•The diversity of plant breeding systems provides the opportunity to study a range of potential reproductive adaptations. Many mechanisms remain poorly understood, among them the evolution and maintenance of male flowers in andromonoecy. Here, we studied the role of morphologically male flowers (‘male morph’) in andromonoecious Passiflora incarnata.
•We measured morphological differences between hermaphroditic and male morph flowers in P. incarnata and explored the fruiting and siring ability of both flower types.
•Male morph flowers in P. incarnata were of similar size to hermaphroditic flowers, and there was little evidence of different resource allocation to the two flower types. Male morph flowers were less capable of producing fruit, even under ample pollen and resource conditions. By contrast, male morph flowers were more successful in siring seeds. On average, male morph flowers sired twice as many seeds as hermaphroditic flowers. This difference in male fitness was driven by higher pollen export from male morph flowers as a result of greater pollen production and less self-pollen deposition.
•The production of male morph flowers in P. incarnata appears to be a flexible adaptive mechanism to enhance male fitness, which might be especially beneficial when plants face temporary resource shortages for nurturing additional fruits.
Andromonoecy, both male (staminate) and hermaphroditic (perfect) flowers in the same individual, occurs in nearly 4000 plant species, comprising 1.7% of flowering plants (Yampolsky & Yampolsky, 1922). Despite its infrequent occurrence, an understanding of the fitness consequences of andromonoecy is important as it can shed light on the evolution of dioecy from hermaphroditism (Barrett, 2002; Charlesworth, 2006; Renner et al., 2007; Torices et al., 2011). In all monoecious species, flower gender is thought to be a plastic response to allocate resources to both male and female function under varying environmental conditions (Klinkhamer et al., 1997; Korpelainen, 1998). Typically, female function is prioritized when resources are ample and male function is expressed more often under stress, limited resource conditions or when female function is already fulfilled (e.g. Diggle, 1993; Steven et al., 1999; Bertin, 2007; Lazaro & Mendez, 2007). Studies of the effect of herbivory on gender expression support this idea with the finding of fewer male flowers when fruit production is reduced by herbivory, possibly to accommodate the need for increased female function (Krupnick & Weis, 1998; Wise & Hebert, 2010). In andromonoecious plants, several hypotheses have been proposed specifically to explain the evolution and maintenance of male flowers in a hermaphroditic system. The central question is: what is gained by possessing the plasticity to produce male flowers? The most likely benefit is through increased male or female function.
It might be rather counterintuitive to imagine a benefit to female function from the production of male flowers. However, there are two mechanisms, not mutually exclusive, to explain how this might take place. First, the resource reallocation hypothesis posits that male flowers are less costly than hermaphroditic flowers, and resources saved by producing male flowers can be reallocated to produce more fruits (Bertin, 1982; Spalik, 1991). This hypothesis has been supported in several taxa in which male flowers are smaller or lighter than hermaphroditic flowers (e.g. Emms, 1993; Cuevas & Polito, 2004; Diggle & Miller, 2004; Vallejo-Marin & Rausher, 2007b; Zhang & Tan, 2009). However, other studies have found that male flowers are similar in size or larger than hermaphroditic flowers (Huang, 2003; Narbona et al., 2008), suggesting that male flowers are not always cheaper to produce in andromonoecious systems. The second mechanism by which male flowers may increase female function is the pollinator attraction hypothesis. Here, the production of male flowers may increase pollinator visits without removing pollen from pollinators, and therefore enhance the fruit production for co-blooming hermaphroditic flowers (Podolsky, 1992). This hypothesis assumes that the floral display is a mixture of male and hermaphroditic flowers, and thus is not applicable to plants that usually produce one flower at a time. Overall, the ability of male flowers to enhance female fitness is somewhat equivocal. Indeed, two well-designed selection analyses on the same species showed conflicting results, with one finding negative selection on male flower proportion via seed production (Elle, 1999) and the other positive selection on male flower number (Vallejo-Marin & Rausher, 2007a).
The production of male flowers in andromonoecious breeding systems might also increase an individual’s male fitness, termed the ‘pollen donation hypothesis’ (Bertin, 1982; Elle & Meagher, 2000). Male flowers may be better than hermaphroditic flowers at donating pollen through various means. First, male flowers may produce more pollen than hermaphroditic flowers. Studies in aquatic Sagittaria have found that male flowers produce four times as much pollen as hermaphroditic flowers (Huang, 2003). However, more frequently, male and hermaphroditic flowers have similar pollen production (e.g. Cuevas & Polito, 2004; Sunnichan et al., 2004; Vallejo-Marin & Rausher, 2007b; Narbona et al., 2008; Zhang & Tan, 2009). Second, male flowers may produce pollen of higher quality. Pollen quality has been shown to be greater in male flowers, in hermaphroditic flowers, or equal in both types (e.g. Emms, 1993; Cuevas & Polito, 2004; Vallejo-Marin & Rausher, 2007b). However, from limited studies that have tested seed siring success, male flowers seem to have equal or higher siring ability than hermaphroditic flowers (Cuevas & Polito, 2004; Sunnichan et al., 2004; Zhang & Tan, 2009). Lastly, male flowers may experience less sexual interference and therefore export more pollen than hermaphroditic flowers. Direct tests of this hypothesis are uncommon. One study has found that male flowers serve primarily as pollen donors and hermaphroditic flowers as pollen recipients (Quesada-Aguilar et al., 2008). By contrast, male flowers were less effective than hermaphroditic flowers in disseminating pollen grains in a hummingbird-pollinated shrub (Podolsky, 1993). However, previous studies have not examined directly the amount of sexual interference in both types of flower. To explore the difference between male and hermaphroditic flowers in pollen donation, an experiment needs to evaluate simultaneously the degree of sexual interference and the amount of pollen export.
Taken together, male flowers may be more successful than hermaphroditic flowers in male function in andromonoecious plants through higher pollen quantity, pollen quality or pollination success via reduced sexual interference. It is largely unknown whether the three mechanisms occur concurrently and what is the relative importance of each in enhancing pollen donation. In addition, very few studies have carried experiments to the fruit stage and compared the number of seeds sired by male and hermaphroditic flowers (but see Elle & Meagher, 2000; Vallejo-Marin & Rausher, 2007b). Therefore, our understanding is limited with regard to how these different mechanisms may affect male fitness in andromonoecious species.
In Passiflora incarnata, the andromonoecious breeding system is characterized by ‘male’ flowers with a different style orientation (‘male morph’) from hermaphroditic flowers (May & Spears, 1988; Supporting Information Fig. S1). Style orientation affects pollen receipt, and therefore fruit production, under natural conditions (Dai & Galloway, 2011). Style orientation also potentially influences the degree of self-pollen deposition and outcross opportunity, namely sexual interference and pollination success, in P. incarnata. Self-pollen deposition is of particular interest in this system because P. incarnata is self-incompatible and thus self-pollen would not result in any seed production. Unlike male flowers in other species that usually have undeveloped or small female organs (ovary and style), male morph flowers in P. incarnata appear to have full-sized female organs. This raises the question of whether the male morph flowers are ‘real’ male flowers with female sterility. Previous work found 4.4% fruit set in male morph flowers and 29% in hermaphroditic flowers in one hand pollination experiment, but very similar fruit set for the two types of flower in another (80% male morph and 81% hermaphrodite; May & Spears, 1988). Therefore, it is unclear whether male morph flowers in P. incarnata are functionally male because of a lack of pollen receipt in nature, but capable of producing fruit if given sufficient pollen and resources. Potential female function in male morph flowers is of interest, as it suggests that the two types of flower do not differ physiologically, and therefore plants may be able to respond to external and internal cues and rapidly switch between male morph and hermaphroditic flowers.
In this study, we investigated the relative cost and reproductive success of male morph flowers in andromonoecious P. incarnata. In particular, we tested whether male morph flowers are less costly than hermaphroditic flowers, and whether male morph flowers are capable of producing fruits. We then focused on male function and compared pollen quality and siring ability between the two flower morphs. To elucidate the siring success of male and hermaphroditic morph flowers, we used floral arrays to quantify the degree of self-pollen deposition (sexual interference), pollen export and siring success of male and hermaphroditic morph flowers. Finally, we explored the seasonal pattern of male morph flower production.
Materials and Methods
Passiflora incarnata L. (Passifloraceae) is a perennial, self-incompatible vine that grows throughout the southeastern USA (May & Spears, 1988). It blooms from July to September in central Virginia, and its primary pollinator is the common carpenter bee (Xylocopa virginica). Flowers open fairly synchronously between 11:00 and 12:00 h, and remain functional for only one afternoon. The flower is composed of a petal platform (corona) on which pollinators land, five downward-facing anthers that form a plane parallel to the petal platform, and three styles from a superior ovary above the anthers (Fig. S1). Carpenter bees are active before flowers open and visits are frequent until dusk. When a carpenter bee lands on the horizontal platform of petals, it pushes its head and thorax in between the anther plane and the petal platform. It then moves in a circle, foraging for nectar in the center of the flower and passively accumulating pollen on the dorsal part of the thorax. The fruit of P. incarnata is a large (up to 7 cm in length), edible, unilocular berry with seeds that are enclosed by fleshy arils. Usually only one flower per plant is open per day. Hence, pollinator attraction is not an advantage for having male flowers in this system.
Passiflora incarnata is functionally andromonoecious throughout its range with both male morph and hermaphroditic flowers on the same plant (May & Spears, 1988; Dai, 2011). Male morph flowers possess all floral structures and look very similar to hermaphroditic flowers, except for the position of their styles. For male morph flowers, the styles are posed upwards for the life of the flower (Figs S1, S2). In this position, it is almost impossible for the stigmas to receive pollen via pollinator visitation. For hermaphroditic flowers, although the styles are initially upwards when the petals open, the styles deflex to a lower position over the duration of 1–2 h, resulting in a higher possibility of contact with the pollen-covered dorsal thorax of the carpenter bees (Dai & Galloway, 2011). Pollination occurs when the pollinator’s back (with pollen) contacts the stigma; there is limited autogamy. Male morph flowers can be produced early, in the middle or at the end of the flowering season. In a previous study, male morph flowers were found to have very low fruit set under natural conditions (0.95%; May & Spears, 1988). However, the results are mixed as to whether male morph flowers can produce fruit if given sufficient pollen, and therefore the breeding system is termed ‘functional andromonoecy’ (May & Spears, 1988).
Trait measurements The experiments were conducted in a natural population of at least 200 P. incarnata genotypes located at the Shadwell Preserve of the Jefferson Monticello Foundation in Shadwell, VA, USA in the summers of 2007, 2008 and 2009. Unless otherwise noted, experiments were conducted during the flowering peak (end of July to the middle of August). Passiflora incarnata plants have rhizomatous growth, and branches emerging from the ground in a clump are generally from one genotype (McGuire, 1999). Experiments and observations were performed at the level of individual flowers and experimental replicates were chosen from clearly different groups of branches. The total sample size for each study is given followed by the number of days over which the sample was taken.
To estimate the morphological and physiological differences between male and hermaphroditic morph flowers, we measured style deflexion, flower size, style length, nectar production and dry weight on randomly selected flowers of the two types. We chose 18 hermaphroditic flowers and 12 male morph flowers to measure style deflexion at c. 15:00 h. Style deflexion was measured as the vertical distance between stigmas and the anther plane. If the stigma is above the anther plane, as is typical of male morph flowers, the value of the style deflexion is positive; if it is below the anther plane, as found in most hermaphroditic flowers, it is negative (see Dai & Galloway, 2011 for an illustration). We also measured the floral size (radius of the flower) and style length from 36 hermaphroditic flowers and 35 male morph flowers (3 d). Nectar production was measured using microcapillary tubes in the morning immediately after flower opening and in the evening immediately before dusk. There were 38 hermaphroditic flowers and 37 male morph flowers included in the morning nectar measurement (3 d). The ovaries from the same flowers were collected and the ovule number was counted. For the evening nectar measurement, 14 hermaphroditic and 25 male morph flowers were chosen in 1 d. The dry weight was determined for 38 hermaphroditic and 38 male morph flowers. These flowers were picked in the field (5 d) and oven dried (50°C) for 2 d before weighing.
To estimate the total pollen production, all five stamens were collected from 30 hermaphroditic and 30 male morph flowers immediately before flower opening (4 d). All anther sacs were dried overnight at 50°C and stored at room temperature before counting. To count pollen, stamens from each flower were rehydrated in 30 ml of 70% ethanol and loosened by sonication. The samples were counted using an HIAC Model 9703 liquid particle counting system (Pacific Scientific, Hach Company, Loveland, CO, USA). To calculate the total pollen production, the pollen counts of nine subsamples of 1 ml each were averaged and multiplied by 30 (the total volume).
To investigate pollinator preference between the two flower types, pollinator visitation was observed for 16 half-hour intervals in the study population (4 d), representing pollinator activity from 11:00 to 18:30 h (roughly the life span of the flower). Before each observation interval, a number of hermaphroditic flowers and male morph flowers were tagged (range 15–39). The total number of pollinator visits to each flower type in every observation session was tallied and divided by the total number of flowers of each type. Therefore, there were 16 pairs of visitation data (‘visits per flower’) for hermaphroditic and male morph flowers.
To compare pollen quality between hermaphroditic and male morph flowers, pollen viability and pollen germination were measured using lactophenol–aniline blue and Brewbaker–Kwack medium with 30% (w/w) sucrose concentration, respectively (Kearns & Inouye, 1993). Newly opened flowers were collected at c. 11:00 h in the field and immediately brought back to the laboratory using coolers. To test pollen viability, we took three pollen samples from each flower and immersed each sample in a drop of lactophenol–aniline blue solution on microscopic slides. Dissecting pins were used to sample with the goal of 200–500 pollen grains in each sample. We first counted the total number of pollen grains in each sample and then counted the number of viable pollen (those that stained dark blue) after 40 min. The percentage of viable pollen was then calculated and averaged across the three samples for each flower. We sampled 40 hermaphroditic and 32 male morph flowers (3 d) for pollen viability. To test pollen germination and growth, 30 plants were selected, and one hermaphroditic and one male morph flower were collected from each plant on different days. The order of collection from each plant was random. Three samples of 200–500 pollen grains from each flower were taken and immersed in a drop of Brewbaker–Kwack medium. We scored the total number of pollen grains in each sample before incubation. All sample slides were then incubated in a covered Petri dish with water-saturated filter paper at 35°C (similar to natural temperature). After 40 min of incubation, the number of germinated pollen grains was counted and the three longest pollen tubes in each sample were measured. The germination percentage and pollen tube length were averaged across samples for each flower.
Fruiting potential and natural fruit set for hermaphroditic and male morph flowers We conducted a controlled pollination experiment to test whether hand pollination and the removal of previous fruits would affect the fruiting probability of hermaphroditic and male morph flowers in the same natural population of P. incarnata in July 2009. The experiment took 6 d during the flowering peak. On each day, the same number of hermaphroditic and male morph flowers were haphazardly chosen and divided into pairs of each flower type of similar rank (one for the first flower on the plant, and so on). One flower in the pair remained untouched, whereas the other was hand pollinated using a pool of pollen from flowers that were at least 20 m away, and all developing fruits on the stem were removed. The manipulation was the same for hermaphroditic and male morph flowers; each stem was used only once. In total, there were 46 pairs of male morph flowers and 45 pairs of hermaphroditic flowers. Flowers were followed until fruit maturation and fruit set.
Siring potential and natural siring success for hermaphroditic and male morph flowers To estimate the siring ability of the two flower types, we used pollen from hermaphroditic flowers and pollen from male morph flowers to pollinate hermaphroditic flowers, and compared the fruits sired by the two types. The experiment was conducted in August (2 d) with 34 flowers sired by hermaphroditic and 34 by male morph flowers each day (68 in each treatment). Before anthesis, a number of flowers were bagged using bridal-veil nylon to exclude pollinator visits. Pollen was then collected from c. 40 hermaphroditic and 40 male morph flowers that were at least 20 m away from the bagged flowers (a fresh pool of pollen was collected each day). Hand pollination of bagged hermaphroditic flowers was conducted using either pollen type. Pollinated flowers remained bagged until they wilted at night. All flowers were then followed through fruit development. Fruits were protected from frugivory by aluminum-screen bags and harvested after maturation. The number of seeds per fruit was counted and the seeds were air dried. Five random seeds per fruit were then weighed (to 0.01 mg).
To investigate the natural siring success for hermaphroditic and male morph flowers, we employed floral arrays with single pollen donors in the summer of 2010. We established a common garden at Blandy Experimental Farm (Boyce, VA, USA) using 87 genotypes from 21 populations of P. incarnata. Carpenter bees naturally occur at Blandy Experimental Farm but P. incarnata is not present. On alternating days, we chose either a plant with a male morph flower or a hermaphroditic flower as the focal individual. Focal plants were not reused to ensure independence. Four or six additional plants with hermaphroditic flowers were selected as the surrounding individuals for each array. We avoided selecting surrounding individuals that were from the same population as the focal individual to ensure compatibility. Flowers on surrounding plants were emasculated; emasculation does not influence pollinator visitation (Dai & Galloway, 2011). Therefore, the focal flower was the only pollen donor for the floral array and focal flowers were not able to produce fruit because P. incarnata is self-incompatible. All the fruits produced by the surrounding individuals were sired by the single focal flower. The floral arrays were placed outside at c. 14:00 h and all other plants in the common garden were covered to prevent pollinator visits for the whole day. Floral arrays were exposed to pollinators for the entire afternoon and all stigmas from flowers on the focal and surrounding plants were collected at c. 19:30 h (previous experiments indicated that cutting the stigmas after 19:00 h had no effect on fruit development). The style deflexion of each flower was also measured on collection. We checked the covered plants to ensure that there were no other sources of pollen; three of 29 arrays were found to have potential pollen contamination and therefore were dropped. To verify that bees do not carry pollen from day to day, several carpenter bees were caught when they first landed on a flower in the morning, and observation with a hand lens found no pollen on their backs. In total, 13 floral arrays with male morph focal flowers (62 surrounding flowers) and 13 with hermaphroditic focal flowers (64 surrounding flowers) were conducted. All surrounding flowers were followed to fruit maturation. The number of seeds per fruit was counted and the seeds were air dried; the weight of five random seeds per fruit was determined.
The difference in self-pollen deposition and pollen export between male morph and hermaphroditic pollen donors was evaluated using the stigmas from focal and surrounding flowers. Specifically, all pollen deposition on the focal flower was from the flower itself, because the focal flower was the only pollen donor in each array. All the pollen deposited on the surrounding flowers was pollen that was exported from the focal flower, and represents siring potential. As a result of the small number of focal flowers and the relatively large number of surrounding flowers, we counted pollen on all stigmas from focal flowers; however, we counted pollen on only one stigma from each surrounding flower (the stigma whose style deflexion was closest to the within-flower mean). To count pollen deposition, each stigma was immersed in a drop of 8 mol l−1 NaOH on microscope slides overnight and mounted with Permount (Fisher Scientific, Fair Lawn, NJ, USA). Extra NaOH was removed with filter paper before mounting. Under an Olympus (IX-70) inverted epifluorescence microscope, pollen grains glowed bright orange, contrasting with pale stigmatic tissues (yellow filter cube with Ex510/20 nm, 525DCLP, and Em590/35 nm). Images of each stigma were taken under 20× magnification using an Orca-2 Hamamatsu CCD camera and stitched together by Photoshop (Adobe Systems Incorporated, 2002). The number of pollen grains per stigma was counted using ImageJ (Rasband, 2009).
Seasonal pattern To determine the seasonal pattern of flower production, the types of flower produced from 213 plants in the same common garden were recorded in the summer of 2009. Starting from 15 July, the total number of hermaphroditic and male morph flowers in the garden was counted each day and all flowers were followed until fruit maturation, with each flower noted as either setting or failing to set fruit. Observations ceased on 5 September when almost all plants had finished flowering. The male morph flower percentage and the percentage of hermaphroditic flowers that set fruit were then calculated for each day. In the common garden, 155 plants produced both hermaphroditic and male morph flowers. For each flower, the flowering rank and number of days from the plant’s first flower (‘flowering days’) were noted and averaged for each flower type per plant.
All analyses were performed using SAS (SAS Institute, 2005). For all trait measurements conducted over > 1 d, ‘day’ was included as a blocking factor.
Analysis of variance was used to compare morphology, physiology, pollinator service, pollen quality and siring ability between hermaphroditic and male morph flowers (PROC GLM or PROC MIXED for continuous variables; PROC CATMOD for fruit set). For pollinator visits, pollen germination traits, and flowering rank and flowering days, the experiments were designed for pair-wise comparisons; therefore, we employed a paired t-test (PROC TTEST) to control for time and/or genotype effects. To meet analysis assumptions, transformations were performed for several traits. Morning nectar was square-root transformed; night nectar, flowering rank and flowering days were log transformed; pollen germination percentage was arcsine square-root transformed.
To test the pollination and siring success, self-pollen deposition, pollen exported and fruit production in the floral array experiment were compared between male morph and hermaphroditic focal flowers. Analysis of variance was employed for all comparisons, with focal gender as a fixed effect and ‘day’ as a covariate (PROC GLM for pollen deposition, seed number per fruit and average seed weight; PROC GLIMMIX for fruiting probability with a binomial distribution and for total seeds sired with a Poisson distribution). For self-pollen deposition, style deflexion was initially included in the analysis as a covariate. However, it was not a significant source of variation and was therefore dropped. For pollen exported, both style deflexion of the recipient flower and the interaction between style deflexion and focal gender were included. For fruit-level traits, both style deflexion and pollen deposition were included as covariates; interactions with the covariates were not significant and were therefore dropped from the models. Pollen deposition was square-root transformed to meet analysis assumptions.
The seasonal pattern of flower and fruit production was analyzed using weighted regression (PROC REG). Male morph flower percentage and fruit percentage were regressed on day in the season. Linear and quadratic models were tested individually. Total flower and total hermaphroditic flower number for each day served as the ‘weight’ for male morph percentage and fruit percentage, respectively.
Trait comparison of male and hermaphroditic flowers
Overall, male morph flowers do not seem to be less costly than hermaphroditic flowers in P. incarnata. However, male morph flowers showed a modest shift from female function to male function with a slightly shorter style (3% difference), lighter dry weight (2% difference) and 13% more pollen than hermaphroditic flowers (Table 1). However, pollen quality (pollen viability and pollen germination, Table 2) and ovule production (Table 1) did not differ between the two types of flower. The pollen germination percentage in our study was relatively low, possibly because of the short incubation duration (40 min). However, we expected the germination to be fast because pollen tubes were observed in styles only 1 h after pollination (C. Dai, pers. obs.). There was also no difference in attractive traits, such as flower size and nectar volume, between hermaphroditic and male morph flowers (Table 1). Probably because of this similarity, pollinators did not discriminate between the two types of flower (Table 1). In P. incarnata, male morph flowers are defined by having a higher style deflexion than hermaphroditic flowers; therefore, the difference in this trait between the two flower types was expected.
Table 1. Morphological and physiological traits, and pollinator service (mean ± SE), of male morph and hermaphroditic flowers of Passiflora incarnata
Male morph flower
*, P <0.05; **, P <0.01; ***, P <0.0001.
Floral size (cm)
2.67 ± 0.03
2.69 ± 0.03
F1,67 = 0.51
Style deflexion (mm)
− 1.7 ± 0.5
6.1 ± 0.4
F1,28 = 171.11***
Style length (cm)
1.00 ± 0.02
0.97 ± 0.02
F1,67 = 5.07*
Nectar morning (μl)
4.19 ± 0.41
4.62 ± 0.45
F1,70 = 0.09
Nectar night (μl)
0.98 ± 0.45
1.18 ± 0.33
F1,37 = 0.65
Dry weight (g)
1.54 ± 0.01
1.51 ± 0.01
F1,70 = 10.34**
62086 ± 3381
69960 ± 3861
F1,55 = 5.88*
90 ± 2.6
91 ± 2.5
F1,70 = 0.21
Pollinator visits/flower/0.5 h
2.4 ± 0.4
2.7 ± 0.4
tpaired = 1.07 (df = 15)
Table 2. Comparison of pollen quality and siring ability (mean ± SE) between male morph and hermaphroditic flowers of Passiflora incarnata (none of the comparisons was significant)
Male morph sire
For χ2 test, n = 136; for paired t-test, df = 29.
Pollen viability (%)
59.8 ± 2.3
61.6 ± 2.1
F1,68 = 0.34
Pollen germination (%)
7.0 ± 1.1
4.9 ± 1.0
tpaired = 1.55
Pollen tube length (0.1 mm)
3.5 ± 0.4
3.2 ± 0.4
tpaired = 0.66
Fruit set (%)
χ2 = 1.37
59.5 ± 6.7
62.1 ± 5.1
F1,22 = 0.05
Average seed weight (mg)
139.2 ± 4.6
137.5 ± 3.6
F1,22 = 0.08
Fruiting potential and natural fruit set for hermaphroditic and male morph flowers
Male morph flowers in P. incarnata were less capable of producing fruit, and this ability did not change even under ample pollination and resource conditions. Natural fruit set for hermaphroditic and male morph flowers was 88.9% and 4.3% (n =45 and 46), respectively. Neither male morph nor hermaphroditic flowers showed an increase in fruit set under the hand pollination and fruit removal treatment (male morph flower: 6.5% vs 4.3%, χ2 = 0.23, n =92, P =0.63; hermaphroditic flower: 88.9% vs 88.9%, χ2 = 0, n =90, P =1). In addition, the overall fruiting probability was significantly higher in hermaphroditic flowers than in male morph flowers (χ2 = 69.68, n =182, P <0.0001). Hence, male morph flowers, although morphologically comparable with hermaphroditic flowers with the exception of style deflexion, are physiologically altered to function mostly as male flowers.
Siring potential and natural siring success for male and hermaphroditic flowers
Male morph and hermaphroditic flowers did not differ in siring ability under controlled pollination. The pollen from the two types of flower had similar success in siring fruits and there was no difference in the number or average weight of seeds (Table 2). The fruit set in this study was low because the experiment was conducted later in the season.
Male morph flowers had higher natural siring success than hermaphroditic flowers. The flowers that were in floral arrays with focal male morph flowers had almost twice the probability of setting fruit as those with focal hermaphroditic flowers (Fig. 1a, Table 3). The total number of seeds sired by focal male morph flowers was also nearly double that of focal hermaphroditic flowers (Fig. 1a, Table 3). However, there was no difference at the fruit level. Fruits that were sired by a focal male morph or focal hermaphroditic flower had similar seed quantity and average seed weight (Fig. 1b, Table 3).
Table 3. Analysis of variance to evaluate the gender differences in self-pollen deposition, pollen exported and siring success in Passiflora incarnata
*, P <0.05; **, P <0.01; ***, P <0.0001.
Gender × deflexion
Total seeds sired
Seed number per fruit
Average seed weight
Male morph flowers also had less sexual interference and more pollen exported, resulting in greater siring potential. There was a more than 10-fold difference in self-pollen deposition between focal male morph and hermaphroditic flowers (Fig. 2, Table 3), indicating that male morph flowers suffered less from sexual interference. Meanwhile, male morph flowers exported more pollen than hermaphroditic flowers to surrounding flowers (Fig. 2, Table 3).
At the plant level, male morph flowers were typically produced after hermaphroditic flowers and later in the season. Male morph flowers were usually produced two nodes later than hermaphroditic flowers (flowering rank: hermaphroditic, 3.0 ± 0.2; male morph, 4.9 ± 0.2; tpaired = 7.27, df = 308, P <0.0001). On average, male morph flowers were 6 d later in the season than hermaphroditic flowers (flowering days: hermaphroditic, 10.1 ± 0.8; male morph, 16.2 ± 0.9; tpaired = 5.98, df = 308, P <0.0001). However, most plants alternated flower production with frequent switches between hermaphroditic and male morph flowers.
The percentage of male morph flowers in a population and the percentage of hermaphroditic flowers that set fruit showed similar seasonal patterns. For male morph percentage, a quadratic model fitted the data better than a linear model (F2,50 = 17.47, P <0.0001). The percentage of male morph flowers peaked shortly after the middle of the season (male morph percentage = − 0.00037Day2 + 0.068Day – 2.55, F2,50 = 55.96, P <0.0001, r2 = 0.69). For fruit set, although the data showed more scatter, a quadratic model was also a better fit (F2,50 = 8.54, P =0.0006) and the highest fruit set occurred shortly before the middle of the season (fruit percentage = − 0.00047Day2 + 0.064Day – 1.60, F2,50 = 8.22, P =0.0008, r2 = 0.25). The peak of fruit set occurred earlier in the season than the peak of male morph percentage (Fig. 3).
Resource cost and female function
In P. incarnata, male morph flowers did not differ significantly from hermaphroditic flowers. Although male morph flowers had a 3% smaller style and 2% less weight, the difference was minimal compared with that found in other species. For example, hermaphroditic flowers have a 19% greater dry weight than male flowers in olive (Cuevas & Polito, 2004). In other andromonoecious species, male flowers also invest much less in female organs than do hermaphroditic flowers, and the difference in style length is usually > 200–300% (Diggle & Miller, 2004; Vallejo-Marin & Rausher, 2007b; Zhang & Tan, 2009). It has been suggested that these morphological differences may be a result of flower rank and inflorescence structure, rather than gender (Diggle & Miller, 2004; Narbona et al., 2008). Although, on average, male morph flowers were produced after hermaphroditic flowers in P. incarnata, and floral traits might be affected by flower rank and seasonal changes, male morph flowers were usually produced between hermaphroditic flowers, especially in the middle of the season. Indeed, we tested the flowering rank in a group of sampled male morph and hermaphroditic flowers and did not find a difference. Hence, our random sampling in the peak flowering season should have removed the possible confounding of flowering rank and flower gender. Comparable trait values from flowers with similar ranks suggest that the production of male morph flowers in P. incarnata is probably not a means of reducing reproductive allocation to flowers, although it could be argued that reproductive allocation is reduced relative to the initiation and maturation, or abortion, of fruit. Furthermore, this system is unusual in having male flowers that only differ in style orientation, which might allow for late developmental decisions on the flower gender in response to variable environments.
As a result of morphological similarity between the flower types, we tested whether male morph flowers retained female function, and found that they rarely set fruit naturally or under increased pollen and resource conditions. This contradicted our expectations because, given that male morph flowers have fully developed female reproductive organs, we expected that they would also have the physiological potential to produce fruit. However, male morph flowers only showed c. 5% fruit set compared with nearly 90% in hermaphroditic flowers. Fruit set in this experiment was higher than in other larger scale surveys performed throughout the reproductive season (fruit set for male morph flowers c. 0.6%; fruit set for hermaphroditic flowers c. 55%), probably because the study was conducted during the early–middle part of the flowering season (see also May & Spears, 1988). Nevertheless, the dramatic difference in fruiting potential still holds, and mirrors a study on an andromonoecious Zigadenus species, in which the natural fruit set for male flowers was 0.3% and, for hermaphrodites, 36% (Emms, 1993). Yet, male flowers in Zigadenus have a reduced gynoecium and fewer ovules. By contrast, May & Spears (1988) cross-pollinated P. incarnata and found an increase in fruit set for male morph flowers, which was not seen in our study. It remains unclear how much fruiting potential male morph flowers can restore under different conditions. Nonetheless, all evidence supports the idea that male morph flowers in P. incarnata, although morphologically appearing to function as females, are less capable of setting fruit than hermaphroditic flowers. Therefore, we refer to these flowers as ‘male’ from this point on. This difference in reproductive potential may represent physiological constraints in P. incarnata, rather than the morphological changes more typical of male flowers in many andromonoecious species.
Male flowers produced c. 13% more pollen than hermaphroditic flowers in P. incarnata. It is interesting that the increase in male gametes was not coupled with a decrease in female allocation. Studies that have found higher pollen production in male flowers have typically also found that the male flowers have vestigial ovaries or no female organs at all (Anderson & Symon, 1989 for Solanum clarkiae; Huang, 2003; Narbona et al., 2005). In these studies, the increase in pollen number was higher (26%, 60% and 300%) than in P. incarnata. It might be the presence of the developed ovary in P. incarnata that constrains further resource allocation to pollen production in male flowers. In spite of the increase in pollen quantity, pollen quality was similar between the two types of flower. Our investigations of pollen quality and siring ability both showed that male flowers did not specialize in male quality.
Male flowers sired nearly twice as many seeds as hermaphroditic flowers under natural conditions. Doubling the seed sired offsets the loss of female function in male flowers. As a result, male flowers can achieve the same fitness as a hermaphroditic flower that sets fruit or an even higher fitness than a flower that fails to set fruit. This supports evolutionary stable strategies (ESS; Charlesworth, 1998; de Jong & Klinkhamer, 2005), which predict that a gender mutant that has double the fitness in one gender is able to invade under comparable resource allocation. Although ESS is typically applied to individual mutants rather than plastic floral gender expression, in P. incarnata, there is usually only one flower at a time, representing a discrete episode in a plant’s reproduction. If a plant lacks resources to produce a fruit, producing a male flower can be more successful in male fitness than the combined male and female fitness of an hermaphroditic flower.
Therefore, in P. incarnata, the production of male flowers when hermaphroditic flowers are present in the population appears to be an adaptation to a temporary resource shortage created by allocation into maturing fruits. A definitive test of this hypothesis would require an evaluation of fitness at the whole-plant level, in addition to the flower level presented here. In particular, male fitness is likely to be frequency dependent, in addition to requiring hermaphroditic flowers with the potential to set fruit. However, results of a phenotypic selection analysis that evaluated female fitness at the whole-plant level support this conclusion, finding no selection on male flower percentage in P. incarnata (C. Dai, unpublished). This result implies that the production of male flowers does not reduce fruit production, whereas the data presented here suggest that it benefits male fitness. Two studies have independently investigated male fitness (number of seeds sired) in andromonoecious Solanum carolinense. One echoes our results, finding that plants with higher male flower proportion, but not total flower number, had a higher male fitness (Elle & Meagher, 2000), whereas the other found similar male fitness between the two flower types (Vallejo-Marin & Rausher, 2007b). One caveat of our study, however, is that the experimental design did not allow pollen competition between two types of flower to contribute to siring success. This may have enhanced the contribution of differences in pollen export to siring success. A design using multiple pollen donors with genetic markers to determine siring success would clarify the ambiguity.
The difference in the number of seeds sired between male and hermaphroditic flowers was a result of a higher fruiting probability from the more successful pollen export in male flowers. This is supported by the fact that the number of seeds sired per fruit did not differ between the two flower types, indicating that the higher total number of seeds sired by male flowers was driven by the higher number of fruits. This, again, supports the finding that pollen quality from the two flower types does not play a role in enhancing seed-level traits. In addition, the difference in pollen exported from male and hermaphroditic focal flowers disappeared when comparing only flowers that set fruit (F1,37 = 1.86, P =0.18). In P. incarnata, the probability of fruit set increases with stigmatic pollen load (Dai & Galloway, 2011), suggesting that additional fruits in arrays sired by male flowers were a result of large pollen loads reaching more flowers.
Reduced sexual interference in male flowers played an important part in enhancing pollen export and, consequently, male fitness. Compared with hermaphroditic flowers, male flowers had higher style deflexion, which reduced the chance of contact between pollinators and stigmas, and therefore more pollen could be exported. Indeed, male flowers had less self-pollen deposition and exported more pollen to other plants. The relative quantities of pollen deposition might be a bit misleading, because self-pollen can only land on one flower (itself), whereas exported pollen is spread over all the recipient flowers in the array. However, Fig. 2 shows stigma-level pollen deposition. From simple calculations, we know that a focal male flower exported 37 times as many pollen grains to other flowers as to itself, whereas a focal hermaphroditic flower only exported 2.3 times as many. The difference in pollen production between the two flower types may partially drive the discrepancy in pollen export. However, focusing only on pollen exported, male flowers exported 32% more pollen than hermaphroditic flowers, even though they only produced 13% more pollen. This additional export is probably a result of reduced sexual interference, given that the two types of flower were visited by pollinators at the same rate. In P. incarnata, the morphological modification of limited style deflexion in male flowers is probably an adaptation to enhance male fitness by reducing sexual interference. By contrast, a study on andromonoecious Solanum, which reduced sexual interference by removing styles from hermaphroditic flowers, showed no increase in siring success (Vallejo-Marin & Rausher, 2007b). However, it is not known whether the rates of pollen exported were altered between the flower types.
Despite the fact that, on average, male flowers were produced after hermaphroditic flowers on a plant, at the population level, male flower percentage peaked in the middle of the reproductive season in P. incarnata. This pattern is unusual compared with other andromonoecious species, most of which have male flowers at distal positions on an inflorescence or a larger proportion of male flowers in later produced inflorescences (e.g. Anderson & Symon, 1989; Emms, 1993; Spalik & Woodell, 1994; Huang, 2003; Miller & Diggle, 2003). However, there are two species in the genus Euphorbia that have male flowers at the first and second inflorescence levels, resulting in a seasonal decline in male flower proportion (Narbona et al., 2005, 2008). In P. incarnata, the intermediate peak of male flower production is probably a result of the alternating, sequential flowering between male and hermaphroditic flowers. The male flower peak follows the fruiting probability peak with a relatively high level of overlap. The pattern of fruit production probably explains why P. incarnata produces male flowers in between hermaphroditic flowers instead of exclusively at the end. When the fruiting probability is low, later in the season, male fitness is also more uncertain. Therefore, male flowers produced in the middle of the season are likely to have higher male fitness than those at the end of the season. If ‘female investment is a sure thing and male investment is a gamble’ (de Jong & Klinkhamer, 2005), the perfect time to gamble is when the sure thing is most sure.
Our research has demonstrated that the production of male flowers in andromonoecious P. incarnata is an adaptive mechanism to increase an individual’s male fitness, probably with limited loss in female fitness. Male flowers were not cheaper to produce and they could barely produce fruit. However, male flowers were more successful than hermaphroditic flowers in siring offspring because of increased pollen export. This greater pollen export resulted from higher pollen production and less sexual interference in male flowers. Based on the seasonal flowering pattern, we propose that the production of male flowers in P. incarnata might be a response to real-time resource levels that predict failure in producing fruits in order to enhance fitness gain. Furthermore, in dimorphic male flowers of other andromonoecious taxa, the reduced sexual interference associated with smaller female reproductive organs might serve as an additional benefit to the more commonly argued shift in resource allocation. Reduced sexual interference is likely to be an overlooked general mechanism favoring andromonoecy.
We are indebted to C. Bennington for sharing her plants and Q. F. Wang, C. F. Yang and J. M. Chen for their help in collecting plants. We thank Q. F. Wang and J. Ren for field assistance, D. Roach and T. Roulston for sharing their fields, glasshouse space and equipment, the W. M. Keck Center (University of Virginia, Charlottesville, VA, USA) for providing imaging facilities, and an International Doctoral Fellowship to C.D. from the American Association of University Women (AAUW) Educational Foundation for financial support. Special thanks are due to C.D.’s dissertation committee for their help throughout this work, especially D. Carr for his advice in the summer. We thank Blandy Experimental Farm for granting C.D. a summer research fellowship with research and logistic support.