How to be a dioecious fig: Chemical mimicry between sexes matters only when both sexes flower synchronously

In nursery pollination mutualisms, which are usually obligate interactions, olfactory attraction of pollinators by floral volatile organic compounds (VOCs) is the main step in guaranteeing partner encounter. However, mechanisms ensuring the evolutionary stability of dioecious fig–pollinator mutualisms, in which female fig trees engage in pollination by deceit resulting in zero reproductive success of pollinators that visit them, are poorly understood. In dioecious figs, individuals of each sex should be selected to produce odours that their pollinating wasps cannot distinguish, especially since pollinators have usually only one choice of a nursery during their lifetime. To test the hypothesis of intersexual chemical mimicry, VOCs emitted by pollen-receptive figs of seven dioecious species were compared using headspace collection and gas chromatography-mass spectrometry analysis. First, fig-flower scents varied significantly among species, allowing host-species recognition. Second, in species in which male and female figs are synchronous, intersexual VOC variation was not significant. However, in species where figs of both sexes flower asynchronously, intersexual variation of VOCs was detectable. Finally, with one exception, there was no sexual dimorphism in scent quantity. We show that there are two ways to use scent to be a dioecious fig based on differences in flowering synchrony between the sexes.

Among the sexual strategies to avoid inbreeding in flowering plants, sexual dimorphism promotes the evolution of dioecy 1 . The transition from monoecy to dioecy in plants is an important event in the evolution of angiosperms 2 , with dioecious species occurring in half of all plant families but representing only 6-7% of all species 3 . In clades with a large number of dioecious species, Vamosi and Vamosi 4 found a higher extinction rate as well as a lower speciation rate than in their non-dioecious sister clades, a difference that may be explained by the strong sexual dimorphism displayed by dioecious species. Indeed, one of the consequences of such sexual specialization is that reward levels to pollinators often differ between male and female flowers as a result of sexual selection 5 . Such specialization leads, in some cases, to cheating by one deceptive sex, which can be either male or female 6 . In such deceptive pollination systems, signals are key factors in maintaining the interaction 7 . For the deceptive sex it is crucial to maintain signals that mimic the signals of the rewarding sex in order to attract pollinators and ensure mating success, thereby maximizing the fitness of both sexes. Among these signals, while floral scents (i.e. bouquets of volatile organic compounds (VOCs) emitted by flowers) are often crucial signals for pollinator attraction 8,9 , they may be among the most variable traits of the plant phenotype 10 . However, some cases of strong stabilizing selection on floral scent exist 11,12 . Despite the potential for variable evolutionary trajectories, floral scents in dioecious species have been much less studied than other floral traits in the context of sexual dimorphism 5 .
In nursery pollination mutualisms, which are highly species-specific interactions where reproduction of the partners is completely inter-dependent, chemical signals are crucial in partner encounter 13 . To ensure effective partner encounter, selection should act both on the specific volatile signals emitted by the host plant and on the capacity of the pollinating insect to detect their message, ensuring effective communication. More than half of the nursery pollination systems described so far involve dioecious species 13,14 . They thus provide excellent systems to study the constraints imposed on plants and pollinators, as well as the strong necessity for inter-sexual resemblance in these signals when the sexes are separate and offer different rewards to their pollinators. In fact, there are many dioecious species where the chemical composition of scents of the rewardless sex -often the females -is similar to those of male plants, supporting the predictions of the intersexual mimicry hypothesis 5,15 . However, sexual dimorphism in scent composition also exists in several dioecious species 5 . In addition, dimorphism in floral scents may also be due to variation in the quantity of emitted VOCs with one sex, very often the males, having higher VOC emission rates 5 .
With more than 300 dioecious species, the genus Ficus (Moraceae) provides an excellent opportunity to test the intersexual mimicry hypothesis of floral scents in nursery pollination mutualisms, since all Ficus display strategies of pollinator attraction strongly based on chemical mediation. Nearly every fig species is usually obligatorily pollinated by a single species of fig wasp (Hymenoptera, Chalcidoidea: Agaonidae), which can only reproduce within the host fig 16 . The female pollinator wasps locate figs of the host by species-specific bouquets of VOC attractants produced by figs at receptivity, i.e. the developmental stage during which the stigmas are ready for pollination [17][18][19] . Dioecious figs represent a special case of pollination by deceit with fatal consequences for female wasps that enter female figs; such wasps will have zero reproductive success, for the following reasons. Dioecious fig species are anatomically gynodioecious, with true female trees and hermaphrodite but functionally male trees. Figs of male trees bear male and female flowers, but their short-styled female flowers serve only as brood sites for pollinators and never develop into seeds 16,20 22 . One explanation for why wasps enter female figs is that, in some dioecious fig species, they have no option other than to enter female figs. In seasonally reproducing species such as F. carica 23 , male trees in a population are simultaneously receptive, but this is well before or after the synchronized receptivity of the female trees. Wasps that exit male figs during the time of synchronous female receptivity have no choice, and due to strong chemically mediated species-recognition mechanisms based on chemical signals 13 , they will be attracted to the deceptive female figs. The wasp population in these species is maintained by the emergence of a few wasps, at the end of the period of female flowering, which enter some of the few male figs that are receptive at that time and in which they can reproduce 20,24 22 . Another possible explanation is the "no preference" hypothesis. Grafen and Godfray 15 , discussing vicarious selection in the dioecious fig-pollinator system, suggested that pollinating wasps cannot choose between male and female figs because there are no detectable differences between them. In fact, female trees need to attract wasps to produce seeds and male trees need to attract pollinators whose offspring will disperse their pollen, i.e. whose offspring will be attracted by female trees. Thus, trees of both sexes are strongly selected to mimic each other. Since at the receptive stage there are no obvious visual differences (in diameter, colour, etc.) between male and female figs and since chemical signals have been shown to be sufficient to attract specific pollinating fig wasps in the absence of other cues 13 , pollination by deceit in dioecious figs can be hypothesized to depend on mimicry in the chemical composition of floral scent.
Yet, the importance of chemical mimicry in maintaining a mutualism based on pollination by deceit in dioecious fig species is still largely unresolved. Previous studies seemed to indicate that different species of Ficus present alternative strategies in intersexual chemical mimicry, whether for pollination or for seed dispersal 23,[26][27][28] . We hypothesize that in fig species where male and female trees flower synchronously, selection will favour a strong resemblance of the chemical signal emitted by figs of the two sexes in order to ensure that pollinators visit female figs. In contrast, in fig species where males and females flower mainly asynchronously, this selection among fig sexes could be relaxed (as female figs will be visited by wasps that do not have a choice between sexes). However, scent divergence between sexes would still be limited because figs are still under strong selection to produce odours allowing host species recognition.
In fig species, dioecy is a derived trait and has evolved at least twice from monoecy 29 ; among the six subgenera of figs, dioecy occurs in three (Ficus, Sycomorus and Sycidium). In the present study, using a large data set (seven fig species from the three dioecious subgenera, Table 1), we investigated the volatile profiles (composition and quantity) emitted by receptive figs of both sexes. We verified the role of these flower scents in host species-specific recognition and in ensuring constancy of pollinators. Comparing fig species in which males and females flower synchronously and asynchronously, we specifically examined the potential of volatile chemical signals in mediating intersexual mimicry in dioecious nursery pollination mutualisms by estimating the degree of intersexual chemical similarity in species with varied flowering phenologies.

Results
Specificity of the signal. We identified a total of about 150 VOCs produced by the receptive figs of the seven dioecious species. The floral scents emitted by these figs were mainly composed of monoterpenes and sesquiterpenes. The blend of compounds was generally dominated by one or a few common terpenoid compounds, the identity of which differed among the species (Fig. 1); for example, β -caryophyllene for F. fulva, germacrene-D for F. fistulosa, or (E)-β -ocimene for F. hispida (see Supplementary Table S1 for more details). In one species, a rare compound dominated, i.e., 4-methylanisole in the scent of F. semicordata.
Scents of individuals from the same species are clearly grouped together based on the ordination (NMDS) of the odour samples from the different species (based on the Bray-Curtis distances) (Fig. 2). Moreover, the PERMANOVA confirmed that the volatile chemical profiles of the receptive figs varied significantly among the different species (see Table 2).
Intersexual variation of scents. We then separated the data by species to test for the effect of sex on the volatile bouquet of VOCs emitted by each species. The species fell into two groups in these analyses (NMDS ordinations presented in Fig. 3). First, the volatile profiles of the sexes cannot be separated in species in which both sexes flower synchronously (namely F. fistulosa, F. fulva, and F. hispida). These results are confirmed by the PERMANOVAs, in which the factor 'sex' is not significant (p > 0.105, see Table 2). The second group consists of species in which the two sexes do not flower synchronously. For this group, volatile profiles showed significant variation between sexes. On the ordination graphs based on Bray-Curtis distances, VOCs emitted by receptive male and female figs are separated in three of the four asynchronous species, F. auriculata, F. semicordata and F. septica (Fig. 3). The PERMANOVA (for all species p ≤ 0.010, Table 2 (Table 2). For the species in which VOC composition significantly differed between sexes, we conducted a one-way Simper analysis to identify the compounds that best explain dissimilarities between sexes. In some species, this analysis revealed that major compounds were responsible for these dissimilarities between the bouquets of male and female figs, as in the case of F. semicordata, where the proportion of 4-methylanisole mainly explains this difference (18% of the variation between sexes). For the other species, this difference was explained not only by some major compounds but also by minor compounds such as decanal (5%), (Z)-3-hexenol, ethyl hexanoate, linalool and β -caryophyllene (all about 4%) for F. auriculata, or linalool, phenylethanol, (Z)-3-hexenol, hexyl acetate, (E)-2-hexenyl acetate, β -bourbonene and eugenol (all at about 3%) for F. septica. Finally, for F. exasperata differences between sexes in VOCs emitted by receptive figs were mainly due to a suite of compounds (both major and minor), although each only contributed to a small proportion (about 2%) of the dissimilarities. In F. exasperata, these compounds included methyl-acetophenone, linalool, methyl salicylate, terpinolene, (E)-β -ocimene, hexadecane, myrcene, perillene and α -copaene.
Variation between sexes in quantity of volatile compounds emitted. Except for one species, F. fistulosa, there was no significant difference in the quantity of VOCs emitted by male and female figs during receptivity (Fig. 4, Table 3). For F. fistulosa, receptive female figs produced about four times more VOCs (total quantity) than did male figs (3.7 ± 1.  Table 1). Among the different species, however, fig diameter (Table 1) is a parameter that contributes to explain differences in the amount of VOCs produced, and not surprisingly species with larger figs, such as F. auriculata and F. septica, produced higher quantities of volatile compounds.

Discussion
Our results represent to the best of our knowledge the first evidence for a strong effect of flowering phenology as a selective pressure on intersexual chemical mimicry. They also reinforce previous results on interspecies variation of flower scent in Ficus 30 . Indeed, they highlight the importance of flower scents as a crucial trait for avoiding host identification mistakes in Ficus species and maintaining highly specific interactions in these and other horizontally transmitted mutualisms 13 . Our findings corroborate the role of flower scents as a pre-zygotic barrier between closely related plant species in nursery pollination systems 31 , which could lead to reproductive isolation during speciation 32 . Moreover, our results confirm that neither potential interpopulation variation nor intersexual variation affects this clear interspecies variation of flower scent in dioecious Ficus species. Furthermore, behavioural experiments conducted on the pollinators of some of the species we studied (F. hispida, F. semicordata, and F. carica) have already shown that pollinating fig wasps use this specific chemical message as a cue to locate their mutualistic host 33 (M. Proffit & C. Soler, unpubl. data). Once again, the "endless" variation in the composition of floral signals allows fig species to elaborate quite complex and specific blends in order to attract their specific pollinators 13 .
Regarding intersexual mimicry in dioecious species, our results provide evidence within a similar phylogenetic background that there are two different answers to the question of 'how to be a dioecious fig species' that depend on flowering phenology. On one hand, in dioecious species in which male and female trees flower synchronously, fig-pollinating wasps loaded with pollen from their natal male figs have a simultaneous choice between receptive male and female figs within the same population. For these synchronous species, it is important to ensure that female pollinating wasps cannot discriminate between sexes in order to maintain both male (pollen and pollinating wasp production) and female (seed production) functions. Consequently, this leads to tighter intersexual mimicry in flower scents, as we have shown for three synchronously flowering dioecious species belonging to two different Ficus subgenera, F. fulva, F. fistulosa and F. hispida.   In only one of these synchronously flowering species did we find a difference in the quantity of compounds released by receptive figs, with the deceptive sex (female figs) producing greater quantities of VOCs. This may be consistent with the need to ensure pollinator visitation even to the fatally deceptive female sex. In other less exclusive but still specialized interactions, and in those without the fatal consequences of choosing the wrong sex, such as the mutualistic interactions between the seed-predating and pollinating noctuid moth Hadena bicruris and its dioecious host Silene latifolia, Dötterl et al. 34 also found no difference in the quality and quantity of VOCs emitted by female and male flowers (but see 35 ). In Salix caprea, an insect-pollinated dioecious species, total scent emission was higher in male flowers, and some sex-specific differences in relative scent composition were detected. However, naïve honey bees were equally attracted to the scent of both sexes 36 . Remarkable qualitative intersexual differences in floral scents were also found by Okamoto and collaborators 37 in their detailed study of sexual differences in floral scent of the monoecious species of the tribe Phyllantheae (Malpighiales, Phyllanthaceae) that are pollinated by Epicephala moths (Lepidoptera, Gracillariidae) in interactions that include some examples of nursery pollination mutualisms. In these examples, differences in scents exist between male and female flowers of the same individual, highlighting the possibility for sexual scent divergence even at the intra-individual level. In addition, these authors show that mated female Epicephala moths preferred the scent of male flowers over that of female flowers. In contrast to the interactions between figs and their pollinating wasps, in the nursery pollination interaction in Phyllanthaceae, it is mandatory for the pollinators to visit first male, then female flowers in order to gain the reward provided by the interaction. Even though the mechanism triggering the visit of female flowers following that of male flowers has not been elucidated in the Phyllanthaceae system, sexual dimorphism of floral scent and its discrimination by pollinators have been interpreted as adaptive for both plant and pollinator 37 .
On the other hand, for fig species in which individuals of the two sexes flower at different times of the year, the short-lived pollinating wasps have no choice and simply enter the available receptive figs within the population based on the host-specific chemical signal. In such situations, the selective pressure on intersexual mimicry should be relaxed. In fact, intersexual variation of floral scents is detectable at the intraspecific level for the species of this category that we studied. However, this intraspecific variation between sexes in flower scents should be less marked than the interspecific variation that allows pollinators to distinguish signals from the specific host fig species to avoid mistakes, especially if fig species occur in sympatry 38 , as for F. hispida and F. exasperata in our study. Our results thus suggest that this sexual dimorphism is not likely to disrupt the species-specific encounter of the two partners, confirming the strong species isolation mechanisms expected in dioecious figs 38 .
The sexual difference in dioecious figs appears to rely on VOC identity rather than on the overall VOC quantity, in contrast to most previously studied angiosperm species, where intersexual differences are mostly due to a higher total quantity of VOCs emitted by male flowers 5 (but see 39 ). In another nursery pollination mutualism characterized by partial asynchrony in the flowering of male and female plants, that between Chamaerops humilis, the dioecious dwarf palm, and Derelomus chamaeropsis, its pollinating weevil, male plants produced four times more VOCs than female plants. However, the qualitative composition of scents did not differ between the sexes, allowing female plants, the deceptive sex, to also attract the species-specific pollinating weevil 40 . These findings suggest that quality and quantity can interact in diverse ways based on particular ecologies and contexts and that no one unique prediction is possible.
For the dioecious fig species in which males and females flower asynchronously, differences in the VOCs emitted may be accounted for by differences in the proportions of some major compounds, such as 4-methylanisole for F. semicordata. However, some minor compounds from different pathways may also play a role in these intersexual differences of VOC bouquets, such as methyl-acetophenone, linalool, methyl salicylate or α -copaene for F. exasperata. The statistical analysis used in the present study, performed as recommended in recent studies comparing flower scents 36,37,41 , minimizes the potential effect of major compounds on their contribution to the total variance.
This study on several fig species provides novel insights into our understanding of the occurrence of intersexual chemical mimicry in dioecious species. In her well-documented review, Ashman 5 reported that the occurrence of sexual dimorphism is usually explained in the literature by i) sexual selection, ii) allometric relations, iii) honest signals, iv) directed pollinator movement, or v) post-pollination partners. In contrast, the absence of sexual dimorphism may be explained by intersexual mimicry or by between-sex genetic constraints. These various explanations have come from different studies conducted on different species, with varied evolutionary and ecological histories, and cannot really be applied in our Ficus case. In dioecious figs we found intersexual mimicry in species with synchronous flowering between the sexes and relaxed intersexual mimicry in species where male and female fig availability is temporally separated. First, given the special biology of dioecious figs, there is no advantage via sexual selection accruing to males that have scent profiles more attractive to pollinators than those of females, since visits (albeit fatal) to female figs (in which seeds are produced) by pollinators carrying its pollen are necessary for the function of a male tree to be performed. Second, for all the fig species we studied, male and female figs have the same size, shape and colour, but they differ greatly in the rewards offered. Since this system is based on deception (in which female fig trees are the fatally deceptive sex), the issue of honesty of signals leading to sexual dimorphism 42 does not arise. Furthermore, since pollinators have only one opportunity to enter a fig  (either a true fig wasp nursery in male trees or a fatally attractive deceptive one in female trees), the question of directed pollinator movement from one sex to another by sexual dimorphism in floral scent also does not apply. The only remaining explanation from Ashman's review 5 is that which takes into account post-pollination visitors. These are effectively different for the two sexes: male figs are parasitized by other wasps, contrary to most female figs 43 . However, this sexual difference in attraction of post-pollination parasites holds not only for asynchronous species of dioecious figs but also for synchronous species: since no sexual variation of scents has been detected in the latter case, it is unlikely that any explanation based on selective pressures exerted by post-pollination interactions will apply here. This therefore means in figs that there is no advantage to sexual dimorphism in floral scent via any of the selection pressures often associated with scent dimorphism 5 . The scent of the fatally attractive Scientific RepoRts | 6:21236 | DOI: 10.1038/srep21236 female must converge towards that of the rewarding male sex in figs, which may imply that females must mimic males, or alternatively that floral scent in each sex must show directional selection towards the other in order for individual trees to achieve fitness and consequently for the mutualism to survive. This is especially true when both sexes are simultaneously available to the pollinator. However, when the sexes are not simultaneously available, as in asynchronous species, it is possible that the host-specific chemical signal (which is likely a blend of volatiles) remains constant between the sexes, but that additional VOCs-that are not sufficient to mask or blur the species-specific host-recognition signal for the pollinator-are unconstrained to vary, resulting in a statistical difference between the scents of the sexes. So far, for these species we do not know the specific compounds in fig scents that are responsible for pollinator attraction. However, we do have some evidence that pollinators of several Ficus species are able to detect not only the major compounds emitted by receptive figs but also some minor compounds 44 (L. Zongbo et al. and M. Proffit et al., unpublished data). Therefore, the possible effect on behaviour (e. g. attraction or repulsion) of these minor compounds should not be neglected in fig/wasp interactions, a precaution that echoes those suggested for other plant-insect interactions 45 . Consequently, thorough analysis of the contribution of the different compounds should be complemented by experimental behavioural studies, or by GC/EAD (gas chromatography/electro-antennographic detection), in order to know precisely which compounds are active in attraction 24,46 .
We have thus shown in this study the existence of two alternative ways to be a dioecious fig species that depend on differences in the synchrony of flowering between sexes. Since dioecious Ficus are in fact anatomically gynodioecious, female and male figs are structurally not identical. Observing intersexual variations of scents (as occur in asynchronous species) is thus not surprising and may be explained by different scent-emitting structures. However, this constitutive difference is thwarted by vicarious selection for intersexual chemical mimicry 15

Material and Methods
Studied species. In order to have a good phylogenetic and geographical representation for this comparative study, we chose species with different flowering phenologies (synchronous versus asynchronous flowering of males and females) from different sections among the three subgenera of dioecious species, and from different geographical regions distributed over tropical, subtropical regions of the Old World (Table 1). No dioecious fig species occur in the New World. From published data or our personal observations, we can divide these species into two groups regarding their flowering phenology 47,48 . In the first group, the two sexes flower in synchrony in our studied populations: F. hispida, F. fistulosa and F. fulva; in the second one, flowering phenology is mainly asynchronous, with receptive figs of male and female trees co-occurring very rarely in time: F. auriculata, F. exasperata, F. semicordata and F. septica 27,28,33 . For both techniques we always performed blank extractions with empty bags to control for contaminant compounds. Odour collection was always performed under natural light and at ambient temperatures between 12:00 and 15:00 hours, corresponding to the insects' period of maximum activity during our field season. Comparisons of the efficiency of the two types of traps (M. Proffit, unpubl. data) allowed us to use both types of volatile collections in this comparative study. Moreover, we always used the same technique to collect the volatiles of one species and straightforward comparison of scents emitted by the two sexes was thus always possible.

Collection of VOCs.
The relative proportions of the compounds emitted by figs were determined for each sample using GC-FID chromatograms. In addition, for each sample, we estimated the total quantities of volatile compounds by using the peak areas of both internal standards in GC-FID chromatograms as a reference quantity. Peak areas of all the compounds were determined, and then divided by the peak areas of the standards and multiplied by the known amounts for the standards.

Analysis of VOCs.
Traps were stored in the dark in a freezer until analysis. For both sampling techniques a Varian column CP-SIL low blend MS (30 m, ID 0.25 mm, film thickness 0.25 μ m) and helium as carrier gas (at 1 ml min −1 ) were used. The injector split vent was opened at a ratio of 1:4. Oven temperature was programmed at 50 °C during 3 min; then increasing by 3 °C min -1 to 100 °C, by 2.7 °C min -1 to 140 °C, by 2.4 °C min -1 to 180 °C and by 6 °C min -1 to 250 °C. For Chromatoprobe ® samples, the injection was done at 40 °C for 1 min, followed by a warm-up for 3 min, until 250 °C, at the rate of 200 °C min -1 . The temperature program for the analysis was: 40 °C for 3 min, ramped from 40 °C to 100 °C at 3.3 °C min -1 to 250 °C, from 100 °C to 140 °C at 2.9 °C min -1 , then 140 °C to 180 °C at 2.7 °C min -1 , and finally from 180 °C to 250 °C at 10 °C min -1 . Temperature was maintained at 250 °C for 8 subsequent minutes.
Compound identification was based on computer matching of the mass spectra with NIST 98 MS library, on retention indices reported in the literature 49 and, whenever available, by injection of reference compounds. By comparing spectra of each sample with the respective control sample (empty bag, same day and conditions of collection), putative contaminant compounds were subtracted. Data analysis. All statistical analyses were performed using R 50 . In order to compare patterns of scent composition both among different species and between sexes within each species, we performed a non-metric multidimensional scaling (NMDS) using the function meta MDS in the package Vegan 51 . For the comparison of volatile profiles among different Ficus species we used the relative proportions of all the compounds emitted by the seven species, whereas for inter-sexual variation of scent each species was analysed separately. Prior to the analysis, data were square-root transformed and then standardized using a Wisconsin double standardization. A data matrix of pairwise comparisons among samples was then calculated using the Bray-Curtis distance index, which ranges between 0 and 1 52 . NMDS was used to find the best low-dimensional representation of the distance matrix. Finally, the solution was scaled by rotating it, so that the largest variance of samples was on the first axis 51,53 . The null hypothesis of no difference in patterns of scent composition among species and between sexes for each species was tested with a permutational multivariate analysis of variance (PERMANOVA) using the function 'adonis' in Vegan. PERMANOVA is a nonparametric (in the case of one-factor models) permutation-based version of the multivariate analysis of variance, and thus with no assumption regarding the distributions of the original variables 54 . However, PERMANOVAs are very sensitive to heteroscedasticity 55 . Therefore, prior to analysis homoscedasticity was confirmed using a multivariate analogue of Levene's test for homogeneity of variance. PERMANOVAs were run on the Bray-Curtis distance index with 1000 permutations per analysis. A similarity percentage (one-way SIMPER, with factor sex) was used to identify the compounds that best explained dissimilarities (up to 30%) between sexes.
For each species, we used a linear model to test if there were significant effects of sex on the total quantities of VOCs emitted by figs. The goodness-of-fit of all models was evaluated by inspecting the models' residuals; the normality of the residuals was tested using a Shapiro-Wilk's test. For data with residuals not normally distributed, a log transformation was used to ensure their normality. Significance (p < 0.05) was assessed by testing the change in deviance after the removal of the factor sex from the model.