Changes in reciprocal herkogamy during the tristyly–distyly transition in Oxalis alpina increase efficiency in pollen transfer

Authors

  • F. BAENA-DíAZ,

    1. Departamento de Ecología Evolutiva, Instituto de Ecología, Universidad Nacional Autónoma de México, México Distrito Federal, México
    Search for more papers by this author
  • J. FORNONI,

    1. Departamento de Ecología Evolutiva, Instituto de Ecología, Universidad Nacional Autónoma de México, México Distrito Federal, México
    Search for more papers by this author
  • P. SOSENSKI,

    1. Departamento de Ecología Evolutiva, Instituto de Ecología, Universidad Nacional Autónoma de México, México Distrito Federal, México
    Search for more papers by this author
  • F. E. MOLINA-FREANER,

    1. Departamento de Ecología de la Biodiversidad, Instituto de Ecología, Unidad Hermosillo, Universidad Nacional Autónoma de México, Hermosillo, Sonora, México
    Search for more papers by this author
  • S. G. WELLER,

    1. Department of Ecology and Evolutionary Biology, University of California, Irvine, CA, USA
    Search for more papers by this author
  • R. PÉREZ-ISHIWARA,

    1. Departamento de Ecología Evolutiva, Instituto de Ecología, Universidad Nacional Autónoma de México, México Distrito Federal, México
    Search for more papers by this author
  • C. A. DOMÍNGUEZ

    1. Departamento de Ecología Evolutiva, Instituto de Ecología, Universidad Nacional Autónoma de México, México Distrito Federal, México
    Search for more papers by this author

César A. Domínguez, Departamento de Ecología Evolutiva, Instituto de Ecología, Universidad Nacional Autónoma de México, Apartado Postal 70-275, México Distrito Federal 04510, México.
Tel.: +52 55 5622 8995; fax: +52 55 56 16 19 76; e-mail: tejada@unam.mx

Abstract

Although the spatial separation of sexual organs within a flower (herkogamy) has been interpreted as a mechanism that promotes efficient pollen transfer, there have been few attempts to relate variation in herkogamy to probabilities of pollen flow. Here, we used a heterostylous species with variation in reciprocal herkogamy to test this hypothesis. We measured legitimate and illegitimate pollen flow with fluorescent dyes in four selected populations of Oxalis alpina corresponding to the extremes of a previously reported evolutionary gradient from tristyly to distyly. After the breakdown of tristyly, the observed increment in reciprocal herkogamy between the long and short morphs was associated with a 30% increase in the proportion of dye received from compatible illegitimate pollinations. In all populations, the most likely effective pollen vectors were two Heterosarus bee species. Our results support the adaptive value of reciprocal herkogamy in promoting efficient pollen transfer in heterostylous species.

Introduction

Flowers are phenotypically integrated structures that evolve to ensure successful pollen donation and receipt, often among individuals within populations (Stebbins, 1970; Bell, 1985; Harder & Barrett, 2006; Ordano et al., 2008; Armbruster et al., 2009). Because hermaphroditic species have both male and female sexual organs within the same flower, self-pollination and interference between sexual functions may reduce female fitness and the opportunities for outcrossing via pollen donation (Barrett, 2002). These two selective forces – selfing and interference – have been recognized as important causes for the evolution of self-incompatibility, and the spatial and temporal separation of sexual functions (herkogamy and dichogamy; Lloyd & Webb, 1986; Webb & Lloyd, 1986; Barrett et al., 2000). Although the adaptive role of these features has been demonstrated in many instances (Motten & Stone, 2000; Jersakova & Johnson, 2007), determining their effects on male and female fitness remains a major challenge in floral evolutionary biology (but see Dai & Galloway, 2011).

Although herkogamy in self-compatible species can simultaneously reduce interference and selfing and promote efficient pollen transfer (Cesaro et al., 2004; Medrano et al., 2005), the functional value of herkogamy in self-incompatible species is more likely associated with the patterns of pollen transfer (male fitness). Heterostylous systems are exemplary in this sense because they usually combine self-incompatibility and reciprocal herkogamy (Ganders, 1979; Lloyd & Webb, 1992). Darwin (1877) hypothesized that heterostylous floral polymorphisms resulted from selective pressures favouring legitimate pollen transfer between anthers and stigmas located at the same level. In other words, reciprocal herkogamy can be interpreted as a mechanical device that simultaneously reduces selfing and sexual interference, while increasing the rate of precise transfer of compatible pollen among the floral morphs (Barrett, 2002). Although the Darwinian hypothesis of the adaptive significance of heterostyly has withstood years of scrutiny (for recent reviews see Barrett & Shore, 2008; Weller, 2009), nobody has explicitly related the variation in reciprocal herkogamy with patterns of pollen transfer to determine the link between floral morphology and successful pollination in heterostylous species (but see Ganders, 1975 and Barrett et al., 2004).

Tristylous species have three floral morphs differing in the positions of the anther and stigmas whorls. In each floral morph, stigmas occur at one level and the two anther whorls occur at the other two levels. This morphology is generally associated with a heteromorphic incompatibility system that only allows seed production after cross-pollination between anthers and stigmas located at the same level (Fig. 1). Moreover, the presence of a floral polymorphism associated with an incompatibility system leads to frequency-dependent fitness for each floral morph (Fisher, 1941). Consequently, when a floral morph disappears or experiences a significant deviation from isoplethic equilibrium (equal frequencies of each floral morph), as commonly seen for the mid-styled morph of tristylous species (Weller, 2009), the remaining or more abundant morphs will suffer a fitness reduction due to the loss of a source of incoming pollen as well as target stigmas (Weller et al., 2007). If reciprocal herkogamy increases precise pollen transfer among floral morphs, a reduction in frequency or the complete loss of a floral morph within a population would alter the selective value of the position of anthers and stigmas in the remaining floral morphs (Mulcahy, 1964; Lewis & Rao, 1971; Eckert & Mavraganis, 1996; Sosenski et al., 2010). Hence, tristylous species expressing variation in the frequency of floral morphs and in the extent of reciprocal herkogamy among populations (Weller et al., 2007; Sosenski et al., 2010; Ferrero et al., 2011) represent ideal study systems to evaluate the influence of reciprocal herkogamy on the probabilities of pollen transfer among floral morphs.

Figure 1.

 Representation of the classical tristylous system with long-styled, mid-styled and short-styled floral morphs, each with two anther levels (within the square). Thick arrows represent examples of ‘legitimate’ crosses (LEG) between anthers and stigmas of the same level. The dashed arrows represent examples of all kinds of illegitimate crosses. Illegitimate heteromorphic crosses (ILHE) are those between anther and stigmas of different levels of different floral morphs. Illegitimate homomorphic crosses (ILHO) occur between anthers and stigmas of flowers of different individuals of the same floral morph. Self-crosses occur between anthers and stigmas of the same flower.

Within the Sonoran Sky Islands region (south-western United States, north-western Mexico), populations of Oxalis alpina range from isoplethic tristylous populations with equal frequencies of short-, mid- and long-styled morphs, to tristylous populations with reduced frequencies of the mid-styled morph and to distylous populations composed only of the long- and short-styled morphs, thus encompassing the evolutionary transition from tristyly to distyly (Fig. 2a) (Weller, 1986; Weller et al., 2007; Pérez-Alquicira et al., 2010). This evolutionary gradient is characterized by two major sequential changes: (i) an initial modification in the incompatibility system allows the production of seeds derived from illegitimate crosses between long-styled and short-styled individuals (Weller et al., 2007; Kutaka et al., 2011); and (ii) a subsequent rearrangement in the position of sexual organs increases the extent of reciprocal herkogamy between long- and short-styled plants (Weller, 1979; Sosenski et al., 2010; Kutaka et al., 2011). Such rearrangements are mainly accomplished by adjusting the length of mid-level anthers to match the position of the stigmas of the complementary morph. These changes correspond to an increase in reciprocity between the long- and short-styled morphs (Sosenski et al., 2010). Hence, instead of the two well-differentiated anther levels typical of tristylous populations, the anther whorls of each floral morph in distylous populations, or in short- and long-styled morphs in tristylous populations with small numbers of mid-styled morphs, tend towards a single anther level (Fig. 2b). These morphological adjustments can be interpreted as an adaptation to increase male fitness through a reduction in pollen wastage during the evolutionary transition from tristyly to distyly (Fig. 2b). This explanation relies on the assumption that increasing the reciprocity between mid-level anthers and short and long stigmas augments the probability of pollen transfer between them. In this study, we tested this hypothesis using four previously characterized populations of O. alpina (Weller et al., 2007; Pérez-Alquicira et al., 2010; Sosenski et al., 2010). Two represent the ancestral tristylous extreme of the evolutionary gradient, whereas the remaining two are close to the derived distylous state (Fig. 2). In each population, we estimated the probabilities of the 18 different paths of pollen transfer (in tristylous) and eight paths (in distylous populations) among floral morphs using fluorescent dyes (selfing excluded). With this data set, we aimed to test the following: (i) the Darwinian expectation that pollen transfer between same-level anthers and stigmas (legitimate crosses) is more frequent than illegitimate pollen transfer, (ii) our prediction that the probability of pollen transfer from mid-level anthers to short and long stigmas increases from tristylous to distylous populations, and (iii) whether differences in the foraging behaviour and/or in pollinator fauna among populations are related with the observed patterns of pollen (dye) transfer.

Figure 2.

 (a) Scatterplot describing the relationship among sexual organ reciprocity of long- and short-styled plants, mid-styled morph frequency and degree of modification in the incompatibility system during the evolutionary transition from tristyly to distyly in Oxalis alpina within the Sky Island region. Values of the mid-styled morph frequency and incompatibility modifications were obtained from Weller et al. (2007) and those of morphological reciprocity from Sosenski et al. (2010). Circles represent O. alpina populations within the Sky Islands, and open circles indicate the four selected populations. Populations MAR and PUR are closer to the ancestral tristyly condition, whereas populations MOR and PC represent those that have already lost the mid-styled morph (i.e. PC) or are closer to the distyly condition (i.e. MOR). (b) Floral morphology of the long- and short-styled morph in ancestral tristylous and distylous populations. Arrows represent legitimate and illegitimate pollen transfer between these morphs. Expected probabilities of each cross are represented by the arrow thickness. Continuous arrows represent legitimate pollen transfer (LEG), between anthers and stigma at the same level (i.e. S × sL), and dashed arrows indicate illegitimate heteromorphic pollen transfer (ILHE), between the mid-level anthers of long and short floral morphs (× mS and S × mL). Illegitimate homomorphic (ILHO) pollen transfer, between two plants of the same morph, is not represented in this figure. Cross notation: the first capital letter represents the style morph of the female parent followed by the anther level (lower case) and style morph of the donor morph (capital letter; i.e. × mS represent a long-styled plant pollinated with pollen from the mid anther of a short-styled plant).

Materials and methods

Study species and populations

Oxalis alpina (Rose) Knuth (section Ionoxalis) is a conspicuous herbaceous species that inhabits the understory of the coniferous forests of the Sonoran Sky Islands. On the basis of previous studies, we selected four populations representing the two extremes of the evolutionary transition from tristyly to distyly. A gradual modification in the incompatibility system, decreasing frequency and eventually loss of the mid-styled morph, and an increase in the reciprocity between long- and short-styled plants characterize this evolutionary gradient (Weller, 1986; Weller et al., 2007; Sosenski et al., 2010; see Fig. 2). Two populations, Mariquita (MAR) and Purica (PUR), are close to the tristylous ancestral condition. Morse Canyon (MOR) is also a tristylous population, but the extent of incompatibility modification and the magnitude of reciprocal herkogamy do not differ from these features in distylous populations (Weller et al., 2007; Sosenski et al., 2010). Finally, the distylous population Pinery Canyon (PC: Fig. 2) was chosen to represent the opposite extreme of this evolutionary gradient (although not included in Weller et al. 2007, PC has been studied extensively elsewhere (Weller, 1981; Sosenski et al., 2010; Fig. 2). Field observations in these populations showed that the main pollinating visitors in O. alpina are two bee species belonging to the genus Heterosarus (Andrenidae; Weller, 1981).

Patterns of pollen transfer

Even though pollen size differs significantly among the anther whorls of O. alpina (Weller, 1979), their size distributions overlap, thus precluding a simple estimation of pollen transfer based on counting pollen on the stigmas. Instead, we used fluorescent dyes to measure the patterns of legitimate and illegitimate pollen flow in each population. Although dye transfer does not provide an accurate measure of the absolute probabilities of pollen flow, this method allows the estimation of the relative proportion of pollen transferred within a population, as well as the comparison of populations (O′Neil, 1992). This technique has been used repeatedly providing good estimates of pollen flow among plants (O′Neil, 1992; Castillo et al., 2002; Adler & Irwin, 2006). In this work, we assumed dye transfer is analogous to pollen transfer.

Proportions of the different sources of pollen transfer were estimated in the four selected populations during two consecutive years (2004–2005). Given that flowering occurs during a short period (< 2 months during July and August) and there are logistic difficulties associated with performing field experiments in isolated mountains (distance, lack of facilities and difficult access), we were unable to perform simultaneous experiments in all populations at the same time. Thus, MAR and PUR were surveyed in 2004, and MOR and PC in 2005. We stayed in each population around 10 days and chose 2–3 sunny days for observations because the activity of flower visitors ceases in cloudy weather (F. Baena-Díaz, personal observation). To estimate the proportion of pollen transfer between all possible combinations of anther levels (short, mid, long) and stigma positions (short, mid, long), we used six different colours of fluorescent dyes. Each colour was randomly assigned to one of six possible anther levels and floral morph combinations (short and mid anthers in long-styled plants, short and long anthers in mid-styled plants, and mid and long anthers in short-styled plants). Individual plants were used as donors for only one dye colour (donor flowers) to avoid between-anther contamination. In this way, we were able to determine both the floral morph and the anther level of the dye source, as well as the morph of the recipient plant. This procedure also permitted us to determine whether pollen transfer was legitimate (hereafter LEG) (transfer between anthers and stigmas located at the same level in plants with different floral morphs), illegitimate heteromorphic (hereafter ILHE) (transfer between anthers and stigmas located at different levels in different floral morphs) or illegitimate homomorphic (hereafter ILHO) (transfer between plants of the same floral morph). Self-crosses were not considered in this study.

For each population, we selected an area of approximately 2500 m2 in which plant density was relatively homogeneous (≈20 plants per m2). Thirty plants of each morph were randomly chosen as pollen (dye) donors, fifteen per anther level. We gently applied the appropriate dye on one anther whorl of one newly opened flower from each pollen donor plant. Dye was applied to the anthers with the blunt end of a toothpick before the peak of insect visitation. After 6 h of exposure to insect visitation (from 10:00 to 16:00 h), we collected all the flowers of the five physically separated individual plants surrounding each donor plant within an area between 1 and 3 m from the focal plant. Although O. alpina is able to grow clonally, preliminary analyses with microsatellite markers confirm the low extent of clonality (O. Tsyusko, personal communication, 2010). In addition, the frequency of the floral morphs surrounding each of the 30 donor plants (within patch scale) was equivalent to the whole population frequency, suggesting low levels of spatial autocorrelation at the scale sampled. Considering this, we assumed that the ILHO pollinations were not biased by spatial autocorrelation. Flowers were dissected, and the presence/absence and colour of fluorescent dyes on the five stigmas of each flower was determined using an UV lamp. A total of 4160 flowers were observed in the four populations. Sample sizes averaged 1026 (SE = 300) flowers per population and ranged from 352 in the distylous population (PC) to 1807 in the tristylous population (MAR). Presence/absence of dye particles was used to estimate patterns of pollen flow rather than number of dye particles because counting particles was not feasible in the field.

Pollinator visitation

To determine whether there were differences in the potential pollinator fauna and the intensity of flower visitation among floral morphs and populations, we performed censuses of floral visitors in the four populations during the summer of 2008. Censuses were conducted between 10:00 and 14:00 h in four different patches per population during two sunny days. The number of flowers per patch ranged between 27 and 49. In each patch, ten observation periods (of 10 min each) were performed every half hour, and the number of visits per flower and the identity of pollinators (genus to family) were scored. The two Heterosarus species could not be identified during the observation and were only identified by genus. All flowers from each patch were previously tagged to facilitate the gathering of data.

Data analyses

Patterns of pollen transfer

To characterize patterns of pollen flow, we conducted three series of analyses. First, we tested for differences in the overall proportion of flowers receiving any colour dye on the stigmas (relative proportion of pollen transfer probabilities) among floral morphs and populations. For this analysis, we pooled legitimate and illegitimate dye transfers into a single category (presence or absence of dye on the stigmas). Because the mid-morph is absent in the distylous population (PC) and we wanted to compare all the four populations, we used a nested nominal logistic model (CATMOD, SAS Institute, 1999) in which the presence/absence of fluorescent dye on stigmas was used as the response variable. Population and morph nested within population were included as explanatory factors. Second, to test the Darwinian expectations that legitimate pollen transfer should be more common than illegitimate pollen transfer, independent χ2-tests were performed for each population. Expected values were estimated considering a 2 : 1 illegitimate-legitimate ratio because there are twice as many illegitimate as legitimate crosses (12 : 6 and 4 : 2, in tristylous and distylous populations, respectively). Third, to describe whether the relative contribution of each cross type changed across populations, a nominal logistic analysis was performed on the observed frequencies of LEG, ILHE and ILHO. Because our main interest focused on crosses between long- and short-styled plants, we wanted to rule out the possibility that pollen contributed by the mid-morph accounted for the relative patterns of pollen flow between the other two morphs. To this end, we calculated the frequency of each cross type with and without the pollen contributed by mid-styled plants. Hence, the presence/absence of dye transfers belonging to the mid-styled morph was included as an explanatory variable as well as population and the interaction between population and the presence/absence of the mid-morph.

After determining that pollen contributed by the mid-morph did not affect the relative patterns of pollen flow among populations, we performed independent nominal logistic analyses for each cross type to determine whether morphological floral modifications increased the likelihood of ILHE crosses between long- and short-styled plants (the crosses that became more reciprocal) and we also explored the consequences for LEG and ILHO crosses.

Finally, to produce a thorough description of the patterns of pollen flow along the tristyly–distyly gradient, we compared the relative contribution of the different crosses included within each category. In the case of LEG, the comparison included the × lS and S × sL crosses; × mS and S × mL for ILHE; and × sL, × mL, × lS and S × mS for ILHO (See Fig. 2b for cross notation). Contingency analyses were run in R 2.10.1 (R Development Core Team, 2008), and contrasts within populations were obtained from model coefficients.

Pollinator visitation

General descriptions of O. alpina visitors were used to determine whether there is a relationship between the extent of movement of fluorescent dye particles and pollinator visitation among populations. We conducted a nominal logistic analysis in which we tested whether the probability of dye transfer depends on the number of visits per plant in a given population. Population was included as an ordinal variable (the ordinal variable described a hierarchy in the intensity of pollinator visitation, PC > PUR > MAR > MOR), and the per flower probability of fluorescent dye occurrence was used as the response variable. A significant positive effect would indicate that the population with the highest intensity of pollinator visitation also had the highest probability of dye transfer. We also tested whether Heterosarus was the most common visitor in the populations to confirm previous observations made in the Chiricahua populations (Weller, 1981). Independent χ2-tests were performed for each population comparing visits by Heterosarus vs. all the other species. To assess differences in the number of visits per plant among morphs and populations, we used generalized linear models with Poisson error distribution and log link function. This analysis was performed for all visitors and repeated for Heterosarus only (JMP, Version 7, SAS Institute Inc., Cary, NC, 1989–2007). Because the distylous population (PC) does not have mid-styled plants, we were unable to analyse all populations at once. Thus, we performed one analysis including the three morphs from tristylous populations, and one with all the four populations but excluding the mid-styled morph. The number of visits per plant was used as the response variable, and the effects of population, morph and patch nested within population were declared as explanatory variables. Finally, the number of flowers per plant was included as a covariate in both analyses.

Results

Patterns of pollen transfer

We found evidence of pollen (dye) transfer in 958 (23.02%) of the 4160 flowers examined in all four populations. The probability of pollen (dye) receipt per flower showed a marked variation among populations (χ23 = 253.39, < 0.001). The proportion of flowers with fluorescent dyes on the stigmas in each population were as follows: PC (0.52) > MAR (0.247) > PUR (0.19) > MOR (0.116).

Likewise, morphs within populations also differed in the number of flowers with fluorescent dyes (log-likelihood: χ27 = 40.69, < 0.001). Long-styled plants received more pollen (22.61% of the flowers with fluorescent dye) than the other two morphs (16.35 and 15.86%, for mid- and short-styled plants, respectively) in tristylous populations (log-likelihood: χ22 = 24.96, < 0.001). In contrast, there was no difference between morphs in the distylous population PC (log-likelihood: χ21 = 0.023, = 0.87). In accordance with the Darwinian hypothesis, the observed occurrence of legitimate pollen transfer was higher than expected, whereas the opposite was true for illegitimate crosses (MAR: χ21 = 63.83, < 0.001; PUR: χ21 = 40.93, < 0.001; MOR: χ21 = 33.93, < 0.001 and PC: χ21 = 10.74, < 0.001). The proportion of illegitimate crosses, however, was relatively high in all populations (ranging from 0.37 to 0.49 of total).

Significant differences occurred among populations in the proportion of LEG, ILHE and ILHO crosses (χ26 = 60.32, P < 0.001). As indicated by the lack of significance of the presence/absence of the mid-styled morph term (χ22 = 5.22, P = 0.073) and the population × presence/absence of the mid-styled morph term (χ26 = 7.06, = 0.31), the pollen contribution of the mid-styled morph did not affect the relative importance of each cross type (LEG, ILHE and ILHO) within populations. Accordingly, further analyses were performed only with the long- and short-styled morphs (see Supporting Information for a complete list of the probabilities of pollen transfer in all four populations).

Independent analyses for each cross type showed no differences among populations in the frequency of LEG crosses (χ23 = 3.7, = 0.29). The analysis of ILHE crosses, testing our prediction for the effects of modified morphology and incompatibility, showed significant differences among populations (χ23 = 17.92, P < 0.001) and also ILHO crosses (χ23 = 33.85, P < 0.001; Fig. 3). Following our expectations, these results showed an increase in the frequency of ILHE crosses in the tristylous population most closely resembling distyly as well as the distylous population, and these increases were associated with a proportional reduction in ILHO crosses (Fig. 3).

Figure 3.

 Proportional contributions of the three cross types (LEG, ILHE and ILHO) in each population from the total stigmas observed with fluorescent dyes. Letters compare each cross type among populations. Different letters indicate significant differences between populations.

Even though differences in the proportion of LEG pollen transfer among populations were nonsignificant, the relative contribution of each of the two legitimate crosses showed marked differences among populations (χ23 = 24.34, P < 0.001). Contrasts within populations revealed that although the × lS cross occurred more frequently than the × sL cross in tristylous populations (MAR: χ21 = 22.7, < 0.05; PUR: χ21 = 4.84, < 0.05; MOR: χ21 = 10.12, < 0.05), no difference was detected in the distylous population (χ21 = 1.28, P > 0.05) (Fig. 4a ). The frequency of ILHE crosses (× mS and × mL) also showed marked differences among populations (χ23 = 30.35, < 0.0001). The frequency of the S × mL cross was higher than that of L × mS in MAR and PUR, whereas the opposite was true in MOR and no difference was detected in PC (Fig. 4b). Finally, the relative contribution of each of the four ILHO crosses did not differ among populations (χ29 = 12.24, = 0.2).

Figure 4.

 Proportional contribution of pollen flow along the tristyly–distyly gradient in (a). Legitimate crosses (LEG) (× lS and S × sL) and (b). Illegitimate heteromorphic (ILHE) crosses (× mL and × mS) within and among populations. Asterisks indicate significant differences within populations (< 0.05). Populations are ordered according to the degree of loss of the mid-styled morph and the extent of incompatibility modification.

Pollinator visitation

Pollinator observations indicated that excepting the MOR population, the most frequent visitors in all populations were two bee species of the genus Heterosarus (Andrenidae) (Table 1). Further analyses revealed a positive relationship between the number of pollinator visits and the probability of pollen transfer per flower in each population (χ23 = 234.6, < 0.001), indicating that fluorescent dyes reflected the variation in pollinator abundance among populations (Table 1). The same pattern was observed when only the Heterosarus visits were included within the analysis (Table 1). This result indicates that Heterosarus bees are the most likely pollen vectors in O. alpina because the relationship between the frequency of other visitors and dye transfer was negative.

Table 1.   Pollinator visitation and pollen transfer.
 Population
  1. MOR, Morse Canyon; PUR, Purica; MAR, Mariquita; PC, Pinery Canyon.

  2. *Populations where Heterosarus bees visit significantly more than all the other visitors (MAR: χ22 = 193, P < 0.00; PUR: χ22 = 485, P < 0.001 and PC: χ22 = 2294, P < 0.001). The opposite result was found for MOR population (χ22 = 10.3, P < 0.005). Note that when Heterosarus visits are excluded, the population with the most visits is MOR. MOR has the least pollen (dye transfer), indicating that Heterosarus bees are the more likely pollen vectors.

Proportion of flowers with dye on stigmasMARPURMORPC
0.2470.190.1160.52
VisitorsPercentage of the visits per population
 MAR*PUR*MORPC*
Heterosarus bees82.990.934.297.27
Other bees14.062.4800.9
Flies1.010.1433.30.19
Syrphids1.454.2832.381.29
Mosquitoes02.2100.27
Other0.58000.08
Total number of visits6867271052572

As indicated by the GLM analysis, the number of visits per plant did not differ among the floral morphs in tristylous populations for all visitors (χ22 = 0.03, P = 0.98) and for Heterosarus22 = 0.14, = 0.93) (Table 2). Also, no preferences for any floral morph were found in the analysis including the PC population and excluding mid-styled plants, for Heterosarus visits (morph: χ22 = 1.74, P = 0.18), nor for the entire set of floral visitors (morph: χ21 = 1.44, P = 0.22). This analysis, however, revealed a marked effect of populations and patches within populations (Table 2). The number of flowers also had a significant effect (Table 2).

Table 2.   Results of the GLM model with Poisson error distribution for pollinator visitation.
Model effectsModels including the three style morphs in the three tristylous populationsModels including only the long- and short-styled morphs
χ2d.f.Pχ2d.f.P
All visitors
 Pop236.820.001747.230.001
 Morph0.0320.981.4410.22
 Patch (Pop)128210.001207280.001
 Number of  flowers per plant21.810.0013810.001
Heterosarus
 Pop343.720.001848.630.001
 Morph0.1420.931.7410.18
 Patch (Pop)163210.001216280.001
 Number of flowers  per plant15.610.00133.710.001

Discussion

Previous studies of O. alpina have shown that the evolution of distyly is associated with a series of gradual changes in compatibility and floral morphology linked with loss of the mid-morph (Weller et al., 2007; Sosenski et al., 2010). Results from this study support the expectation that these changes also increase the proportion of compatible pollen transfer (Mulcahy, 1964; Lewis & Rao, 1971; Weller 1992; Eckert & Mavraganis, 1996; Sosenski et al., 2010). Here, we further demonstrate that both ILHE and homomorphic pollen transfers are significantly altered during the evolutionary transition towards distyly, whereas legitimate transfers were independent of these changes. Specifically, the gradual adjustment in the length of mid-stamens between long- and short-styled morphs increased the extent of reciprocity resulting in a higher proportion of pollen transfer (corresponding to the ILHE flow in a tristylous species with typical incompatibility reactions). In a population with modified incompatibility reactions, these crosses lead to successful seed production and hence should be categorized as legitimate. This evolutionary process appears to have ended in a distylous population (PC) in which the probabilities of pollen reception and compatibility of legitimate (LEG) and formerly illegitimate pollen (ILHE) are equivalent. Moreover, we observed a concomitant reduction in the probability of ILHO crosses. Our results for O. alpina provide strong support for the Darwinian hypothesis that reciprocal herkogamy promotes legitimate crosses among floral morphs of tristylous and distylous populations. Although nearly 50% of all dye transfers in the nonmodified tristylous populations are incompatible (all the ILHE and ILHO crosses), the evolution of distyly reduced this figure to almost 20% (only the ILHO crosses). Therefore, compatible pollinations increased from about 50% to almost 80% in the distylous population (see Fig. 3). Overall, findings from this study demonstrate that, after the incompatibility system modifications, morphological adjustments within the floral morphs increase the efficient transfer of pollen during the evolutionary transition to distyly.

Pollinator shifts among populations have been reported as an important cause of the breakdown of heterostyly (Pérez-Barrales & Arroyo, 2010). Our observations of the pollinator fauna in present populations of O. alpina support observations made 30 years ago by Weller (1981) that Heterosarus bees are the main pollinators in all populations. In addition, no differences in the visitation rate among floral morphs were observed in our study. Thus, neither the variation in flower morphology among populations (Sosenski et al., 2010) nor differences in the pattern of pollen transfer within populations (in this study) were related to differences in pollinator fauna or visitation rates. However, we cannot rule out that a pollinator shift in the past favoured the mutations responsible for the loss of the incompatibility reactions. If Heterosarus functioned as the principal pollinator during the breakdown of tristyly in O. alpina, our results indicate that even in the unmodified tristylous population (MAR) significant levels of illegitimate pollen transfer (50%) was detected. This occurred despite reciprocal herkogamy that is close to the maximum possible value, suggesting that Hererosarus bees do not restrain the spread of a mutation that altered the incompatibility reactions. It is likely that a combination of illegitimate pollen flow and variation in the extent of incompatibility resulted in cascading effects upon the stability of tristyly (Charlesworth, 1979). Once a mutation allows seed production from illegitimate crosses, the initial disadvantages associated with pollen wastage on incompatible stigmas became an advantage expressed in terms of increased male competitive ability (Barrett & Husband, 1990). This modification, in turn, results in a new selection regime on the floral phenotypes to increase the transfer success of these new sources of compatible pollen (Sosenski et al., 2010). Overall, these changes reduced the mid-style morph frequency and led to the evolution of distyly (Charlesworth, 1979; Eckert & Mavraganis, 1996; Weller et al., 2007).

The Darwinian hypothesis (1877) states that reciprocal herkogamy promotes the probability of legitimate pollen transfer in heterostylous species. This expectation has proved to be true in several heterostylous species where legitimate pollen transfer occurs at a higher frequency than that expected by chance (Swamy & Bahadur, 1984; O′Neil, 1992; Nishihiro et al., 2000; Lau & Bosque, 2003). In this study, we have demonstrated that this is also the case of O. alpina in both tristylous and distylous populations. Moreover, our results also revealed that the reduction in the extent of heteromorphic incompatibility and the increase in reciprocity between long- and short-styled plants (Weller et al., 2007; Sosenski et al., 2010) are associated with higher probabilities of pollen transfer for the formerly illegitimate S × mL and L × mS crosses. Interestingly, because increased reciprocity between the long- and short-styled floral morphs also resulted in a longer distance between anthers and stigmas within a flower, it is likely that the probability of ILHO crosses was reduced. Hence, in addition to enhancing the probability of legitimate pollen flow, the evolution of distyly in O. alpina also reduced the occurrence of illegitimate homomorphic pollen transfer (ILHO).

In general, our results for the proportion of stigmas with dye in each floral morph were consistent with classical studies of pollen flow in tristylous and distylous species (Darwin, 1877; Barrett & Glover, 1985; Cesaro et al., 2004, Weller, 1980). In Pontederia sagittata (Glover & Barrett, 1983) and Lythrum salicaria (O′Neil, 1992), more pollen was received by the long- than the short-styled morph. This pattern has been attributed to the exposed position of both the styles of the long morph and the anthers of the short morph (Glover & Barrett, 1983; O′Neil, 1992). In the tristylous MAR and PUR populations, the × mL cross was more frequent than the reciprocal × mS cross. Such asymmetry may results from differences in the distance between the two anther whorls of the long- and short-styled plants (Sosenski et al., 2010). The mid anthers of long-styled plants are close to their short anthers, whereas in short-styled individuals, the mid and long anthers are further apart. Hence, the mid anthers of long-styled individuals are more likely to donate pollen to short styles simply because their position is more similar to the position of the legitimate anthers, thus increasing the chances for ILHE × mL relative to × mS pollen flow. The reversal of this asymmetrical pattern in MOR might result from the changes in the reciprocity between the long- and short-styled flowers where the mid anther level of the short-styled flowers gets closer to the long anther level, that is the better donator whorl, explaining the excess of the × mS cross. However, this result is not conclusive because the reduced number of effective pollinator visits produced a very small number of pollinations, particularly in the × mL cross in this population. The absence of asymmetric pollen flow in the distylous population (PC) may be a likely result of a reduction in pollen wastage due to the increase in reciprocity (Sosenski et al., 2010).

Changes in the patterns of pollen transfer can have important consequences for the evolution of heterostyly and related ancillary sexual traits, including pollen size. Pollen from the long anthers whorl is larger than that from the mid and short whorls, respectively, and this variation is associated with the length of the target styles (Richards & Barrett, 1992). Consequently, changes in the incompatibility reactions and in the patterns of pollen transfer could generate novel competitive scenarios for the pollen landing in a given stigma. Provided the incompatibility system allows seed production after illegitimate crosses, larger pollen grains are expected to be better competitors (Mulcahy, 1979; Torres, 2000). Under this scenario, pollen from mid anthers is expected to outcompete legitimate pollen in short styles and have a size related disadvantage when competing in long styles. Hence, during the evolution of distyly, pollen from mid anthers will face two different competitive scenarios: one in which it will have a competitive disadvantage (× mS crosses), and one where it might have a competitive advantage (× mL crosses). Therefore, experiments are being conducted to examine whether the loss of the mid-styled morph and the evolution of distyly have produced novel selective scenarios favouring the adaptive and divergent adjustment of pollen traits.

Previous studies with O. alpina suggest that modifications in the incompatibility system disturb the isoplethic equilibrium of tristyly, thus initiating the evolutionary transition towards distyly (Weller et al., 2007; Kutaka et al., 2011). Such a transition involves modifications in floral morphology that increase reciprocity between the long- and short-styled morphs, and eventually, loss of the mid-styled morph. Here, we have further demonstrated that adjustments in the lengths of the mid-stamen whorls produce a more efficient transfer of compatible pollen in populations with modified incompatibility, which in turn has the potential to alter the evolution of pollen traits.

Acknowledgments

We thank Sergio Ramírez for access to La Púrica (Sonora), the U. S. Forest Service for permits, Gustavo Escobedo and the staff of the Guillermo Haro Observatory (INAOE) for housing facilities and access to Sierra la Mariquita and Jessica Pérez Alquicira, Adriana López Villalobos and Lluvia Flores Rentería for field assistance. F.B. thanks the Posgrado en Ciencias Biológicas, UNAM and CONACYT for support and grant. This research was supported by grants from the University of California Institute for Mexico and the United States (UC MEXUS), CONACYT (47858-Q), and the Universidad Nacional Autónoma de México (PAPIIT IN217803).

Ancillary