1. Introduction

The visual signals emitted by flowers, including color, size, shape and display size, play a crucial role in shaping plant–pollinator relationships (Willmer, 2011), resulting in the conceptualized pollination syndrome based on flower color together with flower shape, scent and nectar trait (Fenster et al., 2004). For example, it is widely accepted that birds prefer red flowers, but bees generally do not (Fenster et al., 2004). Furthermore, the contrasting color patterns on petals, usually found in close proximity to the position of rewards, are considered to function as visual guides to enhance pollinator attraction (Johnson and Dafni, 1998; Goodale et al., 2014; de Jager et al., 2017). The evolutionary consequences of flower color and the underlying mechanisms have garnered considerable interest in recent years (Rausher, 2008; Rieseberg and Blackman, 2010; Ortiz-Barrientos, 2013; Trunschke et al., 2021), emphasizing the importance of flower color in plant ecology and evolution.

Despite extensive research on flower color and its role in pollination, the majority of studies have focused on visible light as detected by the human eye. However, flowers can appear quite different to animal pollinators than they do to human observers (Trunschke et al., 2021). For instance, ultraviolet (UV) patterns, which are only visible in UV light, can be detected by bees but not by humans (Peitsch et al., 1992; Chittka et al., 1994), underscoring the need to study the UV patterns in plant-pollinator interactions. UV patterns on flowers have received much attention recently (Horth et al., 2014; Koski and Ashman, 2014; 2015a; Papiorek et al., 2016; Zhang et al., 2017b; Chen et al., 2020a). Understanding these patterns could provide valuable insights into plant–pollinator relationships, although UV patterns could also protect reproductive organs from abiotic damage (Koski and Ashman, 2015b; 2016; Koski et al., 2020). However, how UV patterns affect male and female reproductive success of plant species remains unclear, and this might impede our further understanding of flower evolution.

Male reproductive success in plant species is maximized by pollinators that visit plants and disperse their pollen many times (Thomson et al., 2000); in contrast, female flowers would be saturated after several pollinator visits. The difference in requirements for achieving maximum male and female reproductive success might explain why male flowers are typically larger than female flowers in dioecious species relying on animal pollinators (Delph et al., 1996; Eckhart, 1999). Undoubtedly, large flower displays attract more pollinators than do small flower displays (Huang et al., 2006; Zhang et al., 2017a). Animal pollinators, when rewarded, tend to continuously visit one plant (Duan et al., 2005; Zhang et al., 2017a), resulting in pollen transfer among different flowers on a single plant. While this within-plant pollen transfer can reduce male and/or female reproductive fitness in hermaphrodite species (Harder and Barrett, 1995; Liao et al., 2009), research on within-plant pollen transfer on reproductive fitness is scarce in dioecious species (but see Vamosi et al., 2006). In addition, as the measurement of both male and female reproductive success of dioecious species is achievable, dioecious species offer an ideal system to study the effect of pollinator movements on plant reproductive success.

In the present study, we investigated variations in UV bullseye size on male and female plants in a dioecious and annual vine, Herpetospermum pedunculosum (Seringe) C.B. Clarke (Cucurbitaceae), and examined how changes in UV bullseye size affect reproductive success. We aimed to address the following questions. (1) How do pollinators respond to flowers with different UV bullseye sizes? (2) How do differences in the size of flower UV patterns affect male and female reproductive success?

2. Material and methods

Herpetospermum pedunculosum is an annual vine native to the Himalayas and grows along thickets and forest margins on mountain slopes with an altitude ranging from 2300 m to 3500 m (Lu et al., 2011). In natural habitats, H. pedunculosum usually exhibits more than three branches per plant, with these slender branches often climbing trees. Similar to many species in Cucurbitaceae (Renner and Schaefer, 2017), H. pedunculosum is dioecious. While both male and female flowers appear yellow under visible light (Fig. 1a), they display a contrasting bullseye under UV light (Fig. 1b). Male plants have racemose inflorescences, whereas female plants have solitary flowers. Our fieldwork was conducted in 2019 at the Shangri-La Alpine Botanical Garden (27°54′36″ N, 99°38′24″ E, 3250 m), situated in the southern region of the Hengduan Mountains, Northwest Yunnan Province, China. In the study population, H. pedunculosum starts flowering in late August.

2.1. Population and floral traits

To determine the sex ratio of the study population, we conducted a survey to count the number of male and female plants. Additionally, to assess the mating environment, we also counted the total number of open male and female flowers within the population on nine different days, irrespective of plant level.

We measured petal relative reflection of 10 male and 10 female flowers by using a portable spectrometer equipped with a PX-2 pulsed xenon lamp (Ocean Optics Maya2000-Pro, Dunedin, FL, USA). Prior to each measurement, we calibrated the spectrometer with a white standard. Due to the variation in reflective properties between the center and periphery of the petal (Fig. 2b), we measured the relative reflectance at both locations for each flower. We also measured the relative reflectance of H. pedunculosum leaves as background. In the laboratory, from the raw data of flowers and leaves, we extracted the relative reflectance from 300 to 700 nm in 5-nm steps and converted them to individual loci in the bee color hexagon (Chittka, 1992). To determine color and achromatic contrasts, we calculated pairwise Euclidean distances between the loci of the center and periphery on the petals. Additionally, we determined the mean Euclidean distance between the species centroids and loci of center and periphery on the petals was also calculated in the bee color hexagon. The Euclidean distance between any two loci indicated the perceived color difference or the contrast between the stimuli, and the threshold value was set as 0.10 (Chittka et al., 1997) for the color discrimination of bees.

To measure the size of the UV bullseye on H. pedunculosum flowers, we photographed 55 male and 57 female flowers on different plants. To capture the UV images, we used a modified camera (Panasonic DMC-G5) and a Nikon EL-80 mm lens equipped with a near-infrared cutoff filter (glass type: S8612) and a visible light cutoff filter (glass type: UG5). These filters permitted UV imaging within the range of 310–420 nm. For scale, we added a measurement device to the background (Zhang et al., 2017b). To ensure that the flowers belonged to different plants, we selected focal plants that were positioned more than 10 m apart from each other. Next, we eliminated the background of the picture. Subsequently, we used ImageJ (Rasband, 2012) to obtain three distinct color channels (red, blue and green), and measured the distance from anthers/stigma to the boundaries of the UV bullseye and petal after applying the background scale for calibration. We labelled the plants to enable further observations on pollinators. Based on a preliminary measurement of different flowers on multiple plants, we concluded that UV bullseye size was not significantly different in flowers from the same plant. Therefore, we only considered variations of UV bullseye size among plants.

To investigate the potential correlations between nectar reward and the size of UV bullseyes and flowers in male and female flowers, we isolated newly opened male and female flowers on different plants. On the following day from 8:00 to 10:00, we photographed the flowers to determine the UV bullseye size and flower size. Then, we collected nectar from each flower using a capillary tube with an inner diameter of 0.5 mm and measured the nectar production using a digital Vernier calliper.

2.2. Pollinator survey and foraging bouts

To assess pollinator diversity and visitation rates in H. pedunculosum, we observed male and female flowers every hour from 10:00 to 18:00 on sunny days without strong wind. We also recorded nocturnal visitors between 20:00 to 22:00. We ceased observations during rain events. In total, we observed 2896 male flowers and 469 female flowers for over 200 h across 23 days. Potential pollinators were defined as those that made contact with the flower's anther or stigma.

In the daytime, we monitored pollinator behaviors on both male and female H. pedunculosum plants on which both UV bullseye size and flower size had been identified as mentioned above. When a pollinator began to visit a flower on the labelled plant, we traced the pollinator and recorded the number of flowers on the same plants in the subsequent visitation. Since bees, including bumblebees and honeybees, were identified as the primary pollinators (see results), we focused on observing their foraging behaviors. Due to the plant's climbing growth habit and the resultant difficulty in discerning different branches on one plant when several plants grow together, we restricted observations to a single branch on each plant. The relationship between UV bullseye size and the number of visited flowers per branch were examined using linear regression.

2.3. Male reproductive fitness

To assess the impact of UV bullseye size and flower size on male reproductive fitness of H. pedunculosum, we collected the fresh leaves of 34 male plants with identified UV bullseye sizes and flower sizes, and kept them separately in silica gel. We then obtained 150 mature fruits (containing 1472 seeds in total) to quantify the number of sired seeds produced by each of the 34 male plants.

In the laboratory, genomic DNA was extracted separately from the leaves of 34 male H. pedunculosum plants using modified CTAB methods. To assess seed siring success of these male plants, seed coats (including episperm and endopleura) as well as the degenerated endosperm of 1472 seeds were removed carefully with fine forceps. The genomic DNA was then extracted from each embryo using TRNzol Universal Kit (DP-424, Tiangen Biotechnology, Beijing, China) since H. pedunculosum seeds contain substantial amounts of oil. PCR amplification was performed using 11 pairs of microsatellite markers (BLG16, BLG34, BLG37, BLG81, BLG82, BLG84, BLG90, BLG108, BLG119, BLG122, BLG126) for both male plants and seeds following methods of Chen et al. (2020b). The reaction mixture consisted of 25 μl volume, containing 20–30 ng DNA, with an initial denaturation at 94 ℃ for 4 min, followed by 35 cycles of 94 ℃ for 90 s, annealing temperature ranging from 45 ℃ to 60 ℃ for 50 s, 72 ℃ for 50 s, and a final extension at 72 ℃ for 7 min. Purified PCR products were sequenced on an ABI, and genotypes based on STR (short tandem repeats) were analyzed using GenMarker v.2.2.0 (Holland and Parson, 2011). We assessed the paternity of collected seeds in Cervus 3.0 (Kalinowski et al., 2007) by employing maximum-likelihood methods and setting the LOD score (natural log of the likelihood ratio) as default relaxed confidence (80%) (Marshall et al., 1998; Kalinowski et al., 2007). We assigned one seed family to one fruit, with 34 male plants as potential paternity. The total sired seeds by each male plant were summed, and male reproductive success of each plant was calculated by dividing all seeds of identified paternity with the sired seeds of the corresponding male plant. The relationships between the UV bullseye size and flower size of male flowers and male reproductive fitness were analyzed by regression analysis.

2.4. Female reproductive success

To assess the seed production of female flowers, we randomly selected 70 flowers on different plants, and separated them into two groups. For one group (35 flowers), we performed supplemental hand pollination. For the other group (35 flowers), we identified the UV bullseye size and flower size of each flower as mentioned above, and left them for open pollination as control. However, due to the unexpected destruction on experimental plants, the sample size was reduced to 17 for hand pollinated flowers and 18 for naturally pollinated flowers. We employed an independent T test to compare seed numbers per fruit between hand pollination and control. Additionally, a regression analysis was used to explore the relationships between seed production of naturally pollinated flowers and UV bullseye size and flower size.

3. Results 3.1. Population and floral traits

In the study population of Herpetospermum pedunculosum, we examined 347 plants in total, including 205 male plants and 142 female plants. The sex ratio of male to female was significantly male-biased (1.44:1) (χ² = 11.44, p < 0.01). The flowering sex ratio was even more male-biased (6.05:1), suggesting that the mating environment was strongly male-biased (χ² = 184.90, p < 0.001).

As the relative reflectance curves between male and female flowers showed no obvious difference, we merged the data for further analysis. Notably, we observed significant differences in relative reflectance between the center and periphery of petals, especially around 340 nm and 500 nm, characterized with two obvious reflectance peaks in the periphery, but not in the center of petals (Fig. 1c). In the bee's color hexagon, the mean Euclidean distance of 20 measured flowers between center and periphery of petals was 0.19 ± 0.02, larger than 0.10 (one sample T test, t = 4.93, p < 0.01). In addition, the mean Euclidean distance between the species centroids and the center of petals was 0.13 ± 0.01, larger than 0.10 (t = 2.75, p = 0.01). The mean Euclidean distance between the species centroids and periphery of petals was 0.17 ± 0.01, larger than 0.10 (t = 5.54, p < 0.01) (Fig. 1d). Taken together, these results suggested that bee pollinators can perceive the UV bullseye that absorbed UV light and the periphery of petals that reflected UV light.

Both bullseye size (t = 1.87, p = 0.06) and flower size (t = 10.04, p < 0.01) of male flowers were larger than those of female flowers. Specifically, UV bullseye size in male flowers ranged from 0.53 cm to 0.81 cm, and flower size ranged from 2.34 cm to 3.86 cm. For female flowers, UV bullseye size ranged from 0.49 cm to 0.80 cm, and flower size ranged from 1.50 cm to 3.44 cm. Further, a positive correlation between UV bullseye size and flower size was observed in both male and female flowers (Fig. 1e).

The nectary of a male flower develops from the degenerated pistil and produces low nectar volume (less than 1 μl), which made it difficult to examine the relationship between flower size and nectar production due to the possible measurement errors. For this reason, we abandoned further analysis. In contrast, no nectar was secreted at all in female flowers, indicating that female flowers are deceptive to pollinators.

3.2. Pollinator survey and foraging bouts

Four types of insects were observed to visit flowers of H. pedunculosum: honeybees (Apis cerana), bumblebees (Bombus festivus), butterflies, and moths. Honeybees, bumblebees and butterflies were identified as the potential pollinators, whereas moths were not observed to visit female flowers (Fig. 2a). Generally, visitation rate was significantly affected by pollinator type and flower gender, but not by their interaction (Table 1). Visitation rates were higher for honeybees and bumblebees than for butterflies (Fig. 2b). Moreover, both honeybees and bumblebees preferred male flowers over female flowers, although butterflies did not (Fig. 2c). Because of the low abundance and visitation rate of butterflies (Fig. 2a and b, Table 1), only honeybee and bumblebee were considered in subsequent pollination observations.

We monitored the foraging activities of 50 honeybees and 38 bumblebees on male plants. We found that UV bullseye size and the number of visited flowers on a single branch were positively correlated (Fig. 3a), suggesting that bees preferred to visit more male flowers on one branch when UV bullseye size was large. In addition, we observed the foraging activities of 20 honeybees and 14 bumblebees on female plants, and found no significant correlation between UV bullseye size and the number of visited flowers on one branch (Fig. 3b). The mean number of visited female flowers on one branch was 1.23 ± 0.07, suggesting a maximum of 2 flowers on a single branch of female plants.

3.3. Male and female reproductive fitness

Our analysis based on 11 microsatellites showed an average of 4.91 alleles per locus in 34 male plants and all 1472 seeds, with an expected heterozygosity of 0.30. The mean exclusionary power was 0.94. Out of the 1472 seeds, 242 seeds showed potential paternity from 25 out of the collected 34 male plants. UV bullseye size of male plants and the male reproductive success in the population were significantly, negatively correlated (Fig. 4a), suggesting that male plants with large UV bullseye size would sire fewer seeds. Similarly, a negative relationship between flower size of male plants and male reproductive success of sired seeds was also observed (Fig. 4b).

For female plants, supplemental hand pollination increased seed production significantly (Fig. 5a), suggesting pollen limitation in naturally pollinated flowers. Furthermore, we observed that flowers with moderate UV bullseye size produced more seeds than those with small or large UV bullseye size in naturally pollinated flowers (Fig. 5b). However, we did not find a significant relationship between seed number and flower size in naturally pollinated flowers (Fig. 5c).

4. Discussion

The color vision of humans differs from that of animal pollinators. In bees, which have trichromatic color vision, UV-sensitive photoreceptors enable the perception of ultraviolet light (Chittka, 1992; Peitsch et al., 1992). Consequently, flowers with contrasting UV patterns are likely to attract bee pollinators. In the present study, we found that Herpetospermum pedunculosum flowers exhibited UV patterns in the center of flowers, and examined the effect of UV bullseye size on the reproductive success of dioecious H. pedunculosum. Our results show that male plants with small UV bullseye size had a high reproductive success, whereas female plants with moderate UV bullseye size had a high reproductive success, indicating differential selection on UV bullseye size via male and female fitness.

Flower traits that attract pollinators are often assigned the role of improving pollen dispersal or male reproductive success (Stanton et al., 1986; Vaughton and Ramsey, 1998; Paterno et al., 2020). However, animal pollinators might continue visiting flowers if there are numerous flowers on a single plant, leading to geitonogamous selfing in hermaphroditic plants (Harder and Barrett, 1995; Duan et al., 2005; Zhang et al., 2017a). The issue of geitonogamous selfing and the concomitant increase in pollinator attraction represent a dilemma for hermaphrodites that depend on large flower display size via many flowers to attract animal pollinators (Klinkhamer and de Jong, 1993). In contrast, in dioecious species, pollen transfer facilitated by continuous visitation of pollinators to a single plant is often overlooked. In the case of the dioecious plant H. pedunculosum, bee pollinators are attracted by large male flowers. A contrasting UV bullseye might further enhance pollinator attraction, as large bee pollinators prefer larger UV bullseye size (Horth et al., 2014). Additionally, male flowers offer nectar rewards to pollinators, which promotes the continuous visitation on one branch of male H. pedunculosum (Fig. 3a). Moreover, the preference of bee pollinators to visit adjacent plants with large flower size (Wu et al., 2022) increases the probability of them visiting adjacent branches of male plants in H. pedunculosum. Continuous movements among flowers within a single plant of male H. pedunculosum may reduce the effectiveness of pollen export, leading to lower reproductive success of male plants with larger UV bullseye size and flower size. Although we did not assess pollen flow among male flowers on one plant, our results offer persuasive evidence that H. pedunculosum males with large UV bullseye size and flower size have reduced male siring success (Fig. 4a), suggesting a negative directional selection away from large UV bullseye size and flower size in male plants.

Pollinator-mediated selection on floral traits via female fitness has been examined widely with an aim to find traits targeted by natural selection (Sapir and Armbruster, 2010; Caruso et al., 2019). Traits associated with pollinator attraction are generally subject to positive directional selection (Harder and Johnson, 2009), with directional selection on flower size and floral display size commonly observed in species pollen-limited seed production (Sandring and Agren, 2009; Parachnowitsch and Kessler, 2010; Sletvold et al., 2010). Although all research on natural selection for flower size is based on visible light, we found that female flowers with moderate bullseye size, instead of visible flower size, produce more seeds in H. pedunculosum, indicating a pattern of stable selection. Furthermore, we did not find significant selection on flower size via female fitness in H. pedunculosum (Fig. 5c). Two possible intrinsic scenarios might explain the unexpected positive directional selection on UV bullseye size via female fitness. Firstly, the large size of UV bullseye might increase floral costs that could potentially offset the benefits of high pollen deposition due to high pollinator attraction (Cresswell, 1998; Teixido, 2014), leading to the reduced seed production of flowers with large UV bullseye size. The second possible explanation is that UV bullseye size could have some level of inheritance from the mother plant (Syafaruddin et al., 2006; Koski and Ashman, 2013), but the large number of progeny obtained from male plants with small UV bullseye size may counteract the inheritance of large UV bullseye size from the mother plants. However, we cannot presently determine the reasons for the possible stable selection targeting UV bullseye size via female fitness.

Bee pollinators can perceive the contrasting UV bullseye on male and female flowers, as it is within their color vision range (Fig. 1d), suggesting that UV pattern might function as a guide for successful pollination for this dioecious species. Although male flowers reward bees with both pollen and nectar, female flowers do not offer any such rewards, indicating that the UV bullseye of female flowers may function as a deceptive visual signal (Wilson and Ågren, 1989; Schäfer and Ruxton, 2009). Pollination by deceit is prevalent in female plants of dioecious species (Renner and Feil, 1993). Our observations on pollinators support the deceptive pollination in H. pedunculosum since bee pollinators prefer male flowers (Fig. 2b) and visit no more than two flowers on female plants (Fig. 3b). Furthermore, in H. pedunculosum, male reproductive success is higher when UV bullseye size ranges from 0.6 to 0.8 (Fig. 4a), which overlaps with the UV bullseye size (0.6–0.7) linked with greater female reproductive success (Fig. 5b). Selection for similarity between male and rewardless female flowers is expected when the reproductive success of rewardless female flowers is dependent on their degree of resemblance to rewarding male flowers (Ågren and Schemske, 1991; Hossaert-McKey et al., 2016). This is fully supported by our results on H. pedunculosum.

In conclusion, although several studies have examined the occurrence of UV patterns in flower plants (Papiorek et al., 2016; Trunschke et al., 2021; Tunes et al., 2021), we have investigated the impact of UV bullseye size on male and female reproductive fitness for the first time in a dioecious vine. The contrasting UV bullseye may act as an honest signpost for dominant bee pollinators on rewarding male flowers, whereas it can function as a deceptive signal for bees visiting rewardless female flowers. The continuous visitation of bees on a single plant with large UV bullseye size results in decreased male fitness, whereas floral costs, conflicting inheritance and selection for the similarity to visual signal of male flowers might have led to the high seed set of female flowers with moderate UV bullseye size. As UV bullseye size was correlated to pollinator visitation and reproductive success much more than flower size in H. pedunculosum, our research underscores the significance of studying flower evolution from the perspectives of pollinators in future studies.

Acknowledgements

We are grateful to Dr. Zhe Chen and Dr. Yang Niu for their help in data analysis and the staff in Shangri-La Alpine Botanical Garden for their kind support in the field. Financial support for this research was provided by National Natural Science Foundation of China (32160261), the Second Tibetan Plateau Scientific Expedition and Research (STEP) program (2019QZKK0502), Science and Technology Program of Xizang Autonomous Region (XZ202001YD0008C).

Availability of data and materials

All data generated during this study are included in this article as supplementary excel file.

Authors' contributions

Y.-W.D., L.-H.M. and Y.-P.Y. designed the research; Y.-W.D. and L.-H.M. wrote the manuscript; J.-F.W., X.-L.W., Y.-L.T., L.-H.M. and Y.-W.D. performed field experiments; Z.-Q.C. performed laboratory experiments; Y.-W.D., J.-F.W. and Z.-Q.C. analyzed data and prepared figures and tables.

Data sharing/archiving

The data that support the findings of this study are in supplemental material.

Declaration of competing interest

The authors declare no competing financial interests.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.pld.2023.06.004.