The Extent of Pollinator Sharing Among Fig Trees in Southern China

Background: The obligate mutualism between g trees (Ficus, Moraceae) and pollinating g wasps (Agaonidae) is a model system for studying co-evolution due to its perceived extreme specicity, but recent studies have reported a number of examples of trees pollinated by more than one g wasp or sharing pollinators with other trees. This makes pollen ow between species and hybridization more likely. We reared pollinator g wasps from gs of 13 Chinese g tree species trees and established their identity using genetic methods in order to investigate the extent to which are they were supporting more than one species of pollinator. Results: Our results showed 1) pollinator sharing was frequent among closely-related dioecious species (where pollinator offspring and seeds develop on different trees), but not monoecious species and 2) that where two pollinator species were developing in gs of one host species there was usually one g wasp that was far rarer than the other. An exception was F. triloba, where its two pollinators were equally abundant. Conclusions: Our results suggest that host expansion events where pollinators reproduce in gs other than those of their usual hosts are not uncommon among g wasps associated with dioecious hosts. Because closely related trees typically have closely related pollinators that have a very similar appearance, the extent of pollinator-sharing has probably been underestimated. Any pollinators that enter female gs carrying heterospecic pollen could potentially generate hybrid seed, and the extent of hybridization and its signicance may also have been underestimated. Frutescentiae are small shrubs. The monoecious species produce largely synchronous crops, whereas both sexes of the dioecious species usually exhibit asynchronous within-tree fruiting, with gs of different developmental stages present for longer periods on the plants.


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 of free-standing trees, shrubs, climbers, and (hemi-)epiphytes (Corner 1965;Berg 1989;. Figs are de ned by their unique enclosed in orescences (the g or syconium) and their associated pollination system which requires entry into gs by highly-specialized g wasps (Hymenoptera, Agaonidae). Pollinator g wasps enter the gs to lay their eggs inside the ovules of the tiny owers they contain. For a long period it was believed that each species of g tree supported its own unique species of pollinator g wasp, which was associated with no other Ficus species (Janzen 1979). When atypical pollinators were detected within gs it was assumed that these were rare mistakes that resulted in the death of the pollinators without the production of their offspring or generation of fertile seeds (Compton 1990;Ware and Compton 1992). More recently, it has been realized that more than one species of pollinator may be associated routinely with a single species of Ficus, and that widespread g tree species can support multiple pollinators (Molbo et al. 2003;Haine et al. 2006;Sun et al. 2011;Chen et al. 2012;Yu et al. 2019). There are also examples of pollinator sharing, where two or more Ficus species are routinely hosts for a single species of g wasp (Lopez-Vaamonde et al. 2002;Wachi et al. 2016;Wang et al. 2016). The one to one relationship that was originally envisaged is now realized to have been the result of the small number of host records available from each Ficus species, and their limited geographical coverage within the plants' distributions, together with the close morphological similarities of closely-related pollinators making their identi cation di cult. Where two or more pollinators have been recorded as the routine pollinators of a single Ficus species they often appear to be associated with different habitats (Michaloud et al. 1996) and have allopatric or parapatric distributions within the ranges of their hosts. However, sampling intensity is again rarely su cient to con rm this pattern of a single species of pollinator routinely servicing each Ficus species at any given location.
The assumption of extreme host speci city in g wasps was based on a combination of the host records available and the apparent specialized coadaptations required for a g wasp to reproduce inside the gs of each Ficus species. Host choice by pollinators is made by the adult females and centres on long-distance plant-speci c and developmental stage-speci c volatile cues released by the gs when they are ready to be pollinated (van Noort et al. 1989;Grison-Pigé et al. 2002;Hossaert-Mckey et al. 2010). Pollinator females that arrive at a g then need to be able to negotiate their way through a narrow ostiole in order to reach the owers where they lay their eggs, and pollinator head shape is linked to the size of the ostiole (van Noort and . Successful oviposition once inside a g depends on the g wasp having an ovipositor that is longer than the styles through which its eggs are inserted (Nefdt and Compton 1996). Finally, successful development of their offspring depends on a galling response by the plant and gall forming insects are typically highly host speci c (Weiblen 2004;Yu and Compton 2012;Ghana et al. 2015).
Although the relationship between g trees and their pollinators is routinely described as a mutualism, the majority of Ficus species in Asia have a dioecious breeding system, where individual trees have gs that either produce only seeds (on 'female' trees) or only pollinator offspring (on 'male' trees) (Janzen 1979;Berg 2003). This situation contrasts with g trees with a monoecious breeding system, where all the trees have gs that can produce both seeds and support the development of pollinator offspring. Monoecious g trees are often large free-standing trees or stranglers (hemi-epiphytes) growing at low densities in forest habitats, whereas dioecious species are typically smaller and shrubby and more likely to have aggregated distributions (Berg 1990;Yang et al. 2015).
Probably re ecting these differences, some pollinators of monoecious species y and transport pollen for long distances between trees (Ahmed et al. 2009), whereas the pollinators of dioecious Ficus species are believed to usually display more limited dispersal (Harrison and Rasplus 2006;Chen et al. 2011;Nazareno et al. 2013). The longer-distance dispersal exhibited by pollinators of monoecious Ficus might be expected to increase the likelihood of two or more pollinator species developing in gs on a single tree, as has been reported in Africa ), but host shifts are also likely to be easier between closely related species (Rasplus 1996) and there are numerous closely-related dioecious g trees in Asia. Fig trees planted outside their normal range may also be more likely to support multiple pollinators, if their routine pollinators are absent locally (Corner 1965;Compton 1990;Patel et al. 1993).
The extent to which g tree species growing within a single location are supporting more than one species of pollinator remains largely unknown, because most recent studies have concentrated on the pollination biology of individual species of g trees (Chen et al. 2012;Darwell et al. 2014;Bain et al. 2016;Rodriguez et al. 2017;Yu et al. 2019). Pollinator sharing resulting in gene ow between closely-related Ficus species has nonetheless been detected (Wang et al. 2016). Here, we describe a Ficus community approach, where gs from southern China were screened for pollinator identity. The communities included mixtures of native and planted species and trees with both monoecious and dioecious breeding systems. We address the following questions (1) is there any difference on the extent of pollinator sharing between monoecious and dioecious Ficus? and (2) where two pollinators are present, does one species predominate? with 80% of annual precipitation concentrated in April-September. The mean annual temperature is 21.8 C in SCBG (Yu et al. 2006) and 21.9°C in DHS (Han et al. 2019), and the coldest mean monthly temperatures (13.1°C in SCBG and 12.6°C in DHS) occur in January.

Study site
More than 13,000 kinds of living plants are preserved in SCBG. The g trees that support pollinators at SCBG include ve monoecious gs, F. microcarpa, F. benjamina, F. subpisocarpa, F. virens, F. altissima, and seven dioecious g trees, F. hirta, F. triloba (one tree), F. auriculata, F. oligodon (two small trees), F. hispida, F. variegata var. chlorocarpa and F. pumila. In DHS, the natural vegetation comprises mainly southern subtropical monsoon evergreen broadleaved forests, re ecting moist local climatic conditions. The Ficus with pollinators present at DHS are F. microcarpa, F. benjamina, F. subpisocarpa, F. hirta, F. triloba, F. hispida, F. stulosa, F. variegata var. chlorocarpa, F. oligodon, F. erecta and F. pyriformis. The F. auriculata in SCBG are planted though the species is naturally distributed locally. The other dioecious Ficus at the two sites had not been planted, whereas the monoecious species had been planted.
The identi cation of the pollinating wasps from each g tree species The pollinators of four monoecious and six dioecious g tree species were identi ed using DNA sequencing (Table 1). Most non-pollinating g wasp species (belonging to families other than Agaonidae) were excluded, but two species were utilized as out-groups. The four monoecious species are big trees with crops of more than ten thousand gs. F. microcarpa and F. benjamina belong to Urostigma subsection Conosycea, while F. subpisocarpa and F. virens belong to the subsection Urostigma. Among the dioecious species, F. hispida and F. stulosa are small trees belonging to subgenus Sycomorus subsection Sycocarpus (Cruaud et al. 2012). F. oligodon and F. auriculata, are two closely related small trees belonging to subgenus Sycomorus, subsection Neomorphe (Berg 2004). F. hirta and F. triloba belong subgenus Ficus, subsection Eriosycea. F. hirta is a shrub whereas F. triloba is a small tree with larger crops. The three species belonging to subsection Frutescentiae are small shrubs. The monoecious species produce largely synchronous crops, whereas both sexes of the dioecious species usually exhibit asynchronous within-tree fruiting, with gs of different developmental stages present for longer periods on the plants. The mitochondrial genetic marker mtCOI was sequenced from an average of 23.2 g wasp individuals reared from male gs of each Ficus species (range 8-35, total 301). Five g wasp genera were represented (Table 1 and Table 2). All the sequenced g wasps were adult offspring and therefore had developed successfully in the g tree species from which they were reared. Genomic DNA was extracted from the whole body of each g wasp using the EasyPure Genomic DNA Extraction Kit (TransGen, Beijing, China). A 435-710 bp fragment of the mtCOI gene for each pollinating species was then sequenced following the protocol used in previous studies (Tian et al. 2015). The reaction was optimized and programmed on a MJ Thermal Cycler (PTC 200) as one cycle of denaturation at 94°C for 5min, 35 cycles of 30 s denaturation at 94°C, 30 s at a 55°C annealing temperature, and 30 s extension at 72˚C, followed by 8 min extension at 72˚C. All ampli ed PCR products were puri ed 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: MW851213-MW851283).
We did not detect any indications 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 5.0 was used to count the number of haplotypes (Librado and Rozas 2009).
Maximum likelihood trees were constructed using MEGA 6.0 (Tamura et al., 2013) for COI, and node supports were assessed based on 2000 bootstrap replicates. Kimura-2-parameter (K2P) distances within and between clades for COI haplotypes were then calculated. The clades with high gene sequence differences (larger than 0.02), were blasted to Genbank with the rst 1-3 sequences sorted by percent identity. Two species of non-pollinating g wasps reared from F. hirta, Sycoscapter hirticola (MG548706) and Philotrypesis josephi (MG548673 and MG548674, both Pteromalidae) were included as outgroups (Yu et al. 2018).

Results
Phylogenetic analyses of the COI sequences detected 13 pollinator species that had reproduced within the gs of the 13 Ficus species, but there was not a 1:1 concordance between them. All the pollinator clades were strongly supported ( Fig. 1; Table 1), with low within-clade and large between-clade K2P distances (Table S1) and the cumulative distribution of K2P distances indicating a marked barcoding gap between clades (Fig. 2). We therefore treat each clade as a distinct species. Based on the sequences downloaded from GenBank and our de novo sequencing we detected numerous examples of pollinators associated with more than one Ficus species and of Ficus species supporting the development of more than one species of pollinator. Up to three different species of pollinators were reared from the gs of a single host species and up to four host taxa were recorded for a single species of pollinator ( Table 2).
The classical 1:1 pollinator and host Ficus relationship was only detected among two dioecious Ficus species (F. hispida and F. stulosa), but it was the norm among the monoecious g trees, where no pollinator-sharing was detected. Ficus subpisocarpa nonetheless supported the development of two closely-related g wasps, rather than one (Table 2, Fig. 1). As reported previously based on morphological identi cations, the same pollinator species (Blastophaga sp. 1) was reared from F. erecta var. beecheyana, F. pyriformis and F. variolosa, but in addition the same species of g wasp was also reared from gs of F. oligodon, an unrelated g tree. F. oligodon was routinely supporting two species of Ceratosolen, both of which were shared with F. auriculata but no other species. The closely related taxa F. hirta and F. triloba also shared a pollinator (Valisia javana hilli) which was not reared from any other hosts.
Each g wasps were generally reared from one or two host species (Table 2; Fig. 1). Ficus species supporting more than one species of g wasp generally had one predominant pollinator that provided between 90-97% of the total reared individuals. The exception was F. triloba where its two pollinators were present in roughly equal proportions ( Fig. 3; Table 1). Around half of the pollinators reared from F. triloba were V. javana hilli, a species routinely associated with F. hirta (V. javana complex sp. 1 in Yu et al. 2019).

Discussion
Our COI screening detected numerous examples of pollinator g wasp species entering and successfully reproducing in more than a single host Ficus in southern China. Host overlap was frequent among pollinators of dioecious species and in most cases involved pairs of g wasp species where one pollinator predominated and a second was reared only rarely. One interpretation of this is that the more rarely encountered pollinator species had other hosts where they were more abundant, but our screening across different Ficus species was not su ciently extensive to con rm this and in some cases the pollinator species may simply be rare within our sampling area. Most examples of g wasps developing in gs of more than one host involved g trees that were closely related, but there were exceptions involving species of Blastophaga and Ceratosolen that were reared from gs normally associated with the other genus of pollinators. Fig wasp offspring developing successfully in unrelated host Ficus has been recorded previously from Africa (van Noort et al. 2013). This ability to develop inside hosts that are phylogenetically distant shows that the host speci city of g wasps may be determined more by the choices made by searching adult females than by any physiological limitations.
Even within the Ficus species we sampled the size of the samples was not extensive and we are unlikely to have detected the full range of Ficus hosts being utilized locally by the g wasps. There was generally one routine pollinator species combined with rarer entries by two or more additional pollinators (Moe et al. 2011;Yang et al. 2015). An exception to the general pattern of pollinator sharing where one pollinator species predominant was provided by F. triloba, where two pollinator species were present in roughly equal numbers of its gs, but more samples, taken throughout the year, will be needed to con rm this pattern. Some of the trees we sampled were planted individuals and this may have increased the extent of pollinator sharing that we detected. Our results nonetheless suggest that exceptions to the 'classical' one pollinator to one tree relationship are routine among sympatric dioecious g tree species in southern China, to the extent that among trees with this breeding system strict speci city is the exception, not the norm.
Fig wasps develop inside gs on male trees of dioecious g trees, but it is likely that similar entry by two or more pollinator species is taking place in both male and female gs. Pollinator host choice, based mainly around species-speci c volatile attractants released by receptive gs, is the major isolating mechanism that helps prevent heterospeci c pollen being deposited on the owers inside female gs, but is not always effective (Souto-Vilarós et al. 2018).
Other isolating mechanisms such as pollen incompatibility appear to be poorly developed in Ficus (Huang et al. 2019), so whenever g wasps species are entering female gs of two more host trees in an area there is the possibility of viable hybrid seed being developed. The extent to which hybrids can mature successfully and can lead to introgression between species is unknown, but mature natural hybrid individuals are known to occur (Wilde et al. 2020). Some arti cially generated hybrids appear to be at no reproductive disadvantage in terms of seed production, but male hybrid offspring can be sterile because pollinators cannot develop inside their gs (Ghana et al. 2015, but see also Yakushiji 2012), so patterns of introgression may be complex.
Sharing of pollinators was not a feature of the monoecious g tree species we sampled. This is not the case elsewhere, where for example up to four different species of pollinators from gs of a single individual tree in Africa ). Monoecious and dioecious g trees differ in numerous ways that may in uence pollinator behavior including growth form (trees versus shrubs), owering phenology (large synchronous crops versus smaller asynchronous crops) and typical local abundance (monoecious species are often more dispersed). Perhaps more signi cantly in our study area and across SE Asia, there an exceptionally high diversity of dioecious species, most of which are pollinated by g wasps that belong to a small number of genera. Opportunities for chance landing on gs of atypical hosts are therefore greater for those insects associated with dioecious hosts, but in addition most of the sharing of pollinators was between closely related dioecious species, which are likely to be generating relatively similar attractant cues (Wei et al. 2014;Wang et al. 2016). The morphology of closely-related pollinator g wasps is often very similar, and our results emphasize that pollinator-sharing is likely to have been underestimated because of this. Barcoding and other molecular identi cation techniques are used increasingly to distinguish between g wasp species, but our results also highlight the need to sequence g wasps from several gs, even if they look alike, in order to detect pollinator species that may be present at low frequencies. More than one morphologically similar species can even be reproducing within the same individual gs (Sutton et al. 2017).

Conclusions
Our survey of the g wasp pollinators associated with local assemblages of Ficus species in Southern China revealed contrasting pollinator relationships between monoecious and dioecious trees. Monoecious trees and their pollinators largely displayed a highly speci c one pollinator for one tree association.
Among dioecious species there was no such speci city, with frequent sharing of pollinators across trees and two or more species of pollinators associated with each tree species. Possible biological traits favoring this breakdown in pollinator speci city among dioecious Ficus include their extended asynchronous owering phenologies and the mixtures of closely-related species that can grow in close proximity. This lack of speci city suggests that the extent of pollen ow between dioecious g tree species is likely to have been underestimated, with unknown consequences.  Figure 1 COI ML phylogenetic tree of the pollinators of the sympatric gs, with sequences of two non pollinators (Pteromalidae) as outgroups. Node support rates are shown. Haplotypes are also listed together with their host gs.