Combining biogeographic, ecological, morphological, molecular and chemical data, we document departure from strict specialization in the fig-pollinating wasp mutualism. We show that the pollinating wasps Elisabethiella stuckenbergi and Elisabethiella socotrensis form a species complex of five lineages in East and Southern Africa. Up to two morphologically distinct lineages were found to co-occur locally in the southern African region. Wasps belonging to a single lineage were frequently the main regional pollinators of several Ficus species. In South Africa, two sister lineages, E. stuckenbergi and E. socotrensis, pollinate Ficus natalensis but only E. stuckenbergi also regularly pollinates Ficus burkei. The two wasp species co-occur in individual trees of F. natalensis throughout KwaZulu-Natal. Floral volatile blends emitted by F. natalensis in KwaZulu-Natal were similar to those emitted by F. burkei and different from those produced by other African Ficus species. The fig odour similarity suggests evolutionary convergence to attract particular wasp species. The observed pattern may result from selection for pollinator sharing among Ficus species. Such a process, with one wasp species regionally pollinating several hosts, but several wasp species pollinating a given Ficus species across its geographical range could play an important role in the evolutionary dynamics of the Ficus-pollinating wasp association.
Ficus species are pollinated by one or a few species of agaonid wasps that breed within the enclosed inflorescences called figs. These wasps are generally host-specific . This led to the assumption that co-speciation of figs and fig-pollinating wasps is prevalent [2,3]. However, accumulating data demonstrate numerous exceptions to parallel cladogenesis [4–7]. Two or more agaonid wasps pollinating the same host Ficus has been documented [8–10], as well as cases of one wasp species pollinating several Ficus species . Studies based on molecular markers have shown at local- , regional-  and in one case broad-scale , the coexistence of two or more genetically closely related pollinators on the same host. In some cases, coexistence of distantly related pollinators has been documented . Genetic data on other species suggest that, while more than one wasp species may occur across the fig-distribution range, a single pollinator species is usually present locally [11,15,16]. In some cases, the two wasp species pollinating the same Ficus species are associated with different environments . In other cases, one of the two sympatric species is a non-pollinating agaonid [17,18].
There are numerous observations of agaonid wasps associated with a Ficus species that sometimes also visit (and pollinate) another Ficus species . The importance of such events in terms of the potential for genetic introgression among Ficus species should not be understated . Data from Panama could correspond to such a case or alternatively to a situation of an agaonid wasp species that is a quantitatively important pollinator of two or more Ficus species [11,19]. At least two cases of a wasp being the major pollinator of two Ficus species have been confirmed by molecular markers: Ceratosolen arabicus is the sole pollinator of both Ficus sycomorus and Ficus mucuso  and Alfonsiella binghami, the sole pollinator of both Ficus petersii and Ficus sthulmannii . Finally, some molecular data suggest that this could also be the case in the species complex of Elisabethiella stuckenbergi–Elisabethiella socotrensis, which is associated with a number of Ficus species in East Africa . Such situations may be evolutionarily relevant. They suggest that transient or persistent pollinator sharing may be an important facet of how the Ficus-pollinating wasp association has evolved and diversified.
Fig pollinators are attracted by volatile compounds emitted by receptive figs of their host Ficus, i.e. by figs at the stage when the wasps enter them to oviposit, and they are usually not attracted by the odours of other Ficus species [22–24]. The chemical message generally comprises several compounds and is differentiated among species [24,25]. The wasps are assumed to recognize a mixture of volatile organic compounds constituting the host species'-specific signature [23,25–27]. Data on two Ficus species have evidenced significant among-population variation of receptive fig floral odour within species . As in other systems, intraspecific variation in the floral blend, forming floral ‘chemotypes’, could allow for more than one species of pollinator to be associated with the same host species [29,30]. Conversely, similarity in the floral blends produced by two Ficus species could facilitate their pollination by the same wasp species.
In this study, we collected and compared wasp morphological and genetic data, fig-visitation patterns and fig volatile chemical profile data within and among localities to document and analyse a situation of non-strict specificity in the Ficus-pollinator mutualism. The African section, Galoglychia of Ficus, provides a unique opportunity to analyse systems in which some wasps use several host species and hosts are pollinated by several wasp species [31,32]. In KwaZulu-Natal, South Africa, E. socotrensis Mayr and E. stuckenbergi Grandi have been reported to be the sole agaonid wasps colonizing F. natalensis Hochst. . Furthermore, E. stuckenbergi is the main pollinator of F. burkei Miq. in KwaZulu-Natal. The E. stuckenbergi–E. socotrensis species complex is not resolved and could represent either a geographically variable complex of species using multiple hosts or a species complex showing host-dependent differentiation in some cases . With the limited data available, we hypothesized that at a local scale, E. stuckenbergi pollinators found in F. burkei and in F. natalensis represent a single genetic lineage. In parallel, we expected that E. socotrensis co-occurring with E. stuckenbergi in KwaZulu-Natal represents a separate evolutionary lineage because the two are morphologically distinct .
In order to address this conundrum of complex host associations, we used molecular tools to clarify the genetic clusters within the E. stuckenbergi–E. socotrensis species complex, and asked whether this clustering follows Ficus host identity or geographical distribution. Then, in KwaZulu-Natal, once the local correspondence between morphotype and genetic identity had been confirmed, we assessed which species colonized individual figs on a number of F. natalensis trees across several locations in order to establish (i) whether the two wasp species are effective pollinators of F. natalensis; (ii) whether the two wasp species frequently co-occur on the same tree, over a broad geographical area; and (iii) whether the two wasp species compete for receptive figs. Finally, we analysed the odours produced by receptive figs from the sampling sites in order to test (i) whether variation in the suite of volatile chemicals produced by F. natalensis could be associated with preferential visitation by one or the other pollinating wasp species and (ii) whether the volatile chemicals of receptive figs were similar between F. burkei and F. natalensis, facilitating wasp sharing.
2. Material and methods
(a) Ficus natalensis and Ficus burkei
Both F. natalensis subspecies natalensis (henceforth, F. natalensis will be designated as F. natalensis subspecies natalensis except in parts of the text where ambiguity between subspecies is possible) and F. burkei belong to subgenus Urostigma, section Galoglychia, subsection Chlamydodorae. They are distributed from South Africa to Uganda and Kenya in East Africa. Ficus natalensis has a more coastal distribution than F. burkei . In southeast Africa, they form a monophyletic species group with Ficus craterostoma, F. petersii and Ficus lingua  and subsection Chlamydodorae is further represented by Ficus burtt-davyi, Ficus fisheri and Ficus ilicina . The two Ficus species studied here are morphologically well-defined and are easily identified . They are not sister species, but are closely related . Although the two species mainly occupy different habitats (forest and rocky habitat for F. natalensis and savannah woodland and wet or dry forest for F. burkei ), they also co-occur locally .
In most of South Africa, E. stuckenbergi is the sole pollinator of F. burkei, with sporadic occurrences of E. socotrensis, but in Zambia, Tanzania, Kenya, Zimbabwe and Botswana, F. burkei is also regularly pollinated by Alfonsiella brongersmai with both wasp species present in the same crop (i.e. a cohort of figs produced by a single tree) . In contrast, F. natalensis is pollinated by at least five fig wasp species . In East Africa, it is pollinated by E. socotrensis, A. brongersmai, Alfonsiella natalensis and Alfonsiella longiscapa, which can all be present in the same crop of figs . Only two of these pollinator species, A. longiscapa and E. socotrensis, are present in KwaZulu-Natal, where most F. natalensis trees are pollinated by E. socotrensis and by E. stuckenbergi. However, in the extreme north of KwaZulu-Natal, extending from Kosi Bay (figure 1) northwards into Mozambique, A. longiscapa replaces these two species as the pollinator of F. natalensis . Further, the figs of both F. natalensis and F. burkei are also entered by the sycoecine non-pollinating wasps Philocaenus barbarus Grandi and Crossogaster odorans Wiebes that gall ovules usually without providing a pollination service , but in the population of F. natalensis pollinated by A. longiscapa, these two sycoecine species are replaced by another two species: Philocaenus medius van Noort and Crossogaster lurida van Noort [37–40].
(b) The Elisabethiella socotrensis–Elisabethiella stuckenbergi species group
Both E. socotrensis and E. stuckenbergi are widely distributed in East Africa and extend into South Africa. They do not seem to form distinct monophyletic groups . In the original descriptions, the length–width ratio of the head is different [34,41], suggesting that they are adapted to different fig ostiole shapes. Indeed, head shape, among other diagnostic differences, was shown to be a function of ostiole length as determined by fig wall thickness for wasps associated with section Galoglychia . Further diagnostic differences recorded for samples collected in South Africa include the proportional length to width shape of the mandibular appendage, number of rows of teeth on the appendage, shape and armature of the mandible itself, shape of the antennal pedicel, and the length of the post-marginal vein relative to the stigmal vein. These are all distinctive morphological traits that delimit species within the family Agaonidae.
(c) Phylogenetic methods applied to pollinating wasps
Phylogenetic relationships within the E. socotrensis–E. stuckenbergi species group were examined using a Bayesian analysis of the mitochondrial gene cytochrome oxidase I (COI). Samples were obtained from multiple sites to assess genetic grouping (see electronic supplementary material, ESM1, for more details, method S1 and table S1). We estimated the mean percentage sequence divergence (p-distance) ± s.d. between and within the clades or lineages of Elisabethiella. A limited number of wasps belonging to the different lineages were examined morphologically to assess whether (i) locally coexisting wasp lineages were correctly separated and (ii) two consistent morphotypes were recognizable across the investigated range.
(d) Local occurrence of the two pollinating wasp species on Ficus natalensis in KwaZulu-Natal
To analyse local and regional co-occurrence of E. stuckenbergi and E. socotrensis in F. natalensis across its distribution in KwaZulu-Natal, wasps were sampled from figs of 28 F. natalensis trees at seven sites (Vernon Crooks, Durban, Mtunzini, Tugela Mouth, St Lucia, Richard's Bay and Kosi Bay; see electronic supplementary material ESM2, table S4, for more details). Sampling was carried out in March–June 2008 and covered a range of 290 km (figure 1). At each site, 20–40 figs per tree were collected, either during fig receptivity when pollinator foundresses were present, or at the end of pollinator development just before wasp emergence. Figs collected at emergence stage were placed in closed vials and wasps were allowed to emerge naturally from their natal figs within the vials. Figs collected at receptivity were opened and pollinators were extracted from within the fig cavity. All figs were stored individually together with their pollinators in 70 per cent ethanol. In the laboratory, fig wasps were sorted to morpho-species under a dissecting microscope. For each sampled receptive fig, the numbers of foundress E. socotrensis and E. stuckenbergi females trapped inside the fig cavity were recorded. For mature figs, presence–absence of the two wasp species was recorded. Furthermore, because receptive figs were collected close to the end of the receptive phase of the crops, but before abortion of unvisited figs, we could not only measure the relative proportion of receptive figs visited by each species but we could also provide an estimate of the absolute proportion, taking into account the proportion of still unvisited figs.
We tested for among-site variation of each of the two variables ‘proportions of figs per tree visited by E. socotrensis only and by E. stuckenbergi only’ using a generalized linear model with a quasi-binomial error distribution (including only the three sites at which at least two trees were sampled: St Lucia, Mtunzini, Durban).
We also tested whether the two pollinator species competed for access to figs of a tree. Interspecific competition for available receptive figs will translate into presence of one species reducing the probability that the fig is visited by the other species. We tested this probability using contingency χ2-tests at the crop level and over the distributional range through KwaZulu-Natal. For all statistical analyses, we used R (v. 2.7.0; R Development Core Team; URL http://www.R-project.org).
(e) The chemical signatures of Ficus natalensis and Ficus burkei
Volatile organic compounds (compounds with usually fewer than 20 carbons) emitted by receptive figs from 15 F. natalensis trees from six different sites and five F. burkei trees from three sites (see electronic supplementary material ESM2, table S4) were collected using the headspace technique and subsequently analysed by GC–MS . Details of methods are described in the electronic supplementary material ESM3, method S2. Preliminary tests confirmed that both pollinators were attracted by the blend of volatile organic compounds emitted by receptive figs of F. natalensis collected using the headspace technique, and eluted using dichloromethane (A. Cornille 2008, unpublished data).
Comparison of the composition of volatile compounds among samples was performed using multivariate analyses  using R with the Vegan package. First, data (using relative amounts of all the compounds with respect to total peak areas) were square root-transformed and standardized, then a data matrix of pairwise Bray–Curtis dissimilarity indexes  between samples was built. In order to scale intraspecific and interspecific variations among our samples, four other Ficus species from KwaZulu-Natal, analysed with the same methods and equipment, were included in the comparison: Ficus sur and Ficus sycomorus (both belonging to subgenus Sycomorus, section Sycomorus, subsection Sycomorus; data published in Proffit & Johnson ), three individuals of Ficus glumosa and three individuals of F. stuhlmannii (both belonging to subgenus Urostigma, section Galoglychia, subsection Platyphyllae, this study). We further compared our data with published Bray–Curtis index data for comparisons between populations and between species for two Asiatic Ficus species: Ficus hispida (subgenus Sycomorus, section Sycocarpus) and Ficus racemosa (subgenus Sycomorus, section Sycomorus) . Then, non-metric multi-dimensional scaling (NMDS) was used to visualize similarities among all the samples from KwaZulu-Natal by finding the best two-dimensional representation of the distance matrix.
Within F. natalensis, permutational multivariate analysis of variance (permanova) of the Bray–Curtis index (100 000 permutations) was used to test variation in scent composition among populations and among trees visited mainly by one or the other pollinator species (grouped as over 70% of figs per tree visited by E. stuckenbergi and less than 40% by E. socotrensis, or vice versa, or both wasp species present in more than 40% of the figs per tree). The same analysis was used to test for significant difference in the volatile profile of receptive figs, between species occurring in KwaZulu-Natal, and within and between the two Asian species. p-Values were adjusted for multiple comparisons using the false discovery rate .
The proportions of the seven main compounds present in the floral blends of the two Ficus species were compared using Student's t-tests. Data were log(x + 10)-transformed to fit the assumption of normality (for all Shapiro tests, p > 0.05). For variables that failed the assumption of normality, a non-parametric Mann–Whitney rank sum test was used. p-Values were adjusted for multiple comparisons using the false discovery rate .
(a) Five molecular clades supported by morphology in the Elisabethiella stuckenbergi–Elisabethiella socotrensis species complex
In the Bayesian COI analysis, wasps identified as E. socotrensis and E. stuckenbergi form a species complex presenting at least five lineages (figure 2). This is in agreement with a posteriori morphological investigation: no simple criterion allowed us to assign all the samples into only two classes of individuals (see electronic supplementary material ESM1, table S3). Nevertheless, within a geographical region, all individuals identified under a same name belonged to a single lineage. For instance, E. stuckenbergi from Zambia, whatever their host, all belonged to clade E even though they were collected 600 km apart. Similarly, E. stuckenbergi from South Africa, whatever their host, all belonged to clade B even though they were collected up to 590 km apart, and E. socotrensis from South Africa, whatever the host collected on, all belonged to clade C even though they had been collected up to 800 km apart. Furthermore, where two molecular lineages coexisted locally, they had been distinguished a priori on morphological grounds and samples were correctly attributed to distinct lineages, before molecular analysis. A posteriori morphological measurements, although performed on a limited number of specimens, provide a morphological basis to facilitate identification within regions (see electronic supplementary material ESM1, table S3). On the other hand, wasps collected from the same host in different parts of its range belonged to different clades.
Maximum intraclade sequence divergences were globally one order of magnitude smaller than the minimum between-clade sequence divergences (see electronic supplementary material ESM1, table S2). Exceptions are the lower divergence between clades B and C, which co-occur in South Africa and are grouped into a monophyletic clade (94% bootstrap, 1.00 posterior probability) and between clades D and E.
(b) Distribution and co-occurrence of pollinating wasp species in KwaZulu-Natal
Figs visited only by E. stuckenbergi or E. socotrensis produced abundant seeds (qualitative records). Figs visited only by Philocaenus barbarus produced no seeds. Both Elisabethiella species are pollinators and Philocaenus is a parasite of the mutualism.
The two wasp species co-occurred within crops in all locations, across a range of 290 km (figure 1). The proportion of figs per tree visited by a single species did not differ among the three sampled sites tested (binomial glm, respectively: E. stuckenbergi—χ2 = 132.5, d.f. = 2, p = 0.69; E. socotrensis—χ2 = 135.2, d.f. = 2, p = 0.29; figure 1 and see electronic supplementary material ESM2, table S5).
(c) Competition between Elisabethiella socotrensis and Elisabethiella stuckenbergi for figs
The two wasp species co-occurred within figs less frequently than expected by chance, within the whole dataset (χ2 = 203.3, d.f. = 1, p = 2 × 10−16), and within the 17 crops with over three figs sampled (see electronic supplementary material ESM2, table S5). Indeed, for receptivity period figs, only 15 per cent of the collected figs contained the two species of wasps while only 2 per cent of the figs were unvisited.
(d) Chemical signature of Ficus natalensis
Receptive figs from the 15 F. natalensis trees emitted 38 volatile organic compounds, which all have fewer than 20 carbons (see electronic supplementary material ESM3, table S6). The mean number of compounds emitted by individual trees was as follows: St Lucia 10.3 ± 2.5, Durban 13.8 ± 6.1, Mtunzini 13.6 ± 6.3, Tugela Mouth 11, one individual, Kosi Bay 16, one individual and Vernon Crooks 11, one individual, and did not differ among sites (glm (Poisson): χ2 = 20.5, d.f. = 5, p = 0.91). We found no variation in the odour emitted by receptive figs of F. natalensis among sites (permanova: F5,10 = 1.05, p = 0.38) and no variation depending on which wasp species was more abundant among the foundresses (permanova: F2,12 = 1.38, p = 0.11).
(e) Comparison of the chemical signatures among species
Receptive figs from the five sampled F. burkei trees emitted 18 volatile compounds (see electronic supplementary material ESM3, table S6). The proportions of the main compounds emitted by receptive figs did not differ between F. natalensis and F. burkei (see electronic supplementary material ESM3, table S7). The number of compounds emitted per individual fig did not differ between the two species (glm (Poisson): χ2 = 26.3, d.f. = 1, p = 0.07): F. natalensis (12.9 ± 4.8), F. burkei (9 ± 2.2).
The NMDS plot suggests that odours of F. natalensis and F. burkei are more similar to each other than the odour of the four other Ficus species included in the analysis (figure 3). The odours produced by the two other species pairs of similar level of taxonomic proximity (subsection) did not overlap. This result was confirmed by the permanova: the composition of volatile organic compounds produced by receptive figs is significantly different among Ficus species (F5,31 = 4.41; padjust = 0.00003) occurring in KwaZulu-Natal. The odour of F. natalensis was significantly different from those of all other African species except F. burkei (see electronic supplementary material ESM3, table S8). The two other within-Ficus subsection comparisons gave significant differences (subsection Sycomorus, between Ficus sur and Ficus sycomorus; subsection Platyphyllae, between Ficus glumosa and Ficus sthulmannii). For the two Asian species, the odours of receptive figs were significantly different between species but also among populations. Bray–Curtis indexes among populations within Asian species were higher than Bray–Curtis indexes between F. natalensis and F. burkei (see electronic supplementary material ESM3, table S8).
Ficus are deeply rooted as a strict highly specialized mutualism albeit few recent molecular studies pinpointed the existence of multiple wasp species per host and their consequences for the evolutionary dynamics in this system [4,5,7,11]. Here, combining original polyvalent methodologies (biogeographic, ecological, morphological, molecular and chemical), we have provided the first consistent demonstration of a set of Ficus sharing their main pollinators.
The molecular data show that the Elisabethiella socotrensis–E. stuckenbergi species complex encompasses at least five lineages (figure 2). No locally cryptic species was discovered and, within each region, specimens were consistently sorted a priori according to morphology into their genetic group. Geography rather than fig-host determined which clade was present in a location. A similar situation may be occurring in west Africa within the same Ficus species complex, but molecular data are lacking [10,34]. It may be noted for clade B that F. n. natalensis and F. n. graniticola are allopatric, and that they occupy different habitats from F. burkei . Similarly, F. lingua and F. burkei also occupy different habitats . This is also the case for another system of one pollinator–two Ficus species, namely the pollination of Ficus mucuso (forest species) and Ficus sycomorus (savannah species) by Ceratosolen arabicus . So it would seem that Ficus species occurring in the same region but in different habitats may share agaonid wasps that are their sole or a major pollinator. In that perspective, the most surprising situation is that of F. petersii (the sister species of F. burkei), which in South Africa is almost exclusively pollinated by the same wasp species (A. binghami) as F. stuhlmannii: the two Ficus species share the same habitat  and may grow side by side in South Africa. However, further north in the range of F. petersii, E. stuckenbergi has also been recorded as the pollinator of this host fig in Zambia (S. van Noort 2006, personal observation). Pending further molecular and morphological investigation of the species group, we retain the names E. stuckenbergi for lineage C and E. socotrensis for lineage B in South Africa.
In South Africa, F. burkei and F. natalensis are both pollinated by E. stuckenbergi clade C, and F. natalensis is also abundantly pollinated by E. socotrensis clade B. However, in other parts of their range, both Ficus species are pollinated by other clades or even by a different genus. Further, clades B and C are sister clades restricted to South Africa. We may speculate that the observed zonation of pollinator distributions corresponds to climatic limits. A pollinating wasp adapted to the South African climate would have allowed the two host species to colonize South Africa and the wasp would be in the process of evolutionary diversification. A similar situation may occur in Australia, where cryptic species associated with a single host, Ficus rubiginosa, show some geographical separation, though the pattern seems less clear . Range expansion following climatic modifications has been shown to shape species' biology in an ant–plant species-specific insect–plant mutualism . Climatic oscillations have also been shown to be important in driving the evolution of the Ficus–fig wasp mutualism, in particular : they could be important for shaping mutualistic interactions in general.
The results show that F. natalensis is abundantly pollinated by two species of wasps over a range of 290 km. We found no evidence for variation in receptive fig odour that would correlate with more abundant visitation by one of the wasp species. Fig odour was also homogeneous among sites throughout KwaZulu-Natal (see electronic supplementary material ESM3, table S6). The odour of receptive figs of F. burkei was similar to that of F. natalensis figs (figure 3). These odours were much more similar to each other than are the odours produced by other pairs of species of similar taxonomic relatedness. They were even more similar than the reported variation among populations within species . The type of analysis of odours presented here is not based on what compounds wasps respond to. It is rather a comparison emphasizing variation in molecules present in intermediate quantities. Nevertheless, the analysis seems to capture a biological reality: the odours produced by the two host species are similar but tend not to be strictly identical. This fits rather well with the biological result that one wasp is a major pollinator of both Ficus species while the other one is only a marginal pollinator of F. burkei. Two complementary processes may explain receptive fig odour similarity: one or both Ficus species may be selected to mimic the other species and/or phylogenetic inertia is involved. As odour variation among populations but within species for two Asian species was larger than the difference observed here between F. burkei and F. natalensis, we may suggest that phylogenetic inertia is not sufficient to explain the data. To answer the question, we will need to investigate Ficus petersii, a close relative of F. burkei [34,48]. It has the same pollinator as F. stuhlmannii  and receptive fig odour of F. stuhlmannii is different from that of F. burkei (figure 3). Interestingly, a single genetic lineage of the fig-entering wasp Philocaenus barbarus is associated with F. burkei and F. natalensis, with another lineage associated with F. petersii and F. stuhlmannii: odour similarity at receptivity may also result in sharing of non-pollinator fig wasp species (S. van Noort 2008, unpublished data).
Our observations that fewer than the expected number of figs within crops were visited by both species suggest strong interspecific competition for access to receptive figs. This raises a question as to how these two species can coexist within the same fig crop, but generally not in the same fig. The origin of species coexistence is a complicated issue [49–51]. In our case, we surmise that competition is mediated by (i) figs rapidly becoming inaccessible to wasps once they have been entered and (ii) heterogeneity over time of arrival of the two different wasp species. Indeed, in most individual figs visited by more than one wasp, the wasps were of the same species, and domination of individual crops by one or other pollinator suggests that interspecific competition for figs could be linked to spatio-temporal constraints. Thus, stable ecological coexistence of competing wasps could be driven by divergent strategies with respect to trade-offs between flight capacity, fig-entering capacity and brood size. Such trade-offs are documented within species for the pollinator of Ficus hispida in China: large wasps disperse more efficiently but often get stuck in the ostiole when trying to enter figs . Further, on Ficus sycomorus, the pollinator Ceratosolen arabicus and the non-pollinator Ceratosolen galili present multiple biological differences, involving among other characters, wasp size, dispersal capacity, timing of dispersal (diurnal/nocturnal) and resistance to heat . Conversely, Molbo  presented evidence suggesting that coexisting wasps presented similar fecundities: biological similarities between coexisting wasps may facilitate unstable but prolonged local coexistence, especially in contact zones.
Alternatively, we propose that the populations of E. stuckenbergi associated with F. natalensis correspond to a slowly emptying sink population that is regularly provisioned with wasps migrating from F. burkei. Indeed, the shorter head shape of E. stuckenbergi suggests adaptation to F. burkei, which presents smaller figs than F. natalensis. The presence of populations of F. natalensis graniticola (a taxon restricted to granite outcrops in northern South Africa and Zimbabwe and pollinated by both E. socotrensis and E. stuckenbergi) in the middle of the distribution of F. burkei supports the contention that E. socotrensis is a pollinator specialized on F. natalensis in South Africa. On the other hand, regular colonization of both F. burkei and F. natalensis by E. stuckenbergi suggests that it may not be optimally adapted to F. natalensis and could be maintained on that host by repeated incidental colonization.
The plasticity of the interaction between agaonid wasps and Ficus evidenced here could have important implications for Ficus evolution. Indeed, in order to reproduce in a Ficus species, wasps must (i) be attracted by the odours of receptive figs , (ii) be capable of getting into the fig through the ostiole , (iii) present an ovipositor length compatible with the length of the styles through which they oviposit [1,54], and (iv) be capable of initiating the transformation of the ovule into a gall on which they will feed [6,55]. Hence, we may surmise that host use by fig pollinators is constrained by a number of factors that are generally more similar among related hosts. As a result, pollinators using several hosts can be predicted to use related hosts, a feature that should favour fig hybridization, and hence genetic introgression among related fig species . Natural interspecific hybridization commonly occurs in flowering plants, and is thought to play an important role in their evolution . It is as yet unclear how much genetic introgression occurs within Ficus, but reticulate evolution could be at work in some cases .
The situation illustrated here for a group of African Ficus may occur in several other generally highly species-specific pollination systems. For instance, in orchids of genera Ophrys and Pseudorchys, situations of flower odour convergence associated with local use of the same pollinator by two plant species have been demonstrated . Similarly, the receptive flowers of Glochidion obovatum and Glochidion rubrum, two parapatric species pollinated by the same two species of Epicephala moths, produced similar odours, more similar than those produced by co-occurring species of Glochidion pollinated by other species of Epicephala . We may suggest that there is sometimes sufficient overlap in the chemical messages produced by different host species and in the chemicals to which different pollinator species respond for selection to lead to odour convergence among species that use the same pollinator(s) species. Imperfect chemical differentiation may facilitate a dynamic reticulate evolution of host–pollinator associations. As such, it may facilitate ecological and evolutionary plasticity of species that were previously considered to be locked into hyper-specialization.
A.Cor. performed acquisition of data, analysis, interpretation of data, drafted the article and incorporated all comments from revisions. M.P. conceptualized and designed the study, performed acquisition of data and analysis, interpretation of data and gave final approval of the version to be published. F.K. and M.H.K. contributed to analysis, interpretation, writing and gave final approval of the version to be published. J.G.U. and A.Cr. did acquisition and analysis of phylogenetic data. S.v.N. did acquisition of phylogenetic data, conceptualized the problem, performed morphological determinations, writing and gave final approval of the version to be published. K.A.T. did acquisition of phylogenetic data and gave final approval of the version to be published. S.D.J. contributed to the writing and interpretation of data and gave final approval of the version to be published.
This research was supported by the GDR Ecologie chimique, by the GDR1 191, by ANR grant NICE Figs, by National Research Foundation of South Africa (NRF) grant GUN 61497 to S.v.N. and by the South African National Biodiversity Institute. We wish to thank staff from the pollination laboratory in Pietermaritzburg for their help with field and laboratory experiments, Marie Charpentier, Tatiana Giraud, Doyle McKey, Marc André Selosse and Allen Herre for their helpful comments on the manuscript, and Ezemvelo KwaZulu-Natal Wildlife for allowing us to work in protected areas. All the experiments comply with the current laws of South Africa.
- Received September 30, 2011.
- Accepted November 8, 2011.
- This journal is © 2011 The Royal Society