Making the most of your pollinators: An epiphytic fig tree encourages its pollinators to roam between figs

Abstract Ficus species are characterized by their unusual enclosed inflorescences (figs) and their relationship with obligate pollinator fig wasps (Agaonidae). Fig trees have a variety of growth forms, but true epiphytes are rare, and one example is Ficus deltoidea of Southeast Asia. Presumably as an adaptation to epiphytism, inflorescence design in this species is exceptional, with very few flowers in female (seed‐producing) figs and unusually large seeds. Figs on male (pollinator offspring‐generating) trees have many more flowers. Many fig wasps pollinate one fig each, but because of the low number of flowers per fig, efficient utilization by F. deltoidea's pollinators depends on pollinators entering several female figs. We hypothesized that it is in the interest of the plants to allow pollinators to re‐emerge from figs on both male and female trees and that selection favors pollinator roaming because it increases their own reproductive success. Our manipulations of Blastophaga sp. pollinators in a Malaysian oil palm plantation confirmed that individual pollinators do routinely enter several figs of both sexes. Entering additional figs generated more seeds per pollinator on female trees and more pollinator offspring on male trees. Offspring sex ratios in subsequently entered figs were often less female‐biased than in the first figs they entered, which reduced their immediate value to male trees because only female offspring carry their pollen. Small numbers of large seeds in female figs of epiphytic F. deltoidea may reflect constraints on overall female fig size, because pollinator exploitation depends on mutual mimicry between male and female figs.

reproductive success of a plant, external selection pressures influencing inflorescence design vary with the specific behavior of pollinators (Endress, 2010). These selection pressures may also act differently on male and female reproductive functions, driving differences between male, female, and bisexual inflorescences (Emms et al., 1997;Fishbein & Venable, 1996;Lau et al., 2008). After pollination, the behavior of any animals responsible for the dispersal of the seeds may also have a significant influence on inflorescence design, because the size and location of inflorescences largely determine infructescence characteristics and which animals are attracted to them (Fleming & Kress, 2011).
Although the basic structure of figs is conserved throughout the genus, within the over 800 described species there is considerable variation in where the inflorescences are produced (underground, from the trunk, or among the leaves), whether both male and female flowers are present, whether male flowers are scattered or aggregated, how many flowers are present, the external diameter of the figs, and whether they ripen to offer a reward to frugivores (Berg & Corner, 2005). The overall structure of figs, and the co-adaptive features they share with their pollinators, has apparently remained essentially unchanged for tens of millions of years, and much of the current variation in the size and location of figs may reflect adaptations to facilitate the attraction of different groups of vertebrate frugivores (Compton et al., 2010).  (Cornille et al., 2012). The specificity of the association is achieved mainly through the release of species-and developmental stagespecific attractant volatiles at the time when the female flowers are ready to be pollinated (van Noort et al., 1989;Wang et al., 2013).
When passing through the ostiole, pollinators lose their wings and parts of their antennae (Kjellberg et al., 2005) and ostiole shape and size are reflected in the head shapes of pollinators (van Noort & Compton, 1996). Once inside, foundresses attempt to oviposit into seeds, but cannot herself reproduce. This is a consequence of the long styles of flowers in the female figs and the structure of their stigmas, which together prevent the wasps from galling the ovules and laying their eggs (Kjellberg et al., 2014).  (Grafen & Godfray, 1991). In particular, conspecific male and female receptive figs release broadly similar blends of attractant volatiles (Proffit et al., 2020). After entry into female figs, foundresses of actively pollinating species  remove pollen from their pollen baskets and deposit it on the stigmas, despite being unable to oviposit. This reflects another example of vicariant selection acting on those foundresses that entered male figs and had offspring, because all foundresses that enter female figs fail to reproduce (Raja et al., 2008a). Pollinators that transport pollen passively lack direct control of pollen deposition, but are likely to shed more pollen the longer they remain active inside a fig, searching in vain for suitable oviposition sites.
Many foundresses never re-emerge from the first receptive figs they enter, but the frequency of pollinator foundress re-emergence after entering a fig varies between species (Gibernau et al., 1996;Suleman et al., 2013) and may be more common among pollinators of dioecious than monoecious fig trees .
Re-emergence depends on the figs having their ostioles open for an extended period and for the pollinator foundresses to exhibit appropriate behavior. Natural selection will favor a willingness to re-emerge if the foundresses that do so have more offspring than those that choose to remain in the first fig they enter and lay all their eggs there (Suleman et al., 2013). Only one offspring can develop in a single flower (Jousselin et al., 2002) (Gibernau et al., 1996). The benefits of re-emergence will also vary according to local environmental factors (Gibernau et al., 1996), including the likelihood of predation by ants (Bain et al., 2014). Re-emergence may also be flexible and responsive to aggression between ovipositing foundresses (Moore & Greeff, 2003) and the extent of competition for the limited number of oviposition sites available in any one fig .  (Peng et al., 2014). This effect is generated largely by females laying more male eggs at the start of an oviposition sequence in combination with increasing competition for oviposition sites resulting in fewer offspring per foundress in figs shared with others (Greeff & Newman, 2011;Raja et al., 2008a). In a Ficus species where foundresses can pollinate and lay eggs in several figs, it was found that under glasshouse conditions brood sizes were larger in the first figs they entered and that later broods contained a higher proportion of male offspring (Greeff et al., 2020;Raja et al., 2008a). Only female fig wasps carry pollen, so less strongly female-biased sex ratios are less productive for male plants, but this may be partially compensated for by more pollen per The dioecious F. deltoidea is highly unusual among Ficus species in that the figs produced by female plants contain only small numbers of flowers, and in some varieties just a single flower (Corner, 1969).
Its seeds are also unusually large, a feature that may be linked to the ability of some varieties to grow as true epiphytes (Corner, 1969).
This growth form is rare among fig trees (in contrast to the many hemi-epiphytic stranglers that start as epiphytes, but eventually send roots to the ground).
One variety of F. deltoidea (var. angustifolia) occurs regularly as an epiphyte in oil palm plantations in peninsular West Malaysia (Mohd Hatta, 2019). Seed production in Ficus species is often pollinatorlimited (Bronstein, 1988;Compton et al., 1994), and a large majority It can only take place on male trees because that is where the fig wasps can successfully reproduce, but its effects can provide vicariant benefits for the female trees, as is the case with the apparent mutual mimicry in attractant volatile production by male and female figs (Grafen & Godfray, 1991). This is analogous to the selection for active pollen dispersal in female figs of other Ficus species, which occurs despite the behavior being of no benefit to the individual foundresses (Raja et al., 2008a).
We sought to understand the pollination biology of one of the varieties of F. deltoidea that has few flowers in its female figs. The specific questions we asked were (1) Do foundresses re-emerge from figs and do rates of emergence differ between male and female figs? (2) How many figs can a single foundress enter? (3) How many seeds are generated in female figs and do pollination rates per fig decline when a foundress enters additional figs? And (4) How many offspring do foundresses generate in their first and subsequent male figs, do females that re-emerge produce more offspring, and do their offspring sex ratios change?

| Study site and study species
The study was carried out in Banting district, Selangor, West

| Natural pollination rates
The behavior of Blastophaga sp. in the plantation was recorded using 200 haphazardly collected figs from 21 male and 10 female trees, each of which was located on a different oil palm trunk. The figs were at early C phase (sensu Galil & Eisikowitch, 1968), the stage of development when the figs contain galled ovules (male trees) or developing seeds (female trees) and the bodies and wings of foundress females that died in the figs were also present. Seed or gall development confirmed that a foundress had entered the figs, and the presence of wings indicated that the fig had been the first one to be entered by at least one of the foundresses. The development of figs where no foundress was present indicated that one or more foundresses had entered, but had then re-emerged.
When the body of a foundress was present, the position and orientation of its head were recorded.

| Data analysis
Statistics were performed in R (1.0.153) and SPSS Statistics 20.
The frequency of figs entered by a single foundress was analyzed using a chi-square test. The total brood size for foundresses that entered different numbers of figs, total brood size of first entered and subsequent male figs, and total seed production in the first and subsequent female figs entered were analyzed using generalized linear models (GLMs) with the Poisson error distributions.
If overdispersion occurred, quasi-Poisson errors were used for counts of seeds and numbers of offspring. Offspring sex ratios of emerged and nonemerged foundresses were analyzed using generalized linear models (GLMs) with a Gaussian error and offspring sex ratios in the first and subsequently entered figs were compared using generalized linear models (GLMs) with a quasi-binomial error distribution and logit link. Spearman's rank correlations were used when examining the relationship between brood size and sex ratio in the experimental figs.

| Entry and emergence after experimental foundress introductions
Evidence of the entry of the lone foundresses that had been introduced into the fine nylon mesh bags was provided by the develop-

| Pollinator offspring in male figs
There was a large difference in the total brood sizes achieved by Blastophaga sp. foundresses that re-emerged and produced offspring in two or more figs and those that only produced offspring in the first figs they entered, with the former generating roughly twice the number of offspring (Figure 2, Table 1). Total brood sizes of females that laid all their eggs in a single fig ranged from 42 to 137 (mean ± SE = 80.70 ± 14.24, n = 6), whereas reemerging females had total brood sizes that ranged from 26 to 236 (mean ± SE = 161.42 ± 16.65, n = 12) (GLM, χ 2 = 100.15, df = 1, p < .05). There was also a significant difference in their combined brood sizes depending on the total number of figs they entered (GLM, χ 2 = 348.35, df = 3, p < .001) with the maximum number of offspring generated among foundresses that had entered three or four figs (Figure 2).
More offspring were present in the first figs entered by foundresses (mean ± SE = 79.11 ± 7.32, n = 18 first figs) than in the figs entered by foundresses that had already entered other figs before then (mean ± SE = 45.32 ± 5.20, n = 22 figs) (GLM, χ 2 = 185.72, df = 1, p < .001), based on all foundresses including those that entered just one fig (Figure 3). The mean ± SE offspring in the first figs of re-emerging and non-re-emerging foundresses were 78.33 ± 8.84 and 80.70 ± 14.24, respectively, but foundresses that re-emerged did not have significantly smaller brood sizes in their   design will nonetheless have favored those foundresses that opted to re-emerge because they were sheltered from predators, including ants, as they were walking between figs. The roughly doubling in offspring numbers that they achieved by re-emergence is therefore likely to be greater than would be achieved under fully natural con-  (Anstett et al., 1998). Entry into male figs of F. deltoidea var. angustifolia by more than one foundress meant that a reduced number of oviposition sites were available to later-entering foundresses, so their clutch sizes were necessarily smaller. Under these circumstances, male offspring become more valuable to these foundresses because they can mate with the female offspring produced by earlier-entering foundresses. Changes in offspring sex ratios were observed in single foundress Blastophaga sp., even though there was no competition for oviposition sites with other foundresses, with offspring sex ratios tending to be more female-biased in the first figs they entered, where their clutch sizes were larger.   The pollinators of these varieties would seem to have less incentive to re-emerge from the first figs they enter, unless competition for oviposition sites is generated by routinely high numbers of foundresses entering each male fig. Phylogenetic relationships within the complex are unclear (Corner, 1969), and a reliable phylogeny, in combination with comparative behavioral studies, will be needed before the interplay between pollinator behavior and their unusual inflorescence design can be fully understood.

| CON CLUS ION
Evolutionary innovations are potentially constrained by the be-

ACK N OWLED G M ENTS
The paper was improved by the valuable comments of two anonymous referees. The authors would like to express our gratitude to Puan Siti for giving us permission to conduct the work in her plan-

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