Sexual specialization in phenology in dioecious Ficus benguetensis and its consequences for the mutualism

Timing of reproductive events has become central in ecological studies linking success in pollination and seed dispersion to optimizing the probability and periods of encounters with pollinators or dispersers. Obligate plant–insect interactions, especially Ficus–fig wasp mutualisms, offer striking examples of fine-tuned encounter optimization as biological cycles between mutualistic partners are deeply dependent on each other and intertwined over generations. Despite fig flowering phenology being crucial in maintaining Ficus–fig wasp mutualisms, until now, the forces of selection shaping the phenological evolution of dioecious fig trees have received little attention. By conducting a 2-year survey of a population of Ficus benguetensis in Northern Taiwan, we assessed whether environmental factors or other selective pressures shape the phenology of male and female fig trees. Constraints by mutualistic pollinating wasps and seed dispersers, rather than climatic factors, appeared to mainly shape fig phenology and allometry in F. benguetensis. We identified a new sexual specialization in dioecious fig trees: the position of fig production. We propose that the continuous male fig production on tree trunks can enhance the survival of pollinating fig wasps through faster localization of receptive figs while reducing the mutualistic conflict between the fig and its obligate pollinators. By contrast, in female trees, fig production is massive in summer, located on the twigs of the foliar crown and seem more related to seed dispersal and germination. Identifying variations in the allometry and phenology of dioecious figs provide valuable insights into how monoecious and dioecious species resolve mutualism conflicts and into the emergence of dioecy in fig trees.


Background
Knowing the phenology of flowering and fruiting is essential for understanding the ecology and evolution of a plant (Forrest and Miller-Rushing 2010). Moreover, phenological shifts in flowering/fruiting periods may severely affect the associated community of a plant , particularly for interacting species such as obligate mutualistic partners. The failure of mutualistic partners to meet at the appropriate time for pollination or seed dispersal can disrupt biological cycles and cause the extinction of both partners. However, understanding the mechanisms underlying mutualistic interactions requires obtaining precise knowledge on phenology. To assess the role of phenology in the maintenance and evolution of mutualisms, fig trees (Ficus, Moraceae) are appropriate models because they have various species, ecologies, reproductive strategies, and phenological cycles (Harrison et al. 2012), and are all involved in obligate pollination mutualism. Generally, each Approximately half of Ficus species are functionally dioecious (Berg 1989) and present physiological (Dumont et al. 2004) as well as phenological (Valdeyron and Lloyd 1979) adaptations that fulfil reproductive functions. Male fig trees bear figs that produce pollen and pollen vectors. The short-styled ovaries of such figs provide suitable oviposition and larval development sites for pollinators . Such figs ensure the viability of fig wasps, but produce no seeds (Kjellberg et al. 1987). Male trees continuously produce figs (Dunn et al. 2008); this adaptation was suggested to fit their short-lived pollinators. By contrast, female trees produce seeds (Berg 1989) (female figs produce no pollen) and maintain the Ficus biological cycle. Pollinating fig wasps fertilize the flowers inside female figs and ensure seed generation, but fail to produce any offspring because long-styled ovaries form a mechanical barrier to prevent oviposition (Kjellberg et al. 1987). Thus, female trees are a population sink for wasps. This conflict between wasp interest (avoiding female figs) and fig tree interests (pollination of female figs and seed production) is considered solved through phenology (Anstett et al. 1997). For instance, some dioecious species exhibit intrasexual synchronized flowering phenologies (Kjellberg et al. 1987;Patel 1996), where trees from one sex crop after the other. In such systems, pollinators have no choice but to enter the receptive fig of whichever sex is available. Thus, alternated phenologies stabilize the mutualism. In other dioecious species, male and female trees simultaneously flower, thus offering a choice to pollinating wasps. However, two studies (Soler et al. 2011(Soler et al. , 2012 have suggested that combined phenological synchrony and intersexual mimicry (resembling scents in receptive figs of both sexes) can reduce the ability of fig wasps to actively choose male figs. In addition, these studies have supported the concept of alternative or additional adaptations to phenology allowing fig trees to "slave" their pollinators into pollinating female figs. For example, some tropical dioecious and most monoecious Ficus species present a synchronous intratree flowering combined with a high intertree asynchrony. This differential phenology ensures the permanent availability of receptive figs and the survival of the pollen carriers (Patel 1996;Bronstein et al. 1990). However, until now, the forces of selection that shape the phenological evolution of dioecious fig trees have been overlooked (Anstett et al. 1997;Patel and McKey 1998). While plant phenology has long been considered driven by climatic factors (Körner and Basler 2010), recent studies have emphasized taking account of additional features such as phylogeny (Davis et al. 2010), genetic diversity (Yang et al. 2014, ecology (Forrest and Miller-Rushing 2010), and physiology (Keller et al. 2011) to disentangle climatic and biological selective pressures. In the dioecious fig-fig  wasp interactions, a strong sexual selection is expected in life history traits that facilitate, synchronize, and coordinate the interaction between different partners (i.e., male and female trees, fig trees and fig wasps, and fig trees and seed dispersers). We predict that male trees will exhibit a continuous production of figs over a year and locate the fig production over a tree to maximize the probability and speed of localization, and hence, the survival of their obligate pollinators. Conversely, we predict that female trees will present a production limited in time to the optimal season for seed germination and seedling growth as well as locate their

Study species
Ficus benguetensis Merr., a functionally dioecious fig species belonging to the subgenus Sycomorus and the section Sycocarpus, is pollinated by Ceratosolen wui (Chen and Chou 1997). This fig species is distributed in the Philippines, Taiwan, and the Japanese Ryukyu Islands (Berg and Corner 2005). It is a small to medium-sized tree (4-10 m height) growing at low altitudes, particularly in valleys with high humidity. Figs grow on branches or directly on the trunk.
According to the terminology for dioecious figs (Galil and Eisikowitch 1968), including the five developmental stages of figs, F. benguetensis exhibits three common stages, with the final stage differing in each sex (Fig. 1).
Seventeen male trees and seven female trees were surveyed weekly, and fig abundance and developmental stages were recorded from March 2011 to March 2013 (25 months). After 10 months of monitoring, noticeable differences in production were observed regarding fig positions on a tree. In addition to the trunk area (between 120 and 150 cm high; Additional file 1: Figure S1), two other areas (30 cm long), the lower part of the first branches and the terminal twigs, were monitored for the second part of the survey. Fourteen male trees and five female trees were then assessed for 15 months from January 2012 to March 2013; three individuals were excluded because their branches and twigs were inaccessible.
Weather data from the Quchi meteorological station of the Taiwan Central Weather Bureau, located approximately 3 km from the study field, were also recorded weekly. During the census period, the weekly mean temperature was 21.3 °C, ranging 12.7-30.0 °C, whereas the annual cumulated precipitation was 3992.25 mm (Additional file 1: Table S1).

Data analysis
To evaluate the synchronization of figs within a tree, the evenness index, modified from Smith and Bronstein (1996) was used. This modification considers the difference in the duration of each developmental phase of a fig for each sex, a parameter critical to calculate the probability of their occurrence. The fig evenness within a tree, E, for each sex was calculated as follows: E = ΣPi × ln [(Pi/Di) + 1], where i represents each developmental stage, Pi represents the proportion of figs at stage i, and Di represents the expected duration proportion of each developmental phase. Values were then calibrated to range from 0 (total asynchrony) to 1 (total synchrony). An average duration for each developmental stage was calculated based on the measures obtained during our 2-year survey.
To explain the observed patterns of fig production in F. benguetensis, an estimator of the total yield of a tree, Y, was designed. Fig abundance was first measured based on the samplings of the three areas of production of a tree: trunk, branch, and twigs. Subsequently, the total yield (Yi) was estimated as follows: Yi = Ti + 2.5·Bi + 150 Wi, where Ti represents the number of figs produced by the sampled area on the trunk of the tree i, Bi represents the number of figs on its sampled branch, and Wi represents the number of figs on its sampled twigs. The coefficients were estimated according to the average number of branches observed per tree (i.e., 2.5), and the average number of twigs observed per tree (i.e., 150).
Statistical analyses were performed using SYSTAT v.12. The Mann-Whitney U test and Kruskal-Wallis test were used to compare pairs of data sets and multiplefactor data sets. Moreover we use simple regressions to verify whether dependent fig phenological variables were explained with independent climatic variables (Pereira et al. 2007). Since phenology data sets generally exhibit temporal autocorrelation (i.e., non-independent error variance) and thus break the assumptions of serial independence required for most inference tests (Pyper and Peterman 1998). Durbin-Watson test was employed to When figs are ready to be pollinated, the receptive stage (B phase) begins. From the outside, the bracts slightly open to permit the mutualistic wasp to enter. After pollination, the bracts close and the interfloral stage begins (C phase). At this stage, pollinating wasp larvae develop exclusively within the ovaries of male figs, which are transformed into galls. The development of male figs and wasp larvae finishes with the wasp-releasing stage (D phase) when the stamens are mature and adult wasps exit their natal galls. In female trees, the final stage is the ripe stage (E phase), coinciding with seed maturation and frugivore attraction check for temporal autocorrelation (Chatterjee and Price 1991). Because our phenology data is structured in time series, we used ordinary least squares (OLS) models with auto-correlated errors (Venables and Ripley 1999). These regression analyses were conducted using the Proc AUTOREG procedure of SAS 9 (SAS Institute Inc. 2002).

Crop number, fig abundance, and evenness
The phenological pattern in fig production of F. benguetensis showed clear differences between both sexes. First, male trees produced significantly more crops (2.08 ± 1.13) than did female trees (1.36 ± 0.90) (Table 1). Furthermore, male trees began producing figs earlier in spring and grew figs continuously, except for a short gap in winter. By contrast, female trees produced fewer crops within a year, most of them being restricted to summer (Fig. 2). Second, the average fig abundance on the trunk of male trees (24.60 ± 16.00) was significantly greater than that on the trunk of female trees (9.74 ± 9.21) for the entire survey period (Table 1).
Trees from both sexes had low values of average evenness (<0.3), but higher asynchrony was exhibited by female trees than by male trees (Table 1). The extent of the evenness values (0.148-0.773 and 0.085-0.657 for male and female trees, respectively) was driven by major variations in the estimated duration of their different developmental phases. Moreover, evenness was similar in the growth areas of male trees: 0.31 ± 0.29 for the trunks and 0.27 ± 0.27 for the branches [Mann-Whitney test: non-significant (NS)], but not in the growth areas of the female trees: 0.26 ± 0.29 on the trunk, 0.17 ± 0.23 on the branches, and 0.16 ± 0.20 on the twigs (Kruskal-Wallis test statistic: 15.9, df = 2, P < 0.001). The Mann-Whitney tests for the three fig-bearing positions showed that the evenness on the twigs was significantly lower than that on the trunk and branches (twig/branch: P < 0.01; twig/ trunk: P < 0.001; trunk/branch: NS; Additional file 1: Figure S2).

Crop allometry
In comparing the production of the three fig-bearing positions (trunk, branch, and twig), we discovered significant differences between the sexes. Because no male trees produced figs on their twigs, the distribution of fig abundance was highly skewed among the fig positions in both sexes. For male trees, figs were more abundant on trunks than on branches: 24.60 ± 16.00 figs and 6.19 ± 6.99 figs, respectively (Mann-Whitney test: P < 0.001). By contrast, female trees produced 9.74 ± 9.21, 6.00 ± 7.25, and 2.96 ± 2.41 figs on trunks, twigs, and branches, respectively (Kruskal-Wallis test: 26.994, df = 2, P < 0.001).
Regarding yields per entire tree, the average estimated fig yield for female trees, Y F , indicated that 98.13, 1.06, and 0.08 % of the fig production was located on the twigs, trunks, and branches, respectively (Fig. 3) (Table 2). No self B-D event occurred on the branches. At the population level, B-D events were slightly more abundant because the sample size was also greater. The matching between male D-phase figs and female B-phase figs (female B-D) occurred 32 times in 102 surveys, and 45 times among different male trees (male B-D) ( Table 2).

Climatic factors
The number of figs in phase A (beginning of the crops) and fig abundance showed time series autocorrelation in both male (Durbin-Watson D Statistic: 0.419, p < 0.001 and 0.343, p < 0.001, respectively) and female (Durbin-Watson D Statistic: 0.461, p < 0.001 and 0.317, p < 0.001, respectively). None of the phenological variables

Discussion
Our study shows that the phenology of fig production in F. benguetensis is the product of biological processes. The variation observed in fig production over time could not be attributed to temperature or rainfalls, whereas male and female trees growing in the same environment exhibited distinct phenologies. Because this mutualism is obligate and because dioecious individuals serve different reproductive functions, distinct spatial and temporal adaptations were selected to enhance the effectiveness of each function. Male  (Patel and McKey 1998;Chou and Yeh 1995;Harrison and Yamamura 2003;Bain et al. 2014). While a few pluriannual and quantitative surveys have already been carried out over 2 (Patel 1996) and 3 years (Corlett 1993), our study is In dioecious species, the necessity to maintain the pollinator biological cycle (Kjellberg et al. 1987) with a constant supply of figs shapes the spatial and temporal distributions of the figs over an individual male tree. The general pattern of cauliflorous male fig production in F. benguetensis (spring and autumn production peaks and higher male fig production) is consistent with that in other dioecious Ficus in Taiwan  or the Asian continent (Yu et al. 2006). Our analysis reveals several mechanisms maximizing the chances of survival of pollinating wasps despite their estimated lifespan of 12 h to 3 days (Dunn et al. 2008)  was expected because seeds from fallen figs are unlikely to germinate if they are too close to their genitor. The change in resource allocation towards the twigs seems more linked to seed dispersal. Similar to other species of the Sycomorus subgenus, F. benguetensis is dispersed by fruit bats (Harrison et al. 2012). In contrast to F. racemosa and F. variegata (Patel et al. 1995), two tall and cauliflorous tree species dispersed by bats, F. benguetensis is a medium-sized tree, with its foliar crown often reaching the canopy in Taiwan forests. Producing female figs on the terminal twigs can facilitate the access and consumption of ripened figs by bats, and then increase the chances of effective seed dispersal. Additionally, the highest intratree asynchrony of female figs growing on twigs can further increase dispersal by ensuring repeated visits by frugivores. This extreme asynchrony seems directly linked to longer fig developmental stages. We estimated the total duration of development at 121 days for female figs and 71 days for male figs. Such extended prereceptive and maturation phases of the female figs are likely due to a greater number of ovaries in female figs and the necessity to be fleshier to attract seed dispersers (Patel and McKey 1998). Furthermore, a high intratree fig asynchrony ensures that female trees experience an extended and greater variety of environments for seed germination (Patel et al. 1993).
Our data support the recommendation by Harrison et al. to reassess the assumption that fig cauliflory is an evolutionary adaptation to bat dispersal (Harrison et al. 2012). Intersexual differences in allometry indicate that cauliflory in F. benguetensis is likely an evolutionary response that enhances pollination, whereas ramiflory is apparently an adaptation to seed dispersion by fruit bats. However, despite such differences between male and female figs, it remains unclear why the pollinator Ceratosolen wui did not "learn" to avoid female figs that entail such a fitness loss for the insect. Two main hypotheses with a few subhypotheses assuming intersexual chemical mimicry (Patel et al. 1995) may explain this absence of pollinator discrimination in synchronous male and female trees. First, the "no preference" hypothesis states that pollinating wasps are simply unable to distinguish between male and female figs. This inability could be due to vicarious selection leading to complete intersexual chemical mimicry (Grafen and Godfray 1991) or to the absence of selection by fig wasps to favour a specific sex (Anstett et al. 1997). An intersexual comparison of the composition and quantities of the volatiles attracting pollinating wasps, followed by insect bioassays, is necessary to choose between these two competing subhypotheses. Second, the "limited partial preference" hypothesis, which is based on imperfect sex mimicry, states that pollinators might develop a partial preference for male figs. However, this partial preference is a frequencydependent mechanism susceptible to disappearing in a male-dominated environment (Getty 1985), particularly if males present individual scents that vary excessively. Such variability, a common trait in Ficus scents (Soler et al. 2012), can increase the risk of a tree not being recognized as male, thus increasing the risk of a pollinator never entering a male fig (Patel et al. 1995). Finally, another subhypothesis of limited partial preference is called "selection to rush" (Patel et al. 1995). Although pollinating wasps may discriminate between male and female figs, their lifespan is a constraint and they cannot afford the reproductive cost incurred by the delay of choosing between male and female figs. Because wasps face limited availability and high competition for male receptive figs, wasps that rush into any receptive fig, whether male or female, are likely to be strongly selected. Therefore, we hypothesize that the combination of the short lifespan and the dispersion range of the pollinator could counter-select any partial preference of the insect. Moreover, such a system could produce an overload of pollen vectors, thereby triggering a scramble competition for a few receptive male figs. Pollinating wasps rush to enter the closest available receptive fig (likely a male fig), and then enter other figs at increased distances. Female figs available on the canopy in an extremely patchy distribution are pollinated because of the number of pollinators and their physical position acting as a pollinator net for the pollinating wasps that escaped the understorey. Further research is necessary to determine the sex ratio in various populations of F. benguetensis and to assess the lifespan and capacity of the dispersion of Ceratosolen wui.
Dioecy in Ficus is considered a characteristic derived relative to monoecy (Berg 1989). Although dioecy was proposed to be particularly adapted to seasonal climate (Kjellberg and Maurice 1989), alternative hypotheses, such as parasitic pressure (Kerdelhué and Rasplus 1996) or ant predation (Harrison and Yamamura 2003), have been subsequently developed to explain the appearance of dioecy in Ficus. A study on F. exasperata and F. hispida observed strong seasonal patterns in fig production (Patel and McKey 1998); however, in the present study, we observed no such strong seasonal patterns in F. benguetensis, except for the patterns of the estimated yield of female figs growing on twigs. Our results do not support the hypothesis that the appearance of dioecy could result from a differential allocation to reproductive functions among seasons (Kjellberg and Maurice 1989). However, our observations agree with the concept proposed by Patel and McKey (1998), that extreme sexual specialization can be an adaptive response to unstable trade-offs in the reproductive traits of monoecious figs.