Genome‐wide sequence data show no evidence of hybridization and introgression among pollinator wasps associated with a community of Panamanian strangler figs

Abstract The specificity of pollinator host choice influences opportunities for reproductive isolation in their host plants. Similarly, host plants can influence opportunities for reproductive isolation in their pollinators. For example, in the fig and fig wasp mutualism, offspring of fig pollinator wasps mate inside the inflorescence that the mothers pollinate. Although often host specific, multiple fig pollinator species are sometimes associated with the same fig species, potentially enabling hybridization between wasp species. Here, we study the 19 pollinator species (Pegoscapus spp.) associated with an entire community of 16 Panamanian strangler fig species (Ficus subgenus Urostigma, section Americanae) to determine whether the previously documented history of pollinator host switching and current host sharing predicts genetic admixture among the pollinator species, as has been observed in their host figs. Specifically, we use genome‐wide ultraconserved element (UCE) loci to estimate phylogenetic relationships and test for hybridization and introgression among the pollinator species. In all cases, we recover well‐delimited pollinator species that contain high interspecific divergence. Even among pairs of pollinator species that currently reproduce within syconia of shared host fig species, we found no evidence of hybridization or introgression. This is in contrast to their host figs, where hybridization and introgression have been detected within this community, and more generally, within figs worldwide. Consistent with general patterns recovered among other obligate pollination mutualisms (e.g. yucca moths and yuccas), our results suggest that while hybridization and introgression are processes operating within the host plants, these processes are relatively unimportant within their associated insect pollinators.


| INTRODUC TI ON
Hybridization between species results in novel genetic combinations derived from divergent parental genomes and can lead to introgression, the transfer of genetic material between species. In many lineages, strong pre-and postzygotic barriers limit hybridization and reinforce species boundaries (e.g. Nosil et al., 2005). Evidence from next-generation sequencing, however, has revealed that hybridization and introgression have occurred throughout the evolutionary history of the tree of life (Mallet et al., 2016;Taylor & Larson, 2019).
The influx of genetic material into a lineage introduces new alleles and multilocus combinations that, depending on the situation, can be either harmful (Rhymer & Simberloff, 1996) or (Dowling & Secor, 1997), and thus impact both the genetic diversity and evolutionary dynamics of a species (Payseur & Rieseberg, 2016).
In plant-insect interactions (pollination or herbivorous), higher levels of host specialization generally correspond to a reduction in the diversity of interspecific interactions, potentially limiting opportunities for plant or insect hybridization (de Medeiros & Farrell, 2020). This is particularly true for brood pollination mutualisms (figs and fig wasps, yuccas and yucca moths, leafflowers and leafflower moths) that often exhibit relatively species-specific host-pollinator relationships (Hembry & Althoff, 2016;Pellmyr et al., 2020;Weiblen, 2002). Therefore, a general question in the evolutionary ecology of host-insect interactions involves identifying the ecological processes that maintain or undermine species boundaries, and generating testable hypotheses concerning factors that affect the evolutionary importance of hybridization and introgression for both host and insect. (e.g. Bronstein, 1987;Jackson et al., 2008;Kerdelhué et al., 1997;Machado et al., 2005;Molbo et al., 2004;Ramírez, 1970;Sutton et al., 2017;Wang et al., 2016;Wiebes, 1995a). For example, Yang et al. (2015) found that about 30% Noort et al., 1989;Wang, Yang, et al., 2021;Ware et al., 1993). When a female wasp arrives at a fig tree, she must enter a syconium through the ostiole (a small terminal pore) that excludes other insects. Once inside, the female wasp (the foundress) pollinates flowers, oviposits into a subset of them inducing gall formation and dies within the syconium (Janzen, 1979). Offspring then develop over several weeks inside the galled, univolute fig flowers (Galil & Eisikowitch, 1968).
Male wasps emerge from their galls, locate galls that contain female wasps and chew holes in the galls to expose females for mating. Importantly, mating among offspring takes place inside the same syconium that was pollinated by the mother. After mating, female pollinators emerge from their galls, gather pollen from male flowers, exit the syconium and disperse. Female wasps routinely travel many kilometres to encounter another receptive fig inflorescence-on typically a conspecific fig tree-to reproduce (Ahmed et al., 2009;Nason et al., 1998).
There are approximately 120 described species of strangler figs (Ficus subgenus Urostigma, section Americanae) in the Neotropics (Berg, 1989). Figs in this group are pollinated by wasps from the genus Pegoscapus (family Agaonidae). Cophylogenetic studies of strangler figs and their pollinators-throughout the Neotropics (e.g. Cruaud et al., 2012) and within Panama (e.g. Jackson et al., 2008;Machado et al., 2005;Satler et  Pollinator host switching appears to be an important mechanism contributing to hybridization and introgression in the Panamanian strangler fig community (Jackson et al., 2008;Machado et al., 2005).
In particular, using genome-scale data coupled with a model-based approach, Satler et al. (2019) demonstrated host switching to be the most important process generating phylogenetic patterns in the community of Panamanian strangler figs and pollinating wasps. The demonstration of pollinator sharing and an evolutionary history of host switching by the pollinators associated with the Panamanian strangler figs (Molbo et al., 2003(Molbo et al., , 2004Satler et al., 2019) is consistent with observed widespread hybridization and subsequent introgression across most of the Panamanian figs (Jackson et al., 2008;Machado et al., 2005). Further, evidence suggests that hybridization occurs in figs more generally and that introgression is a potentially important process in the evolutionary history of Ficus (Bruun-Lund et al., 2017;Compton, 1990;Compton et al., 2009;Cornille et al., 2012;Renoult et al., 2009;Van Noort et al., 2013;Wang et al., 2016;Wilde et al., 2020). all pollinator wasp offspring within the syconium will be her direct descendants-haploid sons and diploid daughters-and mating will be between siblings. In contrast, if two or more foundresses enter and oviposit within the same syconium, there are opportunities for nonsibling mating among their offspring. If these foundresses represent different species, there is the opportunity for hybridization. Since figs vary substantially in their characteristic foundress numbers (Herre, 1985(Herre, , 1989 In addition to evidence of hybridization and introgression in the host strangler figs, hybrid pollinator wasps have been detected between two species within the Panamanian community. For example, Molbo et al. (2003)   .
Including the outgroup, we generated sequence data from 176 wasp samples representing 20 pollinator species, with an average of 8.8 individuals per species ( Table 1, Table S1).
We used pollinator species names when applicable as described by Wiebes (1995aWiebes ( , 1995b. As several pollinators in this community are undescribed, we denote these species as Pegoscapus sp. followed by a unique identifier and their host fig species name. In addition, although a single pollinator species (P. hoffmeyeri) was identified morphologically as being associated with F. obtusifolia (Wiebes, 1995a), the pollinators comprise two cryptic sister species (Molbo et al., 2003). Consistent with previous studies (Jackson et al., 2008;Molbo et al., 2003Molbo et al., , 2004Satler et al., 2019), we denote these two species as P. hoffmeyeri sp. A and P. hoffmeyeri sp. B.
For this, we took our aligned loci (before the use of Gblocks) and mapped cleaned sequence reads for each individual to the UCE locus set using BWA-MEM (Li, 2013) in bwa v0.7.17 (Li & Durbin, 2010).
We then phased the data with samtools v1.9 (Li et al., 2009) using the phase command, calling two alleles per individual per locus. Loci were then realigned and cleaned as described above, with loci sampled for a minimum of 70% of individuals retained for downstream analysis.

| Phylogenetics
We used both concatenation and coalescent-based species tree methods to estimate phylogenetic relationships among the fig wasps. We first estimated a concatenated phylogeny using maximum likelihood (ML) in IQ-TREE v1.6.12 (Chernomor et al., 2016;Nguyen et al., 2015). This approach allows us to test species monophyly as individuals are treated as tips in the tree. The data set was partitioned by UCE locus, with each partition estimated under the GTR + γ substitution model. Nodal support values were generated through 1000 repetitions of the ultrafast bootstrap approximation (Hoang et al., 2018).
We used two coalescent-based approaches to estimate a species tree. Rather than assuming all genes evolved under the same tree topology, these methods allow for discordance between the gene histories and species tree by explicitly accounting for the biological process of incomplete lineage sorting. First, we used the program SVDQuartets (Chifman & Kubatko, 2014) as implemented in PAUP* v4.0a166 (Swofford, 2003). SVDQuartets uses site patterns in the nucleotide data to estimate a phylogeny under the multispecies coalescent model. We used SVDQuartets in two ways. We initially estimated a lineage tree (SVDQ LT ), where all individuals are represented as tips in the tree, to confirm species assignment and test species monophyly. We then assigned individuals to species a priori and estimated a species tree (SVDQ ST ). For both SVDQuartets analyses, we evaluated all quartets and used standard bootstrapping to generate nodal support values. Second, we used the program ASTRAL-III v5.6.3 (Zhang et al., 2018) to estimate a species tree.
Instead of using site patterns in the nucleotide data, ASTRAL-III uses gene trees as input to estimate a species tree. Maximum-likelihood Foundress information, when present, is from Herre (1989) Table 2 in Herre (1989) for details). For figs that share a pollinator species, foundress numbers were averaged between the host figs. Lines separate the three host-sharing systems from the remaining species (with one-to-one fig-wasp association) found in this community.
TA B L E 1 Pollinator wasp sampling gene trees were first estimated in IQ-TREE. For each locus, IQ-TREE selected the substitution model of best fit with ModelFinder (Kalyaanamoorthy et al., 2017) using Bayesian Information Criteria (BIC). We assigned individuals to species a priori (Rabiee et al., 2019) and then used the ML gene trees as input to estimate a species tree.
Previous work in the Panamanian system has recovered pollinator species as monophyletic and has not brought into question species validity (Jackson et al., 2008;Machado et al., 2005;Satler et al., 2019). Given the typical pattern of small intraspecific divergence coupled with large interspecific divergence for these wasps, we wanted to ask what proportion of loci recover each species as monophyletic. If the pollinators have been evolving in isolation over evolutionary time, we would expect the species to be monophyletic for all or nearly all sampled loci. In contrast, if interspecific gene flow has been an important process in the evolution of this system, we would expect interacting species to show a lack of monophyly for numerous loci. We used DendroPy v4.4.0 (Sukumaran & Holder, 2010) to count the proportion of gene trees (as estimated above in IQ-TREE) for which a species was monophyletic. We required that at least two individuals were sequenced for a species for a given gene tree to assess monophyly.

| Population genetics
In addition to estimating phylogenetic relationships and testing for monophyly, we were interested in understanding the population genetics of the pollinator species. In particular, we wanted to test whether genetic diversity within a pollinator species varied with average foundress number per host fig. If outbreeding is common in pollinator species that contain multiple foundresses per fruit, we would expect to see a positive correlation between genetic diversity and average foundress number. We used DendroPy to calculate nucleotide diversity (π), number of segregating sites and Watterson's theta (per site) for all pollinator species. For species with foundress number data (see Table 1), we used Spearman's rank correlation test in r v3.5.2 (R Core Team, 2018) to test for a correlation between average number of foundresses and summary statistic.

| Testing for monophyly with mitochondrial DNA
To complement analyses and inference with the genome-wide sequence data from the nuclear genome, we used mitochondrial DNA (mtDNA) to test whether pollinator species were monophyletic in the mitochondrial genome. Even when not targeted, mitochondrial DNA is often collected using sequence capture approaches (Barrow et al., 2017;do Amaral et al., 2015). We used NOVOPlasty v3.8.3 (Dierckxsens et al., 2017) to identify mitochondrial reads and generate haplotypes from the sequencing files. Since no Pegoscapus mitochondrial reference genome is available, we used a P. hoffmeyeri (AY148119) cytochrome oxidase I (COI) mtDNA sequence as a seed with which to align the reads. We used default settings with an initial kmer value of 39. For samples that either did not generate any matches or produced obvious sequencing errors, we lowered the kmer value to 23. Using a lower kmer value can help with low coverage data, appropriate here since we were only targeting regions of the nuclear genome. Finally, for samples that did not recover any mtDNA with the initial seed sequence, we aligned reads (using default settings) to a longer COI sequence of a different pollinator species (Pegoscapus sp., JN103329) to see whether a different seed sequence could recover mtDNA data.
We aligned the mitochondrial sequences with MAFFT and edge trimmed to match the primary seed sequence (AY148119) to reduce missing data. We used DendroPy to calculate several summary statistics (as described above) to characterize genetic variation within species. In addition, we used PAUP* to calculate GTR corrected genetic distances for both within-species and between-species comparisons. Finally, we estimated a maximum-likelihood gene tree in IQ-TREE, using ModelFinder to select the substitution model and 1000 repetitions of the ultrafast bootstrap approximation to generate nodal support. If patterns of monophyly differ between the nuclear and mitochondrial genomes, the cytonuclear discordance would suggest introgression or processes other than genetic drift generating these discordant patterns.     (1491) Note: Summary statistics include nucleotide diversity (π), number of segregating sites (SS) and Watterson's theta per site (θ w ). Monophyly shows the proportion of gene trees for which a pollinator species is monophyletic. A gene tree was tested for monophyly when two or more sequences for a given species were sampled, with total trees tested per species in parentheses. Lines separate the three host-sharing systems from the remaining species (with one-to-one fig-wasp association) found in this community.

| Testing for admixture and gene flow
axes of variation to visualize genetic clustering of individuals. Second, we used Structure v2.3.4 (Pritchard et al., 2000) to cluster individuals into genetic groupings. Structure clusters individuals by maximizing Hardy-Weinberg equilibrium within clusters and minimizing Hardy-Weinberg equilibrium among clusters. Evidence for recent hybridization and introgression would be reflected by individuals containing multilocus genotypes sampled from multiple clusters. Our choice of the number of clusters (K) was informed by previous work identifying the wasp species within these systems (Jackson et al., 2008;Molbo et al., 2003). For the F. americana/F. colubrinae (AC) and F. bullenei/F. popenoei (BP) host sharing systems, we used a K value of 3; for F. obtusifolia (O), we used a K value of 2. We used the admixture model and allowed allele frequencies to be correlated among populations. For each analysis, we used a burnin of 100,000 steps, followed by 500,000 MCMC reps, and completed 10 replicates. We processed results in the r package pophelper v2.2.7 (Francis, 2017).
Genetic data were processed independently for pollinators in each host sharing system to generate phased data sets. By processing each system independently, we retained loci with a minimum of 70% of individuals for the taxa of interest only. We followed the same procedure as outlined above for generating the phased data sets and used unlinked SNPs for both the PCA and Structure analyses. To generate these data sets, we first scanned our phased aligned taxon-specific UCE loci for variable sites. Within a locus, we then selected the SNP that had the highest sample coverage. If a locus had multiple SNPs with equal sample coverage, one of those SNPs was selected at random.
Tests of hybridization among co-pollinators provide biologically motivated hypotheses where opportunities for interspecific mating are most likely. If gene flow among these species has been recent or ongoing, we would expect to detect that signal and infer a history of hybridization. This approach, however, is limited to specific sets of taxa based on present-day associations. Given the history of host switching in this system, and accounting for a potential dynamic process of host-pollinator associations through time, it would be useful to test for hybridization among all sampled taxa.
To test whether historical introgression has been an important process in this community of pollinators, we used TreeMix v1.13 (Pickrell & Pritchard, 2012). TreeMix is a maximum-likelihood approach that uses allele frequency data to construct a population graph and places hybridization events on that graph for populations with the least fit to a tree model. Since the number of hybridization events is specified a priori, we can test models varying the number of hybridization events to determine a model of best fit. We used the total variance explained by the model to inform the number of hybridization events that best characterizes this community of pollinator species. We used allele frequency data representing unlinked, biallelic SNPs from the phased data set from all pollinator species.
To remove effects of missing data, we first identified biallelic SNPs within each locus that were sampled for at least one individual per species. For loci with multiple SNPs, we selected a single SNP with the highest coverage. If a locus had multiple SNPs with equal sample coverage, one SNP was selected at random. We then estimated a maximum-likelihood population graph in TreeMix, allowing between zero and five migration events.
Although additional approaches are often used to test for hybridization and introgression, they were not applicable for our data set. For example, D-statistics (Durand et al., 2011;Green et al., 2010), often referred to as ABBA/BABA tests, have been widely

| Phylogenetics
When individuals are treated as tips in the tree, phylogenetic analyses recover each pollinator species as monophyletic with strong support. We find this result with both concatenation (Figure 1) and the coalescent-based lineage tree analysis with SVDQuartets (SVDQ LT , Figure S1). To further investigate, we tested species monophyly for each sampled UCE locus. All species were monophyletic for all or nearly all loci (    (Figures 1 and 2, Figures S1 and S2).
Pollinator species associated with F. bullenei and F. popenoei show phylogenetic relatedness intermediate to these two systems, where they are less divergent from one another than the pollinators associated with F. americana and F. colubrinae yet are not sister taxa within the tree. We find this result with both concatenation (Figure 1) and species tree approaches (Figure 2, Figure S2), highlighting a range of evolutionary relatedness among the systems containing cooccurring pollinator species.

| Population genetics
Population genetic statistics were variable across pollinator species, with generally wide ranges and large variances (  (Table S2).
In the gene tree, all wasp species are recovered as monophyletic with strong support, with all but one species (Pegoscapus lopesi, bootstrap value of 99) having a bootstrap value of 100 ( Figure S3). There is, however, little support for phylogenetic relationships among species, as most nodal support values are low. Consistent with how individuals cluster within species with the nuclear data, the same clustering of individuals within species is recovered with data from the mitochondrial genome, providing strong support for what we consider a pollinator species in this community. Given the low support values for interspecific relationships, we make no comparisons between the structure of the nuclear genome phylogeny and the mitochondrial gene tree. Comparing how individuals cluster within species in the nuclear phylogeny and mitochondrial gene tree, however, these data provide no evidence of recent cytonuclear discordance.

| Testing for admixture and gene flow
There was no evidence of admixture or introgression in the three systems (AC, BP and O) containing co-occurring pollinator species.
Distinct clusters corresponding to species are recovered in the PCA analyses, with no signal of hybridization ( Figure 4). For both AC and BP, wasp species plot in distinct PCA space with little intraspecific variance-individuals within species appear as a single point-in comparison with the much larger interspecific variance (Figure 4a,b). The two wasp species in O are well differentiated in PC1, but show some intraspecific spread in PC2, although this axis only comprises 0.35% of the variance (Figure 4c). These results were further supported in Structure, as individuals were assigned to their respective species with no evidence of admixture (Figure 4). In contrast to seeing shared ancestry indicative of admixture and introgression, each individual maps unambiguously to its respective species. Thus, in systems where opportunity for hybridization is present, we detect none.
TreeMix estimated a population graph consistent with the other phylogenetic approaches ( Figure S4). Some differences were seen, mainly at places with short internal branches and low support along the backbone of the tree. Adding migration events to the tree only incrementally improved the proportion of variance explained by the model (Table 3) We used these data to estimate phylogenetic relationships and conduct detailed tests for hybridization and introgression among all 19 pollinator wasp species. We analysed the data within the context provided by this ecologically and evolutionarily well-characterized strangler fig (Herre, 1989;Machado et al., 2005).  (Jackson et al., 2008;Machado et al., 2005). In the pollinator wasps, F1 pollinator hybrids have been previously detected using a combination of COI and microsatellites in one pair of intensively sampled pollinators associated with one of the hosts in this community (Molbo et al., 2003(Molbo et al., , 2004. Based on our sampling of genome-wide UCE data, we found all 19 pollinator species to be well-delimited Hybridization followed by successful introgression is strongly suggested to be a relatively frequent occurrence in the Panamanian fig hosts (Jackson et al., 2008;Machado et al., 2005). The processes of hybridization and introgression also appear to be more important for figs in general, as they have been detected in numerous fig systems (Compton, 1990;Compton et al., 2009;Cornille et al., 2012;Van Noort et al., 2013;Wang et al., 2016;Wilde et al., 2020).
In particular, cytonuclear discordance between plastid and nuclear genomes provides additional support that hybridization and intro- single-foundress figs will be severely restricted ( Table 1).
Among the pairs of Panamanian wasp species that share the same host fig species, there appears to be no clear pattern to their degree of phylogenetic relatedness. Only P. hoffmeyeri sp. A and P. hoffmeyeri sp. B associated with F. obtusifolia are recovered as sister species in our phylogeny (Figures 1 and 2, Figures S1 and S2), and these are the only two species with evidence supporting occasional F1 hybrid offspring (Molbo et al., 2003(Molbo et al., , 2004. The other co-occurring pollinators span the phylogenetic breadth of our sampled community, with wasps associated with F. bullenei and F. popenoei being relatively closely related (but not sister taxa), while wasps associated with F. americana and F. colubrinae are distantly related (Figures 1 and 2,   Figures S1 and S2). These Panamanian observations are consistent F I G U R E 3 Correlation between population genetic summary statistics and average foundress number per pollinator species. Summary statistics include (a) nucleotide diversity, (b) number of segregating sites and (c) Watterson's theta (per site). The Spearman's rank correlation was used to test significance. Fourteen wasp species were included for which we had average number of foundress information (see Herre, 1989 (Yang et al., 2015). Importantly, despite previous evidence for occasional F1 hybridization events between the two F. obtusifolia pollinators (0.9%, Molbo et al., 2003Molbo et al., , 2004, we found no evidence for F1 or backcrossed hybrids between these two species, or introgression in any of the species. and would be expected to greatly limit differentiation (F ST < 0.03) between species (Wright, 1949). In fact, all of the wasp species in our study are highly genetically differentiated from each other (Figures   1-3). This indicates that if undetected F1 hybrids are being formed, they are not a bridge to effective gene migration between species.
While our power to detect recent hybrid formation may be limited by the number of wasps per species that we genotyped in this study, the large number of UCE loci we assayed gives us substantial power to detect historical introgression. If hybridization has occurred in evolutionary time, resulting in even low levels of recombination between parental species genomes, with our comprehensive data set of UCE loci and the statistical analyses we have employed, we should have been able to detect that signal. We did not. Given our observations, how might different processes that af-   (Yang et al., 2015). range between one and four-and-a-half (Herre, 1989 mean foundress numbers will also determine wasp population structure which in turn will affect many aspects of these wasp species (sexual competition and sex ratios, heterozygosity) (Herre, 1985(Herre, , 1989(Herre, , 1993Molbo et al., 2003Molbo et al., , 2004. Notably, we found a significant positive correlation between the average number of foundresses and genetic diversity of UCE loci across pollinator species (Figure 3) Several postzygotic barriers potentially play a role in restricting successful hybridization and introgression among the pollinator wasps. Greater genetic divergence usually corresponds to greater degrees of reproductive isolation (Coyne & Orr, 1989, 1997. We expect that the considerable divergence among the 19 pollinator species in this community ( Figure 1) might preclude successful hybrid formation due to functional incompatibilities among either nuclear-nuclear or nuclear-cytoplasmic components derived from different parental species in the genomes of hybrid individuals. For example, the presence of diploid males-reported in some hybrid pollinators associated with F. obtusifolia (Molbo et al., 2004)-suggests a breakdown in the haplodiploid sex determination mechanism. In a mating between the males of one species and the females of another species, only the diploid F1 daughters will be hybrid offspring; haploid males develop from unfertilized eggs, so will inherit only portions from their mother's nuclear and cytoplasmic genomes and therefore will not repre-  (Breeuwer & Werren, 1995;Gadau et al., 1999;Koevoets et al., 2012). Because co-occurring fig wasp pollinators are usually distantly related species (Satler et al., 2019;Yang et al., 2015), these effects could be exacerbated in hybrid individuals that would contain mixes of more divergent genomes. Importantly, when F1 hybrid females are single foundress, all potential subsequent matings between their sons and daughters will be between F2 hybrids. Thus, the ability of those F2 daughters to mate will be strictly determined by the development and reproductive capability of their F2 hybrid male siblings.
An additional potential postzygotic barrier to hybridization and introgression in the wasps is Wolbachia. These maternally inherited cytoplasmic bacteria are known to cause drastic reductions of hybrid formation in Nasonia wasps (Bordenstein et al., 2001). Wolbachia are commonly found across insect species (Werren, 1997)  examined six species pairs of figs and wasps found along an elevational gradient in New Guinea and found an increased rate of genetic divergence with elevational distance among the wasps relative to their host figs. They suggest pollinators have an increased speciation rate, with populations of wasps diverging in isolation faster than their hosts, effectively decoupling speciation processes in the two mutualist taxa Machado et al., 2005;Peng et al., 2008;Satler et al., 2019;Van Noort et al., 2013).
Despite the ability of pollinating wasps to disperse many kilometres (Ahmed et al., 2009;Nason et al., 1998), broad geographic sampling suggests that multiple distinct pollinator species-in allopatric and parapatric distributions-characterize many, if not most, host fig species across their ranges. Molecular data also suggest a previously underappreciated prevalence of host switching in Neotropical strangler figs, as well as figs in general Jackson et al., 2008;Satler et al., 2019;Yang et al., 2015).
We suspect that combinations of pre-and postzygotic barriers contribute to the reproductive isolation of pollinator wasps (Molbo et al., 2003(Molbo et al., , 2004Satler et al., 2019). At the prezygotic level, the opportunities for interspecific matings are fairly limited due to small chances that heterospecific foundress wasps oviposit successfully in the same individual fig syconium and due to possible unknown premating barriers. At the postzygotic level, low hybrid fitness can result from intrinsic genetic incompatibilities that would limit not only the ability of hybrids to produce viable and fertile offspring, but could also affect the capacity of hybrid wasps to locate, enter, pollinate and oviposit in a receptive syconium. Furthermore, the prevalence of Wolbachia in fig wasps can also play a major role in postzygotic isolation. When pollinator species experience secondary contact, we suspect that, relative to their plant hosts, these barriers contribute to less fit hybrids and reinforce species boundaries (Nosil et al., 2003). We suggest that the still poorly understood processes of pollinator speciation and extinc-

| CON CLUS IONS
We have focused on detecting evidence for historical introgres- More generally, we suggest that studies across host-pollinator mutualisms will support the suggestion that the evolutionary history of the host plants is relatively more influenced by successful hybridization and introgression than in their insect pollinators.

ACK N OWLED G EM ENTS
We thank Brant Faircloth and the Faircloth laboratory for generously providing time, space and resources for learning sequence capture approaches for generating UCE data. We thank Nick Davis, Charlotte Jandér, Devin Molnau, Finn Piatscheck and Kevin Quinteros for providing some of the pollinator specimens. We thank members of the Heath laboratory and Nason laboratory for discussion and comments.
We also thank AE Jeremy Yoder, Jeff Feder, George Weiblen and three anonymous reviewers for comments that helped improve this manuscript. Funding was provided by the National Science Foundation

B EN EFIT-S H A R I N G
Raw sequence data are available from the NCBI Sequence Read Archive (SRA) under BioProject ID: PRJNA789240 (BioSample accessions: SAMN23703750-SAMN23703925). NCBI BioSample accession numbers for individual wasps are included in Table S1. All data sets and custom scripts are available on Dryad (https://doi. org/10.5061/dryad.fbg79 cnwk).

B EN EFIT-S H A R I N G
Benefits Generated: Benefits from this research accrue from the sharing of our data and results on public databases as described above.