2021 Volume 86 Issue 4 Pages 329-338
The evolution of dioecy from hermaphroditism allows for avoidance of self-pollination, and its evolutionary background has been investigated both experimentally and theoretically since it was first proposed by Darwin. To reproduce this evolution, we screened hermaphroditic mutants of Silene latifolia using heavy-ion beam or γ-ray irradiation and characterized the phenotypes of their floral organs. Our scatterplots indicate severe deviations from the trade-off relationships between pollen and ovule numbers and between seed and germinated pollen numbers in hermaphroditic mutant S. latifolia. These deviations presumably led to promotion of dioecy from the ancestral state of S. latifolia. To infer the likely flower phenotypic characteristics of the ancestral plant of S. latifolia before evolving dioecy, the flowers of Silene viscosa, a naturally hermaphroditic plant related to S. latifolia were also characterized. S. viscosa exhibits both spatial separation of stamens from pistils within the flower (reverse herkogamy) and temporal separation of stamen and pistil maturation (dichogamy), raising the question of whether hermaphroditic mutant S. latifolia, which is thought to be the ancestral state, would possess these functions. We show that two hermaphroditic mutants of the dioecious plant S. latifolia exhibit signs of protogyny (reverse dichogamy) and approach herkogamy, as pistils were constantly longer than stamens. These findings illustrate the evolution of dioecy from hermaphroditism as a self-pollination avoidance mechanism and to balance the investments into male and female functions.
The evolution of dioecious plants from hermaphroditic ancestors was first explained by Darwin (Darwin 1877), and subsequent theoretical and empirical studies have demonstrated that dioecy in plants such as Silene latifolia evolved from the hermaphroditic condition via gynodioecy (involving male-sterility mutations), followed by creation of males via female sterility factors. Most hermaphroditic plants have various mechanisms to avoid self-pollination, including herkogamy, dichogamy, and self-incompatibility (Lloyd and Webb 1986, Webb and Lloyd 1986, Bertin 1993, Barrett 2003, Takayama and Isogai 2005, Brys et al. 2013, Torang et al. 2017). These widespread features raise the question of whether nearly ubiquitous mechanisms for avoiding self-pollination evolved in hermaphroditic plants before the development of dioecy.
Sex in the dioecious plant S. latifolia is determined by X and Y sex chromosomes, and consistent with the hypothesis outlined above, males showing loss of female sterility (hermaphrodites carrying a Y chromosome) occasionally arise in S. latifolia. (Desfeux et al. 1996, Charlesworth 2006, Marais et al. 2011). Based on research into deletions of parts of the Y chromosome, the flower characteristics of males (XY) have been shown to depend on three factors on the Y chromosome (Westergaard 1958); the gynoecium is suppressed by one or more gynoecium-suppressing factors (GSFs), deletion of which leads to hermaphroditic mutants; the stamens are promoted by a stamen-promoting factor (SPF); and anther and pollen grain maturation is promoted by a male fertility factor (MFF) (Lardon et al. 1999, Lebel-Hardenack et al. 2002). Hermaphroditic mutants of S. latifolia are not self-incompatible; they have both fertile stamens and fertile pistils and can form seeds through self-pollination (Lardon et al. 1999, Miller and Kesseli 2011). In S. latifolia, trade-offs between male and female functions have been inferred from the stamen and pistil sizes of hermaphroditic mutants (Fujita et al. 2012).
A second aim of the present study was to test whether other flower characteristics evolved due to changes in sex chromosome genes. To this end, we used S. viscosa, a hermaphroditic plant closely related to S. latifolia, to infer the likely flower phenotypic characteristics of the ancestral plant before the evolution of dioecy in S. latifolia. S. viscosa exhibits both spatial separation of stamens from pistils within the flower (herkogamy) and temporal separation of stamen and pistil maturation (dichogamy), as observed in some other hermaphroditic and gynodioecious plants in the genus Silene (Kephart et al. 1999, Buide and Guitian 2002, Davis and Delph 2005). Whether the mechanisms for avoiding self-pollination that are present in hermaphroditic plants of the genus Silene have been lost in dioecious S. latifolia (where synchronized male and female functions would maximize fitness) is an interesting question, as is whether the evolutionary changes involved occurred due to changes in genes on the sex chromosomes. One method to test this possibility is to determine whether the same mechanisms re-appear in hermaphrodite S. latifolia Y-chromosome deletion mutants, which would suggest that the Y chromosome evolved a factor that controls the altered phenotype.
In the present study, we aimed to quantify the trade-off relationships between male and female functions by counting pollen grains, ovules, and seeds in hermaphroditic mutants of S. latifolia. The floral characters of hermaphroditic mutants, wild-type (WT) S. latifolia, and WT S. viscosa were compered to reproduce the evolutionary progression from hermaphroditism to dioecy.
S. latifolia seeds were obtained from an inbred line (K-line) in our laboratory (Kazama et al. 2003). The K-line was propagated for 16 generations through brother-sister mating to obtain a genetically homogeneous population. S. viscosa seeds were obtained from Professor Boris Vyskot (Academy of Sciences of the Czech Republic). Plants were grown from seeds in pots within a growth chamber at 23°C with a 16-h light/8-h dark cycle.
Screening for hermaphrodite mutants obtained through heavy-ion beam or γ-ray irradiationAnthers were collected from male flowers of S. latifolia and irradiated with a heavy-ion beam or γ-rays. Female flowers were pollinated with irradiated pollen grains and the resulting seeds were sown to obtain M1 seedlings. Both partial Y-chromosome deletion mutants and mutants with unusual phenotypes were screened exhaustively through polymerase chain reaction (PCR) and flower phenotype screening.
Pollen grain, ovule, and seed numbersFor determination of pollen grain and ovule numbers, mature buds were used just before flower opening. To determine the pollen grain number, anthers were opened under a stereomicroscope and half of the pollen grains contained in the anther were spread onto agar-covered microscope slides. The number of pollen grains was counted on all ten stamens of each flower under a fluorescence microscope (BX60; Olympus, Tokyo, Japan). To determine the ovule number, the ovary was opened, and ovules were counted at 40× magnification under a stereomicroscope. Each hermaphroditic mutant was emasculated and pollinated with WT male pollen. Then, the numbers of seeds in the resulting capsules were counted. The number of germinated pollen grains was calculated as the number of pollen grains multiplied by the pollen germination rate.
Stereomicroscopic observationsFlowers, anthers, ovaries, and capsules were observed under a stereomicroscope (MZ16; Leica Imaging Systems, Cambridge, UK).
Measurement of floral organ lengthTo determine the length of the floral organ, the calyx was partially dissected, and the inner floral organs were measured without removing the flower. The lengths of the pistil, stamen, petal, calyx, ovary (longitudinal), ovary (horizontal), and gynophore were measured in each hermaphroditic mutant every 6 h using a ruler.
Pollen viability and in vitro germination determinationTo determine pollen viability, double staining with fluorescein diacetate (FDA) and propidium iodide (PI) was performed (Mandaokar and Browse 2009). A stock solution of 2 mg mL−1 FDA was prepared in acetone and added dropwise to 17% sucrose (Suc; w/v) until the solution became cloudy. PI was diluted to 1 mg mL−1 in water and then to 100 µL mL−1 in 17% Suc (w/v). Equal amounts of these FDA and PI solutions were mixed together and added to freshly isolated pollen on glass slides. The pollen was covered with a coverslip and viewed under a fluorescence microscope. A pollen germination assay was also performed (Jolivet and Bernasconi 2007). Pollen was isolated from mature flowers by gently releasing it from the anther locules onto microscope slides coated with pollen germination medium (Aonuma et al. 2013). One hundred grains were examined from each of the three anthers and their average value was used as the measurement for the plant.
In vivo pollen germination and tube growthStigma staining was performed with aniline blue (Fang et al. 2010). Previously pollinated pistils were excised from flowers 6 h after pollination and fixed for 24 h in acetic alcohol (glacial acetic acid: ethanol, 1 : 3, v/v), then cleared with 8 M NaOH overnight and rinsed thoroughly before staining with decolored 0.1% (w/v) aniline blue in 0.1 M K3PO4. The preparation was conducted under a fluorescence microscope for measurement of pollen germination and pollen tube growth down the style. Pollen grains were scored as germinated when pollen tubes were apparent between the papillate cells of the stigma, and the number of pollen tubes, the number that reached the end of the style, and the number that barely elongated were recorded.
The dioecious plant S. latifolia has male and female flowers, with stamens and pistils located on separate plants, whereas the hermaphroditic plant S. viscosa has hermaphroditic flowers containing stamens and pistils on the same plant (Fig. 1a–c). S. viscosa exhibits both herkogamy and dichogamy; the pistils extend straight, while the stamens hang down vertically (herkogamy, Fig. 1c). To study the time course of flower development, we previously divided 10 stamens into two groups based on their lengths, characterizing the five long stamens as early-maturing stamens (Em-stamens) and the five short stamens as late-maturing stamens (Lm-stamens) (Aonuma et al. 2013). The lengths of eight floral organs were measured every 6 h (Fig. 1d–f, Fig. S1). In S. latifolia, Em-stamens and pistils extended above the petals at 6 h after flowering (haf) (Fig. 1d, e), while Lm-stamens extended above the petals at 30 haf.
In S. viscosa, both Em- and Lm-stamens extended above the petals 6 h before flowering, while pistils extended above the petals at 30 haf (Fig. 1f). Although the stamen and pistil lengths of S. latifolia synchronously reached petal length after flowering, the pistil length of S. viscosa equaled petal length more than 1 day later than its stamen length. In S. viscosa, both Em-stamens and Lm-stamens were consistently longer than the pistils throughout the flowering period. In addition, the stamens hung down vertically from the petals, likely to maintain physical distance from the pistils (Fig. 1c), indicating the occurrence of herkogamy in S. viscosa. Herkogamy has been divided into two types: approach herkogamy, in which the pistils are located above the stamens and therefore contact pollinators first upon entry into flowers, and reverse herkogamy, in which the stamens are located above the pistils (Barrett 2003). S. viscosa was classified as showing reverse herkogamy.
The maturation states of the stamens and pistils were measured every 6 and 12 h, respectively, based on in vitro pollen germination rates on pollen medium and in vivo observations of stigmas (Fig. 1g–i). The maturation periods of Em-stamens and Lm-stamens in S. latifolia male flowers were 6 h before flowering to 24 haf and 18 to 48 haf, respectively (Fig. 1g). The maturation period of pistils in S. latifolia female flowers was 6 h before flowering to 96 haf (Fig. 1h). Although the maturation periods of Em-stamens and Lm-stamens were the same in S. viscosa as in S. latifolia male flowers, the maturation period of pistils in S. viscosa was delayed, occurring at 42 to 138 haf (Fig. 1i), which was long after stamen maturation. Thus, S. viscosa is dichogamous and protandrous. These findings raised the question of whether the hermaphroditic mutant of S. latifolia, which is thought to represent the ancestral state, also possessed these features.
Phenotypes of partial Y-deletion mutants used in this studyIn total, 38 mutants with partial deletions of the Y chromosome were isolated through screening of M1 generation plants generated with C-ion beam, Fe-ion beam, or γ-ray irradiation (Table 1, Kazama et al. 2016, Krasovec et al. 2019). All mutants used in this study are presented in order of stamen development from left to right in Fig. 2. The hermaphroditic mutants (EGP: early-stage gynoecium promoted mutant) formed 10 complete stamens and a gynoecium composed of five fused carpels with five styles. EGP5, EGP6, and EGP7 had petals so small that the corolla did not open fully (Fig. 2a). Asexual mutants (ESS: early-stage stamen-suppressed mutant) formed neither a gynoecium nor stamens. The corolla of ESS3 did not open fully (Fig. 2b). The asexual flowers are arranged in order of the degree of petal development from left to right. The anther defect mutants (ISS: intermediate-stage stamen-suppressed mutant) had stamens that did not elongate fully. The ISS mutants are arranged in order of the degree of stamen development from left to right. ISS9 had stamens in which the filaments were very short, and a gynoecium composed of two fused carpels with two styles (Fig. 2c). The pollen defect mutants (LSS: late-stage stamen-suppressed mutant) had sterile stamens. The LSS mutants are arranged in order of the degree of gynophore development from left to right (Fig. 2d). The XY female mutant (GPSS: gynoecium promoted and stamen-suppressed mutant) did not form stamens but formed a gynoecium composed of five fused carpels with five styles. Although the XY female mutant had a Y chromosome, it had the same phenotype as the WT female flower (Fig. 2e). WT male and female flowers of S. latifolia are shown in Fig. 2f.
| Genotype | Phenotype | Mutagen Source |
|---|---|---|
| R025 | Hermaphroditic | C-ion 100–120 Gy |
| EGP4 | Hermaphroditic | C-ion 20 Gy |
| EGP5 | Hermaphroditic | γ 40 Gy |
| EGP6 | Hermaphroditic | C-ion 20 Gy |
| EGP7 | Hermaphroditic | γ 40 Gy |
| EGP8 | Hermaphroditic | C-ion 20 Gy |
| EGP9 | Hermaphroditic | C-ion 20 Gy |
| EGP10 | Hermaphroditic | C-ion 10 Gy |
| EGP11 | Hermaphroditic | C-ion 20 Gy |
| EGP12 | Hermaphroditic | C-ion 20 Gy |
| EGP13 | Hermaphroditic | C-ion 20 Gy |
| EGP14 | Hermaphroditic | C-ion 20 Gy |
| EGP15 | Hermaphroditic | γ 20 Gy |
| K034 | Asexual | n/a |
| ESS1 | Asexual | C-ion 20 Gy |
| ESS2 | Asexual | C-ion 20 Gy |
| ESS3 | Asexual | C-ion 40 Gy |
| ESS4 | Asexual | C-ion 20 Gy |
| ESS5 | Asexual | C-ion 20 Gy |
| ESS6 | Asexual | C-ion 100 Gy |
| ESS7 | Asexual | C-ion 20 Gy |
| ESS8 | Asexual | C-ion 20 Gy |
| ESS9 | Asexual | C-ion 20 Gy |
| ISS1 | Anther defectiveness | C-ion 100 Gy |
| ISS3 | Anther defectiveness | C-ion 100 Gy |
| ISS5 | Anther defectiveness | C-ion 20 Gy |
| ISS6 | Anther defectiveness | C-ion 20 Gy |
| ISS7 | Anther defectiveness | C-ion 20 Gy |
| ISS8 | Anther defectiveness | C-ion 20 Gy |
| ISS9 | Anther defectiveness | C-ion 20 Gy |
| ISS10 | Anther defectiveness | C-ion 20 Gy |
| ISS11 | Anther defectiveness | C-ion 120 Gy |
| LSS1 | Pollen sterility | γ 40 Gy |
| LSS2 | Pollen sterility | C-ion 20 Gy |
| LSS3 | Pollen sterility | C-ion 100 Gy |
| LSS4 | Pollen sterility | C-ion 20 Gy |
| LSS5 | Pollen sterility | Fe-ion 4 Gy |
| GPSS1 | XY Female | γ 20 Gy |
The hermaphroditic mutants with both complete stamens and pistils enabled quantitative investigation into the possibility of a trade-off between stamens and pistils in the hermaphroditic state of S. latifolia. The numbers of pollen grains and ovules in nine hermaphroditic mutants (R025, EGP4, EGP9, EGP10, EGP11, EGP12, EGP13, EGP14, and EGP15) and one XY female mutant (GPSS1) were counted. The hermaphroditic mutants are arranged in order of increasing number of ovules from left to right in Fig. 3a. The numbers of pollen grains and seeds in hermaphroditic mutants are listed in Table 2. Ovule numbers in these hermaphroditic mutants ranged from 129 in EGP9 to 334 in EGP13 and were all lower than the number in the WT female (367). The hermaphroditic mutants could not produce as many seeds as the WT female plants, and EGP11 produced a maximum of only 69 seeds. EGP14 did not produce any seeds, as its flower withered and dropped after pollination (Fig. 3c). The pollen grain numbers in hermaphroditic mutants ranged from 910 in EGP13 to 2193 in EGP9 (Fig. 3d). The pollen grain number in WT males was 1735, which was lower than in EGP9, EGP10, EGP12, and EGP15.
| EGP9 | EGP12 | EGP14 | EGP4 | R025 | EGP15 | EGP10 | EGP11 | EGP13 | GPSS1 | WT♂ | WT♀ | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Ovule | 129 | 144 | 150 | 155 | 210 | 248 | 270 | 290 | 334 | 277 | 0 | 367 |
| Seed | 15 | 12 | 0 | 28 | 57 | 19 | 22 | 69 | 63 | 111 | 0 | 203 |
| Pollen | 2193 | 1803 | 1271 | 1534 | 1654 | 2149 | 1944 | 1366 | 910 | 0 | 1735 | 0 |
| Pollen germination rate | 0.4 | 0.16 | 0.6 | 0.48 | 0.2 | 0.47 | 0.17 | 0.03 | 0 | 0 | 0.7 | 0 |
| Germinated pollen | 877 | 288 | 763 | 736 | 331 | 1010 | 330 | 41 | 0 | 0 | 1215 | 0 |
A pollen–ovule scatterplot was drawn to investigate resource allocation to male versus female functions in hermaphroditic mutants (Fig. 4a). With the exception of GPSS1, all of the mutants generally produced large numbers of pollen grains and ovules and were located to the upper right of a straight line connecting the WT male and WT female in the pollen–ovule scatterplot (Fig. 4a). Determining whether the pollen grains and ovules produced are fertile is also important. A scatterplot of germinated pollen grains and seeds was drawn to investigate the fitness allocation to male versus female functions in hermaphroditic mutants (Fig. 4b). Pollen germination rates were lower in all hermaphroditic mutants than in WT male flowers (70%) (Fig. S2) (Aonuma et al. 2013). The number of germinated pollen grains was determined as the number of pollen grains multiplied by the pollen germination rate (Table 2). All of the hermaphroditic mutants were located to the lower left of a straight line connecting WT males with WT females in the germinated pollen grain–seed scatterplot (Fig. 4b). The hermaphroditic mutants exhibited poor ability to produce germinated pollen grains and seeds; therefore, the fitness of the WT male and WT female appear higher than the fitness of any the hermaphroditic mutants. In addition, a trade-off relationship was revealed between germinated pollen and seeds in the scatterplot.
In terms of temporal development, the S. latifolia hermaphroditic mutants differed from S. viscosa, as the pistils of the mutants grew prior to the stamens, rather than pistil growth being delayed until after stamen growth. These results are shown in Fig. 1g–i and Fig. 5. In all hermaphroditic mutants, the Em- and Lm-stamens matured at 6 haf and 30 haf, respectively, coinciding with the maturation of pistils, as occurs in WT S. latifolia. These findings suggest that the Y-chromosome deletions did not cause the loss of any factors controlling the relative timing of stamen and pistil maturation, which differs greatly from S. viscosa. In all hermaphroditic mutants of S. latifolia, the pistils matured synchronously with the Em- and Lm-stamens, rather maturing after the stamens withered, as occurs in S. viscosa (Fig. 1i). In two mutants, R025 and EGP15, pistil maturation began very early, at −18 haf, which is 12 h earlier than in the other mutants or WT females (Fig. 5e, f).
To investigate whether the flowers of any S. latifolia hermaphroditic mutants exhibit herkogamy, as observed in S. viscosa, the lengths of eight floral organs were measured every 6 h during the period from −18 to 96 haf (Fig. 6). The lengths of each floral organ are defined in Fig. S1. After flowering, the pistil length, Em-stamen length, and Lm-stamen length exceeded petal length in that order. The pistils were longer than stamens in most of the hermaphroditic mutants, with differences of 2–10 mm between the two organs.
By contrast, in four hermaphroditic mutants (EGP9, EGP14, EGP4, and EGP10), the stamen lengths reached the pistil lengths during the flowering period (indicated by black arrows in Fig. 6a, c, d, g). Although the stamen lengths were temporarily equal to the pistil lengths in these mutants, the pistils again exceeded the stamens in length later (Fig. 6a, d, g). The Lm-stamen lengths did not reach the pistil lengths in any of the hermaphroditic mutants. Outlined arrows in Fig. 6 indicate the time points when stamen lengths most closely approached pistil lengths after flowering. In five hermaphroditic mutants (EGP12, R025, EGP15, EGP11, and EGP13), the pistils were longer than the stamens throughout the flowering period, that is, the plants approached herkogamy (Fig. 6b, e, f, h, i). This finding differs from S. viscosa, which exhibits reverse herkogamy.
In the current study, floral characteristics were compared among hermaphroditic mutants, WT S. latifolia, and WT S. viscosa (Fig. 7). The hermaphroditic mutants of S. latifolia exhibited synchronous maturation of both stamens and pistils shortly after flowering, as occurs in WT S. latifolia, rather than delayed maturation of the pistils relative to the stamens by more than 40 h, as occurs in S. viscosa (Fig. 7). Furthermore, the hermaphroditic mutants tended to show accelerated maturation of the pistils, as observed in R025 and EGP15. This appearance of protogyny implies that the Y chromosome has lost a factor or factors that delay pistil maturation during evolution from an ancestral state similar to S. viscosa.
While reverse herkogamy was observed in S. viscosa, five of the nine hermaphroditic mutants of S. latifolia showed approach herkogamy. Herkogamy is nearly ubiquitous in hermaphroditic plants (Webb and Lloyd 1986, Barrett 2003). Stamen length equaled pistil length during the period of 0–12 haf in four hermaphroditic mutants. In all hermaphroditic mutants, the Em-stamen began to mature at −6 haf, and the stamen length reached the pistil length after the beginning of stamen maturation (Fig. 7). In most hermaphroditic plants, stamen–pistil contact occurs during the later period of flowering (Kephart et al. 1999, Duan et al. 2010, de Vos et al. 2012, Jia and Tan 2012, Luo and Widmer 2013). Delaying stamen–pistil contact has the advantage of increasing the possibility of cross-pollination and assuring reproduction (Lloyd and Schoen 1992, Fenster and Martén-Rodríguez 2007, Ruan and Teixeira da Silva 2012). The characteristic of delayed autonomous selfing has been reported in Silene (S. acutifolia, S. noctiflora, and S. douglasii) (Kephart et al. 1999, Buide and Guitian 2002, Davis and Delph 2005). In S. latifolia, the timing of pollen arrival significantly affects siring success among competing pollen donors, and paternity success of the second-arriving pollen was significantly below 50% after a 2-h delay (Burkhardt et al. 2009). The hermaphroditic mutants of S. latifolia appear to have ample likelihood of outcrossing due to the delay in stamen–pistil contact.
Dioecy may be favored over hermaphroditism if the trade-off in the investment of resources to male and female functions results in increased fitness, or return on investment, in a single sexual function (Charlesworth and Charlesworth 1981, de Laguerie et al. 1991, Muyle et al. 2021). The dioecious plant S. latifolia is assumed to have evolved from the hermaphroditic condition via gynodioecy (Delph and Wolf 2005). A model for the evolution of dioecy and gynodioecy suggests that the occurrence of ovule production modifiers reduces ovule production in hermaphrodites, and can also increase pollen production based on the compensation principle (Charlesworth and Charlesworth 1978). Several fieldwork-based studies have attempted to quantify the allocation of resources to male and female fitness, and phenotypic trade-off relationships between stamens and pistils have been found in hermaphrodites of some species (Davis 2002, Ashman 2003, Rosas and Dominguez 2009). Our scatterplots showed that the trade-off relationship is stronger between germinated pollen grain and seed numbers than between pollen grain and ovule numbers in hermaphroditic mutants (Fig. 4a, b). The shape of the investment curve predicted from the pollen–ovule scatterplot in the hermaphroditic mutants was slightly convex, as the mutants were located to the upper right of a straight-line connecting WT males and WT females (Fig. 4a). This tendency indicates that the capacities for both ovule and pollen grain production were not severely limited, and the trade-off between the numbers of ovules and pollen grains was weakened. By contrast, the shape of the fitness curve predicted from the scatterplot of germinated pollen seeds in hermaphroditic mutants was slightly concave, as the mutants were located to the lower left of a straight-line connecting WT males and WT females (Fig. 4b), indicating that the production capacities of mature pollens and seeds were strictly limited by the energy investment in the hermaphroditic state. In other words, excess investment in the hermaphroditic mutants did not result in greater productivity than in WT male and female plants. Therefore, our results indicate that dioecy will be favored in populations consisting of both dioecious and hermaphroditic S. latifolia.
The results of the present study, which are based on the numbers of pollen grains, ovules, and seeds in radiation-induced hermaphroditic mutants of a dioecious plant, provide guidance for the design of similar manipulative experiments in other plants and contribute to the emerging consensus that floral traits influence several aspects of plant mating, extending the traditional research focus on patterns of self-and cross-fertilization.
With two Supplementary Figures (Figs. S1, S2).
We thank Dr. Boris Vyskot for his generous gift of S. viscosa seeds. This work was supported by a Grant-in-Aid from the Japan Society for the Promotion of Science (JSPS) Fellows (256876) and by a Grant-in-Aid for Exploratory Research (to SK 24657046) from the Ministry of Education, Culture, Sports, Science and Technology, Japan.
