Pollen resistance to water in 80 angiosperm species: flower structures protect rain-susceptible pollen

Authors


Author for correspondence:
Shuang-Quan Huang
Tel: +86 27 68753553
Email: sqhuang@whu.edu.cn

Summary

  • • Flowers exhibit adaptive responses to biotic and abiotic factors. It remains unclear whether pollen susceptibility to rain damage plays a role in the evolution of floral form.
  • • We investigated flower performance in rain and compared pollen longevity in dry conditions, pure water and solutions with different sucrose concentrations in 80 flowering species from 46 families with diverse floral shapes and pollination modes.
  • • A pollen viability test showed that pollen longevity in all studied species was greatly reduced by wetting. We found that pollen of species with complete protection by flower structures was susceptible to water damage and a high proportion of resistant pollen occurred in unprotected species. Flowers whose structures expose pollen to rain may also reduce rain damage through temporal patterns of pollen presentation. This prediction was supported by our direct measurement of pollen presentation duration on rainy days.
  • • Our observations showed that variation in pollen performance in water was associated with differences in floral forms. Water-resistant pollen and extended pollen presentation duration were favored by selection via rain contact in species in which pollen was not protected from rain. These findings support the functional hypothesis that flower structures protect susceptible pollen from rain, demonstrating that rain acts as a force shaping floral form.

Introduction

Successful transfer of viable pollen from one flower to another is essential for sexual reproduction in plants. Pollen experiences a period of water loss within anthers before release in most angiosperms (Heslop-Harrison, 1987), and later a process of rehydration in a humid atmosphere restores the organization of the membranes and their osmotic properties (Shivanna & Heslop-Harrison, 1981). This dehydration and rehydration process means that ambient water conditions are crucial both within flowers and in transit between flowers (Nepi et al., 2001). Although some pollen grains are able to withstand contact with free water (Hagerup, 1950; Eisikowitch & Woodell, 1975; Huang et al., 2001), for others this is fatal (Percival, 1955; Corbet & Plumridge, 1985; Dafni, 1996; Burke, 2002; Huang et al., 2002; Sun et al., 2008). Water has been used as an emasculation tool for hybrid cotton (Gossypium hirsutum L.) production, as cotton pollen grains burst a few minutes after exposure to water (Burke, 2002). Taken together, previous observations indicate that water plays an important role in the fate of pollen.

Sprengel (1793) first recognized that rain may wash away pollen grains and dilute nectar. Since then, other observers have noted that flower structures may function to protect pollen from rain damage (see page 376 in Darwin, 1876). For example, some flowers were proposed to use shielding ‘umbrellas’ or surface patterns of hydrophobicity to protect pollen from wetting (Hagerup, 1950). Although deleterious effects of rain on pollen viability have been recognized, experimental studies of rain effects on floral traits are surprisingly few (Corbet, 1990; Galen, 2005; Sun et al., 2008). In an early study, flower morphology was compared with the resistance of pollen to moisture in two areas of Costa Rica, and it was found that the water resistance of pollen was higher in seven species from the wet area than in eight species from the dry area (Jones, 1967). One species, Muntingia calabura (Elaeocarpaceae), where no flower structures protect pollen, had more moisture-tolerant pollen in the wetter area. Similarly, a comparison of pollen performance in water between Primula vulgaris, which has erect flowers in which pollen regularly gets wet, and Primula elatior, which has secund (sideways) flowers in which pollen remains dry, showed that pollen resisted water longer and germinated later in P. vulgaris, indicating resistance against pollen germination in a rain-filled flower (Eisikowitch & Woodell, 1975). Further observations, based on a comparison of animal-pollinated species between two seasons with different rainfalls in the Mediterranean (Dafni, 1996; Aronne et al., 2006), suggested that flower structures were likely to protect pollen if the species had pollen that was susceptible to rain damage. Also, Aizen (2003) found that the number of plant genera with downward-facing flowers increased along a precipitation gradient across the Andes irrespective of pollination mode, suggesting that this type of flower orientation could constitute an adaptation against the deleterious effects of rain. Finally, plants may reduce pollen damage by rain via flower behaviors, such as closing the corolla (Bynum & Smith, 2001), changing the flower orientation (Huang et al., 2002) or gradual pollen presentation (Percival, 1955). Direct observation of the changes of flower structure and pollen viability in rain may shed light on how plants respond to rain damage.

Here we examined pollen performance in water and observed flower performance and pollen presentation patterns following rain in 80 flowering plant species to investigate whether the evolution of floral form involves adaptive strategies to protect pollen from rain. Rain may interfere with pollen presentation duration (Thomson & Thomson, 1992). To our knowledge, the effects of rain on plant reproductive strategies have not been incorporated into our understanding of floral longevity (Ashman & Schoen, 1994) or pollen presentation (Thomson, 2006). It has been proposed that flower structures protect pollen from rain in species with rain-susceptible pollen (Tadey & Aizen, 2001; Sun et al., 2008). We specifically investigated whether water-resistant pollen is associated with a lack of floral protective structures and whether prolongation of pollen presentation is found in rainy conditions. The results suggest rain as a selective factor acting on floral form and function.

Materials and Methods

Study species

We selected spring-flowering plants within or around the campus of Wuhan University (30°32′N, 114°21′E) and nearby Wuhan Botanical Garden, central China, to test pollen viability in different solutions. Average monthly precipitation is usually c. 180 mm in spring (March–May) in Wuhan. Some plants in the campus were transplanted but we selected those that set seeds naturally, that is, plants with fertile pollen. To identify an optimized solution for pollen germination in each species, in spring 2004 and 2005 we examined pollen performance for 42 and 60 flowering plants, respectively, in solutions with various concentrations of sucrose, namely 0, 2.5, 5, 7.5, 10, 15, 20 and 25% by mass, with or without four ionic components (0.001% HBO3, Ca(NO3)2, MgSO4 and KNO3) (Dafni et al., 2005) and measured pollen germination rates. From the tests performed in these two years, we knew that pollen performance varied in solutions with different sucrose concentrations and was also influenced by ionic components. Such records permitted us to examine a total of 80 spring-flowering plant species in 46 families with diverse floral shapes and sizes and pollination modes in 2006 (Supporting Information Table S1).

Pollen performance

To test the effect of rain on pollen performance, we used two solutions, distilled water and an ionic solution, to simulate rain. The effects of rain on pollen performance may derive either from the effects of water or from the effects of both water and its typical ionic components. We only used ionic components (0.001% HBO3, Ca(NO3)2, MgSO4 and KNO3) which are likely to influence pollen performance in water (Dafni et al., 2005). Pollen grains of most species were collected from at least 10 individuals, with the exception of six uncommon trees for which pollen was collected from at least five individuals (Ligustrum japonicum Thunb., Phellodendron chinense Schneid., Symplocos Sumuntia Buch., Elaeagnus pungens Thunb., Cinnamomum pauciflorum Nees. and Laurus nobilis L.). Pollen viability was tested using 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl-tetrazdium bromide (MTT) following the method of Rodriguez-Riano & Dafni (2000). Because pollen of some species burst soon after immersion in a dilute solution, the test solution for these species consisted of 1% MTT in 20% sucrose. Viable pollen stained deep purple after 10 min of immersion in this MTT solution, distinguishing it from dead pollen.

As a quantitative index of pollen resistance to water, we determined the time for which pollen of each species could remain viable in water. This was measured as a half-life to take account of the differences in pollen viability among individuals (Song et al., 2001; Castellanos et al., 2006). We defined the half-life as the time at which 50% of initially viable pollen grains immersed in distilled water or ionic solution would still stain. To test the initial pollen viability, we collected pollen from newly opened flowers from 08:00 to 09:00 h; pollen was combined for all individuals of a given species, and tested in the MTT solution within half an hour. After 10 min of immersion, we scored viable and nonviable pollen grains under a light microscope. For each species, 10 drops of pollen sample from the collected flowers were evaluated, and for each drop we counted at least 200 pollen grains in each of three random visual fields (the pollen counting mentioned below was all carried out following this procedure). To test the resistance of pollen to water, fresh pollen was immersed in three types of solution, namely distilled water, ionic solution and the optimized solution identified for each species (Dafni et al., 2005), and was stored in centrifuge tubes at room temperature (c. 20°C). To determine whether the ungerminated pollen was still viable, we placed pollen drops from these solutions on glass slides at intervals, and examined pollen viability using MTT. We repeated this process until we found the time (in hours) after which < 10% of the pollen immersed in distilled water or ionic solution could be stained (termed the ‘pollen viability time’). From this preliminary test, we knew how long pollen retained viability for each species. The pollen viability times for each species in distilled water and in ionic solution were very close (within 1 h), so we used the former as the pollen viability time for a species. We then divided this viability period into three equal stages, and defined the end of each stage as a check-point.

With this information we repeated the process, and at each check-point we placed the pollen drops from each solution onto glass slides, and stained the pollen with MTT. Stained and germinated pollen was counted, respectively. Germinated pollen was stained in the grain or at the tip of the pollen tube depending on the stage of germination. Germinated pollen was not counted as stained pollen because pollen germinating before landing on a stigma showed low resistance to water (see Eisikowitch & Woodell, 1975). Regarding the data for stained fresh pollen, we have four sets of stained pollen data for each species, differing in the time for which pollen was immersed in the solutions. To obtain the half-life, these four sets of stained pollen data for each species were fitted with a GaussAmp curve using Origin 6.0 Professional (OriginLab Corporation, Northampton, MA, USA). Through computation, we obtained the half-life inline image from parameter ω, which was obtained by nonlinear curve fitting. The GaussAmp equation for describing the relationship between the time for which pollen was immersed in solutions and the percentage of pollen that could still be stained is:

image(Eqn 1)

When inline image,

image(Eqn 2 )

(See Table 1 for definitions of variables in Eqns 1 and 2.) In seven species (Commelina communis, Calystegia sepium, Diospyros lotus, Cinnamomum pauciflorum, Laurus nobilis, Vicia faba and Nelumbo nucifera), pollen burst quickly (within 1 min) in distilled water and ionic solution, so we defined their half-lives as zero.

Table 1.  Definition of parameters used to calculate pollen half life
VariableParameterUnitValueDefinition
  1. MTT, 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl-tetrazdium bromide.

P 1Data from MTT testThe percentage of stained pollen at each checkpoint time t
t hData from MTT testThe time for which pollen was immersed in solutions or kept in dry conditions
 P10The percentage of stained pollen at time t (infinity)
 P01Data from MTT testThe percentage of stained original pollen at the start time t0
 t0h0The start time at which we immersed the pollen in the solution or placed it in dry conditions
 ωhUnfixedThe range of times

We used the pollen half-life as the index of pollen resistance of a species to water, and performed cluster analysis based on K-means to classify the species into two groups. Pearson correlation analysis showed that the correlation between pollen half-lives in distilled water and ionic solution was strong (R = 0.964, P < 0.0001). Thus, we only used pollen half-life in distilled water as the index to simplify this analysis.

To estimate pollen longevity without water contact, the half-life of pollen in dry conditions was determined. We kept fresh pollen from each species at room temperature and c. 50% relative humidity, and examined pollen viability with MTT over hours or days. Then, we estimated the half-life.

Flower morphology and flower performance in rain

To estimate the ability of a flower to protect pollen against rain, we recorded flower response to rain. Ten inflorescences of each species were randomly labeled before rain and observed on rainy days. We examined whether the pollen was wetted or washed away by rain, and whether floral form, such as flower orientation or movement (the closing of the corolla), changed with rain. These observations permitted us to classify species into three types: (1) species in which the anthers were fully exposed to rain, whose flowers were designated ‘no-protection’ flowers; (2) species in which the perianth partly protected the pollen, depending on the rain intensity and the angle of the flowers, whose flowers were designated ‘partial-protection’ flowers; and (3) species in which the perianth completely covered the fertile organs and protected the pollen from rain, whose flowers were designated ‘complete-protection’ flowers. To determine whether flower protection is associated with pollen water resistance, we performed G tests of the null hypothesis that rain resistance is independent of form of protection. To test whether the response of flower types differed across solutions, we performed a two-way ANOVA of pollen germination and pollen half-life among species using solutions and flower types as fixed factors. Data for pollen germination rate and pollen half-life were arcsine transformed and square-root transformed, respectively, to improve normality.

Pollen presentation

To estimate the relationship between the duration of pollen presentation and weather, we observed the process of anther dehiscence in the 80 species on rainy days and in fine weather. In our early observations, the presentation of small flowers with few anthers was always too brief to be measured, particularly in some wind-pollinated species. Furthermore, the inflorescence size varies considerably with species and the flowers in an inflorescence did not develop synchronically in some species. We therefore defined a pollen presentation unit, which comprises all the synchronically blooming flowers in an inflorescence, as a standard number of flowers in each species. We measured the duration of 10 pollen presentation units in rainy periods and 10 in fine weather for each species. From the time at which the first anther dehisced in a unit, we checked twice on the first day (after 4 and 8 h) and then once every 24 h, until all the anthers had dehisced. We estimate the plasticity of pollen presentation by calculating the extended times of pollen presentation duration as a quotient obtained by dividing the average pollen presentation duration in rain by the duration in fine weather. Some anthers were irreversibly damaged by long periods of rain, that is, their pollen was lost without anther dehiscence, making the timing of pollen presentation uncertain. We removed these data from our data set. We used one-way ANOVA to determine the significance of differences in the plasticity of pollen presentation among species with different flower types. A Pearson correlation analysis was performed to determine whether there was a relationship between plasticity and pollen half-life in distilled water.

Results

Pollen germination

The MTT test of fresh pollen indicated that initial pollen viability was high in all 80 species. The mean percentage of stained pollen ranged from 76.2 to 98.5%, with an average of 91.0 ± 0.4% (mean ± SE). Pollen germination tests showed that 14 species could not germinate in any of the solutions that we used and three species germinated with no significant difference among the three solutions. The remaining 63 species germinated better in the optimized solution than in distilled water, of which 13 showed equally good germination in the ionic solution and in the optimized solution, 15 showed equivalent germination in the ionic solution and in distilled water, and 35 showed significant differences in germination among the three solutions (Table S1). The pollen germination rate varied significantly with the nature of the solution (Table 2, Fig. 1). The mean rate of germination was highest in the optimized solution (75.1 ± 3.5%), while it was higher in the ionic solution (36.7 ± 4.0%) than in distilled water (16.0 ± 2.7%) for the 80 species (at the 0.05 level; Tukey test in two-way ANOVA; Table 2). However, the pollen germination rate was not different among species with different flower types (Table 2, Fig. 1), and the response to solutions was similar across flower types (Table 2, Fig. 1).

Table 2.  Two-way analyses of variance for pollen germination and pollen half-life among the three floral protection types in different solutions
Source of variationPollen germination ratePollen half-life
dfMSFP-valuedfMSFP-value
  1. Pollen germination rate data were arcsine transformed and pollen half-life data were square-root transformed.

Flower type  20.01 0.150.86  27.6521.64< 0.0001
Solution  25.6862.11< 0.0001  20.98 2.770.07
Flower type × solution  40.03 0.290.89  40.08 0.230.92
Error2170.09  2170.35  
Figure 1.

Pollen germination rate (mean ± SE) (a) and half-life of pollen viability (b) in three flower type groups (open bars, no protection; dotted bars, partial protection; hatched bars, complete protection) in different solutions. Pollen germination rate was significantly different among solutions and pollen half-life was significantly different among flower types (Tukey test in two-way ANOVA; see Table 2).

Half-life of pollen viability

The GaussAmp equation described the relationship between the time for which the pollen was immersed in the solution and the percentage of viable pollen. We observed that R2 ranged from 0.676 to 0.996 (0.956 ± 0.006) in distilled water and from 0.781 to 0.997 (0.949 ± 0.005) in the ionic solution, indicating that the GaussAmp equation can be used here. The half-lives of the 80 species ranged from zero to 14.1 ± 0.5 h in distilled water and from zero to 10.0 ± 0.4 h in the ionic solution (Table S1). Because of the very long time for which pollen of two Rhododendron species could resist water contact, the differences among the other species were obscured. Therefore, we conducted a 95% winsorization on the data for pollen half-life in distilled water. The cluster analysis classified half-lives ≥ 3.1 h together, forming a high-resistance group, and the remaining half-lives together, forming a low-resistance group (Fig. 1). In distilled water the mean half-life of high-resistance species was 5.8 ± 0.8 h. Compared with these 15 species, the mean half-life in the low-resistance group, which contained 65 species (65/80 = 81.3%), was much shorter, only 1.3 ± 0.1 h. Consequently, pollen from more than three-quarters of the species exhibited low resistance against water, indicating that pollen of most species is susceptible to water (Table S1).

The half-life of pollen in dry conditions ranged from 6 h to 23 d in 76 species. Although the half-life of pollen in dry conditions in 71 species significantly correlated with the half-life in water (R = 0.42, P < 0.001) and in the ionic solutions (R = 0.38, P = 0.001), water contact greatly shortened pollen longevity by 2.9 to 863 times. Pollen longevity in 94.4% of species (67 of 71) was shortened by more than 10 times (Table S1).

Flower morphology and pollen protection

Of the 80 species we observed, 44 species exposed pollen completely (no protection), 20 species protected pollen completely (complete protection), and the remaining 16 species partly protected pollen (partial protection). In the no-protection group the half-life of pollen ranged widely while in the complete-protection group the mean half-life was always less than 3.1 h (Fig. 1, Table S1). The mean half-life of pollen viability in the no-protection group (1.6 ± 0.06) was significantly higher than that in the partial-protection (1.2 ± 0.07) and complete-protection (1.0 ± 0.05) groups in all three types of solution (at the 0.05 level; Tukey test in two-way ANOVA; Table 2).

Species with complete-protection flowers all had low water resistance pollen; no species with high-resistance pollen were observed in this flower type (Fig. 2). We found that a higher proportion of species (13/44; 29.5%) in the no-protection group producing highly resistant pollen than species (2/16; 12.5%) in the partial-protection group (G = 7.1, df = 1, P < 0.01) (Fig. 2).

Figure 2.

Classification of the 80 species studied with high-resistance (hatched bars) or low-resistance (open bars) pollen among three flower types (no, partial, and complete protection).

Pollen presentation duration

The duration of pollen presentation in all species varied with the weather. Rain extended pollen presentation to 2.0 to 10.6 times the duration on sunny days among the 80 studied species. The plasticity of pollen presentation duration was highest in no-protection flowers (6.9 ± 0.2), and it was significantly higher in partial-protection flowers (5.0 ± 0.3) than in complete-protection flowers (3.5 ± 0.3) (F2,77 = 39.96, P < 0.0001). Pearson correlation analysis showed that pollen presentation plasticity and pollen half-life in distilled water were significantly correlated only in the no-protection group (Fig. 3). No significant correlation between pollen presentation plasticity and pollen half-life in distilled water was observed in either the partial-protection (r = −0.26, n = 16, P = 0.34) or the complete-protection groups (r = −0.39, n = 20, P = 0.09).

Figure 3.

Negative relationship between pollen presentation duration and half-life of pollen viability in distilled water in 44 species with no-protection flowers (r = −0.42, P = 0.005).

Discussion

Our results indicate that the pollen of most species was susceptible to damage by water. Less than one-quarter of the species had pollen with a relatively high resistance against water, implying that rain may directly reduce the fertility of unprotected pollen. Indeed, we found that pollen of species with complete protection by flower structures showed low resistance to water, supporting the functional hypothesis that flower structures protect pollen from rain in species with rain-susceptible pollen. Water-resistant pollen was more frequent in flower structures that did not protect pollen from rain (Fig. 2). The finding of a high proportion of resistant pollen in no-protection species suggests that selection by rain contact favors pollen resistance to water. An extreme condition of pollen resistance to water occurs in some aquatic plants with submerged flowers that release pollen under water which is then transferred by water currents (see Huang et al., 2001). We discuss below how terrestrial plants adapt to rainy weather and reduce the unfavorable effect of rain, particularly considering diverse pollen presentation strategies (Thomson & Thomson, 1992).

Many factors may affect the fate of pollen before successful pollination (Harder & Routley, 2006). Pollen may fail to be removed from the anthers, or may be consumed by animals, or lost during transfer, or may lose viability. Generally, c. 1–2% of pollen is successfully transferred to conspecific stigmas in animal-pollinated plants, suggesting a high proportional pollen loss. To date, we do not know how much pollen loss is attributable to rain in any plants that expose pollen in rain. In one case, when the bracts of Davidia involucrata were removed during anthesis, rain stripped away c. 80% of the pollen of the exposed capitula (Sun et al., 2008). Clearly, rain may aggravate the process of pollen loss. We found that pollen grains had high viability when they were released from the anthers and high germination rates in optimized solutions, but free water contact reduced pollen viability (Fig. 1). Pollen survival in rain is enhanced by flower structures or production of water-resistant pollen (Figs 1, 2).

It has been suggested that the highly plastic nature of pollen performance provides the potential for genotypes to respond differently to environmental variation (Delph et al., 1997). To reduce pollen loss caused by rain, plants with pollen-protecting flowers would be favored, given that pollen from many species is susceptible to water. We found that pollen half-life was consistently higher in species with no-protection flowers than in species with partial- or complete-protection flowers in three types of solution, indicating that prolongation of pollen longevity has evolved in species without structures protecting pollen (Fig. 1). Under these circumstances, rain selection would favor plants with high resistance to water and auxiliary devices that prevent pollen from being easily washed away. For example, pollen can be held in the anther by devices such as pollenkitt, tryphine or elastoviscin, viscin threads or sporopollenin filaments (Pacini et al., 1997). Only two of the 80 species studied employed these two approaches to avoid pollen damage by rain. Pollen from Rhododendron simsii and Rhododendron mucronulatum (Ericaceae) remained viable in water for relatively long times (Table S1). Furthermore, the poricidal anthers and the viscin strands of Rhododendron pollen reduce the probability of pollen being washed away by rain.

Pollen of exposed flowers can be easily dispersed by wind or removed by animals, but suffers the associated risk of pollen loss by rain damage. The capacity of pollen to respond to relative humidity differs among species and is usually associated with the intrinsic hydration state at the time of pollen dispersal (Nepi et al., 2001). Although animal-pollinated species are generally considered to have longer pollen viability because pollen must wait for pollinators, mature pollen exposed to air usually remains viable during anthesis in both wind- and animal-pollinated species (Pacini et al., 1997; Table S1). In our study, there was no significant difference among seven wind-pollinated and 73 animal-pollinated species in pollen half-life in either distilled water (F1,78 = 0.06, P = 0.80) or ionic solution (F1,78 = 0.002, P = 0.97) or in plasticity in pollen presentation duration (F1,78 = 2.19, P = 0.14). These results suggest that both wind- and animal-pollinated species have experienced rain selection and employed similar strategies to reduce rain effects.

Theory suggests that temporal staggering of pollen presentation would be selected because pollen from sequentially opening anthers or flowers in one plant can donate to many recipients through frequent pollinator visits (Harder & Thomson, 1989). By contrast, infrequent visits favor simultaneous presentation of all pollen (for example in orchids), to avoid losing the chance of pollen removal. Visitation rate and efficiency of pollen removal by pollinators provide a paradigm that aids our understanding of various pollen presentation patterns in animal-pollinated plants (Castellanos et al., 2006). However, a low visitation rate is sometimes coupled with gradual pollen presentation (Thomson & Thomson, 1992), suggesting that pollinator visitation alone could not fully explain the pattern of pollen presentation. We observed that the duration of pollen presentation in a flower or inflorescence varied greatly with weather conditions. Pollen presentation could range from 1 d to more than 1 wk in species with numerous anthers in one flower, such as six species of Rosaceae (Table S1). In all flower types anther dehiscence was delayed on rainy days, and staggered pollen presentation reduced pollen loss by rain. Nevertheless, pollen grains retained for too long in undehisced anthers because of humid weather may also lose their viability (Linskens et al., 1989). One may expect that species with pollen that has low resistance to water in no-protection flowers would reduce pollen exposure to rain, delaying dehiscence and extending pollen presentation. This prediction has been supported by our finding of a correlation between the half-life of pollen viability and the plasticity of pollen presentation duration (Fig. 3).

Our study provides insights into floral longevity, pollen presentation strategy and the evolution of flower structures. Unpredictable weather has been considered as an important factor limiting sexual reproduction in many flowering plants as a result of deficient pollen transfer (Corbet, 1990; Dafni, 1996). Pollen is enclosed in anthers and anthers in flowers, which may protect pollen from loss by rain damage to some extent. Yet pollen needs to be removed or presented to pollinators. In unfavorable environments, viable pollen may adopt a ‘sit-and-wait’ strategy, resulting in an extension of floral longevity which involves additional costs of maintaining reproductive structures (Ashman & Schoen, 1994). Thus, temporal staggering of pollen presentation reduces the chance of pollen removal by infrequent visitors but also reduces the risk of all the pollen being lost to rain.

We have demonstrated that flower structures can protect susceptible pollen from rain, indicating that rain could interact with other factors governing the evolution of floral traits. Studies have demonstrated floral evolution under selection through mutualistic pollinators and antagonistic herbivores as well as physical environments (Armbruster, 1996; Galen, 2005). Understanding plant reproductive strategies for protecting pollen from rain may provide a clue to the diversity of plant–pollinator interactions. In high-rainfall places or seasons, a high proportion of species were observed with nodding, pollen-concealed flowers or upward-facing flowers capable of temporary closure through perianth movements to avoid pollen loss (Dafni, 1996; Bynum & Smith, 2001; Huang et al., 2002; Hase et al., 2006). In bowl-shaped flowers exposed to rain, the petals are often separate or fused only at the base, so that raindrops flow quickly away between the petals, comprising a water drainage system. Sympetalous flowers are often pendulous and rarely erect. Erect sympetalous flowers usually have a narrow corolla. Morning glory flowers (Convolvulaceae) seem exceptional, with a wide open corolla, but they are generally ephemeral and secund. We observed that the corolla of Calystegia sepium closed on rainy days. Such floral changes may function to shelter pollen from rain damage before pollinators arrive. Extended pollen presentation duration in rain and the restoration of flower performance in fine weather involve reproductive costs that could otherwise have been invested in seed production. If we recognize that rain affects floral evolution, it is apparent that increased variability of precipitation may have potential effects on plant diversity and distribution in a given area.

Acknowledgements

We thank Ying-Zhuo Chen, Jing Du, Xiao-Xin Tang, Xuan Tang, Qing-Hua Wu, Fang Yang, and Dan Zhu for their help with data collection; Lawrence Harder for advice on data analysis; Sarah Corbet, Susanne Renner and Xiao-Lin Zhang for helpful discussions; and the editor Laura Galloway, Paul Wilson and two anonymous reviewers for providing useful comments on the manuscript. This work was supported by the National Science Foundation of China (Grants 30825005 and 30770135 to SQH).

Ancillary