Sexual and temporal variations in floral scent in the subdioecious shrub Eurya japonica Thunb

Abstract In many flowering plants, floral scents are a significant trait for visitors, playing an important role in attracting pollinators and/or detracting herbivores. The evolution of flowering plants from hermaphroditism to dioecy is often accompanied by sexual dimorphism in floral scent. In this study, floral scents emitted by different sexual morphs of the subdioecious shrub Eurya japonica Thunb. were collected using a dynamic headspace method, and sexual and temporal variations were evaluated by gas chromatography–mass spectrometry (GC–MS). Two volatiles, α‐pinene and linalool, were identified as the major components of floral scents in females, hermaphrodites, and males. The males emit higher amounts of floral scents, particularly α‐pinene, compared to females or hermaphrodites. Floral scents emitted by males generally decrease as flowers enter senescence, whereas those from females or hermaphrodites do not significantly differ. Intraspecific variations in floral scents of subdioecious species provided by this study would contribute to better understanding of sexual dimorphism in floral scent.

The majority of previous investigations have focused on dioecious species and compared the differences in floral scents between females and males (Dötterl et al., 2014;Dufa, Hossaert-McKey, & Anstett, 2004;Milet-Pinheiro et al., 2015;Tollsten & Knudsen, 1992;Tsuji & Sota, 2010). Ashman, Bradburn, Cole, Blaney, and Raguso (2005) investigated emission rates and floral scent composition in a gynodioecious plant (Fragaria virginiana) in females and hermaphrodites. In this study, floral scent differences in a subdioecious sexual system (females, hermaphrodites, and males) were investigated. Resource reallocation favors the evolution of sexual dimorphism (Charlesworth, 1999). The sexual selection theory (Bateman, 1948;Dötterl et al., 2014) predicts that "males are limited in their reproductive success by access to mates, whereas females are more limited by resources" (Waelti, Page, Widmer, & Schiestl, 2009). Thus, males in the majority of species emit more scents per flower than females to attract visitors (reviewed in Ashman, 2009). Therefore, it is of much interest to investigate whether the profile and the temporal pattern of floral scents of hermaphrodites are similar to that in males or females in subdioecious species. This will improve our understanding of the patterns of sexual dimorphism in floral scent.
To better understand the patterns of sexual dimorphism in floral scent, this study aimed to i) chemically characterize its floral scents, and to assess ii) qualitative and iii) temporal variations in floral scents in female, hermaphrodite, and male flowers.

| Flower scents collection
Scents were collected in the laboratory with a pump extractor using a dynamic headspace method (Dötterl & Jürgens, 2005; Figure 1). To maintain freshness, the flower or leaf branches were harvested and immersed in water at room temperature (shown in Figure 1), immediately followed by scent collection. The branches were enclosed in a polyethylene oven bag (340 mm × 240 mm), and the emitted volatiles were trapped in an adsorbent tube using a membrane pump (SIBATA, Inc., Akashi, Japan). The flow rate was adjusted to 200 ml/min using a flow meter. Samples were collected for 1 hr. An adsorbent tube was constructed using a PTFE tube (Φ3 × 5 mm, 100 mm) that was filled with 60 mg of Tenax-TA (60-80 mesh). The adsorbents were fixed in the tubes F I G U R E 1 Schematic diagram of floral scent collection device using dynamic headspace method using glass wool. Room air was simultaneously collected and used as control.
The volatiles trapped in the adsorbent tube were dissolved and washed with diethyl ether (5 ml × 3) and collected into a test tube. Docosane (0.1 g/L, 0.1 ml) was used as internal standard. The collected liquid was concentrated to approximately 1.5 ml by N 2 and stored at 4°C.
GC-MS-operating conditions were as follows: injector temperature, 250°C; oven temperature program, 35°C held for 5 min, 35°C→180°C (5°C/min), then 180°C→200°C (10°C/min), and then held for 10 min, and finally to 280°C (20°C/min) and then held for 5 min; carrier gas, He; flow rate, 1.6 ml/min; interface temperature, 250°C; and ion source temperature, 200°C. The quantity of each volatile compound was calculated by comparing the GC data with the internal standards.
F I G U R E 2 Total ion chromatogram (TIC) of the collected volatiles from room air (a), leaf branch (b), and flower branch (c). The chemical structures of peaks 1 (α-pinene) and 2 (linalool) are shown. I.S., internal standard (docosane) F I G U R E 3 Amounts (μg·hr −1 ·branch −1 ) of major volatiles emitted by flower and leaf branches of different sexual morphs at the entire flowering stages. Maximum and minimum values for each sample are shown at the upper and lower ends of the vertical bars, respectively. The 75% and 25% points are given by the upper and lower ends of the box, respectively. The middle bar indicates the median (female: n = 21, 3 stages × 7 individuals; hermaphrodite: n = 18, 3 × 6; male: n = 21, 3 × 7). Different letters beside the bars indicate significant differences in the results of multiple comparisons in which family-wise errors were adjusted using Tukey's method at p = 0.05

| Statistical analysis
The Kruskal-Wallis test was used to assess differences in the amount of each volatile between the flower branch and leaf branch of females, hermaphrodites, and males. The generalized linear mixed models (GLMMs) were used to examine the effects of sexual morphs (female, hermaphrodite, and male) and flowering stages (stage 1, 2, and 3) on the amount of floral scents (Bolker et al., 2009). In the models, the sexual morphs, flowering stages, and their interactions were set as fixed effects, and target individuals were set as random effects. To assess the statistical significance of each fixed factor, the changes in deviance when each factor was removed from the full model were compared with the F-test for Gaussian error distributions with Identity link functions (Bolker et al., 2009) Figure 2 shows the total ion chromatograms (TICs) of the collected volatiles and room air (control). The peak assignments of the main compounds were based on the mass spectral data of previous studies (Adams, 2007;Motooka et al., 2015)  (n = 21, 3 stages × 7 individuals), 86.9% ± 12.0% (n = 18, 3 × 6), and 94.3% ± 8.2% (n = 21, 3 × 7) in females, hermaphrodites, and males, respectively.

| Sexual variations in floral scent
The amount of floral scents (α-pinene and linalool) was significantly affected by the sexual morphs (female, hermaphrodite, and male), flowering stages (stage 1, 2, and 3), and their interaction (Table 1).
At stage 1 (Figure 4a), males emit significantly higher amounts of α-pinene (μg·hr −1 ·flower −1 ) compared to females or hermaphrodites in post hoc comparisons. In addition, males emitted marginally TA B L E 1 Effect of sexual morphs (female, hermaphrodite, and male), flowering stages (stage 1, 2, and 3), and their interactions on the amount of α-pinene and linalool. To test the statistical significance of explanatory variables, the changes in deviance when each variable was removed from the full model were compared with F distributions for Gaussian distributions. Boldface indicates statistical significance F I G U R E 4 Amounts (μg hr −1 flower −1 ) of main volatiles of floral scents emitted by different sexual morphs or flowering stages (female: n = 7; hermaphrodite: n = 6; male: n = 7). Different letters beside the bars indicate significant differences in the results of multiple comparisons in which family-wise errors were adjusted using Tukey's method at p = 0.05. F, female; H, hermaphrodite; M, male; S1, stage 1; S2, stage 2; and S3, stage 3 higher amounts of linalool (0.13 ± 0.09) than females (0.05 ± 0.02) or hermaphrodites (0.05 ± 0.02), although this difference was not statistically significant (Figure 4b). However, no significant difference in the amount of α-pinene/linalool between females and hermaphrodites was observed. A similar situation was observed at stage 2 ( Figure 4).

| Identification of major floral scents
Floral scents are generally composed of dozens, even hundreds of volatile chemicals (Knudsen et al., 1993;Miyazawa et al., 2016;Motooka et al., 2015;Tollsten & Knudsen, 1992). Motooka et al. (2015) and Miyazawa et al. (2016) prepared extracts from the flowers, vegetative parts, or flower buds of E. japonica in organic solvents, which were then analyzed using GC-MS. More than 50 compounds were detected in the essential oils. In the present study, only four major volatiles were detected. Differences in the results may be attributable to the use of various methods in these studies.
Under consideration of the results of the control (room air) and leaf branches (Figures 2 and 3

| Sexual variations in floral scent
The evolution from hermaphroditism to dioecy is coupled to sexual dimorphism in floral scent (Ashman, 2009). In this study, the floral scents of different sexual morphs show only quantitative (quantity of volatile compounds) differences and no qualitative (blend composition) differences (Figures 2-4). These findings indicate that the observed differences in floral scents between sexual morphs cannot be explained by the emission of additional pollen-or stigma-specific compounds in flowers (Ashman et al., 2005;Mayo, Bogner, & Boyce, 1997;Miyake & Yafuso, 2003;Vogel, 1990).
In this study, the production of characteristic compounds of floral scents (particularly α-pinene) in males is generally higher than that in females or hermaphrodites (Figure 4). Previous researches have reported similar results in some dioecious species (Dötterl & Jürgens, 2005;Dötterl et al., 2014;Waelti et al., 2009). Sexual selection theory predicts differential resource investment among different sexual morphs to attract pollinators (Bateman, 1948;Dötterl et al., 2014;Waelti et al., 2009). Accordingly, males should thus be selected to invest more resources in floral scents than females or hermaphrodites to enhance pollination success in subdioecious species, E. japonica.

ACK N OWLED G M ENTS
The and LetPub for linguistic assistance in preparing our manuscript.

CO N FLI C T O F I NTE R E S T
None declared.

AUTH O R CO NTR I B UTI O N S
HW, PZ, TM, and MN conceptualized and designed the study; HW and PZ conducted floral scent collection; HW, PZ, DA, SY, YM, and KF performed chemical analysis; and PZ, HW, TM, and MN wrote the manuscript. All authors have reviewed and approved the final manuscript.

DATA ACCE SS I B I LIT Y
Data are available from the Dryad Digital Repository.