Asymmetric sharing of pollinator fig wasps between two sympatric dioecious fig trees: a reflection of supply and demand or differences in the size of their figs?

Host specificity among pollinator fig wasps (Agaonidae) depends on host plant specific volatile cues, but fig wasps must also pass through a narrow physical barrier (the ostiole) if they are to pollinate and oviposit. Across South East Asia the dioecious shrub Ficus hirta is associated with at least ten pollinator species allied to Valisia javana. Ficus triloba has a single recorded pollinator, Valisia esquirolianae. Receptive figs of F. hirta are usually much smaller than those of F. triloba, but at a mainland site where F. hirta has atypically large figs we identified both V. esquirolianae and V. javana from both Ficus species using COI and ITS2 sequencing. To investigate whether this host overlap was exceptional we reared fig wasps from the two trees elsewhere and recorded features that may facilitate host transfer between them, including attractant volatiles, reproductive phenology and the sizes of their figs and fig wasps. The two Ficus species were found to support both Valisia species at several of the sites, suggesting that the differences we detected in volatile profiles, ostiole sizes and pollinator head sizes are not strict barriers to host sharing. Valisia javana colonised F. triloba more frequently than V. esquirolianae colonised F. hirta. This asymmetric sharing of pollinators may reflect the relative abundance of the two species of fig wasps and differences in host reproductive phenology. Asynchronous flowering of individual F. hirta may favor local retention of pollinators, in contrast to the tree-wide synchrony of F. triloba figs, which can generate local shortages of V. esquirolianae. If the pollinator sharing by male figs of F. triloba and F. hirta also occurs in female figs then this could result in gene flow between them.


Background
Ficus (Moraceae) is one of the most species-rich genera of woody plants in tropical and subtropical regions of the world (Harrison 2005), with more than 800 described species. Ficus shares a mutually beneficial symbiotic relationship with pollinating fig wasps (Agaonidae) that has remained structurally unchanged for tens of millions of years (Compton et al. 2010). As a result, fig trees and their pollinators have been a model system for the study the co-evolution and co-speciation between plants and insects (Janzen 1969;Weiblen 2002;Rønsted et al. 2005;Cruard et al. 2012). It was long believed that each species of Ficus is pollinated by a single species of fig wasp associated with a single host species, and that it was this specific obligate fig wasp pollination that has promoted both reproductive isolation and speciation in the genus (Ehrlich and Raven 1964;Segar et al. 2019). It is now becoming apparent that this apparent extreme specificity was in part at least a consequence of low sampling intensity in Open Access *Correspondence: yuhui@scib.ac.cn 5 South China Botanical Garden, The Chinese Academy of Sciences, Xingke Road 723, Tianhe district, Guangzhou 510650, China Full list of author information is available at the end of the article what is a largely tropical mutualism, which often resulted in records that were based on samples of a single species of pollinator from one or a small number of locations (Yu et al. 2021). Widely distributed species of fig trees are increasingly becoming known to support two or more species of pollinators. The insects rarely seem to have sympatric distributions, so in any single location only a single pollinator species tends to be present, except in contact zones (Sun et al. 2011;Chen et al. 2012;Yu et al. 2019). More rarely, examples are being found of pollinators that are not host-plant specific and routinely develop in the figs of two or more host Ficus (Lopez-Vaamonde et al. 2002;Cornille et al. 2012;Wei et al. 2014). Host specificity nonetheless remains the norm, and studies of exceptions to a 1:1 relationship provide an opportunity to examine how this specificity is achieved.  (Galil and Eisikowitch 1968). Female fig wasps are attracted to the figs by what are typically species specific-and developmental-stage specific blends of volatile compounds released from receptive figs when they are ready for pollination (van Noort et al. 1989; Grison-Pigé et al. 2002;Raguso 2008;Soler et al. 2011;Gu et al. 2012;Wang et al. 2021). Entry is achieved via a small hole called the ostiole which opens for a brief period during receptivity. Successful entry through the ostiole depends on foundresses having an appropriate head shape and body size (van Noort and Compton 1996;Cook and Segar 2010;Liu et al. 2013) and for successful oviposition the female must also have an ovipositor that is longer than the styles through which they reach the ovules where they lay their eggs (Nefdt and Compton 1996;Weiblen 2004). Females of some species re-emerge from the first figs they enter, but they can then only attempt entry into other figs on the same plants, because their wings become detached when they first enter the ostiole (Suleman et al. 2013;Mohd Hatta et al. 2021).
Although inter-specific variation in the physical structure of receptive figs and differences in the volatile cues they emit generate the specificity of the relationship (van Noort et al. 1989;Grison-Pigé et al. 2002), they do not entirely prevent the arrival at receptive figs of pollinators that routinely reproduce inside figs of other species (van Noort et al. 2013) and some of these pollinators manage to penetrate through the ostiole and even successfully reproduce (Compton 1990). Such 'mistakes' appear infrequent (Cook and Segar 2010) but may be more common in boundary locations or where a tree's normal pollinators are rare or absent (Ware and Compton 1992;Compton et al. 2009) (Yu et al. 2019). Based on the data currently available, it appears that only a single species of pollinator is present in each part of their range (Bain et al. 2016;Yu et al. 2019), but there are exceptions (Chen et al. 2012;Rodriguez et al. 2017) and sampling intensities remain low, so views on this may change in the future.
Factors that facilitate sharing of pollinators are likely to include having a complementary flowering phenology, have receptive figs that are of a similar size offering similar visual cues (Lomásco and Levey 2020) and with ostioles of a similar width, and figs that release attractant volatile blends that share common components that the fig wasps use as cues (Liu et al. 2015). Such combinations seem particularly likely to occur in the Asian tropics and sub-tropics, where Ficus communities are rich in closelyrelated species (Berg and Corner 2005), many of which share congeneric pollinators. This was confirmed by our survey of fig wasps reared from closely-related Ficus species in Guangdong Province, China where several Ficus species were found to be hosts of two or more pollinator species (Yu et al. 2021). Further samples of an atypical population of F. hirta Vahl., where the trees produce unusually large figs that are similar in size to those of the related F. triloba Buch.-Ham. ex Voigt found that both trees could support two species of Valisia fig wasps at the site, one of which we had previously recorded as associated with F. hirta, while the other is recorded as the pollinator of F. triloba in Taiwan (Chen and Chou 1997). In

Focal species
The mutualism between F. hirta and its fig wasp pollinators does not follow the classic one pollinator: one Ficus pattern. Recent molecular studies detected nine different closely related Valisia species breeding in figs of F. hirta in different parts of its extensive range (Yu et al. 2019). Most of the species have not been formally described, and there is uncertainty about which of the currently available names should be applied. Valisia javana javana was described from F. hirta collected in Java (Mayr 1885) and Valisia javana hilli (as Blastophaga javana hilli) was described from figs of F. hirta collected in Hong Kong (Wiebes 1994;Cruaud et al. 2010). Here for convenience, we use the name 'V. javana sp. 1' for what is one of several species in this aggregate. In addition to the nine pollinator species already recorded from F. hirta, we record here a tenth species, Valisia esquirolianae, which has previously been reared from figs of F. triloba (as F. esquirolianae) in Taiwan (Chen and Chou 1997). V. esquirolianae differs morphologically from V. javana (Wiebes 1994;Chen and Chou 1997). . They differ in growth form, with F. hirta usually a shrub or a small tree not higher than 5 m and F. triloba forming a small tree up to 15 m tall (Berg et al. 2011). The figs of both species are produced in the leaf axils and are more variable in shape and larger in F. triloba, reaching about 30 mm in diameter, whereas mature figs from most populations of F. hirta reach a diameter of 15 mm. However, at the northern edge of its distribution the diameters of F. hirta figs can reach 20 mm (Kuaraksa et al. 2012;Yu et al. 2018). Production of figs by F. hirta is asynchronous, with figs of different ages routinely present on the same individuals (Kuaraksa et al. 2012). This allows pollinators emerging from its figs on male plants to potentially re-enter figs on the same trees. In contrast, F. triloba individuals usually produce discrete crops of similar-aged figs, which forces its pollinators to fly to other trees though there is also some synchrony among trees. There is nonetheless occasional asynchronous fig production within trees, especially by male plants (Kuaraksa et al. 2012). F. triloba can produce up to three crops a year (Kuaraksa et al. 2012). F. hirta produces figs year-round, but in the south of China fig production peaks during the wet season from April to October (Yu et al. 2006(Yu et al. , 2008.

Source populations
Figs of F. hirta in Dapu City (N24°14′19″, E116°49′13″; DAP) of Guangdong province, were collected after it was noted that the figs (of both sexes) on the plants at this site had noticeably larger figs than is typical for this species, closer in size to figs of F. triloba. Site DAP has a seasonal climate with a mean annual temperature of around 21.1 °C and January as its coldest month (Liu et al. 2016). F. triloba is also present in the area (Fig. 1).
Further samples of F. hirta, F. triloba and their pollinators were collected from three sites in Guangdong province where the two trees can be found growing close together (Fig. 1). The South China Botanical Garden (N 23° 10′ 46″, E113° 21′ 06″; SCBG) is about 90 km east of Dinghu Mountain (N 23° 09′ 21″, E112° 30′ 39″; DHS). Shimen Mountain (N 23° 38′ 7″, 113°46′ 20″; SHI), is a further 60 km northeast of SCBG. Edaphic and climatic conditions at the three sites are similar with mean annual temperature around 21.5 °C and January the coldest month with a mean of 12.4-13.1 °C (Yu et al. 2006;Han et al. 2019). Ficus hirta is an abundant and naturallyoccurring species in and around the South China Botanical Garden, whereas we only know of a single (apparently naturally occurring) male individual of F. triloba there. Dinghu Mountain has semi-natural vegetation, described as southern subtropical monsoon evergreen broadleaved forest (Han et al. 2019). Both F. hirta and F. triloba are common there, whereas F. triloba is more common than F. hirta at Shimen Mountain, an area of mainly secondary forest. The three sites therefore represent a spectrum from F. hirta being predominant (South China Botanical Garden) to being the less abundant of the two species (Shimen Mountain).
F. triloba was also sampled from Gaoxiong city in Taiwan (N22°53′54″, 120°43′45″; TAI), an island where F. hirta is absent (Fig. 1;Tzeng 2004;Bain et al. 2015). Taiwan has a subtropical monsoon climate with a distinct wet season from April to September (Fig. 1). . We included representatives of these two genera (two species of Ceratosolen and 11 species of Kradibia) in our phylogenetic analysis of the Valisia pollinators (see Yu et al. 2019).

Ficus species reproductive phenology
The pollinators of dioecious figs appear to disperse less widely than those of monoecious species, reflecting their hosts' small stature, high local population densities and often asynchronous within-tree fruiting (Harrison and Yamamura 2003;Ahmed et al. 2009;Yang et al. 2015).

Morphological comparisons of pollinator species and their host figs
The diameters of whole receptive male figs of F. hirta from the South China Botanical Garden and Dapu City and F. triloba from Dinghu Mountain were measured across their equator (at right angles to the ostiole) to the nearest 0.1 mm using vernier calipers. For each species, at least five figs from each individual and five or more individuals were measured. An indication of the size of the ostioles through which the insects needed to crawl was obtained by halving receptive male figs from the centre of the ostiole to the centre of attachment of the basal peduncle. The distance between the descending bracts at their widest part (Fig. 2) was measured to the nearest 0.1 mm using photographs taken using a Leica E24W camera. Female fig wasps entering the figs pass through the centre of the descending bracts and this measurement provides an indication of the extent to which the bracts could be pushed aside. The calibrated digital images were then recorded using Imagej (version 1.48) software. Flowers were removed haphazardly from male figs to measure their styles, with the lengths of the styles measured from where the style joins the ovary to the tip of the stigma (Nefdt and Compton 1996;Yu and Compton 2012). Photographs and Imagej software were used for this. For each species, we measured the female flowers and pollinators obtained from at least five figs from each of five or more individuals. Head shape and size in female fig wasps is likely to be under selection pressure generated by the size of the ostioles through which they need to crawl if they are to oviposit (van Noort and Compton 1996;Liu et al. 2013). We measured the size of the heads of female fig wasps reared from F. hirta and F. triloba figs that had been stored in 70% ethanol. The maximum head lengths and widths (across the compound eyes) were measured under a Leica E24W microscope and again using Imagej software (version 1.48). After successful passage through the ostiole, female fig wasps can only lay eggs if their ovipositors are of sufficient length to reach the ovules where the eggs are placed (Nefdt and Compton 1996). The lengths of ovipositors (first and second valvulae) of fig wasps reared from the two host species were measured using the same software after they had been dissected out and placed under a cover slip.

Volatiles emitted by receptive male figs
Floral volatiles released by receptive figs were collected in situ from eight male individuals of F. hirta and five male individuals of F. triloba growing in Dinghu Mountain using the chromatoprobe head-space technique, following methods developed for other fig tree species (Proffit and Johnson 2009;Soler et al. 2012). Samples were analyzed with gas chromatography-mass spectrometry (GC-MS) using a GC-MS QP2010Plus (Shimadzu Instrument Co., Ltd; Japan) following the procedures of Soler et al. (2012). Compounds were mainly identified by matching of mass spectra with the FFNSC1.3 and NIST14S libraries, and by comparison of Kovats retention indices with that reported in the NIST chemistry Web Book (http:// webbo ok. nist. gov) and published data. The peak area of each compound was used for quantification of the relative quantities of each ) and the same PCR amplification reaction as for the COI gene. The amplified PCR products were purified using QIAquick spin columns (Qiagen) and were sequenced in an ABI 3730xl capillary sequencer using BigDye Terminator V 3.1 chemistry (Applied Biosystems). All unique haplotype sequences were deposited in GenBank (accession numbers: MT145612-MT145639 for COI sequences and MT218428 for V. esquirolianae ITS2 sequence). There was only one ITS2 haplotype recorded for each species. The ITS2 haplotype recorded here under 'V. javana' was the same as that of V. javana complex sp.1 published in Yu et al. (2019) in Genbank with accession number MF467418.
We did not detect any indication of pseudo-genes, such as multiple peaks in chromatograms, stop codons or frame shift mutations (Song et al. 2008). Sequences were aligned using MUSCLE (Edgar 2004) implemented in MEGA 6.0 (Tamura et al. 2013) with manual corrections. DnaSP was used to count the number of haplotypes (Librado and Rozas 2009). Neighbour-joining trees were constructed using MEGA 6.06 (Tamura et al. 2013) for both COI and ITS2, and node supports were assessed based on 2000 bootstrap replicates. Finally, Kimura-2-parameter (K2P) distances within and between clades for both COI and ITS2 haplotypes were calculated.

Statistical analysis
Between-site and -species differences in fig diameter, ostiole diameter, style length, and their pollinating wasp ovipositor lengths, head width and head length were examined (after testing for equality of variance) with independant sample t-tests in SPSS 21.0.
Statistical analyses of volatile data were performed using past 3.0. In order to compare patterns of odour composition between two species we performed a nonmetric multidimensional scaling (NMDS). For the comparison of volatile profiles between two species we used the relative proportions of all the compounds emitted by them. A data matrix of pairwise comparisons among samples was then calculated using the Bray Curtis distance index. NMDS was used to find the best low dimensional representation of the distance matrix. The null hypothesis of no difference in patterns of odour composition between species was tested with a permutational multivariate analysis of variance (PERMANOVA) with 9999 permutations.

Reproductive phenology
Small numbers of figs of varying developmental stages were present throughout the year on the figs of F. hirta at Dinghu Mountain. The nine male F. triloba trees at Dinghu Mountain all had figs releasing adult fig wasp offspring during short periods in June-July and again in January (Fig. 3a). In contrast, receptive figs on trees of both sexes of F. triloba were present for extended periods, especially in the wet season from April to September (Fig. 3a). The extended periods when un-entered receptive figs were present suggests that only limited numbers of pollinators were finding and entering the figs of F. triloba. Consequently, receptive male and female figs of F. triloba were often available when both V. esquirolianae and V. javana sp. 1 were emerging from host figs of both F. triloba and F. hirta.

Morphology of receptive figs and fig wasp foundresses
The receptive phase male figs of F. hirta in South China Botanic Garden were much smaller in diameter than those of F. triloba (Table 2, T = 23.35, P < 0.001) and as expected, the diameter of receptive male figs of F. hirta from Dapu City were larger than those from South China Botanic Garden (Table 2, T = 2.96, P = 0.004). The median diameter of receptive male F. hirta figs from Dapu City was nonetheless still smaller than the figs of F. triloba from the same location (Table 2, T = 9.83, P < 0.001), although there was a small overlap in their diameters not seen elsewhere ( Table 2). Measurements of the ostiole bract separation showed a similar pattern as the overall diameters of the figs (Table 2; Fig. 3b), with figs of F. triloba having significantly longer ostiole widths than those of F. hirta at both South China Botanical Garden and Dapu City, despite the ostioles of F. hirta figs being wider at Dapu City than the Garden (T tests, all P < 0.05). This might have made it easier for female V. esquirolianae that had developed in F. triloba figs to enter the figs of F. hirta at Dapu City, if they were significantly larger than those of V. javana sp. 1. However, although V. esquirolianae females were found to have a slightly wider median head width than the V. javana sp. 1 reared from F. hirta at the same site there was almost complete overlap in head sizes between the two species (Table 3; Fig. 3b).
The median style lengths in male figs of F. hirta were significantly shorter than those of F. triloba, but with considerable overlap between the two Ficus species (Table 2, T = 22.11, P < 0.001; Fig. 3c). This difference seems unlikely to differentially impact their associated fig wasps, because the ovipositor lengths of the two species showed complete overlap (Table 3; Fig. 3c).

Volatiles emitted by receptive male figs
The volatile profiles of F. hirta and F. triloba both contained more than 50 compounds, with sesquiterpenes predominant in both species (Table 4 and Additional file 1: Table S1). Fully 48 of the compounds were shared by the two Ficus species and although several of the most abundant volatiles in the profile from F. hirta were relatively less abundant among the volatiles detected from F. triloba, and vice versa, there were three common abundant compounds shared by the   (Table 4). (E)-Caryophyllene was the most abundant compound released by F. hirta and was the second by F. triloba. Alpha copaene was the second most abundant compound released by F. hirta and was the most by F. triloba and has been detected previously among the volatiles released by figs of F. fulva Reinw. ex Blume (Grison et al. 1999) which is in the same section as F. hirta and F. triloba. We do not know which of the 50 + compounds are attractive (or repellant) to the two pollinators of F. hirta and F. triloba, but it is clear that with their almost total overlap in the volatiles they were releasing, there is considerable scope for cross attraction of the two fig wasp species. The volatile profiles of F. hirta and F. triloba display overlap using NMDS but there was significant difference between them (PERMANOVA, F = 5.944; P = 0.0009; Fig. 3d). One of the species was also reared from F. triloba in Taiwan and was identified as V. esquirolianae (Chen and Chou 1997), where F. hirta is absent. These individuals combined with those in the mainland from both hosts formed a single clade that nested with the V. esquirolianae reared previously from F. triloba in Xishuangbanna (Jiang et al. 2006;Figs. 1, 4; Additional file 1: Tables S2,  S3). The second pollinator reared from both host species was identified as being the V. javana sp.1 recorded by Yu et al. 2019. Ficus hirta therefore supports the development of a tenth pollinator species in addition to the nine species in the V. javana complex that have been recorded    previously (Fig. 4). V. esquirolianae is closely related to the members of V. javana complex (Fig. 4), but the relationships among V. javana sp.1, V. esquirolianae and other species in this complex are unclear because of low bootstrap values (Fig. 4). At sites where both Ficus species were sampled the two species of fig wasps were usually associated with male figs of both host plants, though their relative frequencies varied between sites and hosts (

Discussion
We reared the same two fig wasp species from male figs of F. hirta and F. triloba at most of the Guangdong province sites we investigated. Our hypothesis that figs of F. hirta at Dapu City were supporting development of V. esquirolianae because its figs were larger there than elsewhere can therefore be rejected, though F. hirta at this site did support more V. esquirolianae than elsewhere. Monitoring of fig wasp entry success into figs of different sizes will be required to determine whether penetration by either Valisia species is influenced by their head widths relative to ostiole sizes (Liu et al. 2013). The shared use of the figs of the two Ficus species is likely to have been facilitated by individuals of the two species sometimes growing in close proximity, their complementary fruiting phenologies and the rather similar volatile blends released by their receptive figs, possibly together with the overlap in sizes of the two fig wasp species. Studies involving electroantennogram and behavioral responses of the two fig wasps to known volatiles are required before an assessment can be made of the extent to which the two Ficus species share functionally significant volatiles. Despite F. triloba producing larger figs, the style lengths of its female flowers also overlapped with those inside the smaller figs of F. hirta, so after they enter male figs the two Valisia species are likely to find oviposition equally easy.
Our results suggest that transfer of V. javana females into F. triloba figs was more frequent than the entry of V. esquirolianae into figs of F. hirta. The overlap in size of the two pollinators suggests that the marginally larger heads of V. esquirolianae are unlikely to be biologically significant. Alternatively, the asymmetry in host transfer could be a simple reflection of the relative abundance of the two fig wasps, rather than indicating differences in their willingness to enter figs of an unusual host. Differences in the trees' flowering patterns may also be significant. The flowering phenology of F. hirta is typical of many dioecious shrubby Ficus species, with asynchronous development on individual trees. Figs at different developmental stages are therefore present in close proximity, with fig wasps being released nearly all year round and providing opportunities for emerging fig wasps to enter receptive figs on the same male trees (Yu et al. 2006). This is not the case with F. triloba, where more synchronous crops will require movements of pollinators between trees. Furthermore, at the single site where we monitored the phenology of this species we found receptive figs were present for extended periods. This suggests a local shortage of V. esquirolianae, a situation that is likely to make entry of atypical pollinator species more likely (Compton 1990).
The distributions of F. hirta and F. triloba overlap across much of their extensive ranges. Ficus hirta is absent from the island of Taiwan, where V. esquirolianae is the only pollinator recorded for F. triloba so far. Whether the distribution of V. esquirolianae extends to the south of China and southeast Asia are unknown, but extensive sampling of 530 fig wasps from F. hirta in at least 34 locations for did not detect any V. esquirolianae (Yu et al. 2019), but did record a further eight Valisia species. It seems possible that one or more of these species is associated F. triloba as well as F. hirta, and might replace V. esquirolianae in the southern part of its range.
In plants pollinated by insects, their host specificity and behavior can have a major role in determining the extent to which pollen is transferred between individuals and consequently the extent of gene flow between them (Peter and Johnson 2014). We only confirmed the ability of the two fig wasp species to successfully reproduce in the male figs of the two Ficus species, and that sharing of the two fig wasps is widespread. It is nonetheless highly likely that female fig wasps that had developed in either host were also able to enter female figs of both F. hirta and F. triloba irrespective of the identity of their natal figs. Whether heterospecific pollen carried into the figs could result in the production of viable seed is unknown, but post-pollination barriers to hybrid production are weak in Ficus (Ghana et al. 2015;Wang et al. 2016). The pollinator sharing of F. triloba with F. hirta therefore has the potential to result in gene flow between them (Arnold 1997; Coyne and Orr 2004).