Genomic evidence of prevalent hybridization throughout the evolutionary history of the fig-wasp pollination mutualism

Ficus (figs) and their agaonid wasp pollinators present an ecologically important mutualism that also provides a rich comparative system for studying functional co-diversification throughout its coevolutionary history (~75 million years). We obtained entire nuclear, mitochondrial, and chloroplast genomes for 15 species representing all major clades of Ficus. Multiple analyses of these genomic data suggest that hybridization events have occurred throughout Ficus evolutionary history. Furthermore, cophylogenetic reconciliation analyses detect significant incongruence among all nuclear, chloroplast, and mitochondrial-based phylogenies, none of which correspond with any published phylogenies of the associated pollinator wasps. These findings are most consistent with frequent host-switching by the pollinators, leading to fig hybridization, even between distantly related clades. Here, we suggest that these pollinator host-switches and fig hybridization events are a dominant feature of fig/wasp coevolutionary history, and by generating novel genomic combinations in the figs have likely contributed to the remarkable diversity exhibited by this mutualism.

Ficus evolutionary history there have been at least one reversal in both traits.The question becomes whether any basal lineages exhibiting basal characteristics still exist in extant Ficus taxa.Currently, some studies show active pollination as basal traits for the extant fig and wasp taxa 4,[10][11][12] .In this study, Astral species trees based on nuclear data supported the pantropical subgenus Urostigma (monoecious, with both active and passive pollination) as sister to all other Ficus species.In contrast, the primary concordance tree (PCT tree) based on nuclear data and both mitochondrial-and chloroplast-based phylogenies support the neotropical section Pharmacosycea (monoecious, passive pollination) as basal to all other figs, as other studies have concluded [13][14][15] .Our results thus strongly support monoecy as the basal state of the sexual system of all extant taxa, but we consider the basal state of pollination mode in extant figs to be still an open question.

Supplementary Note 3: Pollinator phylogenies used for cophylogenetic analysis in JANE
Two phylogenies of pollinator wasps associated with the 15 studied fig species, extracted from published wasp phylogenies 12,14 , were used in cophylogenetic analysis in JANE.The Machado et al. 14 tree is based on only one mitochondrial protein-coding gene (COI) fully sequenced for all 15 wasp species that is completely sequenced for all species considered.In contrast, the Cruaud et al. 12 tree is based on six genes (two nuclear protein-coding genes (EF1a, Wg); two mitochondrial protein-coding genes (COI, Cyt b), and two ribosomal genes (28S rRNA, 18S rRNA)) although not all genes were successfully sequenced in all of the 200 wasp species.The denser species sample and bigger gene matrix in the study by Cruaud et al. 12 should make their phylogeny relative more robust.
Although neither of these phylogenies of pollinator wasps is as strong as those for Ficus trees generated here, they represent two alternative views, both considered plausible, of fig wasp evolution (Supplementary Fig. 5, Supplementary Table 6 (last line)).The most notable difference between them is which wasp clade is placed as sister to all others.Machado et al. 14 treated Tetrapus wasps (whose hosts are section Pharmacosycea in Neotropics) as sister to all other agaonids, a placement that better matches the primary concordance tree of Ficus, which places section Pharmacosycea as sister to all other figs.This pair of phylogenies represents a long-held view on the evolutionary history of fig-fig wasp mutualisms.However, in the phylogeny of Cruaud et al. 12 , Tetrapus is inferred to lie in the middle of the phylogeny.In the Astral species tree for Ficus, section Pharmacosycea is also inferred as being placed near the middle of the tree.
Pollinators of three of the fig species studied here (F.gasparriniana, F. vasculosa, F. triloba) were not included in Cruaud et al.'s phylogeny 12 .However, their study included pollinators of fig species closely related to these three (F.chapaensis, F. nervosa, and F. fulva, respectively), each pollinated by a wasp belonging to the same genus as the pollinator of the studied fig (Blastophaga, Dolichoris, Valisia, respectively).As the three pollinator genera are all monophyletic 12 , and only the genus-level phylogenetic topology was used in JANE, these replacements do not influence the analyses of phylogenetic discordance between figs and pollinators.
Two pollinator clades associated with fig species we studied were not included in the phylogeny reconstructed by Machado et al.We inserted them artificially into that phylogeny, following their placement in the phylogeny of Cruaud et al. 12 .These were (i) a Ceratosolen clade (whose hosts are section Adenosperma, represented by F. microdictya in our study), added as sister to another Ceratosolen clade (hosts section Sycomorus, represented by F. hispida); and (ii) a Blastophaga clade (hosts section Ficus, represented by F. carica), added as sister clade to Dolichoris (hosts section Oreosycea, represented by F. vasculosa).

Ficus Astral tree and Cruaud et al.'s pollinator phylogeny
Results of cophylogenetic reconciliation based on the Astral tree for Ficus and the most recently published pollinator tree 12 (Fig. 5B) inferred nine pollinator host-switch events.In detail, one pollinator host-switch event is inferred to have occurred near the root of the wasp phylogeny, from subgenus Sycomorus to subgenus Sycidium; four other switch events are inferred around the middle of wasp phylogeny.One is inferred from the Old-World gynodioecious section Eriosycea (subgenus Ficus) to the Neotropical monoecious section Pharmacosycea (subgenus Pharmacosycea), another from section Eriosycea (subgenus Ficus) to the F. carica-clade (subgenus Ficus), a third from the F. carica-clade to one clade of subgenus Urostigma (section Americana -Galoglychia clades), and a fourth from the Americana -Galoglychia clades to section Urostigma.Finally, three other switch events are inferred near the terminal branches of wasp phylogeny: (i) within gynodioecious clades (from subgenus Ficus to subgenus Synoecia), (ii) within monoecious clades (subgenus Urostigma, from section Americana to section Conosycea) and (iii) from the gynodioecious F. carica-clade (section Ficus) to the monoecious section Oreosycea (subgenus Pharmacosycea) (Fig. 5B).All those inferred host-switch events match with corresponding hybridization events (Supplementary Table 7).6: Frequency of evolutionary events inferred with cophylogenetic analysis in JANE.The optimal cost setting, frequency (mean ± SD) of each evolutionary event, and cost setting selection results in JANE 4.01 are given.Frequency of both "Duplications" and "Failure to diverge" events are zero for all phylogeny pairs, and thus do not appear in the table.The optimal cost setting for each phylogenies pair was chosen with both the significant P-value (two sided T-test) and lowest total cost value in both random tip mapping (test1) and random parasite tree permutation tests (test2).Two nuclear phylogenies of Ficus (Astral tree and primary concordance tree (PCT)), chloroplast and mitochondrial phylogenies produced by our analyses, and the two main published phylogenies of pollinator wasps 12,14 (Supplementary Fig. 5  NB: "C0D1DH1L1F1" indicates cost setting with zero cost for cospeciation and one for all other events, "C0D1DH2L1F1" indicates cost setting with zero cost for cospeciation, a cost of two for associate switches, and a cost of one for all other events.Results of cophylogenetic analyses between the two wasp phylogenies (bottom row of the table) just show that their phylogenetic topologies differ significantly, and do not indicate any other biological meaning.* indicate number of reconciliation solutions inferred by JANE under the optimal cost setting.Each solution with give an inferred value of each five event (codivergence, duplication, associate switches, associate losses, and fail to divergence).Most or even all solutions may show same pattern.Based this data, the mean and SD of each event under each phylogeny-pairs were calculated.
Meanwhile, the reconciliation pattern confirmed with highest number of solutions were shown in supplementary figure 7-15.

Supplementary Fig. 10 :
One of the equally most parsimonious reconciliations between Ficus chloroplast and mitochondrial phylogenies inferred with JANE.The black tree represents the chloroplast genomic phylogeny and the blue tree the mitochondrial genome phylogeny.Hollow red circles on nodes represent codivergence events; filled color dots and arrows on the blue lines represent events of associate host switch following duplication, while filled orange dots indicate the existence of another potential switch location with equal cost, and a filled red dot means that all other potential switch locations have higher cost.For the optimal cost settings used, refer to Supplementary Table6.Supplementary Fig 11: Geographic clustering pattern of Ficus chloroplast genomes on the 59-species chloroplast phylogeny.The published 59-species chloroplast phylogeny 13 supported none of the six Ficus subgenera as a monophyletic clade, but revealed a pattern of geographic clustering, in which sympatric fig species from different subgenera were frequently found in the same clade.For example, specific Afrotropical (black symbol) figs of subgenera Sycomorus (light-blue names) and Urostigma (green names) clustered in the same clade, but the two subgenera are shown to be distantly related clades, and each monophyletic, in nuclear phylogenies (Fig. 1).Furthermore, gynodioecious figs from different subgenera co-distributed from Eurasia to Australasia were also repeatedly found to cluster in multiple chloroplast clades.Subgeneric classification is shown by tip name colors and geographic distribution is shown by tip symbol color.

Supplementary Table 1: Sample information for the 17 Moraceae species selected for genomic resequencing. Species Subgenus Section Pollination mode Sexual system Distribution Sources* Voucher**
Fig classification and traits followCruaud et al. 12.All trees sampled from the plant collection originated from seeds or seedlings collected from their natural forest habitat nearby or introduced from other countries.Voucher**: all herbarium specimens listed are stored in the Herbarium of Xishuangbanna Tropical Botanical Garden, CAS., with detail samples information and photos could be accessed online (http://hitbc.xtbg.ac.cn/) with voucher number.

Table 2 : Sequencing summary for the 17 Moraceae species.
Number of contigs is shown in parentheses when mitochondrial genomes could not be completely assembled (circular).

Table 3 : Summary of genomic-windows datasets used for inference of species trees.
Physical distribution of genomic sliding windows in chromosomes of the reference genome is shown in Source data 2. Fixed number of SNPs indicates that each window is the genomic fragment including 1000 SNPs.Size of the windows can thus vary, while window size in each of the other three windows types is fixed.

Table 4 : Information on fossil calibrations for Ficus divergence time estimation in MCMCTree
Supplementary

Table 5 : All potential basal Ficus inferred with BUCKy based on the windows1000SNPs dataset.
Primary and alternative minor splits indicating all main potential basal clades of Ficus phylogeny.The mean and 95% credibility interval of the split in the primary concordance tree are shown.Gynodioecious clade: including all four subgenera dominated by gynodioecious species.Section Pharmacosycea obtained the highest CF values (0.406) as a signal clade.However, subgenus Urostigma or its sections (highlighted in bold) are also repeatedly inferred as the basal Ficus clade, although with CF values.

Table 8 . Summary of evidence for hybridization events among main clades of Ficus.
, respectively.CP and MT: indicate Ficus phylogenies based on chloroplast and on mitochondrial genomes, respectively.fig_C and fig_M: indicate pollinator phylogenies extracted from Cruaud et al. 12 and Machado et al. 14 , respectively.There are three columns of direct evidence for hybridization detection, and 13 columns of indirect evidence based on cophylogenetic reconciliation using JANE, when associate switches were inferred to have occurred between related figs groups.Yes or No: indicates whether each hybridization relationship was supported (Yes) or not (No) by the method in question.a: indicates that the hybridization detected involved an ancient clade of one group.fig_A and fig_P: indicate that the Astral tree or the primary concordance tree was used to represent Ficus nuclear phylogeny, respectively.CP and MT: indicate Ficus phylogenies based on chloroplast and on mitochondrial genomes, respectively.fig_C and fig_M: indicate pollinator phylogenies extracted from Cruaud et al. 12 and Machado et al. 14 , respectively. phylogeny

Table 9 : Genome ploidy estimated with nQuire program 17 .
All species are inferred as diploid species.Code as: create -b bam -o prefix -f 0.3 && nQuire lrdmodel bin.