems (Charlesworth, 1979). Such a mechanism is reinforced by a tubular or closed funnel-shaped perianth that constrains the way in which pollinators enter the flower (Ganders, 1979), thus increasing the likelihood of pollen transfer between same-level sexual organs (legitimate crosses, sensuDarwin, 1877). Hence, selection for features promoting legitimate pollen flow should produce high covariances among floral traits, and therefore relatively high levels of floral integration.

image

Figure 1.  Hypothetical changes in the reciprocal placement of sexual organs among floral morphs during the evolution of distyly from tristyly in Oxalis alpina. Horizontal arrows indicate legitimate pollen transfer between mid-level anthers and mid stigmas in a hypothetical ancestral tristylous population (a). During the evolutionary transition towards distyly, modifications of incompatibility in short- and long-styled morphs result in a reduction in the frequency of the mid-styled morph. This, in turn, reduces the availability of mid styles and, consequently, the probability of legitimate pollen transfer from mid-level anthers (b, c). The arrow thickness represents the probability of successful pollen transfer from mid anthers to target stigmas. Although the shift from tristyly to distyly involves the loss of the mid-styled morph and, consequently, mid styles in the population, both short- and long-styled plants retain mid-level stamens. Thus, natural selection should increase the efficiency of pollen transfer between the mid-level stamens of the short- and long-styled morphs through changes in the length of the mid-level stamens, such that they match the positions of short and long stigmas. See vertical arrows above mid level anthers in (c). Consequently, the two stamen whorls of the short or long floral morph should converge to the length of the style in its reciprocal morph (d). A new phenotype in which each floral morph has all its stamens at a single level is expected to evolve, thus making pollen transfer more efficient.

Download figure to PowerPoint

Altered patterns of pollen transfer and incompatibility modifications have been shown to account for the breakdown of tristyly through the loss of one floral morph (Charlesworth, 1979; Weller, 1986; Weller et al., 2007). This evolutionary transition is expected to shift the adaptive peak of a tristylous floral phenotype towards a new adaptive peak in which only two floral morphs coexist (distyly; see Fig. 1 for details). Given that the breakdown of tristyly generally involves the loss of the mid-styled morph (Ornduff, 1972; Charlesworth, 1979; Weller, 1986, 1992; Eckert & Barrett, 1994; Barrett et al., 2004; Weller et al., 2007), its gradual loss would reduce the availability of target stigmas for pollen from the mid anthers of short- and long-styled plants (Fig. 1b,c). Thus, natural selection should favour any modification of the floral phenotype of the short- and long-styled morphs that increases the rates of pollen transfer from their mid stamens towards their new target styles (Mulcahy, 1964; Lewis & Rao, 1971; Weller, 1992; Eckert & Mavraganis, 1996). Such modifications should produce a derived dimorphism in which both anther whorls of short- and long-styled plants would have the same length as the style of their complementary morph (long and short styles, respectively; see Fig. 1d). In other words, increased reciprocal placement between sexual organs of the short- and long-styled floral morphs is expected during the evolution of distyly from tristyly (i.e. during the loss of the mid-styled morph).

If flowers constitute integrated modules with correlated floral traits, mid stamens as well as other floral traits (e.g. other sexual organs or perianth traits) should respond to the selective pressures that promote an increased reciprocity between the long- and short-styled morphs. At some point during the evolution of a new distylous floral morphology, correlations among floral traits may constrain further evolutionary changes (Merilä & Björklund, 2004; Campbell, 2009). Thus, natural selection is expected to reduce temporarily the magnitude and modify the pattern of floral integration during the transition from tristyly to distyly. Although some studies have shown that the breakdown of tristyly generally involves modifications in the length of the mid stamens of both short- and long-styled morphs (Mulcahy, 1964; Ornduff, 1964; Lewis & Rao, 1971; Lewis, 1975; Weller, 1979), no studies have examined whether such changes involve an increase in the extent of reciprocity between formerly illegitimate sexual organs during the transition from tristyly to distyly, whether other floral traits are involved in this process, and how these modifications affect the magnitude and pattern of floral integration.

Within the Sonoran Desert Sky Islands (Arizona, New Mexico, USA; and Sonora, Mexico), the heterostylous species Oxalis alpina shows a marked variation in the frequency of floral morphs, ranging from isoplethic tristylous (equal frequency of the three morphs) to distylous populations in which the mid-styled morph is absent (Weller et al., 2007). The loss of the mid-styled morph in this species results from incompatibility modifications (defined as the relative loss of incompatibility between illegitimate crosses) in the short- and long-styled morphs (see details in Fig. 2; Weller et al., 2007). Accordingly, the presence of populations representing different steps in the evolutionary transition from tristyly to distyly offers a unique opportunity to explore changes in flower morphology associated with this evolutionary shift.

image

Figure 2.  Relationship between the frequency of the mid-styled morph and the extent of incompatibility modification in 10 populations of Oxalis alpina (modified from Weller et al., 2007).

Download figure to PowerPoint

In this study, we first describe the floral variation among morphs and populations of O. alpina representing the evolutionary transition from tristyly to distyly. Second, we test whether the variation in flower morphology among these populations fits the hypothesized pattern of increased reciprocity between short- and long-styled plants associated with changes in the position of sexual organs within the flower. Third, we assess how these changes affect the magnitude and pattern of floral integration during this evolutionary transition.

Materials and Methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Species and study populations

Oxalis alpina (Rose) Knuth (section Ionoxalis) is a widespread perennial species ranging from Guatemala to the southwest USA (Denton, 1973) that blooms during the rainy season (July–September) producing bright-pink, funnel-shaped flowers with radial symmetry. Two species of bees Heterosaurus bakeri and H. neomexicanus (Andrenidae) (Weller, 1981), account for 94.5–100% of the floral visits among populations (F. Baena et al., unpublished).

Twelve populations (three distylous and nine tristylous) from the Sky Islands region, representing the evolutionary gradient from tristyly to distyly, were selected (Fig. 3; see also Weller et al., 2007). In this region, incompatibility modifications of short- and long-styled morphs are strongly correlated with the frequency of the mid-styled morph (Fig. 2; Weller et al., 2007). For the populations included in our study, we used modifications of incompatibility to describe the evolutionary gradient between tristyly and distyly, because the frequency of the mid morph may be influenced by other factors (i.e. founder effects; Eckert & Barrett, 1992; Agren & Ericson, 1996). Modifications of incompatibility were obtained from a previously published index of incompatibility modification based on hand-crossing experiments performed for each of 10 populations (Weller et al., 2007). For both short- and long-styled plants in each population, the mean number of seeds produced by the illegitimate cross (S × m/L and L × m/S) was divided by the mean number of seeds produced by the corresponding legitimate cross (S × s/L and L × l/S; the first capital letter represents the morph of the recipient plant and the latter the anther level of the donor morph; i.e. m/L is the mid level anther of the long-styled morph). The index of incompatibility modification for each of nine tristylous and one distylous population was then obtained by averaging these two ratios (see Weller et al., 2007). High values of the index correspond to populations in which illegitimate crosses produce as many seeds as legitimate crosses. These populations were those with a lower frequency, or even absence, of the mid-styled morph (Fig. 2).

image

Figure 3.  Distribution of 12 distylous and tristylous populations of Oxalis alpina in the Sky Island region of the Sonoran Desert. The number in parentheses after each population name corresponds to the extent of incompatibility modification of each population. For frequencies of the short-, mid- and long-styled morphs, see Weller et al. (2007). Modified from Marshall (1957) and McLaughlin (1995).

Download figure to PowerPoint

Phenotypic measurements

In the summer of 2005, we measured five flowers per plant of 30 plants per morph from nine populations of O. alpina. Given the low availability of plants in three additional populations, their sample sizes per morph were reduced to 10 plants. Because we had estimates of the extent of incompatibility modification for only 10 populations, analyses based on this index excluded the remaining two populations. Plants from each population were collected in 20012004. In all cases, we avoided the collection of plants less than 1 m apart in order to minimize the probability of sampling genetically related individuals. Bulbs were planted in soil-less mix at the University of California-Irvine glasshouse (Weller et al., 2007). Given that plants from all populations were grown in the glasshouse in similar conditions, most of the observed morphological variation should have a genetic component. Eight morphometric measurements were taken from each flower (Fig. 4). These attributes were chosen because they have been shown to be related to pollinator attraction and flower handling (corolla traits) and pollen donation and deposition (style and stamen lengths) in various species (Conner & Rush, 1996; Conner et al., 2003). The three different reproductive whorls of each morph are referred to as the short (so), mid (mo) and long (lo) sexual organs (see Fig. 4). Depending on the floral morph, the short, mid and long sexual organs will correspond to the length of the stigma or one of the two stamen levels. Except for the length of the reproductive whorls (which were measured under the microscope, STEMI SR Zeiss, Oberkochen, Germany), all measurements were performed with a Mitutoyo (CD-6, CSX, Mitutoyo Corp., Kawasaki, Kanagawa, Japan) digital calliper to the nearest 0.01 mm. Trait means were calculated for each plant and used as independent observations in statistical analyses.

image

Figure 4.  Morphological floral traits measured in the three floral morphs of Oxalis alpina: flower width (fw), petal length (pl), petal width (pw), flower length (fl), flower tube width (tw), length of the short (so), mid (mo) and long (lo) sexual organs (whose identity changes depending on the floral morph; see ‘Phenotypic measurements’ section).

Download figure to PowerPoint

We tested whether differences in flower morphology among O. alpina populations, representing the evolutionary transition from tristyly to distyly, fitted the hypothesized pattern of increased reciprocal positioning of sexual organs of short- and long-styled plants (Fig. 1). In order to increase the reciprocity between two morphs of a formerly tristylous system, both within- and between-morph floral modifications should occur. The within-morph component of reciprocity (RW) refers to the convergence of the two stamen whorls of a given morph to the same length (the mid and long stamens of short-styled plants, or the mid and short stamens of long-styled plants), whereas the between-morph component of reciprocity (RB) measures how well the stamen length of a given morph matches the height of the style of its reciprocal morph. RW and RB were calculated for each morph in every population. A value of unity for RW and RB indicates that the two stamen whorls of short- or long-styled flowers have converged to the same length, and that the mid stamens of either short- or long-styled plants have reached the length of their new target styles, respectively. Differences in flower size among populations were accounted for by dividing RW and RB of each floral morph by the length of the longest whorl in each population, the long styles.

To address changes in reciprocity along the evolutionary gradient, we integrated RW and RB into one composite index describing the extent of reciprocity (RI) within a population. RI ranges from zero to unity, with unity expressing perfect reciprocity between long and short morphs (see formulae for the calculation of RW, RB, RI in the legend of Fig. 6).

Statistical analyses

Patterns of variation in flower morphology  Because we were interested in determining whether the flower morphology of short- and long-styled morphs changed during the breakdown of tristyly, we first looked for differences in overall flower morphology among morphs and populations (and their interaction) by means of multivariate analysis of variance (MANOVA) (JMP version 5.1, SAS Institute Inc, 2005). Univariate analyses were also performed to determine which traits were modified during this evolutionary transition.

To describe the relationships among floral traits and how they changed during the shift from tristyly to distyly, we performed a principal components analysis (PCA) including all 12 populations (JMP version 5.1, SAS Institute Inc, 2005). Plant means of each floral trait were used as the raw data for PCA. Sexual organs and corolla traits were included together in this analysis. Because principal components (PCs) are linear combinations of the original variables, they can be used as integrated descriptors of the independent covariance groups that make up the flower (Kleinbaum et al., 1988; Marcus, 1990; Reyment & Jöreskog, 1993; Ordano et al., 2008). In order to determine whether these covariance groups changed during the breakdown of tristyly, we performed two independent analyses of covariance (ANCOVAs) in which we evaluated the effect of morph (long and short), the extent of incompatibility modification (covariate) and the interaction term (morph × extent of incompatibility modification) on the scores derived from the first two PCs. Only the values of the short- and long-styled morphs for each population were included in these analyses because these are the floral morphs that were expected to change.

Reciprocal morphological adjustments between sexual organs during the breakdown of tristyly  According to our hypothesis, both RW and RB should show a positive relationship with the extent of incompatibility modification. We tested whether these two dependent variables followed the predicted trend by means of two independent covariance analyses. The extent of incompatibility modification (the covariate), the effect of morph (long and short) and the interaction of morph × extent of incompatibility modification were included in these models as independent variables. A significant effect of the covariate would indicate that both RW and RB changed during the tristyly–distyly transition, whereas a significant effect of the morph term would indicate that both floral morphs differed in the extent of reciprocity. Finally, a significant interaction term (morph × extent of incompatibility modification) would indicate that the rate of change in either RW or RB varied between long and short plants. In order to evaluate whether the extent of reciprocity (RI) increased along this evolutionary gradient, we then regressed RI per population against the extent of incompatibility modification. A Shapiro–Wilk’s test was performed on residuals of RW, RB and RI to evaluate whether they were normally distributed.

Relationship between floral integration and the breakdown of tristyly  Given our expectation that the increment in reciprocal placement between sexual organs of short- and long-styled plants may reduce the extent of integration of the whole flower, we first searched for a correlation between the magnitude of floral integration and the extent of incompatibility modification by means of regression analysis. The magnitude of floral integration was estimated for each population and floral morph through the calculation of the variance among the eigenvalues of the correlation matrix of floral attributes (INT = Var[λi], i = 1, …8; Wagner, 1984). A high variance among eigenvalues results from phenotypes characterized by strong correlations among traits, and thus most of the phenotypic variation will be accounted for by the first PC (i.e. high levels of phenotypic integration). Because correlation matrices from different populations were estimated using different numbers of plants, the integration value for each population was corrected by subtracting the expected integration value [expected(INT) = (number of traits − 1)/number of plants; Wagner, 1984; Herrera et al., 2002]. Each short- and long-styled plant was considered as an independent observation in these calculations. In order to determine whether INT values were significantly different from zero, 95% confidence intervals were calculated by bootstrapping (S-PLUS 2000, Mathsoft, Inc., 1999). Previous analyses showed that INT estimates did not differ among floral morphs; thus, we only show results for population estimates.

Second, we explored whether the pattern of floral integration (characterized by the covariance structure of the floral matrix) had a significant relationship with the extent of incompatibility modification. Because only one phenotypic matrix can be calculated for each level of analysis (morph and extent of incompatibility modification), we used the procedures outlined by Roff (2002), relying on the jackknife to produce a set of pseudovalues of the covariances between floral traits. These values were used in a multivariate ANCOVA (MANCOVA) in which morph, extent of incompatibility modification (covariate) and the interaction term (morph × extent of incompatibility modification) were included as independent variables. A morph effect would indicate that at least some of the covariances of the floral matrices differ between long and short plants. A significant effect of the extent of incompatibility modification would indicate that changes in the pattern of covariation among floral traits were related to the tristyly–distyly transition. Finally, a significant interaction term (morph × extent of incompatibility modification) would show that the rate of change in the pattern of covariation among floral traits differed between floral morphs during the evolutionary transition from tristyly to distyly.

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Patterns of variation in flower morphology

MANOVA indicated that flower morphology differed significantly among morphs (Wilks’λ = 1.95, F8,476 = 116.02, < 0.0001) and populations (Wilks’λ = 0.08, F88,3130 = 16.43, < 0.0001), and that the interaction between morphs and populations was also significant (Wilks’λ = 0.40, F88,3130 = 5.21, < 0.0001). Univariate tests showed a significant effect of morph, population and the interaction term on all floral attributes ( 0.04), except for the morph term for three floral traits and the interaction term for five floral traits [see Fig. S1(Supporting Information) for a description of attribute variation among morphs and populations].

PCA indicated that the first two PCs accounted for 72.56% of the total variation in flower morphology (Table 1). Loadings of PC1 were all relatively high and positive (Table 1), indicating that this component represents the variation in flower size. Positive loadings on the second component (PC2) were related to the length of the short, mid and long reproductive whorls (so, mo, lo; Table 1). However, the lengths of the short (so) and mid (mo) sexual organs had higher loadings than those of the long sexual organs (lo). Thus, differences in the magnitude of loadings for each floral morph indicated that changes in short (so) and mid (mo) sexual organs were larger than those in long sexual organs (lo). Accordingly, PC2 was interpreted as the ‘relative position of sexual organs’. These results suggest that O. alpina flowers are composed of two relatively independent sets of correlation pleiades (sensuBerg, 1960): flower size and the relative position of the sexual organs.

Table 1.   Results of a principal components analysis (PCA) using the correlation matrix among eight morphological floral traits from 12 populations of Oxalis alpina from the Sonoran Sky Islands
 PC1PC2
  1. Loadings of individual traits on the first two principal components (PCs) are presented.

% of explained variance57.2615.29
Accumulated variance57.2672.56
Petal length0.43−0.17
Petal width0.36−0.13
Flower length0.37−0.07
Flower width0.42−0.20
Flower tube width0.39−0.24
Short sexual organs0.220.62
Mid sexual organs0.200.64
Long sexual organs0.330.18

ANCOVA performed on the scores from the first PC revealed that the evolutionary transition from tristyly to distyly had a significant negative effect on flower size (scores from PC1, R2 = 0.26, Table 2). No differences in flower size were detected between floral morphs, nor was the interaction term significant (morph × extent of incompatibility modification; Fig. 5a).

Table 2.   Results from analysis of covariance (ANCOVA) on the scores for principal component 1 (PC1) (flower size) and PC2 (relative position of sexual organs) derived from a principal components analysis (PCA) on short- and long-styled morphs of 10 populations of Oxalis alpina representing the evolution from tristyly to distyly
VariableSource of variationdfFP
  1. Error degree of freedom for both analyses was 16.

PC1Morph10.240.6278
Extent of incompatibility modification15.380.0339
Morph × extent of incompatibility modification10.080.7756
PC2Morph15.420.0332
Extent of incompatibility modification16.720.0196
Morph × extent of incompatibility modification13.470.0806
image

Figure 5.  Relationship between PC1 (flower size; a) and PC2 (relative position of sexual organs; b) of short- and long-styled floral morphs and the extent of incompatibility modification in 10 populations of Oxalis alpina. Filled circles and broken lines, short-styled floral morph; open circles and full lines, long-styled floral morph.

Download figure to PowerPoint

ANCOVA performed on the scores derived from PC2 indicated that 49% of the variance in the relative position of sexual organs was explained by differences between morphs and by the extent of incompatibility modification (Fig. 5b). The significant effect of morph (Table 2) indicates that sexual organs of short- and long-styled morphs are organized differently. In addition, the significance of the extent of the incompatibility modification term demonstrates that changes in the relative position of sexual organs of short- and long-styled morphs are related to the tristyly–distyly transition. However, the interaction term (morph × extent of incompatibility modification; Table 2) was not significant (slopes of lines similar; Fig. 5b). Overall, changes in flower size and in the relative position of sexual organs account for most of the variation in flower morphology, and indicate that such variation can be explained by differences between morphs and the extent of incompatibility modification.

Reciprocal morphological adjustments between sexual organs during the breakdown of tristyly

In accordance with our expectation, the within-morph component of reciprocity (RW) was positively related to the extent of incompatibility modification (F1,16 = 17.84, < 0.0006, R2 = 0.77). This result indicates that the two anther whorls of the short- and long-styled morphs converged to almost the same length during the evolution from tristyly to distyly (Fig. 6a). The analysis also detected a significant morph, but not interaction, effect, indicating that both floral morphs differed in the relative difference in length between anther whorls across the tristyly–distyly transition (F1,16 = 35.22, < 0.0001), and showed a similar pattern of change in reciprocity (RW) (F1,16 = 3.28, = 0.088; Fig. 6a). The between-morph component of reciprocity (RB) also showed a positive relationship with the extent of incompatibility modification, thus indicating that the difference between mid stamens (of short- and long-styled plants) and long and short styles decreased during the evolution from tristyly to distyly (F1,16 = 30.03, < 0.0001; R2 = 0.87; Fig. 6b). Again, a significant effect of floral morph was detected (F1,16 = 74.66, < 0.0001), but not for the interaction term (F1,16 = 3.43, P = 0.082; Fig. 6b). Regression analysis, evaluating the overall changes in the extent of reciprocity (RI) along the evolutionary gradient, showed a strong positive relationship (F1,8 = 23.74, = 0.0012; R2 = 0.74; Fig. 6c). Finally, a Shapiro Wilk’s test showed that RW, RB and RI were all normally distributed (W ≥ 0.94,  0.58). These results confirmed our hypothesis that, during the evolutionary shift from tristyly to distyly, both anther whorls of the short- and long-styled floral morphs converged to the length of the target style (long and short styles, respectively).

image

Figure 6.  Relationship between different estimations of reciprocity among sexual organs of long and short floral morphs and the extent of incompatibility modification in tristylous and distylous populations of Oxalis alpina. Filled circles, short-styled floral morph; open circles, long-styled floral morph. (a) Within-morph component of reciprocity (RW) (RWshort = 1 −  [mS − lS/L]; RWlong = 1 − [mL − sL/L]). (b) Between-morph component of reciprocity (RB) (RBshort = 1 − [mS − L/L]; RBlong = 1 − [mL − S/L]). (c) Reciprocity index (RI = [(RWshort) + (RWlong) + (RBshort) + (RBlong)]/4).

Download figure to PowerPoint

Relationship between floral integration and the breakdown of tristyly

Floral integration ranged from 0.99 (± 0.28 SE, hereafter) in the San Luis population to 2.82 (± 0.30) in San Jose. These values correspond to only 12.37–35.25% of the maximum level of floral integration that can be obtained with a matrix of eight attributes. Bootstrapping showed that all values were significantly different from zero (see Table S1).

Regression analyses, evaluating whether or not the magnitude of floral integration diminished during the transition from tristyly to distyly, failed to detect a significant relationship (F1,8 = 0.01, = 0.92). In contrast, MANCOVA performed on the pseudovalues produced by the jackknife MANOVA (Roff, 2002) detected significant effects of morph and incompatibility modification, indicating that the pattern of floral integration differs between floral morphs and that this pattern changes during the evolution of distyly from tristyly (Table 3). In summary, the tristyly–distyly transition involved modifications in the pattern, but not in the magnitude, of floral integration.

Table 3.   Results from a multivariate analysis of covariance (MANCOVA) to determine the effect of morph, extent of incompatibility modification (covariable) and their interaction on floral phenotypic matrices (the pattern of floral integration) of Oxalis alpina
Source of variationWilks λApprox. FdfP
  1. Replicates of the phenotypic matrices for each population were obtained using the jackknife approach proposed by Roff (2002; see section on Statistical analyses).

Morph0.263.1836, 433< 0.0001
Extent of incompatibility modification0.121.4436, 4330.0490
Morph × extent of incompatibility modification0.070.8636, 4330.6984

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

In this study, we have shown that the loss of the mid-styled morph in O. alpina is associated with modifications in the floral phenotypes that increase the morphological reciprocity between short- and long-styled plants. Such changes, in turn, seem to be moving the floral phenotypes towards a new distylous adaptive peak. Although this adjustment entails modifications in both the flower size and the relationships among sexual traits, no evidence of changes in the magnitude of floral integration was found. Our analyses further revealed that the pattern of floral integration (the structure of the phenotypic floral matrix) changed during the evolution of distyly from tristyly, probably reflecting modifications in the allometric, functional or pleiotropic relationships among floral traits.

Previous studies analysing the consequences of the breakdown of tristyly on floral morphology have reported changes in the length of the mid stamens (Mulcahy, 1964; Ornduff, 1964; Weller, 1979; Eckert & Barrett, 1994; Eckert & Mavraganis, 1996). Other important features, however, such as modifications in flower size (Barrett et al., 2004; Hodgins & Barrett, 2006, 2008), other sexual organs or the entire variance–covariance matrix have received much less attention (but see O’Neil & Schmitt, 1993). As predicted, the loss of the mid-styled morph in O. alpina resulted in modifications in the floral phenotype, increasing the reciprocity between formerly illegitimate sexual organs of the short- and long-styled floral morphs. This process involved both the within- and between-morph components of reciprocity and, as revealed by PCA, morphological adjustment entailed modifications in the length of both the stamens and styles, and in the flower size. Our results indicate that the selective pressures favouring a higher reciprocity between short- and long-styled plants affect the whole floral phenotype.

Even though the breakdown of tristyly involved changes in both the flower size and relative position of the sexual organs, the magnitude of floral integration was apparently independent of this evolutionary process. The average magnitude of floral integration among populations of O. alpina (20.37% ± 2.10) falls near the average observed for flowering plants (21.5% ± 1.72, Ordano et al., 2008). This result is unexpected because a high magnitude of floral integration was predicted on the basis of the selective value of reciprocal morphology between floral morphs in heterostylous species. The magnitude of floral integration varied greatly among populations, with a threefold difference between the population having the lowest and highest value. It is likely that the magnitude of floral integration in O. alpina has played a minor role during the evolution of a distylous floral phenotype.

Although other studies have tested the adaptive role of the magnitude of floral integration, the evidence of its putative adaptive value is controversial (Fornoni et al., 2009; Harder, 2009). For example, experimental modifications of the extent of floral integration in Lavandula latifolia had no significant effects on pollination success (Herrera, 2001). Ordano et al. (2008) estimated the intensity of selection acting on individual floral integration in three Rosaceae species, and showed that this attribute is selectively neutral. Studies based on the comparative method, or looking for differences in floral integration among populations with dissimilar pollinator faunas, have found no support for the adaptive value of floral integration (Herrera et al., 2002; Armbruster et al., 2004). Pérez-Barrales et al. (2007), but Pérez et al. (2007), in contrast, reported that phenotypic floral integration responded to pollinator-mediated selection in Narcissus papyraceus and Schizanthus species, respectively. Anderson & Busch (2006) found that outcrossing species of the family Brassicaceae had higher levels of floral integration than conspecific selfing species. Higher levels of phenotypic integration in floral relative to vegetative traits in Dalechampia scandens implicitly assume the importance of floral integration in pollination biology (Hansen et al., 2007). Thus, more empirical studies are necessary before we can definitively conclude that there is an adaptive role for the magnitude of floral integration (Fornoni et al., 2009).

In contrast with the magnitude of floral integration, we found a significant relationship between the pattern of floral integration and the extent of incompatibility modification, a result in agreement with other studies showing that the pattern of phenotypic integration is generally responsive to natural selection (Herrera et al., 2002; Ashman, 2003; Olsson, 2004). Our analyses also revealed that the pattern of covariation among floral traits changed during the breakdown of tristyly, and this was consistent with our adaptive hypothesis. Hence, it is likely that natural selection modified the strength of covariation among specific floral traits in order to reshape the flower of short- and long-styled plants according to the new selective regime imposed by the loss of the mid-styled morph. These adjustments resulted in an increment in the reciprocal morphology during the evolution of a derived distylous floral phenotype.

Our analyses assume that the floral morphology in populations of O. alpina has evolved independently. A preliminary phylogeographical study suggests that geographical aggregation of distylous populations of O. alpina in the northwestern Sky Islands (Fig. 3) resulted from several independent evolutionary events (J. Pérez-Alquicira, unpublished). Future studies should explore how the ancestry of these populations may have influenced the geographical variation in flower morphology in O. alpina.

In addition to the increase in reciprocity between short- and long-styled plants, the breakdown of tristyly involved a marked change in flower size. This was an unexpected result because flower size, at least in theory, evolves in response to the selective pressures exerted by the most common/efficient pollinators (Galen, 1989; Mitchell, 1994; Conner & Rush, 1996). The reduction in flower size in distylous populations suggests that the magnitude of the selective pressures favouring increased reciprocity may have negative consequences for the pollination of O. alpina. Given that the identity of the principal pollinators [two species of small bee (Heterosaurus; Andrenidae) contributing to 94.5–100% of the floral visits] does not change among the populations studied, the reduction in flower size and the increase in reciprocity cannot be explained by changes in pollinator species. The similarity in pollinator fauna of tristylous and distylous populations of O. alpina suggests that the fitness consequences of a reduction in flower size should be explored.

Our PCA indicates that more than half of the morphological floral variation is a result of differences in size, with an additional 15% accounted for by differences in the relative position of sexual organs. Although the shortest path to evolve a distylous floral phenotype in O. alpina would require modifications only in the relationships among sexual organs, it is the variation in flower size that corresponds to the axis of maximum variation. Because this axis is one of minimum resistance to morphological change, we suggest that the reduction in flower size is a byproduct of the selective pressures acting on the rearrangement of the sexual organs (Björklund, 1996; Merilä & Björklund, 2004). Results from our study support this interpretation because flower size was reduced by 0.78 SD (the difference between the means of the standardized scores (PC1) of tristylous and distylous populations), whereas the relative position of sexual organs (PC2) changed by only 0.40 and −0.33 SD (for short- and long-styled morphs, respectively), a twofold difference between these two components. These results suggest that, although natural selection is apparently driving the floral phenotype towards a distylous adaptive peak, the constraints imposed by the structure of the phenotypic matrix have forced the evolution of smaller flowers. This explanation relies on the assumption that the phenotypic variance–covariance matrix represents a fair estimate of the genetic floral matrix (Roff, 1997; Waitt & Levin, 1998; Steppan et al., 2002).

Taken together, results from this study indicate that the breakdown of tristyly in O. alpina is accompanied by major changes in the morphology of short- and long-styled plants. Most of the observed floral modifications were associated with flower size, suggesting that the evolutionary trajectory is driven chiefly by the axis of maximum phenotypic variation. Accordingly, the evolution of flower morphology in O. alpina seems to have been conditioned (but not prevented) by the structure of the phenotypic floral matrix. In spite of this restriction, reciprocal herkogamy between formerly illegitimate sexual organs of short- and long-styled plants increased during the evolution from tristyly to distyly. Thus, the available evidence suggests that natural selection operating on flower morphology is strong enough to overcome any potential constraint resulting from the correlations among traits.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

We thank Rubén Pérez for technical and field assistance, and Mariano Ordano for statistical advice. We are grateful to Ann Sakai, Diane Campbell, Luis Eguiarte, Sonia Sultan and three anonymous reviewers for helpful comments on the manuscript. P.S. acknowledges the academic support received from the Posgrado en Ciencias Biomédicas and the Instituto de Ecología of the Universidad Nacional Autónoma de México. She was supported by PhD fellowships from CONACYT and UNAM-DGEP, and by an investigation fellowship from CONACYT. This study was also supported by grants from UC-MEXUS, UNAM (DGAPA IN217803) and CONACYT (47858-Q).

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information
  • Agren J, Ericson L. 1996. Population structure and morph-specific differences in tristylous Lythrum salicaria. Evolution 50: 126139.
  • Alexandersson R, Johnson SD. 2002. Pollinator-mediated selection on flower-tube length in a hawkmoth-pollinated Gladiolus (Iridaceae). Proceedings of the Royal Society of London Series B 269: 631636.
  • Anderson IA, Busch JW. 2006. Relaxed pollinator-mediated selection weakens floral integration in self-compatible taxa of Leavenworthia (Brassicaceae). American Journal of Botany 93: 860867.
  • Armbruster WS, di Stilio VS, Flores TC, Velásquez Runk JL. 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: 3955.
  • Armbruster WS, Pélabon C, Hansen TF, Mulder CPH. 2004. Floral integration, modularity and accuracy: distinguishing complex adaptations from genetic constraints. In: PigliucciM, PrestonK, eds. Phenotypic integration: studying the ecology and evolution of complex phenotypes. New York, NY, USA: Oxford University Press, 2349.
  • Arnold SJ. 1992. Constraints on phenotypic evolution. American Naturalist 140: S85S107.
  • Ashman TL. 2003. Constraints on the evolution of males and sexual dimorphism: field estimates of genetic architecture of reproductive traits in three populations of gynodioecious Fragaria virginiana. Evolution 57: 20122025.
  • Badyaev AV. 2004. Integration and modularity in the evolution of sexual ornaments. An overlooked perspective. In: PigliucciM, PrestonK, eds. Phenotypic integration: studying the ecology and evolution of complex phenotypes. New York, NY, USA: Oxford University Press, 5079.
  • Barrett SCH, Glover DE. 1985. On the Darwinian hypothesis of the adaptive significance of tristyly. Evolution 39: 766774.
  • Barrett SCH, Harder LD, Cole WW. 2004. Correlated evolution of floral morphology and mating-type frequencies in a sexually polymorphic plant. Evolution 58: 964975.
  • Berg RL. 1960. The ecological significance of correlation Pleiades. Evolution 14: 171180.
  • Björklund M. 1996. The importance of evolutionary constraints in ecological time scales. Evolutionary Ecology 10: 423431.
  • Campbell DR. 2009. Using phenotypic manipulations to study multivariate selection of floral traits associations. Annals of Botany 103: 15571566.
  • Charlesworth D. 1979. The evolution and breakdown of tristyly. Evolution 33: 489498.
  • Chernoff B, Magwene PM. 1999. Morphological integration: forty years later. In: OlsonEC, MillerRL, eds. Morphological integration. Chicago, IL, USA: University of Chicago Press, 319353.
  • Conner J, Rush S. 1996. Effects of flower size and number on pollinator visitation to wild radish, Raphanus raphanistrum. Oecologia 105: 509516.
  • Conner J, Via S. 1993. Patterns of phenotypic and genetic correlations among morphological and life-history traits in wild radish, Raphanus raphanistrum. Evolution 47: 704711.
  • Conner J, Franks R, Stewart C. 2003. Expression of additive genetic variances and covariances for wild radish floral traits: comparison between field and greenhouse environments. Evolution 57: 487495.
  • Darwin C. 1877. The different forms of flowers on plants of the same species. London, UK: John Murray.
  • Denton MF. 1973. A monograph of Oxalis, section Ionoxalis (Oxalidaceae) in North America. Publications of the Museum. Michigan State University Biological Series 4: 455615.
  • Domínguez CA, Avila G, Vázquez-Santana S, Márquez J. 1997. Morph-biased male sterility in the tropical distylous shrub Erythroxylum havanense (Erythroxylaceae). American Journal of Botany 84: 626632.
  • Eckert CG, Barrett SCH. 1992. Stochastic loss of style morphs from populations of tristylous Lythrum salicaria and Decodon verticillatus (Lythraceae). Evolution 46: 10141029.
  • Eckert CG, Barrett SCH. 1994. Tristyly, self-compatibility and floral variation in Decodon verticillatus (Lythraceae). Biologial Journal of the Linnean Society 53: 130.
  • Eckert CG, Mavraganis K. 1996. Evolutionary consequences of extensive morph loss in tristylous Decodon verticillatus (Lythraceae): a shift from tristyly to distyly? American Journal of Botany 83: 10241032.
  • Fornoni J, Ordano M, Boege K, Domínguez CA. 2009. Phenotypic integration: between zero and how much is too much. New Phytologist 183: 248250.
  • Galen C. 1989. Measuring pollinator-mediated selection on morphometric floral traits: bumblebees and the alpine sky pilot, Polemonium viscosum. Evolution 43: 882890.
  • Ganders FR. 1979. The biology of heterostyly. New Zealand Journal of Botany 17: 607635.
  • Hansen TF, Pélabon C, Armbruster WS. 2007. Comparing variational properties of homologous floral and vegetative characters in Dalechampia scandens: testing the Berg Hypothesis. Evolutionary Biology 34: 8698.
  • Harder LD. 2009. Questions about floral (dis)integration. New Phytologist 183: 247248.
  • Herrera CM. 2001. Deconstructing a floral phenotype: do pollinators select for corolla integration in Lavandula latifolia? Journal of Evolutionary Biology 14: 574584.
  • Herrera CM, Cerdá X, García MB, Guitián J, Medrano M, Rey PJ, Sánchez-Lafuente AM. 2002. Floral integration, phenotypic covariance structure and pollinator variation in bumblebee-pollinated Helleborus foetidus. Journal of Evolutionary Biology 15: 108121.
  • Herrera CM, Castellanos MC, Medrano M. 2006. Geographic context of floral evolution: towards an improved research programme in floral diversification. In: HarderLD, BarrettSCH, eds. Ecology and evolution of flowers. Oxford, UK: Oxford University Press, 278294.
  • Hodgins KA, Barrett SCH. 2006. Mating patterns and demography in the tristylous daffodil Narcissus triandrus. Heredity 96: 262270.
  • Hodgins KA, Barrett SCH. 2008. Geographic variation in floral morphology and style-morph ratios in a sexually polymorphic daffodil. American Journal of Botany 95: 185195.
  • Johnson SD. 2006. Pollinator-driven speciation in plants. In: HarderLD, BarrettSCH, eds. Ecology and evolution of flowers. Oxford, UK: Oxford University Press, 311325.
  • Kirkpatrick M, Lofsvold D. 1992. Measuring selection and constraint in the evolution of growth. Evolution 46: 954971.
  • Kleinbaum DG, Kupper LL, Muller KE. 1988. Applied regression analysis and other multivariate methods. Boston, MA, USA: PWS-Kent.
  • Lewis D. 1975. Heteromorphic incompatibility system under disruptive selection. Proceedings of the Royal Society of London Series B 188: 247256.
  • Lewis D, Rao AN. 1971. Evolution of dimorphism and population polymorphism in Pemphis acidula Forst. Proceedings of the Royal Society of London Series B 178: 7994.
  • Lloyd DG, Webb CJ. 1992. The selection of heterostyly. In: BarrettSCH, ed. Evolution and function of heterostyly. Berlin, Germany: Springer-Verlag, 179207.
  • Marcus LF. 1990. Traditional morphometrics. In: RohlfFJ, BooksteinFL, eds. Proceedings of the Michigan Morphometrics Workshop. Special Publication No. 2. Ann Arbor, MI, USA: Museum of Zoology, University of Michigan, 78122.
  • Marshall JT. 1957. Birds of pine–oak woodland in southern Arizona and adjacent Mexico. Pacific Coast Avifauna Number 32. Berkeley, CA, USA: Cooper Ornithological Society.
  • Mathsoft Inc. 1999. S-Plus 2000 guide to statistics. Seattle, WA, USA: Mathsoft.
  • McLaughlin S. 1995. An overview of the flora of the Sky Islands, southeastern Arizona: diversity, affinities, and insularity. In: DeBanoLF, GottfiedGC, HamreRH, EdminsterCB, FolliotPF, Ortega-RubioA, eds. Biodiversity and management of the Madrean Archipielago: the Sky Islands of the southwestern United States and northwestern Mexico. Fort Collins, CO, USA: USDA Forest Service, General Technical Report RM-GTR-264, 6070.
  • Merilä J, Björklund M. 2004. Phenotypic integration as a constraint and adaptation. In: PigliucciM, PrestonK, eds. Phenotypic integration: studying the ecology and evolution of complex phenotypes. New York, NY, USA: Oxford University Press, 107129.
  • Mitchell RJ. 1994. Effects of floral traits, pollinator visitation, and plant size on Ipomopsis aggregata fruit production. American Naturalist 143: 870889.
  • Mulcahy DL. 1964. The reproductive system of Oxalis priceae. American Journal of Botany 51: 10451050.
  • Nilsson LA. 1988. The evolution of flowers with deep corolla tubes. Nature 334: 147149.
  • O’Neil P, Schmitt J. 1993. Genetic constraints on the independent evolution of male and female reproductive characters in the tristylous plant Lythrum salicaria. Evolution 47: 14571471.
  • Olsson K. 2004. Population differentiation in Lythrum salicaria along a latitudinal gradient. PhD thesis, Umeå, Sweden: Umeå University.
  • Ordano M, Fornoni J, Boege K, Domínguez CA. 2008. The adaptive value of phenotypic floral integration. New Phytologist 179: 11831192.
  • Ornduff R. 1964. The breeding system of Oxalis suksdorfii. American Journal of Botany 51: 307314.
  • Ornduff R. 1972. The breakdown of trimorphic incompatibility in Oxalis section Corniculatae. Evolution 26: 5265.
  • Pérez F, Arroyo MTK, Medel R. 2007. Phylogenetic analysis of floral integration in Schizanthus (Solanaceae): does pollination truly integrate corolla traits? Journal of Evolutionary Biology 20: 17301738.
  • Pérez-Barrales R, Arroyo J, Armbruster WS. 2007. Differences in pollinator faunas may generate geographic differences in floral morphology and integration in Narcissus papyraceus (Amaryllidaceae). Oikos 116: 19041918.
  • Reyment RA, Jöreskog KG. 1993. Applied factor analysis in the natural sciences. Cambridge, UK: Cambridge University Press.
  • Roff D. 1997. Evolutionary quantitative genetics. New York, USA: Chapman and Hall.
  • Roff D. 2002. Comparing G matrices: A MANOVA approach. Evolution 56: 12861291.
  • SAS Institute Inc. 2005. JMP Version 5. Cary, NC, USA: SAS.
  • Schlichting CD, Pigliucci M. 1998. Phenotypic evolution: a reaction norm perspective. Sunderland, MA, USA: Sinauer.
  • Schluter D. 1996. Adaptive radiation along genetic lines of least resistance. Evolution 50: 17661774.
  • Smith RA, Rausher MD. 2008. Selection for character displacement is constrained by the genetic architecture of floral traits in the ivyleaf morning glory. Evolution 62: 28292841.
  • Stebbins GL. 1970. Adaptive radiation of reproductive characteristics in angiosperms. I: Pollination mechanisms. Annual Review of Ecology and Systematics 1: 307326.
  • Steppan SJ, Phillips PC, Houle D. 2002. Comparative quantitative genetics: evolution of the G matrix. Trends in Ecology and Evolution 17: 320327.
  • Wagner GP. 1984. On the eigenvalue distribution of genetic and phenotypic dispersion matrices: evidence for a nonrandom organization of quantitative character variation. Journal of Mathematical Biology 21: 7795.
  • Waitt DE, Levin DA. 1998. Genetic and phenotypic correlations in plants: a botanical test of Cheverud’s conjecture. Heredity 80: 310319.
  • Weller SG. 1979. Variation in heterostylous reproductive systems among populations of Oxalis alpina in southeastern Arizona. Systematic Botany 4: 5771.
  • Weller SG. 1981. Pollination biology of heteromorphic populations of Oxalis alpina (Rose) Knuth (Oxalidaceae) in south-eastern Arizona. Biological Journal of the Linnean Society 83: 189198.
  • Weller SG. 1986. Factors influencing frequency of the mid-styled morph in tristylous populations of Oxalis alpina. Evolution 40: 279289.
  • Weller SG. 1992. Evolutionary modifications to tristylous breeding systems. In: BarrettSCH, ed. Evolution and function of heterostyly. Berlin, Germany: Springer-Verlag, 247272.
  • Weller SG, Domínguez CA, Molina-Freaner FE, Fornoni J, LeBuhn G. 2007. The evolution of distyly from tristyly in populations of Oxalis alpina (Oxalidaceae) in the Sky Islands of the Sonoran Desert. American Journal of Botany 94: 972983.