•Staminate flowers of andromonoecious species are thought to be produced to increase reproductive success through enhancing male function or diverting resources from unneeded pistils to fruits. This does not explain why andromonoecy occurs within genera with monoecy, since staminate flowers of monoecious plants can also serve these functions.
•Here the male allocation of staminate and perfect flowers was measured in the annual herb Sagittaria guyanensis ssp. lappula , an andromonoecious species in a predominantly monoecious genus. Staminate flowers contained anthers that were larger and more numerous than those of perfect flowers, and their petals were also larger. This contrasts with most investigations where staminate flowers typically have equal or lower investment in male structures than perfect flowers.
•Seed set was not significantly different between bagged and open-pollinated flowers. Pollinator observation and a pollen-dyeing experiment indicated visits to the flowers rarely occurred.
•The presence of functional stamens in perfect flowers could be selected to allow reproductive assurance in case of inadequate pollination in andromonoecious species, rather than as a mechanism for optimal resource allocation.
These explanations do not explain why andromonoecy and monoecy exist within the same genus, since monoecy can provide these benefits as well. For such systems, one must focus less on why staminate flowers are produced and more on why perfect flowers rather than pistillate flowers are produced. Bertin (1982) discussed four reasons for retaining functional stamens within flowers: self-fertilization may be possible if pollen vectors are inadequate, stamens may act as pollinator attractants, pollen may be a reward for pollinators, and combining male and female functions within a flower makes better use of fixed costs (e.g. perianth and nectar). Charlesworth & Morgan (1991) had asked why andromonoecious plants produce staminate flowers, rather than simply maturing fruits from a few flowers. They deduced that hermaphroditism might be a ‘bet-hedging’ strategy in the face of unreliable pollination, but this argument has not been answered yet (Charlesworth & Morgan, 1991). Genera that contain both andromonoecious and monoecious species are suitable for investigating the fitness benefits of producing functional stamens in perfect flowers.
In a preliminary study of Sagittaria guyanensis, which is andromonoecious, we found that the pollen production of staminate flowers was higher than that of perfect flowers (Huang et al., 2000). To explore mechanisms for the maintenance of andromonoecy in this species, floral allocation to male function and what limits reproductive success for female function was further investigated. It was found that female function is partially limited by pollinator activity; hence, reproductive assurance may play a role in maintaining the production of perfect flowers.
Materials and Methods
Study species and populations
Sagittaria guyanensis H.B.K. ssp. lappula (D.Don) Bojin is an annual aquatic herb, distributed mainly in the region south of the Yangtze River in China ( Chen, 1989 ). It is a common weed in rice fields and can rarely be found in shallow pools and ditches. It flowers from June to October. One plant can potentially produce several racemes up to 10 cm long. Each raceme generally has between one and five perfect flowers at the base of the inflorescence and up to three staminate flowers at the top. Flowers are bowl-shaped with three white membranous petals ( Fig. 1 ) that sometimes have fuchsia spots on their bases. Some secretory cells were observed in the aborted pistils of staminate flowers, but floral nectar was not seen in open flowers ( Wang, 1997 ). The corolla width ranges from 10 to 20 mm. The pedicel of the staminate flower is thinner but longer than those of perfect flowers; and the yellow anthers of staminate flowers are distinctively larger than those of perfect flowers ( Huang et al., 2000 ). Pollen grains are spherical, inaperturate and echinate as in monoecious species of Sagittaria ( Wang, 1997 ). The gynoecium of perfect flowers consists of hundreds of tiny pistils densely crowded on a globose receptacle. Pistils of staminate flowers are aborted early in development ( Wang et al., 1999 ). Each pistil of perfect flowers has a single ovule with a pear-shaped stigma. Stigmas are receptive when the flowers open ( Huang et al., 2000 ). Flowers open in sequence from the bottom of the raceme to the top. A few individuals (1.7%) produce one perfect and one staminate flower in the same day. Anthesis of a single flower lasts 4–5 h. Staminate flowers usually opened at 10 : 30, about 0.5 h earlier than the perfect flowers, and they both closed at 14 : 30–15 : 00 in populations ( Huang et al., 2000 ).
Owing to the use of herbicides and other weeding techniques, S. guyanensis is a rare plant in natural habitats (Huang et al., 2000). We conducted pollination observations in Wuyishan, Fujian Province (27°40′ N, 117°30′ E) in September 1998 and 1999 and Dongxiang, Jiangxi Province (28°14′ N, 116°36′ E), China in September 1999. In Wuyishan, S. guyanensis was growing in ditches and rice fields but in Dongxiang it was almost always growing in rice fields. The 230 cultivated plants in this study were grown from seeds gathered in Dongxiang in late September 1998. The seeds were germinated in pots with loam soil in the Garden of Wuhan University. More than 50 seedlings each of S. potamogetifolia Merr., S. trifolia L. and S. pygmaea Miq. were transplanted to the Garden to provide comparative observations of pollinator visits. These three species are monoecious emergent plants, and commonly colonize rice fields in South China (Chen, 1989). All pots were positioned near a 1-ha rice field in the garden so that the cultivated plants might share a similar pollinator assemblage with the field populations.
To compare floral allocation of the two types of flowers, four floral parameters in the cultivated population were recorded. Petal width of 60 staminate and 90 perfect flowers was measured at noon when the folded petals spread to their full size, and the number of stamens in 50 staminate and 90 perfect flowers was counted. In the early morning flowers were randomly collected just as they were opening. The pollen from all anthers in a flower was suspended in 50 drops of water, and pollen production per flower based on scoring pollen grains contained in five drops under a light microscope was subsequently estimated. To estimate pollen size, the diameter of 20 pollen grains per flower was measured with an ocular micrometer on a microscope.
Five-day pollinator visits at each field population from 10 : 00 to 15 : 00 were recorded. At the same time, pollen movement was measured by dyeing pollen grains. This method was used successfully in S. trifolia to track pollen movement and pollinator discrimination against stained flowers was not observed (Huang et al., 1999). Pollen grains in newly dehiscing anthers were carefully stained. Pollen grains in staminate flowers were dyed red with 1% safranin solution and pollen grains in perfect flowers were dyed green using 1% methyl green solution. Five staminate and 10 perfect flowers were stained and the spacing between stained flowers was at least 2 m in each population. At about 14 : 00, when flowers began to close, all stained flowers and 10 nearby un-manipulated perfect flowers were collected. These flowers were observed soon thereafter with a stereomicroscope for dyed pollen, and pollen deposited on stigmas was estimated. In August 1999, the cultivated population was visited daily and flower visitors to S. guyanensis and to the three monoecious Sagittaria species were observed. This observation was taken over 80 h in the Garden of Wuhan University.
Fruit production and pollination treatments
Pollination is required for development of the aggregate fruit of S. guyanensis. Three types of achenes were recognized according to size (Huang et al., 2000). Undeveloped achenes were of the same size as pistils that did not receive pollen. Partly developed achenes with a slightly developed ovary presumably received pollen but were aborted. Fully developed achenes were larger than those partly developed. The developmental status of the achenes allowed discrimination between pollen or resource limitation of fruit development. Before anther dehiscence, 11 perfect flowers were emasculated, 14 were bagged to isolate visitors and 11 flowers were supplementally hand-pollinated using pollen donor from a different plant in the cultivated population. Two weeks later when fruit development had ceased, the achenes of each fruit were categorized into the three types (not pollinated, aborted, mature) for fruits collected from the cultivated and field populations. As control, 131, 92 and 67 aggregate fruits of open-pollinated flowers were collected from the cultivated population, Dongxiang Population and Wuyishan Population, respectively. A total of 47, 171 achenes in 327 aggregate fruits was examined.
One-way ANOVA analysis was used to compare floral allocation, the four measured floral parameters of staminate and perfect flowers, and to compare fruit set between open-pollinated flowers and three treatments of manipulated flowers followed by a Scheffe test (SAS, 1998).
Staminate and perfect flowers were significantly different in several aspects. The petals of staminate flowers were larger than those of perfect flowers (F1,148 = 4.62, P < 0.05; Fig. 2a). Staminate flowers produced more stamens (F1,138 = 38.46, P < 0.0001; Fig. 2c) and larger pollen grains (F1,598 = 83.89, P < 0.0001; Fig. 2b) than perfect flowers. The pollen production of staminate flowers was 4.1 times greater than that of perfect flowers (F1,44 = 133.04, P < 0.0001; Fig. 2d).
Few insects visited S. guyanensis flowers, either in the cultivated or field populations. Syrphid flies visited flowers occasionally. In the population from Wuyishan, six visits were recorded to a patch of 50 plants during 5 d of observation. The visitors foraged for pollen grains for several seconds on each flower. Four visits were to staminate flowers and two to perfect flowers. In the Dongxiang population, both 1998 and 1999, no visits were observed and the dyed pollen grains were not moved from stained flowers to unmanipulated flowers for 5-d observation of each year. In Wuyishan during one of the 5-d observation periods in 1999, however, three of 10 perfect flowers located in a ditch received 5, 12, 28 red pollen grains from staminate flowers, respectively. Dyed pollen comprised less than 1% of the hundreds of pollen grains usually deposited per flower.
The stamens of both staminate and perfect flowers usually dehisced rapidly in 10 min of flower opening and some dehisced before the flower opened. The height of the stamens of a perfect flower commonly approached the height of the carpels when they were splitting and some stamens extended beyond the height of the carpels. Pollen grains were able to fall directly onto the adjacent stigmas when the anthers dehisced. Pistils at the top of gynoecium and those not located close to the anthers seldom received pollen grains. Thus, only some stigmas are able to be autonomously pollinated.
By contrast with the low frequency of visits to S. guyanensis, bees, flies and butterflies frequently visited the flowers of S. potamogetifolia, S. trifolia and S. pygmaea, both in rice fields and in the Garden.
Fruit set and self-fertilization
The percentage of developed achenes per fruit in open-pollinated and bagged flowers varied widely (Fig. 3). For example, developed achenes per fruit of 131 aggregate fruits from the cultivated population ranged from 6.5% to 100%. Based on the observation of 290 open-pollinated and 14 bagged flowers, undeveloped achenes were common at the top of the aggregate fruit, and sometimes in bands running from top to bottom.
The cultivated plants produced many more ovules (achenes) per flower (fruit) than did plants in the two field populations (F5,321 = 25.0, P < 0.0001; Fig. 3a), indicating that ovule production was influenced by environmental conditions. Emasculated perfect flowers yielded very few mature achenes, but bagged flowers produced as many mature achenes as open-pollinated flowers did (Fig. 3b). The percentage of mature achenes in the two field populations was not significantly different, although the percentage from Dongxiang was lower than that of the cultivated population. Partly developed achenes that received pollen but in which development was constrained were less than 4% of the total and did not differ significantly among treatments (F5,321 = 1.10, P = 0.36; Fig. 3c). Except in emasculated and hand-pollinated flowers, 20–30% of achenes were undeveloped and presumably not pollinated (Fig. 3d).
Pollination observations and flower manipulations to investigate pollinator service and self-fertilization demonstrated that Sagittaria guyanensis ssp. lappula received very few pollinator visits and was self-compatible, usually autonomous autogamy at present habitats. Sagittaria guyanensis is a colonizing weed of rice fields. The subspecies guyanensis in southern Africa occupies similar habitats and possesses the same reproductive characters of lappula (S. Barrett, pers. comm.). The inflorescences of S. guyanensis emerge only several centimeters from the water, compared with 20–120 cm in the monoecious species S. pygmaea, S. potamogetifolia and S. trifolia. These monoecious species secrete floral nectar and are visited by generalist pollinators. For example, pollination rate and seed set per flower in S. trifolia were higher than 96% in both field and cultivated populations (Huang et al., 2002). The lower height and lack of nectar in S. guyanensis is probably related to low levels of pollinator visitation. Additionally, monoecious species of Sagittaria are perennial with clonal reproduction but andromonoecious species are annual with only sexual reproduction.
The maintenance of andromonoecy in Sagittaria guyanensis
Pollination treatments indicated the seed set of bagged flowers did not decrease but emasculated perfect flowers did. Partly developed achenes occupied a low proportion of ovules and were not significantly different between treatments, indicating seed production in this species is less likely to be resource limitation. Hand-pollinated flowers included 86.7% developed achenes, which was not significantly different from open-pollinated flowers (76.5%) in the cultivated population. This is probably because in the hand-pollinated flowers 11.5% of the stigmas were, on average, not completely pollinated (Fig. 3d). Pollination could be a major factor limiting seed production in this species.
In habitats with scarce pollinators, fruit production in S. guyanensis depends mainly on autonomous self-pollination. This finding supports the prediction that retention of stamens in perfect flowers has been selected to provide reproductive assurance (Barrett et al., 2000). Since the absence of pollinators decreases both seed production and male reproductive opportunities, plants that can self-pollinate benefit both through female and male functions (Lloyd, 1979).
Investment in male function in staminate flowers of S. guyanensis was much higher than investment in male function in perfect flowers, as staminate flowers contained more stamens, resulting in greater pollen production and larger pollen grains. Wang (1997) also observed larger pollen grains of staminate flowers. This result contrasts with most previous observations and theoretical predictions on andromonoecy, which have shown that staminate flowers usually have equal or relatively lower investment in male structures than perfect flowers (Solomon, 1986; Emms, 1993; Podolsky, 1993; Spalik & Woodell, 1994; Schlessman & Graceffa, 2002).
Aborted pistil development and late appearance in an inflorescence or in the flowering season are typical of staminate flowers in andromonoecious species. This developmental sequence may result in low allocation to staminate flowers (Thomson, 1989). Based on a resource allocation model, Spalik (1991) predicted andromonoecy would be promoted if pollen production were relatively low in staminate flowers. Consistent with this prediction, staminate flowers of Anthriscus sylvestris had significantly less pollen per flower than perfect flowers (Spalik & Woodell, 1994). The pedicel, petals, stamens and pistil of staminate flowers were all lighter than those of perfects in Zigadenus paniculatus (Emms, 1993). If staminate flowers promote pollinator attraction, one might expect more staminate flowers with fewer stamens (Bertin, 1982). If the number of stamens is stable, individual anthers might produce less pollen. This was evident in Besleria triflora (Podolsky, 1993). However, the number of stamens in staminate and perfect flowers is equal in many andromonoecious plants, for example, Leptospermum species (Primack & Lloyd, 1980; Andersen, 1990; O’Brien, 1994), Solanum species (Solomon, 1986; Anderson & Symon, 1989; Diggle, 1993; Elle, 1998), Passiflora incarnata (May & Spears, 1988), Prosopis species (Hoc et al., 1994), Cneorum tricoccon (Traveset, 1995), Gagea chlorantha (Wolfe, 1998), Isomeris arborea (Krupnick & Weis, 1998), Trevoa quinquenervia (Medan & D’Ambrogio, 1998), and Pseudocymopterus montanus (Schlessman & Graceffa, 2002). Solomon (1986) found no difference in pollen size between two flower types of Solanum carolinense. To my knowledge, this study is the first to show that staminate flowers have higher pollen production per flower with larger pollen grains and petals than perfect flowers in an andromonoecious species.
Differential sex allocation in andromonoecious species
Unlike most reported andromonoecious plants, staminate flowers of S. guyanensis have a higher male investment than perfect flowers. Similar differential staminate/perfect pollen production occurs in some androdioecious species. The ratio of pollen production in staminate to hermaphroditic flowers was 3.3 in Datisca glomerata (Philbrick & Rieseberg, 1994) and 5–10 in Mercurialis annua (Pannell, 1997). Males are expected to produce much more pollen than hermaphrodites in androdioecious species because the maintenance of this sexual system requires a large male-fitness differential between males and hermaphrodites (Lloyd, 1975; Charnov, 1982; Charlesworth, 1984).
Differential sex allocation in andromonoecious species may provide a clue to the shift of breeding systems from hermaphroditism to monoecy via andromonoecy. During this transition, staminate flowers of monoecious species may be selected to enhance their male function because the male function is lost from perfect flowers as they become pistillate (‘female’). In the extensively studied genus Solanum, which has a variety of breeding systems including hermaphroditism, andromonoecy and functional dioecy (Whalen & Costich, 1986; Anderson & Symon, 1989), staminate flowers tend to produce more pollen than perfect flowers in andromonoecious species (Anderson & Symon, 1989). Indeed, in Solanum clarkiae staminate flowers contained about 1.6 times as many pollen grains as did perfect flowers (Anderson & Symon, 1989). The functional gender of andromonoecious S. carolinense ranged from completely female to completely male, because staminate and perfect flowers differed in their ability to contribute to male and female success (Elle & Meagher, 2000). Two alternative transitions between andromonoecy and monoecy could be considered in Sagittaria. First, self-pollination is selected against when the progeny exhibit inbreeding depression (Harder & Barrett, 1996; Barrett, 2002; and references therein); this could promote sex differentiation from andromonoecy to monoecy. Both morphological characteristics (Chen, 1989) and molecular data (Du et al., 1998) of Sagittaria species from China support the hypothesis that hermaphroditism is an ancestral condition. Yet, identification of the evolutionary transition of these breeding systems awaits explicit phylogenetic analysis at a worldwide base. Second, andromonoecy may be derived from monoecy by returning functional stamens within flowers under selection of inadequate pollination (Bertin, 1982), since most Sagittaria species are monoecious. Given the possibility that self-pollinating perfect flowers are potentially under selection to reduce investment in male function (Cruden, 2000), the possible transition may provide an explanation to the presence of higher male investment in staminate flowers in S. guyanensis. This research suggests that the differential male investments of staminate and perfect flowers, such as in S. guyanensis and Solanum species, which may be overlooked in other species, should be considered for understanding the maintenance and evolution of andromonoecy.
In conclusion, staminate flowers in S. guyanensis produced larger petals and more numerous anthers containing larger pollen than perfect flowers. Larger petals of staminate flowers, as suggested by Delph et al. (1996), may potentially have two functions: protection of reproductive structures and attraction to pollinators. Larger pollen of staminate flowers may potentially exhibit faster pollen tube growth than pollen of perfect flowers, and this needs further study. Retention of stamens in perfect flowers in this annual species may have been selected because they provide reproductive assurance in times and places where pollinators are scarce. The loss of functional stamens of female flowers in monoecious Sagittaria species may have occurred in environments with effective pollen dispersal, minimizing the need for reproductive assurance. Monoecy would also replace andromonoecy if male and female functions in perfect flowers interfere with each other (Bertin, 1982).
The author wishes to thank Gui-Hua Liu, Ni Song, Quan Wang, Lu-Lu Tang, Xiao-Ming Wang, Qing-Feng Wang, and Hai-Yang Wang for their help in the field and laboratory, Yasuyuki Ide for drawing the graphs of flowers, Prof. Spencer Barrett and You-Hao Guo for their helpful suggestions regarding this work, Prof. Lynda Delph for correcting the English and providing valuable comments on an earlier draft, and three anonymous reviewers for their improvements to the manuscript. This study was supported by a grant from National Science Foundation of China.