Differential effects of nematode infection on pollinating and non- pollinating fig wasps: Can shared antagonism provide net benefits to a mutualism?

1. Species pairs that form mutualistic associations are also components of broader organismal community networks. These interaction networks have shaped the evolution of individual mutualisms through interspecific interactions ranging from secondarily mutualistic to intensely antagonistic. Our understanding of this com plex context remains limited because characterizing the impacts of species in teracting with focal mutualists is often difficult. How is the fitness of mutualists impacted by the co- occurring interactive network of community associates? 2. We investigated this context using a model interaction network comprised of a fig and fig wasp mutualist, eight non-


| INTRODUC TI ON
Mutualisms, or reciprocally beneficial interspecific interactions, are ubiquitous in nature and strongly influence ecological processes that, in turn, have shaped the trajectories of organismal evolution . Therefore, understanding the ecology and evolution of mutualistic associations is a crucial component to understanding ecosystem function (Bronstein, 2015). To date, the majority of theoretical (Archetti, 2019;Ferriere et al. 2002) and empirical (Heil et al. 2009;Nelson et al. 2018;Paterson et al. 2010) studies have focused on the pairwise interaction between obligate mutualistic partners. Virtually all mutualistic species pairs, however, are members of more complex networks of organismal interactions that may range from secondarily mutualistic to neutral, or strongly antagonistic in nature (Fath, 2007;Melián et al. 2009). This context is often lacking (though expanding; see Nuismer et al. 2018 andArroyo-Correa et al. 2019), but necessary for a clearer understanding of the ecological and evolutionary dynamics of mutualistic systems (Hall et al. 2020;Levine et al. 2017).
Generally, ecological theory predicts that interaction with community-level associates stabilizes or enhances mutualism fitness (Banerjee et al. 2020;Chagnon et al. 2020;Jones et al. 2009; Morris et al. 2003). However, works estimating negative (Bachelot & Lee, 2020;Ferriere et al. 2002;Mougi & Kondoh, 2014) and neutral (Arizmendi et al. 1996;Bronstein, 2001) effects also exist, showcasing a presumed role for context dependence in the diversity of species assemblages. The body of empirical research evaluating the role of interaction networks on mutualism fitness is limited, but growing (Song et al. 2020; Thompson & Fernandez, 2006). One impediment to investigating the effects of community-level antagonism on mutualism fitness is that lifetime fitness in many systems is difficult to quantify (Bronstein, 2015;West et al. 1996). This can be alleviated by focusing on model systems in which all intimately interacting species are known, ecological roles as mutualists and exploiters are well-understood and key components of lifetime reproductive success are easily estimated.  (Janzen, 1979). Pollinators have short adult life spans (<60 hr; Kjellberg et al. 1988), but excellent dispersal capabilities, exploiting wind currents to reach receptive trees that are often located many kilometres from their natal trees (Harrison & Rasplus, 2006;Nason et al. 1998).
In addition to obligate mutualistic relationships with pollinating wasps, individual species of Ficus are subject to exploitation by a diversity of non-pollinating fig wasp (NPFW) genera (multiple Families; Bouček, 1993). Each fig species typically supports at least one, and often several, NPFW species (Compton & Hawkins, 1992) and, like pollinating wasps, many are host fig specific (though exceptions exist, see Farache et al. 2018) and appear to be attracted to receptive figs by the same volatile blends produced to attract pollinators (Proffit et al. 2018). In contrast to the pollinator, which enters and  (Borges, 2015;Zhang & Li, 2020 Davies et al. 2017;Martin et al. 1973;Susoy et al. 2016;Vovlas & Larizza, 1996;Woodruff & Phillips, 2018).
Entomopathogenic nematodes of the genus Parasitodiplogaster are not so great as to prohibit successful dispersal to trees bearing receptive stage figs (Gupta & Borges, 2019;Herre, 1995;Van Goor et al. 2018). Despite this constraint, the virulence of nematode infection varies across species as a function of host-wasp species population density (Herre, 1993) and can range from avirulent or commensal (Shi et al. 2019;Van Goor et al. 2018,) to virulent (Herre, 1993(Herre, , 1995, Pollinator wasp hosts enter figs to lay their eggs, granting infective nematodes access to the next generation of emerging hosts. a behaviour that should be strongly selected against (Giblin-Davis et al. 1995;Krishnan et al. 2010;Vovlas & Larizza, 1996). Surprisingly, however, Parasitodiplogaster has been reported to infect multiple Further, the tightly co-occurring nature of this model system allows unprecedented ability to evaluate the complex role of interacting antagonists of varying ecology on focal mutualist-partner reproductive success over space and time.
Here, we investigate how Parasitodiplogaster nematode infection may limit NPFW fitness and, in turn, potentially benefit the mutualistic partnership between figs and wasps ( Figure 2). If nematode infection of NPFWs is widespread and significantly reduces NPFW reproductive ability, we can hypothesize a previously un- The directionality and significance of the association observed between species offspring production in these models can allow for the inference of ecology and potential antagonism against figs and pollinators, but should be viewed in the appropriate ecological context (Raja et al. 2015). In general, larger, more productive figs may produce more pollinators, seeds and NPFW offspring, making positive correlations between interacting associates a meaningful null hypothesis. However, significant positive associations could also predict kleptoparasitism or parasitoid infection, while significant negative associations could suggest competition for resources. In all models, the predictor variables site, tree nested within site and season were treated as random effects because of the over-dispersion of pollinator offspring and seed counts observed between trees within sites over time. The glmer function does not report p-values for random effect variables or associated nested terms.

| Frequency of interaction between nematodes and NPFWs
We sampled F. petiolaris-associated wasps of each NPFW species emerging from mature figs across each study site and through time. For wasps from nematode-infested figs, the thoracic and abdominal cavities were dissected using 0.25-mm diameter tungsten

| Does nematode infection influence NPFW life history?
Previous

| Which NPFWs antagonize the F. petiolaris mutualism?
The four field collections conducted from 2012 to 2014 yielded a total of 2,187 mature, wasp-producing figs. We obtained an aver- Interestingly, pollinator offspring production had a range of significant (both positive and negative) and non-significant associations with NPFWs (Table 1).
Of the mature figs collected, a total of 120 (60 from two sites) were examined to investigate potential antagonistic effects of  (Table 1).

| Frequency of interaction between nematodes and NPFWs
Parasitodiplogaster infestation was observed in 36% (780 of 2,187) of all mature figs sampled, varying between 12% and 80% depending on individual study site and collection trip. From the infested figs, a total of 2,791 emerging pollinators and NPFWs (range of 4 to 1,182 individuals depending on species) were dissected to determine the presence and number of infective juvenile nematodes. With the exception of Sycophila (presumably due to very low sample sizes), all NPFW species were found to be infected by Parasitodiplogaster, with the incidence of infection in individual wasps varying substantially among species, ranging from 6.7% to 39.6% (Table 2). Interestingly, the number of nematodes per infective event also varied substantially among NPFW species. Idarnes flavicollis and Heterandrium 1, in particular, experienced the highest incidences of nematode infection (39.6% and 27.8% respectively) and also highest average nematode loads (2.52 and 2.32 nematodes per host). These nematode loads were significantly greater than in other NPFW species, though not as high as in pollinators (Table S3).

| Does nematode infection influence NPFW life history?
Nematode infection of NPFWs appears to be widespread and may impact their fitness through reduced longevity and dispersal ability, but this effect has not been previously examined. A total of 281 NPFWs of various species were collected as they were arriving at receptive fig trees. Although we were able to sample six of the eight NPFW species associated with F. petiolaris (Table S4), only Idarnes flavicollis and Idarnes carme 1 and 2 (the most common NPFWs, Table S2) were collected frequently, with sample sizes of 111, 86 and 71 individuals respectively. In each of these Idarnes species, the incidence of nematode infection in individuals arriving at receptive fig trees (3.6%, 1.2% and 2.8% respectively) was lower than in the population as a whole (Table 3). Further, arriving wasps had fewer infective nematodes per individual than did wasps departing nematode-infested figs (poisson. test, p-values = <0.001, 0.005 and <0.001 respectively; Table S3).
Indeed, virtually all arriving infected wasps were infected by only a single juvenile nematode (Table S4), compared to a mean of 2.5 nematodes per infected emerging Idarnes flavicollis wasps (Table 2). Thus, in these three NPFW species we found that both the rate of infection and the number of nematode individuals observed per infective event were lower in wasps that successfully dispersed to receptive figs as compared to wasps emerging from infested figs (Table 3).   Figure S2a and b respectively). Further, we did not find a significant association of the number of nematodes extracted per wasp host on the hour in which that wasp died (see Tables S5 and S6; Figure S3).

| D ISCUSS I ON
Nematode infection of F. petiolaris pollinators, while common and significantly negatively associated with offspring production, is of relatively benign effect, limiting pollinator production by less than 1% each generation (consistent with Gupta & Borges, 2019, Shi et al. 2019and Van Goor et al. 2018). In F. petiolaris, it has been previously observed that nematode infection does not appear to limit pollinator longevity, dispersal ability or offspring production except infrequently when many (10 or more) nematode individuals infect the same host (Van Goor et al. 2018). Here, we found a significant positive association between nematode infestation and seed production (an increase of 10%). This surprising, apparently beneficial effect has not been previously reported and the mechanism responsible for increased seed production is not currently understood. Further,

| Which NPFWs antagonize the F. petiolaris mutualism?
Biological communities are structured and modulated through organismal network interactions ranging from mutualistic to antagonistic (De Andreazzi et al. 2019). Individual species may enact profoundly higher influence on focal species than others, which has been predicted to either stabilize or destabilize network structure (Bachelot & Lee, 2020;Heil et al. 2009;Montesinos-Navarro et al. 2017). A more robust understanding of network-level interactions can provide essential nuance to the evolutionary history and trajectory of whole communities (Paterson et al. 2010). Like the vast majority of monoecious fig systems, the NPFW community associated with F. petiolaris is speciose (Bouček, 1993). Interestingly, in F. petiolaris, the total production of NPFWs per fig typically outnumbers the production of pollinating wasp mutualists (Table S2). We observed that 7% (160) of all figs surveyed produced NPFW offspring in the absence of a pollinating foundress, and that all NPFW genera were produced in these figs. The fact that Parasitodiplogaster nematodes were not observed in any of the zero-foundress figs reinforces our contention that NPFW species are not vectors for nematode transmission to the interior of receptive figs, as has been previously suggested (Giblin-Davis et al. 1995;Jauharlina et al. 2012;Vovlas & Larizza, 1996).
Within the NPFW community of F. petiolaris, three wasps of the genus Idarnes were particularly common, consisting of nearly TA B L E 3 Infection and dispersal data for the most abundant NPFWs associated with Ficus petiolaris. To estimate the percentage of the total wasp population infected with nematodes, we multiplied percentage of infected wasps emerging from infested figs by the percentage of all figs found to be infested with nematodes (36%). To estimate the rate of failure of nematode-infected wasps to disperse, we divided the percentage of individuals arriving with nematodes by the percentage emerging with nematodes. Finally, to estimate the percentage of NPFW individuals removed from the population due to nematode infection, we multiplied the total percentage of individuals infected by the failure to disperse rate

F I G U R E 3 Survivorship of Idarnes flavicollis participants in the
Ficus petiolaris longevity experiments for nematode infected and uninfected individuals. Nematode infection was found here to significantly reduce wasp longevity when compared to uninfected individuals 50% of all wasps collected (similar to findings in Farache et al. 2018 andWest et al. 1996), making them particularly relevant antagonists for network-level evaluation. Of the F. petiolaris-associated Idarnes, Idarnes flavicollis accounted for 24% of all wasps sampled (Table S2). Indeed, in the GLMM analyses that were performed, we found strong negative associations between Idarnes flavicollis offspring production and both pollinator offspring production and fig seed production through competition for ovule oviposition sites (Table 1)

| Frequency of interaction between nematodes and NPFWs
Intriguingly, nearly all NPFW species associated with F. petiolaris were found to be subject to infection by Parasitodiplogaster nema-  (Patel et al. 1995) and herbivorous insects (Silva & Clarke, 2019). Most figs in F. petiolaris are visited by a single pollinating foundress (Duthie & Nason, 2016;Van Goor et al. 2018), which, if infected, will likely only introduce a single lineage of highly inbred nematodes. The infection of NPFWs instead of definitive pollinating wasp hosts could be explained through kinship theory (Hamilton, 1964) related to selection to infect but not over-infect dispersing pollinators (as described in Gupta & Borges, 2020) and to prevent them from doing nothing (and perishing anyway), but the genetic data needed to support this hypothesis are not currently available.
Pollinating fig wasps are the 'appropriate' hosts for Parasitodiplogaster nematodes because these wasps enter receptive figs and secure reproductive space for nematodes. Indeed, we observed here that pollinators are infected more frequently (Table 2) and with significantly higher nematode loads (Table S3)

| Does nematode infection influence NPFW life history?
NPFWs that fail to arrive at receptive figs are incapable of reproduc-  Figure S3); it appears as if the nematodes that enter a NPFW remain there at least until the time of wasp mortality.
It was previously estimated that pollinators average four nematodes per infection event and lose 2.8% of the general population each generation due to Parasitodiplogaster exploitation (Van Goor et al. 2018). Here, we find that the NPFW species associated with F. petiolaris are not only subject to nematode infection, but they are likely sensitive to infection by even a single nematode. With data from the more abundant arriving Idarnes wasps, we similarly estimated net losses due to infection, finding that they are similar to substantially higher than those suffered by pollinators (Table 3) ing that nematode impacts on NPFW communities are much more widespread than previously recognized (including in two separate Ficus subgenera). In particular, NPFWs associated with F. popenoei may be infected more frequently than pollinators (Table S7) and may experience more detrimental fitness limitations (Tables S8 and S9).
In aggregate, this suggests that nematode infection may remove a sizeable proportion of the total wasp-antagonist community in each generation. This may be explained as an indirect effect (Gillespie & Adler, 2013;Guimarães et al. 2017) or may represent a novel densitydependent facultative mutualism between Parasitodiplogaster nematodes, Pegoscapus pollinating wasps and F. petiolaris. Ultimately, this may present a mechanism through which antagonist communities are modulated over shared network resources in other non-fig systems as well, especially those characterized by multiple arthropod species co-occurring in ephemeral environments or within the same host. Such 'the enemy of my enemy is my friend' scenarios may be much more common in nature than currently appreciated, but requires careful future evaluation.  (Michailides et al. 1996). The mechanism underlying this secondarily mutualistic behaviour of Parasitodiplogaster for fig systems is currently unknown, but will be the target of future research. Together, these interactions suggest more ecologically profound roles that Parasitodiplogaster (and perhaps other nematodes) may provide for their hosts. Similar facultative mutualisms may potentially act as hidden drivers for community-network dynamics elsewhere, especially in lesser-studied, invertebrate-rich assemblages.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data and code used within this manuscript will be publically avail- Derek D. Houston https://orcid.org/0000-0002-2730-1159