doi:10.1111/j.1420-9101.2005.00968.x Patterns of diversification of Afrotropical Otiteselline fig wasps: phylogenetic study reveals a double radiation across host figs and conservatism of host association E. JOUSSELIN,* S. VAN NOORT, J.-Y. RASPLUSà & J. M. GREEFF* *Department of Genetics, University of Pretoria, Pretoria, South Africa Natural History Division, South African Museum, Iziko Museums of Cape Town, Cape Town, South Africa àInstitut National de la Recherche Agronomique, Centre de Biologie et de Gestion des Populations, Campus International de Baillarguet, Montferrier sur Lez, France Keywords: Abstract coevolution; cospeciation; Ficus; host-use; oviposition; phylogeny; plant/insect interaction; radiation. We studied the phylogenetic relationships of Otiteselline fig wasps associated with Ficus in the Afrotropical region using rDNA sequences. African fig species usually host two species of Otiteselline fig wasps. Phylogenetic analyses reveal that this pattern of association results from the radiation of two clades of wasps superimposed on the fig system. Within each clade, wasp species generally cluster according to their host classification. The phylogenies of the two clades are also generally more congruent than expected by chance. Together these results suggest that Otiteselline wasp speciation is largely constrained by the diversification of their hosts. Finally, we show a difference in ovipositor length between the two Otiteselline species coexisting in the same Ficus species, which probably corresponds to ecological differences. The diversification of ecological niches within the fig is probably, with cospeciation, one of the key factors explaining the diversification and maintenance of species of parasites of the fig/pollinator system. Introduction The processes that govern the evolution of host-plant use by phytophagous insects have been much debated and are the subject of many text books and review papers (e.g. Mitter & Farrell, 1991; Mitter et al., 1991; Strauss & Zangler, 2002). The 700 plus species of figs (Ficus, Moraceae) and the wasps that reproduce in their inflorescence represent a very interesting system on which to study the diversification of insects on hostplants. Each fig species hosts a very diverse assemblage of wasp species, belonging to the superfamily Chalcidoidea (Bouček, 1993). This assemblage creates a well-defined community and offers a unique opportunity to compare the evolution of host-plant use and modes of speciation across several lineages of wasps that share a limited resource. The inflorescence of Ficus, known as a fig or Correspondence: Emmanuelle Jousselin, Institut National de la Recherche Agronomique, Centre de Biologie et de Gestion des Populations, Campus International de Baillarguet, CS-30 016, 34 988 Montferrier sur Lez, France. Tel.: +33-4-99623326; fax: +27-12-3625327; e-mail: ejousselin@yahoo.com syconium, consists of an enclosed receptacle lined by uniovulate female flowers (Berg, 1989). These flowers are used by the wasps as egg laying sites. Some of these wasps are pollen vectors. They are attracted by the volatiles emitted by receptive figs (Barker, 1985; van Noort et al., 1989; Ware et al., 1993) and enter the fig cavity via the ostiole (a slit formed by bracts situated at the apex of the fig). Their larvae complete their development in the galled flowers. Pollination seems to be restricted to the wasps belonging to the Agaonidae (sensu Rasplus et al., 1998; but see Jousselin et al., 2001). The biology of nonpollinating (parasitizing) fig wasps is less well known: they can be gallers (like the pollinators), but also parasitoids (Bronstein, 1991; West et al., 1996; Kerdelhué et al., 2000). They either enters into the fig cavity like the pollinators or oviposit in the flowers through the fig wall. The stage of the fig development at which they lay their eggs is also very variable (Kerdelhué et al., 2000). Recently, the evolution of the mutualistic association between figs and their pollinators has received much attention (Herre et al., 1996; Kerdelhué et al., 1999; Machado et al., 2001; Weiblen, 2001; Weiblen & Bush, J. EVOL. BIOL. 19 (2006) 253–266 ª 2005 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY 253 254 E. JOUSSELIN ET AL. 2002; Molbo et al., 2003; Jousselin et al., 2003; Jackson, 2004). The relationship is generally thought to be species-specific: one species of pollinating wasp is associated with one species of fig (Ramirez, 1970; Wiebes, 1982). However, exceptions to this ‘one to one rule’ are discovered as more thorough taxonomical and molecular studies are conducted (Michaloud et al., 1985; Rasplus, 1994; Kerdelhué et al., 1999; Lopez-Vaamonde et al., 2002; Molbo et al., 2003; Cook & Rasplus, 2003). The mapping of the evolution of host association on the phylogenies of both partners at a broad taxonomical level is consistent with a global pattern of cocladogenesis (Herre et al., 1996; Weiblen, 2000; Machado et al., 2001; Weiblen, 2001; Jousselin et al., 2003) and the one study that uses statistical tests shows that the pollinator phylogeny and the fig phylogeny are more congruent than expected by chance (Weiblen & Bush, 2002) but the test is limited to 19 species among the 180 species in the Ficus subgenus Sycomorus. While the relationship between figs and their pollinators has been the subject of many studies, little is known about the patterns of diversification of other lineages of wasps that mostly act as parasites of the interaction. Taxonomic studies suggest a close fit between parasitic wasp classification and their host fig classification and a ‘one-to-one’ rule similar to the situation observed in the pollinator/fig association is generally assumed (Berg & Wiebes, 1992). However, fig wasp taxonomy has been influenced by the classification of the host figs, often having been established a posteriori. Fig wasps collected from different fig tree species are assumed to be species-specific; therefore wasps collected from different taxonomical groups of figs are automatically classified into different taxa. The interaction between figs and nonpollinating fig wasps could be a lot less specific than is generally assumed and the diversification of nonpollinating wasps could be independent of their host speciation. Chemical and physical barriers precluding host shift in nonpollinating fig wasps are supposedly not as tight as in the pollinating wasps (Lopez-Vaamonde et al., 2001; Weiblen & Bush, 2002, Jackson, 2004). For example, most nonpollinating wasps do not enter the fig cavity; they do not necessarily lay their eggs at fig receptivity and thus either cannot or need not rely on the volatiles emitted by their host fig to attract the pollinators to find their host. Only three studies so far have explored the phylogenetic relationships of selected groups of nonpollinating fig wasps to look at the evolution of host association and patterns of diversification (Machado et al., 1996; Lopez-Vaamonde et al., 2001; Weiblen & Bush, 2002). Two studies showed a nonrandom association between parasitic fig wasps of the genera Idarnes (Machado et al., 1996) and Sycoscapter (Lopez-Vaamonde et al., 2001) and the associated pollinators, suggesting some cospeciation between these wasps and their host figs. Weiblen & Bush’s study (2002) indicates that the nonpollinating wasp phylogeny of the genus Apocryptophagus is far from congruent with their host fig phylo- geny (figs of the Sycomorus subgenus) and suggests that nonpollinating wasps often speciate on their host fig, probably through the occupation of different ecological niches. In this paper, we investigate the diversification and the patterns of host-use of wasps belonging to the Otitesellinae genera Otitesella Westwood and Philosycus Wiebes (Pteromalidae) in the afrotropical region. They lay their eggs into fig flowers by inserting their ovipositor through the fig wall from the outside of the fig. This process takes place around fig receptivity, roughly at the same time as pollinating wasps enter the fig cavity. Once the eggs have been deposited, the flowers are transformed into galls in which the wasp larvae develop. Otitesella are restricted to the old world tropics and have been subdivided into two species groups the Otitesella africana group and the Otitesella digitata group (Wiebes, 1969), corresponding to the two sections of Ficus from which they have been recorded. Species of the O. africana group have been reared from figs of section Galoglychia (occurring in Africa) while species of the O. digitata group have been recorded from section Urostigma (found in the IndoMalayan region and Africa) (van Noort & Rasplus, 1997). Species in the genus Philosycus are also associated with fig species in section Galoglychia. The characters that differentiate Philosycus from Otitesella are only based on male morphology and the Philosycus clade may not warrant generic distinction (Van Noort & Rasplus, 1997). By establishing a molecular phylogeny, this study will first try to solve the taxonomic problems in Otiteselline. Second, the use of molecular markers will help to detect cases of host specificity breakdowns. Description of Otitesella and Philosycus species are still scarce and female morphology is very conservative which makes it difficult to assess whether different fig species host the same wasp species. The host figs of Otiteselline are often sympatric and we could expect some wasp species to move between host species. Then, by mapping host association onto the wasp phylogeny, we determine whether host association is evolutionarily conserved within Otiteselline fig wasps, i.e. whether related wasps are associated with related fig species. Finally, studying the patterns of diversification of these wasps is also particularly interesting as extensive collections within Africa revealed that two morphospecies of otiteselline could occur per fig species (wasps can be classified into morphogroups that we termed ‘uluzi’ and ‘sesqui’, see material and methods and Table S1). This pattern of association indicates that either the two groups form two clades that have separated and then diversified across all the African figs or that each pair of species occurring on one host has originated through independent duplication events. The latter would imply that the same morphological divergence (long/short ovipositor valves) has evolved repeatedly. The main contribution of our phylogenetic reconstruction will be to test which of these two scenarios is the most likely. We also investigate more thoroughly morphological divergence between J. EVOL. BIOL. 19 (2006) 253–266 ª 2005 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY Diversification of Otiteselline fig wasps four pairs of Otiteselline wasp species associated with the same fig species in order to understand how two congeneric species coexist in the same fig. Materials and methods Sampled species Sampled fig wasps and their collection details are listed in Table S2. We sampled Otitesella wasps from 22 of the 98 recognized species and subspecies of Galoglychia figs (75 fig species have been described in this section) scattered in the different subsections of the classification and from two species of Urostigma figs. When possible we tried to include the two morphospecies of Otitesella for each fig species of the Galoglychia section sampled. The two Otitesella species associated with Ficus burtt-davyi, Otitesella uluzi and Otitesella sesquianellata (van Noort & Compton, 1988) have been described; they are readily distinguishable based on morphology of the wings, ovipositor valves and clypeal margin in the females and mandibles in the males. When two Otitesella morphospecies were found in one host, one species shared characters with O. uluzi, whereas the other shared characters with O. sesquianellata. Based on morphology, we therefore classified Otitesella wasps as belonging to the ‘uluzi species-group’ or the ‘sesqui species-group’. Ficus ingens (section Urostigma) also hosts two species of Otitesella: Otitesella longicauda and Otitesella rotunda (Van Noort & Rasplus, 1997), those are also characterized by differences in ovipositor length, but morphologically do not correspond with the two species-groups we recognized from section Galoglychia. As mentioned in the introduction, the definition of the genus Philosycus is not clear, we thus included Philosycus wasps into our analyses, to investigate the validity of this genus. Another Otiteselline fig wasp, Comptoniella vannoorti, was chosen as an outgroup and we also included sequences available on Genbank for rooting the phylogenetic tree. DNA extraction and PCR Total genomic DNA was extracted from adult male and female wasps preserved in 100% ethanol using the quiaquick kit (Quiagen, Valencia, CA, USA). Each sequence was obtained from the DNA of a single wasp. The ITS2 ribosomal DNA region was targeted for PCR. The ITS2 was amplified and sequenced with primers ITSF (5¢ATT CCC GCA CCA CGC CTG GCT GA) and ITSR (5¢CGC CTG ATC TGA GGT CGT GA) (Campbell et al., 1993) as it proved useful to solve insect phylogenetic relationships at the genus level in many fig wasp taxa and also more generally in insects (Lopez-Vaamonde et al., 2001;Young & Coleman, 2004). We also used a more conserved marker; we sequenced the D3 and D2 of the nuclear large ribosomal subunit 28S rDNA to resolve deep nodes of the phylogenetic tree using primers D1F 255 (5¢ACC CGC TGA ATT TAA GCA TAT) and D3R (5¢TAG TTC ACC ATC TTT CGG GTC) (Harry et al., 1998). As 28S sequence divergence between species was low, we only sequenced 31 specimens for this particular marker. For both markers, PCR amplifications were performed in a 50 lL reaction volume containing 2 mM MgCl2, 250 lM of each dNTP, 1 lM of each primer and one unit of promega polymerase. Following an activation step of 4 min at 94 °C, the PCR mixture underwent 30 cycles of 1 min at 92 °C, 1 min at 50 °C and 1 min at 72 °C. To remove excess primers and DNTP after amplification, PCR products were gel-purified (QiaQuick, Qiagen). Sequencing was performed on both strands using the ABI prism Dye terminator Cycle sequencing Ready reaction Kit (Perkin-Elmer, Foster City, CA, USA) in a 10 lL volume containing 10 ng of purified DNA and 1.6 pmol of amplification primer. Sequencing reactions underwent 25 cycles of 30 s at 96 °C, 30 s at 50 °C and 4 min at 60 °C. The PCR products were purified and sequenced with the ABI Prism dye terminator (Perkin-Elmer). Sequence alignment and Phylogenetic analyses Sequences are available on Genbank (see Table S2). The PCR amplification of ITS2 always yielded single bands and multiple sequences from the same individual were identical which suggests that paralogues were not divergent or not amplified. Sequences were first aligned with the program ClustalW using the default settings. The alignment obtained for 28S sequences was unambiguous. On the other hand ITS2 sequences showed variation in length across species and some parts of the alignment obtained with ClustalW default settings were ambiguous. As outgroups sequences were too highly divergent to be properly aligned, we excluded them from further ITS2 analyses. We then used several strategies to obtain alignment of Otitesella ITS2 sequences: (1) We proceeded to several multiple alignments using ClustalW under different gap opening and gap extension costs, as the gap cost ratios are recognized as the most important alignment parameters (Wheeler, 1995). (2) We delimited three ambiguous parts of the alignment and excluded them from the analysis. (3) A typical ‘by eye’ corrected alignment was also made by modifying a default ClustalW alignment and manually inserting a minimum number of gaps in the ambiguous zones. (4) Finally, ambiguous parts of the alignment were recoded using INAASE (Lutzoni et al., 2000): each ambiguously aligned region was coded as a new character and each of the newly coded character was subjected to a specific step matrix for use in maximum parsimony (MP) analyses. All methods produced an alignment from which we built MP trees using PAUP* (Swofford, 2000), we used J. EVOL. BIOL. 19 (2006) 253–266 ª 2005 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY 256 E. JOUSSELIN ET AL. heuristic searches involving TBR branch swapping with 1000 random additions of taxa and gaps were treated as missing data. Sequences of species associated with Urostigma figs were designated as outgroup sequences in these analyses, as suggested by results obtained with 28S sequences. Node support for MP analyses was assessed through bootstrapping with heuristic searches of 500 replicates and ten random taxon addition sequences For the single alignment obtained from 28S sequences and the ‘by eye’ ITS2 alignment, phylogenies were also reconstructed using maximum likelihood (ML) optimality criterium and Bayesian phylogenetic inference. The ML analyses were conducted with PAUP* (Swofford, 2000). The most appropriate model of nucleotide substitution for ML analyses was chosen by comparing nested models with likelihood ratio tests (Posada & Crandall, 1998). Using a randomly chosen MP tree, we examined the fit of the data to 56 models of substitution and also tested the addition of parameters for proportion of invariant sites (I) and for heterogeneity of substitution across sites (C); this was done using the program Modeltest 3.06 (Posada & Crandall, 1998). The MP tree with the greatest )Ln score was used to estimate the model parameters (gamma shape, base frequencies, transition matrix). A ML heuristic search using TBR branch swapping was then run. We used bootstrapping to provide measures of clade support; heuristic searches of 200 replicates and ten random taxon addition sequences were conducted. Bayesian phylogenetic analyses were conducted using MrBayes 3 (Huelsenbeck & Ronquist, 2001). We used the GTR + C model of molecular evolution and a random starting tree for both data sets (28S and ITS2). We ran four chains of the Markov Chain Monte Carlo, sampling every 100 generations for 200 000 generations. We also ran the full analysis five times to avoid getting trapped in local optima. The chain appeared to reach the stationary phase by about the 20 000th generation, thus the first 200 trees were ‘the burn in’ of the chain and phylogenetic relationships are based on the subsequent 1800 trees. Cospeciation tests Phylogenetic reconstructions showed that the two otiteselline wasps coexisting in the same host fig are not sister species, but that they belong to two distinct clades. We thus tested whether the two clades have diversified in parallel. We tested the null hypothesis that the two phylogenetic trees (the ‘uluzi’ species-group and the ‘sesqui’ species-group) are not more congruent than expected by chance, using the maximum cospeciation analysis implemented in Treemap 1 (Page, 1995). The ‘uluzi’ phylogeny was mapped onto the ‘sesqui’ phylogeny to maximize the number of cospeciation events. Hence the ‘uluzi’ were considered as the symbionts and the ‘sesqui’ as the hosts. In each subtree we deleted taxa that did not have a correspondent in the other tree (a limitation of Treemap). The probability of obtaining the observed number of cospeciation events was then estimated by randomizing both the ‘uluzi’ and the ‘sesqui’ tree 10 000 times using the proportional to distinguishable model and generating a null frequency distribution. Ecological divergence between congeneric wasps associated with the same host fig We measured ovipositor length of four pairs of species: the two Otitesella species associated respectively with F. ingens, Ficus stuhlmannii, F. burtt-davyi and Ficus natalensis. Ten to fifteen wasps from each species were chosen at random. In the Otiteselline, the ovipositor is curled within the abdomen. Hence for ovipositor measurements, each wasp was dissected to reveal the entire length of the ovipositor (from the basal plates to the tip), placed in a drop of water between a slide and a cover slip and measured to the nearest 0.01 mm under a microscope. An estimate of the size of the wasps (hind tibia length) was also measured. The ovipositor length and the tibia length of the ‘sesqui’ and ‘uluzi’ wasps sharing the same host were compared using Mann–Whitney nonparametric tests. We also compared the mandible length and the clypeal margin length of the males (8–12 males depending on species) with Mann–Whitney nonparametric tests. Results Alignment and Phylogenetic relationships The aligned 28S sequences were 771 bp in length. Pairwise sequence divergence between any two taxa (excluding outgroup taxa) ranged from 0 to 5.6%. Heuristic searches yielded 16 136 most parsimonious trees based on 49 parsimony informative characters (length ¼ 218, CI ¼ 0.789). Likelihood ratio tests showed that the general reversible model with rate heterogeneity (GTR + C) was the most appropriate model for analysing the data [model parameters: empirical base frequencies with rate heterogeneity; gamma shape parameter ¼ 0.344, (A–C) ¼ 1, (A–G) ¼ 2.5698, (A–T) ¼ 1, (C– G) ¼ 1, (C–T) ¼ 1, (G–T) ¼ 5.9402]. The MP consensus tree, ML tree ()Ln ¼ 2412.51) and consensus tree obtained through Bayesian analysis had all the same topological structure. Analyses of 28S sequences confidently indicated that the wasps associated with Galoglychia figs formed a robust clade distinct from the Otitesella wasps associated with Urostigma figs (Fig. 1). Otitesella wasps associated with Urostigma figs were thus chosen as outgroups for ITS2 sequence analyses. Summaries of parsimony searches on alignments obtained through different strategies are given in Table 1. We compared the topologies of MP consensus trees obtained with the different alignment strategies: we found that in most cases a node was either present or unresolved. The main discordance between the four topologies was the placement of the Philosycus specimens, J. EVOL. BIOL. 19 (2006) 253–266 ª 2005 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY Diversification of Otiteselline fig wasps 257 Fig. 1 Consensus phylogram from a bayesian analysis of 28S sequences. Numbers above branches are posterior probabilities. Table 1 Summary of results of MP searches conducted on ITS2 alignments. ITS2 alignment Number of MP trees Number of parsimony informative characters/number of characters Tree length Consistency index GOP ¼ 1/GEP ¼ 5 GOP ¼ 10/GEP ¼ 1 GOP ¼ 5/GEP ¼ 5 ‘By eye’ corrected alignment Ambiguous zones excluded from the ‘by eye’ corrected alignment Ambiguous zones recoded with Inaase 18 36 36 2400 54 000 235/527 235/527 233/541 218/553 171/482 804 802 783 749 581 0.634 0.635 0.639 0.676 0.697 1728 174/485 637 0.719 GOP, gap opening penalty; GEP, gap extension penalty. in some of the analyses they did not form a monophyletic clade and in others analyses; even if they formed a clade its position was unstable (e.g. Fig. 2). The positions of the clade formed by Otitesella sp. 42, Otitesella sp. 43 and Otitesella sp. 45 [(either at the base of the uluzi clade, or unresolved (Figs 2 and 3)] and the wasps associated with Ficus lutea [(either at the base of the uluzi clade (Fig. 2), or at the base of the sesqui clade (Fig. 3)] were also sensitive to alignment strategies. The MP consensus trees obtained when excluding the highly variable zones or recoding them was similar to the one obtained with the manually corrected alignment. J. EVOL. BIOL. 19 (2006) 253–266 ª 2005 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY 258 E. JOUSSELIN ET AL. Fig. 2 Bootstrap MP tree (500 replicates) obtained from the analysis of the noncorrected ClustalW ITS2 alignment (settings: GOP ¼ 5, GEP ¼ 5). We chose to retain the alignment obtained ‘by eye’ for further analyses (ML and Bayesian) as it gave the shortest and best resolved MP tree amongst all other alignments. The general reversible model with rate heterogeneity (GTR + C + I) was chosen by Modeltest for analysing the data [model parameters: empirical base frequencies with rate heterogeneity; gamma shape parameter ¼ 1.673, proportion of invariant sites ¼ 0.2032, (A–C) ¼ 0.6070, (A–G) ¼ 2.0696, (A–T) ¼ 1.0992, (C–G) ¼ 1.2183, (C–T) ¼ 2.6654, (G–T) ¼ 1]. The MP consensus tree, the ML tree ()Ln ¼ 4488.4783) and the consensus tree obtained through Bayesian analysis were well resolved and almost perfectly congruent with each other (Fig. 3). Is the phylogeny of Otitesella wasps congruent with fig classification? Both markers, all alignments and tree-building methods (Figs 1–3) indicated that the wasps associated with Galoglychia figs formed a robust clade distinct from the Otitesella wasps associated with Urostigma figs. Another constant topological structure within the group J. EVOL. BIOL. 19 (2006) 253–266 ª 2005 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY Diversification of Otiteselline fig wasps 259 Fig. 3 The phylogram obtained in a ML analysis of corrected alignment of ITS2 sequences. Numbers above branches are, in order: percentage bootstrap support (500 replications) obtained for the same nodes in MP analyses/percentage bootstrap support (200 replications) for the ML analyses/posterior probabilities of the Bayesian analyses. Arrows indicate nodes that were different in the MP strict consensus tree. associated with Galoglychia figs, was the division into two strongly supported clades: the ‘sesqui’ group, the ‘uluzi’ group. In most topologies, the genus Philosycus including the Otitesella wasp associated with Ficus tettensis also formed a clade. Analyses of 28S sequences left the relationships between those three clades unresolved. The intra clade relationships were also unresolved with this marker due to very low sequences divergence. MP, ML analyses and Bayesian inference of ITS2 sequences based on the ‘by eye alignment’, placed the Philosycus group at the base of the ‘sesqui’ clade. However this position was not strongly supported and this node did not appear in the MP analyses based on noncorrected alignments (straightforward alignment obtained with ClustalW under different gap opening and extension penalty). In all reconstructions based on ITS2 sequences, within each of the ‘sesqui’ and the ‘uluzi’ clades, species clustered roughly according to the taxonomy of their host figs (Fig. 4). For example, within each species group, with the exception of some of the wasps associated with Ficus burkei, wasps associated with figs belonging to subsection Chlamydodorae formed a clade. Similarly, with a few exceptions, wasps associated with figs belonging to J. EVOL. BIOL. 19 (2006) 253–266 ª 2005 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY 260 E. JOUSSELIN ET AL. Fig. 4 The association of Otitesella and Philosycus fig wasps with fig subsections mapped on the MP strict consensus tree (ITS2). subsections Platyphyllae and Crassicostae formed monophyletic groups. Are Otitesella wasps specific to their hosts? In most cases, there was significant ITS2 sequence divergence between wasps associated with different fig species (from 1.2 to 56%). However, based on ITS2 sequences, the ‘sesqui’ wasps associated with F. stuhl- mannii and F. natalensis appeared to be the same (sequence divergence: 0–0.5%). Visual inspection of several individuals from each populations supported conspecific status of the three populations (van Noort & Rasplus, in preparation). The ‘sesqui’ wasps associated with Ficus louisii and Ficus elasticoides had very similar haplotypes (1.2% divergence) and the ‘uluzi’ wasps associated with F. natalensis and F. burkei (Otitesella sp. 46) were also very similar (1.4% divergence). J. EVOL. BIOL. 19 (2006) 253–266 ª 2005 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY Diversification of Otiteselline fig wasps 261 Among the wasps sampled for phylogenetic inquiry, with the exception of F. burkei, Ficus cyathistipuloides and F. elasticoides, each fig species hosted a species belonging to the ‘uluzi’ species-group and a species belonging to the ‘sesqui’ species-group, or a species from the ‘uluzi’ species-group together with a Philosycus species. Sampling of several wasps within either the ‘uluzi’ or the ‘sesqui’ species-groups, that were associated with the same fig species, but from different localities, usually showed little ITS2 sequence divergence (from 0 to 7.5%) and they always clustered together in the phylogenetic trees. For example there was only 1.2% sequence divergence between the two sampled populations of the ‘sesqui’ species-group associated with Ficus trichopoda, though one wasp was collected in South Africa and the other in Ivory Coast (c. 5000 km apart). The sequence divergence between the two ‘sesqui’ wasps associated with Ficus glumosa was also only 1.2% (one wasp was collected in South Africa the other in Tanzania: c. 2000 km apart). Similarly the three populations of the ‘sesqui’ wasp’s species associated with F. lutea grouped together; the two samples from South Africa had very similar sequences (divergence 0.3%), but were quite divergent from the one from Ivory Coast (7.3–7.5% divergence). Molecular data however revealed that there were two species of ‘sesqui’ wasps associated with F. burkei (species 26 is divided into two groups, Fig 2 and 3). reflect the absence of the wasp species in our collections. Ficus burkei hosted two species of ‘uluzi’ and a species of ‘sesqui’, the combination of which depended on collection. Only collection 3 and collections 4 and 5 had both an ‘uluzi’ and a ‘sesqui’ (see Table S2). For each topology, we conducted tests on phylogenetic trees including the two pairs of species found in the F. burkei collections with all the associations that were observed. According to topologies tested (eight combinations of trees were tested), the numbers of cospeciating events estimated with Treemap varied from 4 to 6. Six out of the eight tests detected significant cospeciation before correction of P values (P < 0.05), three tests were highly significant (P < 0.001), three tests were siginificant (0.001 < P < 0.05) two tests approached significance (P ¼ 0.06). After a conservative correction was applied following Lopez-Vaamonde et al. (2001), four tests were still significant (P < 0.05). One of the highest estimates of cospeciating events (six cospeciating events out of a maximum of eight) was obtained with one of the topologies including Philosycus in the ‘sesqui’ group (Fig. 5a), the number of cospeciation events observed was significantly greater than expected by chance (uncorrected P < 0.001, corrected P value, P ¼ 0.015). One of the lowest numbers of cospeciating events was obtained for one of the topologies when Philosycus species were excluded from the trees (Fig. 5b). The number of cospeciation events approached significance before correction (P ¼ 0.06), but was not significant after a conservative correction was applied (P ¼ 0.29). Have the ‘uluzi’ and ‘sesqui’ lineages diversified in parallel? Ecological divergence between congeneric wasps associated with the same host fig Our analyses were based on alternative topologies obtained trough the analyses of ITS2 data (Figs 2 and 3). We first compared the topology of the ‘uluzi’ clade to the topology of the clade including the ‘sesqui’ and Philosycus wasps. As the position of Philosycus was not strongly supported and also varied according to ITS2 alignment strategies, we also conducted tests excluding this group: we compared the topology of the ‘uluzi’ clade to the topology of the clade including only the ‘sesqui’ wasps. Apart from the uncertainty concerning the position of Philosycus wasps, the only differences between alternate topologies (topologies obtained through the analyses of different ITS2 alignments or through different different optimality criteria) were the positions of the ‘sesqui’ species associated with F. glumosa and the ‘uluzi’ species associated with F. natalensis, we thus tested topologies with alternate positions. All associations considered represented cases where the wasps in each species group were collected on the same fig tree. Any taxon that did not have its correspondent in the other species-group was pruned from the tree, as these situations would be interpreted as extinction events in Treemap, though they probably only In F. natalensis and F. burtt-davyi, the ‘sesqui’ wasps had a significantly longer ovipositor than the ‘uluzi’ wasps (F. natalensis: Mann–Whitney U test, P < 0.001; F. burttdavyi: Mann–Whitney U test, P < 0.001). Similarly the ovipositor of O longicauda (F. ingens) was significantly longer than the ovipositor of O. rotunda (Mann–Whitney U test, P < 0.001) (Fig. 6). On the other hand, the ovipositor lengths of the ‘sesqui’ and the ‘uluzi’ wasps found in F. stuhlmannii were not significantly different (Man–Whitney U test, P ¼ 0.31). In all cases, the ‘uluzi’ female wasps were bigger than the ‘sesqui’ female wasps sharing the same fig (Mann–Whitney U test: F. burtt-davyi P < 0.05, F. stuhlmannii P < 0.001, F. natalensis P < 0.05). Similarly O. rotunda females were bigger than O. longicauda females (Mann–Whitney U test, P ¼ 0.05). These size differences were also found in males, the mandible length and the length of the clypeal margin of ‘sesqui’ males were always significantly smaller (Table 2). This size difference is congruent with the data presented in the species description of the wasps associated with F. burtt-davyii (Van Noort & Compton, 1988). We also found a difference between O. rotunda and O. longicauda males. In both species, there are two Are fig trees associated with one or several wasps of each species-group? J. EVOL. BIOL. 19 (2006) 253–266 ª 2005 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY 262 E. JOUSSELIN ET AL. Fig. 5 Phylogenies for ‘sesqui’ wasps [and Philosycus wasps for (a)] and their associated ‘uluzi’ wasps. (a) The topologies are based on one of the MP trees. Treemap identified 1 optimal reconstruction with 6 cospeciation events, 3 duplication events, 0 host switch and 12 sorting events. (b) The topologies are based on one the MP trees with alternative positions for the ‘uluzi’ wasp associated with F. natalensis, and excluding the Philosycus wasps and their associate in the ‘uluzi’ tree. Treemap identified 2 optimal reconstructions with 4 cospeciation events, 1–3 duplication events, 0 host switches and 14 sorting events. Fig. 6 Ovipositor length of Otitesella wasps coexisting in the same fig species. conditionally determined morphotypes of males: a fighting morph (religiosa) and a smaller dispersing morph (digitata) (Greeff & Ferguson, 1999; Pienaar & Greeff, 2003). For each morph, the males of O. longicauda were smaller than the males of O. rotunda (Table 2). Discussion Otitesella taxonomy Based on two sources of molecular data, our analyses result in a stable phylogenetic hypothesis with several well-supported nodes. Our reconstruction of the phylogeny of Otitesella species is consistent with the broad level taxonomy of their host figs. Otitesella species associated respectively with Ficus species in sections Galoglychia and Urostigma, form two distinct clades. These two clades correspond to the two taxonomic groups: O. africana and O. digitata (Van Noort & Rasplus, 1997). Survey of the literature, our personal collections and additional sampling revealed that most African fig species within section Galoglychia host two species of Otiteselline, each belonging to a different morphogroup (Table S1). Both molecular markers suggest that these two groups correspond to two clades: the ‘uluzi’ wasps and the ‘sesqui’ wasps. Most analyses suggest that species in Philosycus also form a clade. Several analyses of ITS2 sequences suggest that Philosycus are basal to the ‘sesqui’ clade, but this position is not strongly supported and is sensitive to alignment strategies. In any case, all analyses suggest that Philosycus are more closely related to the Otitesella associated with Galoglychia figs than the O. digitata clade. This position means that Philosycus cannot be recognized as a genus, as the maintenance of this taxonomic status would result in Otitesella being paraphyletic. Future taxonomical revision will synonymize the genera (Rasplus & van Noort, in preparation). In summary all analyses indicates that the ‘sesqui’ species-group (either including or excluding the Philosycus clade) and ‘uluzi’ species-group form two radiations superimposed on the fig tree/pollinator system. This differs from the other study of the patterns of diversification of a genus of parasitic fig wasps: in Apocryptophagus congeneric wasp species occurring on the same J. EVOL. BIOL. 19 (2006) 253–266 ª 2005 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY 263 Diversification of Otiteselline fig wasps Table 2 Measurements of Otitesella males. F. burtt-davyii Mandible (lm) Clypeal margin (lm) F. natalensis Mandible (lm) Clypeal margin (lm) F. stuhlmannii Mandible (lm) Clypeal margin (lm) F. ingens Digitata morph Mandible (lm) Clypeal margin (lm) Religiosa morph Mandible (lm) Clypeal margin (lm) O. ‘sesqui’ O. ‘uluzi’ 3.43 ± 0.25 (n ¼ 8) 4.13 ± 0.6 (n ¼ 8) 3.96 ± 0.62 (n ¼ 12) 4.84 ± 0.8 (n ¼ 12) 0.01 0.02 4 ± 0.29 (n ¼ 9) 4.88 ± 0.4 (n ¼ 9) 4.42 ± 0.36 (n ¼ 12) 5.5 ± 0.69 (n ¼ 12) 0.007 0.03 3.98 ± 0.51 (n ¼ 8) 4.77 ± 0.67 (n ¼ 8) O. longicauda 4.45 ± 0.64 (n ¼ 11) 5.23 ± 0.77 (n ¼ 11) O. rotunda 0.09 0.04 0.96 ± 0.2 (n ¼ 13) 3 ± 0.3 (n ¼ 13) 1.67 ± 0.34 (n ¼ 15) 3.61 ± 0.44 (n ¼ 8) <0.001 0. 002 3.08 ± 0.28 (n ¼ 15) 6.21 ± 0.57 (n ¼ 15) <0.001 0.002 2.5 ± 0.26 (n ¼ 15) 5.4 ± 0.45 (n ¼ 15) P Mann–Whitney U tests were used to analyse differences between ‘uluzi’ and ‘sesqui’ wasps and O. rotunda and O. longicauda. fig seem to have originated on their host (Weiblen & Bush, 2002). The evolution of the fig/Otitesella association: is the association species-specific and is Otitesella wasp diversification constrained by their host association? Morphological descriptions of Otitesella species are often lacking, but in most cases, in each morphogroup, the molecular data support the existence of different genetic entities of wasps associated with different fig species: each wasp species is specific to its host. However there was one definite case where one species of wasp was associated with two fig species. In South Africa at least, F. natalensis shares the same ‘sesqui’ species as F. stuhlmannii. In several other cases, the presence of a barrier to gene flow between wasps associated with different figs, but with very similar sequences (e.g. only 1.2% between ‘sesqui’ wasps associated with F. louisii and F. elasticoides) cannot be discarded based solely on the data we present here. In these particular cases, more variable markers will be necessary to rigorously test the specificity of the relationship. Further, within each group (‘sesqui’ or ‘uluzi’), over a wide geographical range, Otitesella species collected from the same host fig species group together in our phylogenetic reconstruction. From Tanzania to South Africa, Otitesella wasps associated with F. glumosa show very little sequence divergence. Otitesella ‘sesqui’ wasps associated with F. lutea exhibit high molecular variability, but the different specimens still cluster together on the phylogenetic tree. This shows that host association does constrain wasp speciation. We also show that within each species-group, species often cluster according to the taxonomy of their host figs. As suggested by cospeciation tests, after the initial splitting event, both clades have to some extent diversified in parallel. There is no direct biological interaction between the ‘sesqui’ and the ‘uluzi’ species associated with the same host fig. Hence, some of the parallelisms detected in their phylogenies suggest that both groups have cospeciated with their host figs or at least that host association is phylogenetically conserved. This is not surprising, as Otiteselline species lay their eggs around the same time that figs are receptive for pollination; hence it is likely that they rely on the same volatiles as the pollinators to find their host tree. As suggested by LopezVaamonde et al. (2001), the existence of a chemical interaction between nonpollinating wasps laying at fig receptivity and their host fig can be a determinant of host specificity and could mediate cospeciation. Nonpollinating wasps laying their eggs early during fig development might be more prone to cospeciation. A real quantification of cospeciation will only be possible when a robust phylogeny of figs belonging to the Galoglychia phylogeny becomes available. The lack of variability of molecular markers at this taxonomic level has hampered our efforts to resolve the phylogeny. Comparison of sequence divergence between pairs of pollinators and pairs of Otitesella associated with the same fig species would also indicate whether cospeciation is a likely scenario. Our results also reveal obvious host shifts. For example, the fact that F. burkei hosts several species of each group and that Ficus cyasthistipuloides hosts two Philosycus sp. could result from several host shifts. But the most obvious case of host shifting is represented by the sesqui wasps shared by F. stuhlmannii and F. natalensis. The taxonomic position of the host figs and the comparison of the ‘sesqui’ and ‘uluzi’ phylogenies suggest the following J. EVOL. BIOL. 19 (2006) 253–266 ª 2005 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY 264 E. JOUSSELIN ET AL. scenario: the ‘sesqui’ species associated with F. natalensis has colonized F. stuhlmannii, while the original ‘sesqui’ species associated with F. stuhlmannii has become extinct. Ecological divergence between ‘uluzi’ and ‘sesqui’ We show that the ‘sesqui’ wasp has a longer ovipositor than the ‘uluzi’ wasp coexisting in the same fig species. Similarly O. longicauda, which inhabits F. ingens figs, has a longer ovipositor than O. rotunda. This difference in ovipositor length might be linked to differences in timing of oviposition (Kerdelhué et al., 2000). The two species probably lay their eggs at different stages of the fig development (Van Noort & Compton, 1988; Compton, 1993). Indeed, nonpollinators with short ovipositors are constrained to lay their eggs early in the fig development when the fig wall is thin. Nonpollinators with longer ovipositors can oviposit through thicker fig walls, later in the fig development. This assumption is consistent with the data on wasps’ size: wasps with longer ovipositors, which can potentially lay their eggs later, are smaller in both sexes. This result is logical as eggs laid later have less time to complete their development. Weiblen & Bush (2002) suggested that parasitic wasp speciation could result from divergence in the timing of oviposition with respect to fig development. They based their assumption on results showing that in the genus Apocryptophagus there has been repeated evolution of divergence in ovipositor length between sister species of wasps that use the same host fig. Similarly, the speciation events that led to the splits between the ‘uluzi’ and the ‘sesqui’ group or between O. longicauda and the group containing O. rotunda could also correspond to ecological speciation (following Johannesson, 2001). Resource competition between wasps from different populations could have led to specialization in time of oviposition and divergence in ovipositor length. Alternatively, the character differences observed (ovipositor length, body size) have not played a causal role in the speciation process of the two groups and have evolved subsequently through competition. In any case, difference in ovipositor length seems to be an important component of the maintenance of the two congeneric species on the same host fig. In this context, the case of the F. natalensis sesqui colonizing F. stuhlmannii is quite perplexing. A priori, this host shift seems to correspond perfectly to a case of invasion of a vacant niche: a ‘sesqui’ wasp is replacing another ‘sesqui’ wasp. However, as a result of the shift, the ‘sesqui’ and ‘uluzi’ wasps coexisting in F. stuhlmannii have ovipositors of the same length. It would be interesting to determine the timing of oviposition and the respective ecological niches of the two species to see if they overlap. in section Galoglychia are divided into two species-groups that have both diversified with their host figs. Contrary to the results obtained on Apocryptophagus, congeneric Otitesella wasps have not evolved in situ on the host, but are the result of two independent radiations. The occurrence of congeneric species associated with the same fig species, but showing differences in ovipositor length is quite common in parasitic wasp lineages (Weiblen & Bush, 2002; personal observation in genus Aprocrypta, Sycoscapter). A parallel can be made between the study of congeneric wasps associated with figs and the study of diversification of complex of species in island archipelagos (e.g. Losos et al., 1998; Shaw, 2002). Such isolated biotas, often shelter, closely related endemic species. This can be the result of single invasions followed by ecological divergence or multiple radiations diversifying across islands. This parallel underlines the fact that our understanding of the diversification of fig wasps must not only focus on cospeciation but also take into account the diversification of ecological niches within the fig. Figs wasps form well defined communities and comparative study of their patterns of diversification can combine ecological data (such as timing of oviposition, host location) and phylogenetic data and allow tests of the ecological factors mediating modes of speciation (in this case, cospeciation, host shift or speciation on the host through ecological differentiation) which makes them ideal model for the study of diversification of insect communities. Acknowledgments We thank Christoff Erasmus, Snowy Bajnath and Anthony Watsham for providing us with specimens. Many thanks to Jason Pienaar for sharing some of his measurements of F. ingens wasps. We are grateful to A.P. Vogler, G. Stone, James Cook for helpful comments on various drafts of this manuscript. This material is based upon work supported by the National Research Foundation under Grant number 2053809 to JMG. Any opinion, findings and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National. E. J was supported by a NRF post doctoral fellowship. Supplementary Material The following material is available from http://www. blackwellpublishing.com/Products/journals/suppmat/jeb/ jeb968/jeb968sm.htm Table S1 Records of otiteselline species-groups associated with Afrotropical Ficus species. Table S2 List of wasps sampled. Conclusions References We show here that otiteselline wasps are often specific to their host fig and that Otitesella species associated with figs Barker, N.P. 1985. Evidence of a volatile attractant in Ficus ingens (Moraceae). Bothalia 15: 607–611. J. EVOL. BIOL. 19 (2006) 253–266 ª 2005 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY Diversification of Otiteselline fig wasps Berg, C.C. 1989. Classification and distribution of Ficus. Experientia 45: 605–611. Berg, C.C. & Wiebes, J.T. 1992. African Fig Trees And Fig Wasps. North Holland, Amsterdam. Bouček, Z. 1993. The genera of Chalcidoid wasps from Ficus fruit in the New World. J. Nat. Hist. 27: 173–217. Bronstein, J. 1991. The non-pollinating wasp fauna of Ficus pertusa: exploitation of a mutualism? Oikos 61: 175–186. Campbell, B.C., Steffen-Campbell, J.D. & Werren, J.H. 1993. Phylogeny of the Nasonia complex (Hymenoptera: Pteromalidae) inferred from an internal transcribed spacer (ITS 2) and 28S rDNA sequences. Insect Mol. Biol. 2: 225–237. Compton, S.G. 1993. One way to be a fig. Afr. Entomol. 1: 151–158. Cook, J. & Rasplus J.-Y. 2003. Mutualists with attitude: coevolving fig wasps and figs. TREE 18: 241–248. Greeff, J.M. & Ferguson, W.H. 1999. Mating ecology of the non-pollinating fig wasps of Ficus ingens. Anim. Behav. 57: 215–222. Harry, M., Solignac M. & Lachaise, D. 1998. Molecular evidence for parallel evolution of adaptive syndromes in fig-breeding Lissocephala (Drosophilidae). Mol. Phylogenet. 9: 542–551. Herre, E.A., Machado, C.A., Bermingham, E., Nason, J.D., Windsor, D.M, MacCafferty, S.S., Van Houthen, W. & Bachman, K. 1996. Molecular phylogenies of figs and their pollinator wasps. J. Biogeogr. 23: 521–530. Huelsenbeck, J.P. & Ronquist, F. 2001. MrBayes: Bayesian inference of phylogeny. Bioinformatics 17: 754–755. Jackson, A.P. 2004. Cophylogeny of the Ficus microcosm. Biol. Rev. 79: 751–768. Johannesson, K. 2001. Parallel speciation: a key to sympatric divergence. TREE 16: 148–153. Jousselin, E., Rasplus, J.-Y. & Kjellberg F. 2001. Shift to mutualism in parasitic lineages of the fig/fig wasp interaction. Oikos 94: 287–294. Jousselin, E., Rasplus, J.-Y. & Kjellberg F. 2003. Convergence and coevolution in a mutualism: evidence from a molecular phylogeny of Ficus. Evolution 57: 1255–1269. Kerdelhué, C., Le Clainche, I. & Rasplus, J.-Y. 1999. Molecular phylogeny of Ceratosolen species pollinating Ficus of the subgenus Sycomorus sensu stricto: biogeographical history and origins of the species – specificity breakdown cases. Mol. Phylogenet. Evol. 11: 401–414. Kerdelhué, C., Rossi, J.-P. & Rasplus, J.-Y. 2000. Comparative community ecology studies on old world figs and fig wasps. Ecology 81: 2832–2849. Lopez-Vaamonde, C., Dixon, D., Cook, J.M. & Rasplus, J.-Y. 2002. Revision of the Australian species of Pleistodontes (Hymenoptera: Agaonidae) fig – pollinating wasps and their host plant affiliation. Zool. J. Linn. Soc. 136: 637–683. Lopez-Vaamonde, C., Rasplus, J.-Y., Weiblen, G.D. & Cook, J.M. 2001. Molecular phylogenies of Fig wasps: partial cocladogenesis of pollinators and parasites. Mol. Phylogenet. Evol. 21: 55–71. Losos, J.B., Jackman, T.R., Larson, A., de Queiroz, K. & Rodriguez-Schettino, L. 1998. Contingency and determinism in replicated adaptive radiation of island lizards. Science 279: 2115–2118. Lutzoni, F., Wagner, D.L., Reeb, V. & Zoller, S. 2000. Integrating ambiguously aligned regions of DNA sequences in phylogenetic analyses without violating positional homology. Syst. Biol. 49: 628–651. Machado, C., Jousselin, E., Kjellberg, F., Compton, S.G. & Herre, E.A. 2001. Phylogenetic relationships, historical biogeography 265 and character evolution of fig pollinating wasps. Proc. Roy. Soc. Lond. B. 268: 685–694. Machado, C.A., Herre, E.A., McCafferty, S. & Bermingham, E. 1996. Molecular phylogenies of fig pollinating and nonpollinating wasps and the implications for the origin and evolution of the fig-fig wasp mutualism. J. Biogeogr. 23: 531–542. Michaloud, G., Michaloud-Pelletier, S., Wiebes, J.T. & Berg, C.C. 1985. The co-occurrence of two pollinating species of fig wasp on one species of fig. Proc. Kon. Akad. Van Wet. (C) 88: 93–119. Mitter, C. & Farrell, B. 1991. Macroevolutionary aspects of insect–plant relationships. In: Insect–Plant Interactions (E. Bernays, ed.), pp. 35–78. CRC Press, Boca Raton, FL. Mitter, C., Farrell, B. & Futuyma, D.J. 1991. Phylogenetic studies of insect-plant interactions: insights into the genesis of diversity. Trends Ecol. Evol. 6: 290–293. Molbo, D., Machado, C., Sevenster, J.G., Keller, L. & Herre, E.A. 2003. Cryptic species of fig pollinating wasps: implications for the evolution of the fig-fig wasp mutualism, sex allocation, and precision of adaptation. Proc. Natl. Acad. Sci. USA 100: 5867–5872. van Noort, S. & Compton, S.G. 1988. Two new species of Otitesella (Hymenoptera, Chalcidoidea, Pteromalidae) from Ficus burttdavyii. Proc. Kon. Neder. Akad. Wetensch. 91: 419–427. van Noort, S. & Rasplus, J.-Y. 1997. Revision of the Otitesellinae fig wasps (Hymenoptera: Chalcidoidea: agaonidae), I: the Otitesella digitata species-group of the Afrotropical region, with a key to afrotropical species of Otitesella Westwood. Afr. Entomol. 5: 125–147. van Noort, S., Ware, A.B. & Compton, S.G. 1989. Pollinatorspecific volatile attractants released from the figs of Ficus burtt-davyi. S. Afr. J. Sci. 85: 323–324. Page, R.D.M. 1995. TREEMAP [WWW document]. URL http:// taxonomy.zoology.gla.ac.uk/rod/treemap.html. Pienaar, J. & Greeff, J.M. 2003. Maternal control of offspring sex and male morphology in the Otitesella fig wasp. J. Evol. Biol. 16: 244–253. Posada, D. & Crandall, K.A. 1998. Modeltest: testing the model of DNA substitution. Bioinformatics 14: 817–818. Ramirez, B.W. 1970. Host specificity of fig wasps (Agaonidae). Evolution 24: 681–691. Rasplus, J-Y, Kerdelhué, C., Le Clainche, I. & Mondor, G. 1998. Molecular phylogeny of fig wasps: Agaonidae are not monophyletic. C.R. Acad. Sci. 321: 517–527. Rasplus, J-Y. 1994. The one-to-one species specificity of the Ficus – Agaonidae mutualism: how casual? In: The biodiversity of sn Plants (L. J. G. Van der Maesen, X. M. Van der Burgt & J. M. Van Medenbach de Rooy, eds), pp. 639–649. Kluwer Academic Publishers, Dordrecht. Shaw, K.L. 2002. Conflict between nuclear and mitochondrial DNA phylogenies of a recent species radiation: what mtDNA reveals and conceals about modes of speciation in Hawaian crickets. Proc. Natl. Acad. Sci. U.S.A. 99: 16122–16127. Strauss, S.Y. & Zangler, A.R. 2002. Plant-insect interactions in terrestrial ecosystems. In: Plant–Animal Interactions: An Evolutionary Approach (C. M. Herrera & O. Pellmyr, eds), pp. 77–106. Blackwell Science, MA, USA. Swofford, D.L. 2000. Phylogenetic Analysis Using Parsimony (*And Other Methods). Sinauer, Sunderlands, MA, USA. Ware, A.B., Kaye, P.T., Compton, S.G. & Van Noort, S. 1993. Fig volatiles; their role in attracting pollinators and maintaining pollinator specificity. Plant Syst. Evol. 186: 147–156. J. EVOL. BIOL. 19 (2006) 253–266 ª 2005 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY 266 E. JOUSSELIN ET AL. Weiblen, G.D. 2000. Phylogenetic relationships of functionally dioecious Ficus (Moraceae) based on ribosomal DNA sequences and morphology. Am. J. Bot. 87: 1342–1357. Weiblen, G.D. & Bush, G.L. 2002. Speciation in fig pollinators and parasites. Mol. Ecol. 11: 1573–1578. Weiblen, G.D. 2001. Phylogenetic relationships of fig wasps pollinating functionally dioecious Ficus based on mitochondrial DNA sequences and morphology. Syst. Biol. 50: 1–25. West, S.A., Herre, E.A, Windsor, D.M. & Green, R.S. 1996. The ecology and evolution of the New World non-pollinating fig wasp communities. J. Biogr. 23: 447–758. Wheeler, W. 1995. Sequence alignment, parameter sensitivity and the phylogenetic analysis of molecular data. Syst. Biol. 44: 321–331. Wiebes, J.T. 1969. Philosycus, a new genus of fig wasps allied to Otitesella westwood (Hymenoptera, Chalcidoidea, Torymidae). Annales Musée Royal de l’Afrique Centrale. Série in 8. Sci. Zool. 175: 439–445. Wiebes, J.T. 1982. The phylogeny of the Agaonidae (Hymenoptera, Chalcidoidea). Neth. J. Zool. 32: 395–411. Young, I. & Coleman, A.W. 2004. The advantages of the ITS2 region of the nuclear rDNA cistron for analysis of phylogenetic relationships of insects: a Drosophila example. Mol. Phylogenet. Evol. 30: 236–242. Received 8 February 2005; revised 22 April 2005; accepted 25 April 2005 J. EVOL. BIOL. 19 (2006) 253–266 ª 2005 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY