1. Introduction

Approximately 70% of all flowering plants bear hermaphrodites, 10–20% are monoecious, 6% are dioecious, and all others have mixed or variable sexual systems (Yampolsky and Yampolsky, 1922; Bawa, 1974, 1979; Renner, 2014). Andromonoecy, the sexual system in which individuals bear both staminate and hermaphroditic flowers, has been observed in ~4000 species, or 1.7% of flowering plants (Yampolsky and Yampolsky, 1922; Bawa and Beach, 1981). Despite its infrequent occurrence, the fitness consequences of andromonoecy are important because they help elucidate the evolution of sexual systems in flowering plants (Barrett, 2002; Torices et al., 2011; Goldberg et al., 2017). Three key factors fundamental to clarifying the evolution and maintenance of andromonoecy include (1) the function of additional male flowers (Zhang and Tan, 2009; Calviño et al., 2014; Marcelo et al., 2021), (2) the absence of stamen reduction in hermaphroditic flowers (Podolsky, 1993; Huang, 2003; Connolly and Anderson, 2003; Tomaszewski et al., 2018), and (3) the genetically fixed absence of male and bisexual functions in separate individuals (Bertin, 1982; Boualem et al., 2008; Ji et al., 2016; Aguado et al., 2020).

Three hypotheses have been proposed to explain the function of additional male flowers. The resource reallocation hypothesis states that the fitness gained from the production of male flowers is greater than the fitness loss associated with the absence of female function (Primack and Lloyd, 1980; Elle and Meagher, 2000; Reuther and Claßen-Bockhoff, 2013; Bertin, 1982). Although male flowers in several taxa have been shown to be less costly than hermaphroditic flowers (Emms, 1993; Cuevas and Polito, 2004; Zhang and Tan, 2009; Casimiro-Soriguer et al., 2013; Marcelo et al., 2021), little is known about whether the loss of female fitness is compensated by the fitness gain mediated by additional male flowers (but see Elle, 1998, 1999; Vallejo-Marín and Rausher, 2007a). The pollinator attraction hypothesis posits that male flowers increase visitation rates without removing pollen from pollinators, which ultimately increases fruit production (Podolsky, 1992; Liao and Zhang, 2008; Zhang and Tan, 2009). The pollen donation hypothesis states that male flower production improves male fitness in andromonoecious plants through superior pollen quantity, quality, and donation (Sutherland and Delph, 1984; Solomon, 1986; Cuevas and Polito, 2004; Zhang and Tan, 2009; Dai and Galloway, 2012; Murakami et al., 2022; Ndem-Galbert et al., 2021). Several studies have found that pollen production and export, as well as siring success, are all greater in male flowers than in hermaphroditic flowers (Quesada-Aguilar et al., 2008; Zhang and Tan, 2009; Zimmerman et al., 2013; Murakami et al., 2022). Nevertheless, some studies have indicated that male flowers performed equal or even worse than hermaphroditic flowers in pollen viability, pollen germination and pollen export (Podolsky, 1992; Cuevas and Polito, 2004; Sunnichan et al., 2004; Narbona et al., 2005, 2008; Peruzzi et al., 2012; Pérez et al., 2019). Each of these hypotheses suggests that production of male flowers is a strategy to improve male or female function in andromonoecious plants (Bertin, 1982; Vallejo-Marín and Rausher, 2007b; Dai and Galloway, 2012).

In autogamous andromonoecious species, the function of additional male flowers still remains unclear. Autogamous hermaphroditic flowers are superior at fertilization (Boaz et al., 1994; Kalisz and Vogler, 2003; Kamran-Disfani and Agrawal, 2014); therefore, additional male flowers seem to be redundant (Huang et al., 2000; Huang, 2003). However, long-term selfing can result in inbreeding depression (Barrett and Charlesworth, 1991; Charlesworth, 2006; Wright et al., 2013; Cheptou, 2019). Thus, the production of male flowers may be a strategy to promote outcrossing (Narbona et al., 2002; Zhang and Tan, 2008; Casimiro-Soriguer et al., 2013). To our knowledge, few studies have quantified the function of male flowers in autogamous andromonoecious species. Furthermore, because andromonoecious plants bear two flower morphs with male function, male fitness ought to be studied at both the parental level (pollen viability, siring ability, etc.; Pérez et al., 2019) and the offspring level (hundred seed dry weight, seed germination rate, etc., Tremayne and Richards, 2000).

In this study, we examined male function in an endangered aquatic andromonoecious herb, Sagittaria guayanensis. Autogamy of S. guayanensis is a mechanism of reproductive assurance in previous studies. Moreover, because pollinators were thought to be scarce in natural populations, male flowers were considered redundant (Huang et al., 2000; Huang, 2003). Since these earlier studies, we have observed pollinators of S. guayanensis at our study site (Zou et al., 2023; Fig. S1). Thus, the function and adaptive advantages of male flowers must be reconsidered. Here, we aim to elucidate the evolutionary significance of male flowers in S. guayanensis. For this purpose, we determined whether male flowers (1) require fewer resources to produce, (2) are more attractive to pollinators, (3) generate and export more and/or higher-quality pollen than do hermaphroditic flowers, or (4) perform better than hermaphroditic flowers in natural siring success. We compared morphology, pollinator preference, pollen production and donation, siring ability, natural siring success, hundred seed dry weight and seed germination of male and hermaphroditic flowers.

2. Materials and methods 2.1. Study materials and sites

Sagittaria guayanensis is an annual, self-compatible aquatic andromonoecious herb that grows in paddy fields or ditches (Chen, 1989). It is currently listed as endangered on the Red List of Biodiversity in China (Higher Plants Volume) due to herbicide overuse (Huang et al., 2000). Unlike other Sagittaria species in China, S. guayanensis cannot be propagated by cloning (Wang et al., 2021). This is an andromonoecious plant, with most individuals producing both hermaphroditic and male flowers. It flowers from June to October and is pollinated by small bees such as Halictus (Vestitohalictus) pulvereus (Morawitz, 1873; Zou et al., 2023; Fig. S1). Individual plants sequentially produce 18.3 ± 2.7 inflorescences over the course of the flowering period (Zou et al., 2023), with each inflorescence lasting 7–10 days. Flowers of a new inflorescence bloom at the end of the previous one. The flowers are sequentially produced from the base (hermaphroditic) to the apex (male) of the inflorescence (Chen, 1989). The average numbers of flowers per inflorescence in the natural population of Duoxing village (26°29′N, 109°65′E) are 2.1 ± 0.1 (n = 75) hermaphroditic and 1.0 ± 0.1 (n = 75) male flowers. In cultivated conditions, each inflorescence contains 2.8 hermaphroditic and 3.0 male flowers (Zou et al., 2023). A single flower lasts 3 to 4 h. For each inflorescence, the two floral morphs rarely open on the same day (Huang, 2003). Hermaphroditic flowers open first, at about 11:00. Male flowers open one or two days later, at about 10:30. Flowers are bowl-shaped and bear three white petals, and each may have a basal purple spot (Fig. 1). The stamens of the hermaphroditic flowers adhere to the gynoecium and grow toward its center, sometimes growing past it (Fig. 1). The gynoecium of the hermaphroditic flowers consists of 274.4 ± 16.8 (n = 57) apocarpous pistils (Lyu S.-T., unpublished data), each with a single ovule. The pistils of the male flowers abort early in development (Chen, 1989). Secretory cells may occur at the base, but they produce no nectar (Huang, 2003). The pollen is viable during blooming, and its viability does not differ between male and hermaphroditic flowers (Zou et al., 2023).

In early July of 2021 and 2022, approximately 100 plants were uprooted in Duoxing village (26°29′N, 109°65′E), Hunan Province, China. Such collection does not influence the regeneration of the field population. Each individual was planted in a 9.7 L cement bucket in the garden of the greenhouse at Wuhan University (30°32′N, 114°21′E), Wuhan, China. All plants received equal amounts of fertilizer every two weeks. All experiments were conducted in this garden from July to September in 2021 and 2022.

2.2. Morphological measurements

In mid-August 2021, the peak flowering time of S. guayanensis, a total of 40 male flowers and 40 hermaphroditic flowers were randomly selected from different individuals. Pedicel length, corolla diameter, petal length and width, stamen number, androecium or gynoecium diameter, and the number of purple spots were measured to distinguish male and hermaphroditic flower morphology. All morphological parameters were measured with a digital caliper at approximately 12:00 when the floral display was at its peak. After measurement, flowers were harvested and placed in centrifuge tubes, and dried in an electric heating constant-temperature drying box (GZX-9030 MBE; Boxun Co., Shanghai, China) at 50 ℃ for 24 h. Dry weight was then measured to compare resource investment.

2.3. Pollinator observation

To compare pollinator preference between male and hermaphroditic flowers, pollinators were observed from 11:00 to 14:00 in the garden of the greenhouse of Wuhan University, where pollinators have previously been noted to frequently visit Sagittaria flowers (Zou et al., 2023). Each observation period of an experimental quadrat lasted for 15 min. When a pollinator landed on a flower and touched the pistils or stamens, the visit was scored as intact. Visits to the male and hermaphroditic flowers were counted and the visit frequency was calculated as follows: total intact visits/time of observation/number of flowers. Visit frequency was calculated only for visits by Halictus (Vestitohalictus) pulvereus, as they are the main pollinators of S. guayanensis. Other flowers in the garden of the greenhouse were removed or bagged to exclude their influence.

In 2021, a quadrat that contained nine buckets with one individual per bucket (Fig. S2A) was monitored. Before each observation, the total numbers of male and hermaphroditic flowers in the quadrat were recorded. The observation was conducted for 375 min (5 d).

In 2022, in order to exclude the influence of flower sex ratio and flower number to pollinator attraction, a quadrat with 5 male and 5 hermaphroditic flowers from 10 individuals was monitored (Fig. S2B). The observation was conducted for 750 min (9 d).

2.4. Siring ability

Hand pollination was conducted to evaluate the pollen limitation of hermaphroditic flowers and compare the differences in siring ability between male and hermaphroditic flowers in June and August of 2021 and 2022. The following treatments were conducted on hermaphroditic flowers: (1) flowers were emasculated and cross-pollinated by pollen from male flowers; (2) flowers were emasculated and cross-pollinated by pollen from hermaphroditic flowers; (3) flowers were bagged in nylon mesh bags (i.e., autogamy); and (4) flowers were open-pollinated. Emasculation was conducted 1 d before the flowers opened. Pollen grains were collected from all individuals that were not used as maternal plants that day and placed on a Petri dish. The pollen was mixed and a soft brush was used for hand pollination. The pollinated flowers were then bagged in nylon mesh bags to exclude pollinators until the flowers were closed. For treatment (3), the flowers were bagged before blooming to exclude pollinators until the flowers were closed; for treatment (4), no action was taken with flowers until they were closed. A total of 25 and 30 hermaphroditic flowers from 25 to 30 individuals were used in each treatment in 2021 and 2022, respectively. The fruits were harvested at maturity and dried in an electric heating constant-temperature drying box (GZX-9030 MBE, Boxun Co., Shanghai, China) at 50 ℃ for 4.5 h. The achenes were counted under a stereoscope (Olympus SZX16; Olympus Co., Tokyo, Japan) and categorized according to the classification system proposed by Huang et al. (2000); namely, well-developed, abnormally developed, or undeveloped.

2.5. Pollen quantity and quality 2.5.1. Quantity

To quantify pollen production and export in male and hermaphroditic flowers, we measured total pollen amount, export of first visit, and total export by Halictus (Vestitohalictus) pulvereus from 40 hermaphroditic and 40 male flowers. Different individuals were used for each measurement. All flowers were bagged with nylon mesh bags before blooming to exclude pollinators. For total pollen production, flowers were harvested at 9:00 before dehiscence. For export of first visit, the bags were removed and flowers were harvested immediately after first pollinator foraging. To evaluate total pollen export, flowers were foraged by pollinators for the whole flowering time and harvested when they were closed. All the flowers were stored in 1 mL of 70% (v/v) ethanol before counting. The flowers were then removed from the ethanol and placed in a 5.5 cm plastic Petri dish. All stamens were removed and pulverized in a 1.5 mL centrifuge tube with a grinding pestle, and the remaining flower parts, as well as the Petri dish, were washed five times with water. All liquid as well as the ethanol was collected in centrifuge tubes. The liquid was centrifuged at 12, 000 rpm for 2 min, after which less than 1000-μL of liquid remained. A 1000-μL pipette (Eppendorf GmbH, Hamburg, Germany) was used to measure the exact volume of the remaining liquid. The liquid was then vortexed and diluted to an appropriate concentration. Subsequently, pollen grains of 5 μL were counted under a light microscope (Nikon E100; Nikon Co., Tokyo, Japan). Counts were averaged for three replicates per flower. The total pollen number was calculated as follows: (average pollen amount of 3 repeats/5 μL) × (dilution times × volume). The residual pollen count was deducted from the average total production to determine the first-visit and total pollen donations.

2.5.2. Quality

The dry weight of 100 seeds is an indicator of pollen quality that allows comparisons to be made between male and hermaphroditic flowers and autogamy. Seeds were collected from fruit that were used to make siring ability comparisons in 2021 (Section 2.4). Well-developed achenes were dried at 50 ℃ for 24 h in a constant-temperature drying box (GZX-9030 MBE; Boxun Co., Shanghai, China) to a stable weight. One hundred seeds per aggregate fruit were randomly selected and their dry mass was measured with an analytical balance (to 0.001 g).

Seed germination rate is another indicator of pollen quality. Hand-pollination was performed again as previously described in section 2.4. In total, 18 flowers from 18 individuals were randomly selected for each treatment. All fruits were harvested at maturity and stored at 4 ℃ in early September 2021. In October 2021, 100 well-developed achenes per fruit were randomly selected and placed on two layers of filter papers moistened with distilled water in a 5.5 cm-diameter Petri dish. The Petri dish was then placed in a plastic bag and incubated in a growth chamber (GZC-300C; Hefei Youke Instrument Equipment Co. Ltd., Hefei, China) under 14 h/10 h, 33 ℃/20 ℃, and 16660 lux/0 lux day/night conditions. Germinated seeds were counted every 8 d for 64 d when germination ceased, and the total germination rate was calculated.

2.6. Natural siring success

Artificial quadrats were established to investigate the natural siring success between hermaphroditic and male flowers. Each quadrat contained three focal emasculated hermaphroditic flowers as recipients and six surrounding male or hermaphroditic flowers as pollen donors (Fig. S2C). Petals of focal flowers were not destroyed during emasculation so that pollinator visitation was not influenced (Fig. S3). Other flowers in the garden of the greenhouse were removed or bagged to exclude their influence. Focal plants were not reused to ensure independence. The quadrat was placed in the garden of the greenhouse from 11:00 to 15:00. Only one flower morphology was used as surrounding flowers for each quadrat and one quadrat was established each alternative day. A total of seven quadrats with male surrounding flowers (21 focal flowers) and seven quadrats with hermaphroditic surrounding flowers (21 focal flowers) were established, respectively. All focal flowers were harvested until mature and the number of well-developed seeds were counted.

2.7. Statistical analysis

Principal components analysis (PCA) was conducted using seven floral traits (pedicel length, corolla diameter, petal length and width, stamen number, androecium or gynoecium diameter, and number of purple spots) to summarize the floral morphology of male and hermaphroditic flowers. Independent-sample t-tests were used to compare morphological parameters, pollinator visitation rate and natural siring success between male and hermaphroditic flowers. After normality testing, one-way ANOVAs were used to test for differences in siring ability, seed germination rate (followed by Games-Howell post-hoc test) and seed weight (followed by Tukey test) among treatments. Seed germination data were log-transformed. All analyses were conducted in SPSS v.28.0 (IBM Corp., Armonk, NY, USA).

3. Results 3.1. Morphological characteristics

Overall, male flowers differed from hermaphroditic flowers both in shape and size (Table 1; Fig. 2). In the PCA of morphological measurements, PC1 explained 40.81% of the total variance and all the measurements were positively correlated except androecium or gynoecium diameter, whereas PC2 explained 15.05% of the total variance. The dry weight of male flowers was only 27% of that of hermaphroditic flowers, indicating male flowers were cheaper to produce (t = 12.435, df = 78, P < 0.001). Male flowers are significantly larger than hermaphroditic ones (in terms of corolla diameter, t = 7.345, df = 78, P < 0.001). Furthermore, male flowers had more stamens than did hermaphroditic flowers (t = 3.177, df = 78, P = 0.002).

3.2. Pollinator preference

All results showed that male flowers were more attractive to pollinators (Fig. 3) and pollinators spent more time foraging in male flowers (Fig. S4). In 2021, pollinator observation of the quadrat with nine individuals showed that pollinators visited two floral morphs at a similar frequency (t = 0.571, df = 8, P = 0.584; Fig. 3A). However, in 2022, observations of the quadrat with equal male and hermaphroditic flowers revealed that pollinators visited male flowers significantly more often (t = 3.562, df = 16, P = 0.005; Fig. 3B).

3.3. Pollen quantity and quality

Male flowers produced significantly more pollen than did hermaphroditic flowers (F_5,234 = 342.856, P < 0.001). In addition, male flowers exported more pollen than did hermaphroditic flowers in both first visit and total export (Fig. 4).

Dry weight and germination rate were lowest in seeds sired by hermaphroditic flowers and highest in seeds sired by male flowers (Table 2). The seeds sired by male flowers were significantly heavier than those sired by hermaphroditic flowers (F_2,72 = 358.992, P < 0.001). Furthermore, seeds sired by male flowers germinated better than those sired by hermaphroditic flowers (F_2,51 = 8.327, P = 0.001). Seeds sired by autogamy performed better than those sired by hermaphroditic flowers both in hundred seed dry weight and germination rate.

3.4. Siring ability and natural siring success 3.4.1. Siring ability

Hand-pollination experiments revealed the pollen limitation of autogamy was due to inadequate pollination, especially on the top of the gynoecium (Table 2; Fig. S5). The seed set between the hermaphroditic flower sire and male flower sire did not differ significantly (Table 2), indicating no difference in siring ability between hermaphroditic and male flowers. The seed set of the open-pollination treatment was significantly higher than that of autogamy, indicating that stigmas could receive extra pollen from other flowers. The results of 2021 and 2022 were consistent (Table 2), and the lower seed set in 2022 may be a consequence of abnormal weather.

3.4.2. Natural siring success

Natural siring success was significantly higher in male flowers than in hermaphroditic flowers; specifically, the total number of seeds sired by surrounding male flowers was triple that of hermaphroditic flowers (t = 3.126, df = 40, P < 0.001; Fig. 5).

4. Discussion

In this study, we compared male function between male and hermaphroditic flowers in an andromonoecious species with autogamous pollination. Huang et al. (2000) and Huang (2003) considered male flowers of Sagittaria guayanensis to be redundant due to a scarcity of pollinators; however, we not only observed pollinators at our study site but also found that compared to hermaphroditic flowers, male flowers exported more pollen and had higher natural siring success and offspring fitness. These differences may be a result of the higher proportion of male flowers per individual (flower level) in our population (around 33%) than in those of previous studies (less than 5%; Huang et al., 2000; Huang, 2003). Our results support the three hypotheses proposed to elucidate the function of male flowers in andromonoecious species, especially the pollen donation hypothesis. In S. guayanensis, andromonoecy appears to be maintained by the superior male function of male flowers and their potential role in promoting outcrossing.

In Sagittaria guayanensis, dry weight was lower but floral display was greater in male flowers than in hermaphroditic flowers; furthermore, pollinators were more attracted to male flowers than to hermaphroditic flowers. The lower dry weight of male flowers may result from pistil abortion during early development (Chen, 1989). Other andromonoecious plants (e.g., Olea europaea and Zigadenus paniculatus) have also been found to invest fewer resources in male than in hermaphroditic flowers (Emms, 1993; Cuevas and Polito, 2004). However, the male flowers of S. guayanensis are the first to have been observed to have lower resource allocation but larger male-function-related structures. This is likely a strategy to improve male function of male flowers. In addition, the larger floral display of male flowers may increase attractiveness of these flowers to pollinators. However, it is important to acknowledge that our observation of pollinators in a common garden may be different from that of field populations. Our results of pollinator visitation revealed that pollinators preferred male flowers. Furthermore, frequent visits also increased the pollen export of male flowers. Due to a lack of clear determinants of increases in male fitness and decreases in female fitness, the resource reallocation hypothesis for S. guayanensis merits further validation to consider the fitness gain or loss caused by producing male flowers.

In most andromonoecious species, male flowers produce the same amount as or less pollen than hermaphroditic flowers (Anderson and David, 1989; Podolsky, 1993; Cuevas and Polito, 2004; Sunnichan et al., 2004; Vallejo-Marín and Rausher, 2007a; Zhang and Tan, 2009; Casimiro-Soriguer et al., 2013; Pérez et al., 2019). Harder et al. (1985) predicted that the low level of pollen in male flowers is efficient because distributing smaller pollen packages over a greater number of flowers can improve pollen dispersal. However, in Sagittaria guayanensis, pollen was produced at significantly higher levels in male flowers than in hermaphroditic flowers. This is consistent with other andromonoecious species, such as Euphorbia boetica and Commelina communis (Narbona et al., 2005; Murakami et al., 2022). There are two reasons why distributing more pollen grains is advantageous for male flowers in S. guayanensis. Firstly, the pollen itself may be the only reward in nectarless S. guayanensis flowers (Huang, 2003). Secondly, because of low pollinator visitation, pollen export (especially after first visitation) is more efficient when male flowers possess higher levels of pollen.

In Sagittaria guayanensis, pollen was exported more from male flowers than from hermaphroditic flowers, possibly due to more frequent visits and longer foraging times (Fig. S4). In the natural siring success experiment, pollen export of hermaphroditic flowers may have been underestimated because surrounding hermaphroditic flowers are also pollen recipients, necessitating that autogamy be considered. Regardless, when we designated three focal emasculated hermaphroditic flowers as recipients for cross-pollination, we found that the number of successfully sired seeds was three times higher in the six surrounding male flowers than in the six surrounding hermaphroditic flowers. Previous studies on other andromonoecious species have also found that pollen export and siring success is higher in male flowers than in hermaphroditic flowers (Zhang and Tan, 2009; Dai and Galloway, 2012; Murakami et al., 2022). In S. guayanensis, male flowers are less likely to deposit self-pollen due to pistil abortion, which enhances pollen export (Chen, 1989). In addition, tripling the seeds sired by male flowers may offset the loss of female function. Consequently, male flowers can achieve the same reproductive function as hermaphroditic flowers or even better than the flowers that fail to set fruit (Dai and Galloway, 2012). Augmented pollen production, export, and the increased siring success of male flowers support the pollen donation hypothesis.

Fitness, dry weight, and germination rates were all higher in male flowers and autogamous flowers than in hermaphroditic flowers. Previous studies have reported that pollen viability and germination differ between male and hermaphroditic flowers (Cuevas and Polito, 2004; Casimiro-Soriguer et al., 2013; Pérez et al., 2019; Marcelo et al., 2021). However, to the best of our knowledge, the present study is the first to demonstrate that the seeds sired by male flowers are more fit than those sired by hermaphroditic flowers. In our study, seeds sired by hermaphroditic flowers were less fit than those sired by autogamous flowers, which may be attributed to the influence of emasculation on fruit set, seed set, and even fruit weight or seed mass (Hedhly et al., 2009; Guerra et al., 2010; Kosiński, 2010; Bamberg, 2020). Regardless, male flowers had larger pollen grains that performed better and offset the influence of emasculation (Huang et al., 2000; McCallum and Chang, 2016). Offspring fitness has been discussed in some sexual systems, such as gynodioecy, in which seeds produced by female individuals perform better than those produced by hermaphroditic flowers (Emery and Mccauley, 2002; Delph and Mutikainen, 2003). In andromonoecious species, pollen quality is revealed by both parentals (e.g., pollen viability, germination, and siring ability) and offspring (e.g., seed weight and gemination rate). In addition, offspring fitness provides new insights into the maintenance and evolutionary advantages of andromonoecy, which might otherwise be overlooked.

Maintenance of andromonoecy in Sagittaria may result from the trade-off of male function that is distributed in different flowers. Strong empirical evidence indicates that hermaphroditism is the ancestral condition in angiosperms (Bertin, 1982; Sauquet et al., 2017; Zhang et al., 2022). Monoecy with separate male and female flowers on the same plant is considered to be derived from hermaphroditism (Mitchell and Diggle, 2005). Monoecy has been traditionally considered to originate from andromonoecy or gynomonoecy (Bawa and Beach, 1981; Bertin, 1982). In Sagittaria, both morphological (Chen, 1989) and molecular data (Du et al., 1998) indicate that hermaphroditism is the ancestral condition. Because Sagittaria is a genus in which most species are monoecious, how andromonoecy persists in this genus is fascinating. Here, we propose two possible explanations. First, the transition from hermaphroditism to monoecy requires that some flowers reduce male function and other flowers reduce female function (Bawa and Beach, 1981). In S. guayanensis, pistillodes are found in male flowers (Lyu S.-T., unpublished data); therefore, it is clear that male flowers transitioned from hermaphroditic flowers. Second, due to low pollinator visitation in S. guayanensis, inadequate pollination utilizes pollen from hermaphroditic flowers through autogamy. Consequently, in S. guayanensis, production of male flowers and hermaphroditic flowers that function in different ways is balanced.

More importantly, the maintenance of andromonoecy in Sagittaria guayanensis may play a vital role in promoting outcrossing. Selfing lineages are mostly short-lived owing to higher extinction rates (Barrett and Charlesworth, 1991; Charlesworth, 2006; Wright et al., 2013; Eddie and Agrawal, 2018) despite short-term advantages (Lloyd, 1979; Barrett, 2010; Buckley et al., 2019). Therefore, outcrossing is required in S. guayanensis to mitigate the risk of inbreeding depression. Male S. guayanensis flowers produced and exported more pollen, in addition to siring more seeds with superior fitness; thus, these flowers may be superior outcrossing donors. Future studies ought to use genetic markers to focus on the relative contribution of male and hermaphroditic flowers in outcrossing.

Extensive herbicide application on rice fields or ditches has severely damaged S. guayanensis habitats (Huang et al., 2000). Consequently, this species is listed as endangered on the Red List of Biodiversity in China. For conservation of this species, the above-mentioned results imply that more male flowers may enhance pollinator attraction and increase outcrossing. The higher nutrient conditions provided by ex-situ conservation are likely able to increase the production of male flowers (Huang et al., 2000).

5. Conclusion

In conclusion, the present study shows that male flower production in Sagittaria guayanensis is a strategy to improve male fitness and that male flowers may be superior pollen donors for outcrossing. Male flowers had greater pollen export and natural siring success than that of hermaphroditic flowers; moreover, seeds sired by male flowers had higher fitness. These findings indicate that the maintenance of andromonoecy in autogamous S. guayanensis results from better male function and potential role in promoting outcrossing in male flowers. Further studies that account for offspring fitness in andromonoecious species may help provide insight into the maintenance and evolution of andromonoecy.

Acknowledgements

We thank Xin Zan for assistance in the fieldwork. We are grateful to Lan-Jie Huang for his valuable advices. The present study was supported by the National Natural Science Foundation of China (No. 31970250).

Author contributions

S.T.L. and X.F.W. conceptualized and planned the study. S.T.L., T.T.Z. and Q.L.J. conducted the study. S.T.L. and T.T.Z. conducted statistical analyses and wrote the manuscript. All authors contributed to the draft versions and approved the final version of the manuscript.

Declaration of competing interest

The authors declare that they have no conflict of interest.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.pld.2023.03.009.