Introduction

The rise of molecular taxonomy has seen an increase in the discovery of cryptic species, often revealed first by analysis of mitochondrial DNA (mtDNA) fragments (for example, Wilcox et al., 1997; Hebert et al., 2004; Fouquet et al., 2007; Murphy et al., 2011). These cryptic genetic lineages may replace each other geographically, but in cases where they coexist locally, they pose a challenge for ecologists. Conventional theory predicts that two or more species should not coexist locally in the same niche and that competitive exclusion should lead to the extinction of all but one species (Gause, 1934; Chesson, 1991; Hubbell, 2001). Explaining the local coexistence of cryptic species is therefore problematic in itself, but also influences our understanding of species interactions (for example, between plants and pollinators, or host and parasites), which were previously interpreted based on morphological data.

While molecular taxonomy may reveal cryptic lineages, the reliability of mtDNA for species delimitation has been criticised for various reasons (Rubinoff et al., 2006), including the persistence of genetic structure after resumption of gene flow due to lack of mtDNA recombination (for example, Lausen et al., 2008; Ruskey and Taylor, 2016), hybrid introgression (for example, Chen et al., 2009; Machado-Schiaffino et al., 2010) and selective sweeps caused by maternally inherited endosymbionts that can drive divergence and homogenisation across species (for example, Graham and Wilson, 2012; Xiao et al., 2012; Jäckel et al., 2013). These issues raise significant concerns when mtDNA is used in isolation, but they are likely to be resolved with further study of nuclear loci.

The arrays of multilocus nuclear markers provided by microsatellite loci provide an effective means of assessing reproductive isolation in coexisting cryptic mtDNA lineages. For example, Lausen et al. (2008) revealed high estimates of gene flow among well-differentiated mtDNA lineages of the little brown bat, Myotis lucifugus. Similarly, Ruskey and Taylor (2016) found that mtDNA lineages of sympatric fish were not supported by microsatellite analyses. Conversely, microsatellite analyses support mtDNA-delimited species boundaries in many taxa. Naughton et al. (2014) found that microsatellite data support mtDNA lineages in Cenolia feather stars, although introgression was observed in a small number of individuals. In addition, Donnelly et al. (2013) demonstrated support for mtDNA species in Lumbricus rubellus earthworms in South Wales. Interestingly though, divergent L. rubellus mtDNA lineages in Poland are not reproductively isolated (Giska et al., 2015). Population genetics can thus help to elucidate the reliability of species boundaries delimited using mtDNA or only a limited number of DNA markers.

Fig trees (Ficus, Moraceae) and their pollinator wasps (Hymenoptera: Agaonidae) share an obligate reproductive mutualism renowned for high levels of partner specificity (Wiebes, 1979). The traditional view was that most fig species had just one pollinator wasp species. However, mtDNA analyses have helped drive the realisation that many fig species may be pollinated by multiple fig wasp species (Cook and Rasplus, 2003). Recently, there have been several reports of cryptic pollinator ‘species’ using the same host fig species both in sympatry and allopatry (Molbo et al., 2003; Haine et al., 2006; Sun et al., 2011; Chen et al., 2012; Darwell et al., 2014). Where morphologically cryptic pollinator species coexist locally, their persistence requires ecological and/or behavioural explanation (Zhang et al., 2004). For example, non-pollinating species may evolve following the rise of cheating within an existing pollinator species (Compton et al., 1991; Jandér and Herre, 2010). If figs often have two or more co-pollinators, then this is of major consequence for coevolutionary studies, both with regard to local ecological dynamics and speciation. Many cases of supposed cryptic fig wasp species, or at least ones that differ only subtly in morphology, have been reported recently based on mtDNA data, and we need to clarify their status to better interpret this key model system.

An extreme deviation from reciprocal partner specificity occurs with the Port Jackson fig (Ficus rubiginosa Desf. ex Vent.), which is pollinated by five ‘species’ of the Pleistodontes imperialis Saunders species complex along the Australian east coast (Haine et al., 2006; Darwell et al., 2014). Following taxonomic revision in 2002, these were considered a single species (Lopez-Vaamonde et al., 2002), but were later found to resolve into distinct lineages (Haine et al., 2006; Darwell et al., 2014). Haine et al. (2006) found support for four species based on mitochondrial cytochrome b (cytb); however, the nuclear 28S rDNA and wingless genes only resolved two lineages. With wider sampling, Darwell et al. (2014), subsequently showed support for five species using nuclear ITS2 and mitochondrial COI and cytb.

For three of the P. imperialis lineages, no morphological differences are known, while the other two differ only slightly, in colour or head length (J.-Y. Rasplus, personal communication). As the host plant geographic range covers a number of biogeographic barriers (notably the Burdekin Gap, St Lawrence Gap, Dawson-McKenzie Gap and the McPherson Range) that are reflected in the mtDNA phylogeography of many animal taxa (for example, Hugall et al., 2002), the interpretation of these lineages remains unclear.

We explored a potential hybrid zone around Townsville in north-eastern Queensland, where three P. imperialis lineages (‘species 2’, ‘species 3’ and ‘species 4’ sensu Haine et al., 2006) coexist in sympatry. We used genetic markers (nuclear DNA sequences, mtDNA sequences and microsatellites) to explore patterns of gene flow between these three lineages and assess their evolutionary status. Specifically, we assessed (1) the opportunity for hybridisation (that is, the frequency of fig fruits containing offspring from different lineages), (2) nuclear microsatellite differentiation between these distinct mtDNA lineages and (3) the prevalence of hybridisation.

Materials and methods

Fig wasp biology

Receptive fig fruits (syconia) attract pollen-bearing female fig wasps through the emission of species-specific volatiles. The female wasp (foundress) enters the receptive syconium through a narrow opening (ostiole), which closes at the end of the receptive phase. Inside, the foundress pollinates the fig ovules, depositing a single egg into some of these, and dies within 24–48 h. Egg deposition initiates gall formation, within which the developing larva feeds on the fig seed. Offspring develop over a period of about 6–8 weeks (Harrison, 2005; Xiao et al., 2008). Mature males exit their galls to locate galls containing females, and chew holes into galls in order to mate with the females within. After mating, the wingless males chew holes through the outer wall of the syconium for the females to exit through, but do not disperse themselves. Females emerge from their galls, collect pollen and exit the syconium to disperse in search of receptive syconia in which to lay eggs (Galil and Eisikowitch, 1968).

Sample collection

Ficus rubiginosa syconia were collected from eight sites around Townsville in north Queensland (19°15′24″ S, 146°49′3″ E) between November 2013 and November 2014. Near-ripe figs (N=611) were collected from naturally occurring trees and placed into individual plastic specimen pots. Emerged wasps were preserved in absolute ethanol prior to molecular analyses. Typically only 1–4 females lay eggs in a single syconium. Because of the high probability that any two wasps from the same syconium are siblings, only a single wasp from each syconium was used for molecular analyses. Samples from different sites and different time points were pooled for analyses.

Assessment of the potential for hybridisation

The potential for hybridisation was quantified as the proportion of syconia that produced both yellow (‘species’ 2) and black (‘species’ 3 or 4) wasps. This method provides an estimate of the minimum frequency of co-occurrence of multiple species within a syconium, and thus represents the minimum frequency of opportunities for heterospecific matings.

DNA extraction and mtDNA analyses

A subset of samples was used for DNA analyses, with sample sizes approximately proportional to the frequency of each lineage (Table 1). DNA was extracted individually from whole female insects (N=139) using a Chelex method (West et al., 1998) to a volume of 100 μl. The cytb gene was amplified and sequenced in all samples using CB1/CB2 primers and the protocol described in (Jermiin and Crozier, 1994). Although COI is the standard barcoding gene in animals, most barcoding efforts for this species complex have utilised the cytb gene (Haine et al., 2006; Darwell et al., 2014) since COI amplifies inconsistently and is compromised as a marker by nuclear pseudogenes in fig wasps. DNA sequences were edited by eye in Sequencher v4.10.1 (Gene Codes, Ann Arbor, MI, USA) and aligned in Clustal X v2.0 (Larkin et al., 2007).

Table 1 mtDNA and microsatellite diversity of three sympatric P. imperialis species
Full size table

A cytb haplotype phylogeny was constructed in MrBayes v3.2 (Ronquist et al., 2012) to assign individuals to lineages according to Darwell et al. (2014). Reference sequences for each of the three lineages (Haine et al., 2006; Darwell et al., 2014) were included in the phylogeny, along with outgroup sequences for Pleistodontes nigriventris (Girault; Lopez-Vaamonde et al., 2001) and another fig wasp, Ceratosolen galili Wiebes (Kerdelhue et al., 1999). jModelTest v2 (Guindon and Gascuel, 2003; Darriba et al., 2012) was used to determine that the HKY+G (Hasegawa et al., 1985) model was the most appropriate for our data. Two independent runs of four Metropolis coupled Monte Carlo Markov chains for 10 million generations were performed with a relative burn-in of 25%, and sampling every 1000 generations. The two runs were assumed to have converged if the mean s.d. of the split frequencies fell below 0.01.

Microsatellite analyses

Each individual was genotyped at nine microsatellite loci (Sutton et al., 2015). These were amplified in multiplex reactions, as detailed in Sutton et al. (2016). All genotyping was performed on a 3500 Genetic Analyzer (Applied Biosystems, Waltham, MA, USA), and electropherograms were visualised and scored using GeneMapper v4.1 (Applied Biosystems). One locus (Pli10) did not amplify in P. imperialis sp. 3 samples, and was therefore excluded from subsequent analyses. The number of alleles and number of private alleles with rarefaction were calculated in ADZE (Szpiech et al., 2008) and deviation from Hardy-Weinberg equilibrium was tested in GENEPOP v4.2 (Raymond and Rousset, 1995; Rousset, 2008). Pairwise tests of genotypic differentiation (Goudet et al., 1996) were performed in GENEPOP and pairwise population FST was calculated in ARLEQUIN.

The Bayesian clustering algorithm implemented in TESS (Chen et al., 2007) was utilised to assess potential gene flow between the cytb-defined lineages. The CAR admixture model (Durand et al., 2009) was implemented and 100 iterations were conducted, consisting of 60 000 sweeps of MCMC with a burn-in period of 10 000 sweeps. The maximum number of clusters (Kmax) was restricted to three as our aim was to assess gene flow between the three cytb-defined lineages. The estimated admixture coefficients were averaged using the Greedy algorithm in CLUMPP v1.1.2 (Jakobsson and Rosenberg, 2007) and the data were subsequently visualised in DISTRUCT v1.1 (Rosenberg, 2004). For individuals whose highest cluster assignment probability from the TESS analysis was <96%, lineage assignment was confirmed by also sequencing the nuclear ITS2 region (Supplementary Material).

Results

At least 13.38% of F. rubiginosa syconia (N=611) contained wasps from multiple lineages, as evidenced by the presence of yellow (species 2) and black (species 3 or 4) P. imperialis wasps. One hundred and thirty-nine female P. imperialis wasps were sequenced for a 372 bp cytb fragment, and each individual was unambiguously assigned to one of the three known P. imperialis lineages found in Townsville (Supplementary Figure S1). Fragments of similar length (396 bp) were previously used by Darwell et al. (2014) to delimit these lineages. We found no evidence for nuclear pseudogenes in comparisons of amino-acid sequences of our samples with those of Darwell et al. (2014). We identified 52 unique haplotypes across three P. imperialis lineages and nucleotide diversity within lineages ranged from 0.007 to 0.015 (Table 1). Intraspecific cytb genetic divergences ranged from 0.0 to 6.5% (mean=0.07%, 1.6 and 1.0% for species 2, 3 and 4, respectively) and interspecific divergence ranged from 8.9 to 16.1% (mean=12.6%).

Microsatellite diversity (NA and HE) was greatest in P. imperialis sp. 2 and lowest in P. imperialis sp. 3, and over 50% of alleles were not shared between lineages (Table 1; Supplementary Figure S2). All lineages were significantly differentiated with pairwise FST values ranging from 0.28 to 0.54 (Table 2). Bayesian clustering analyses revealed three distinct microsatellite genetic clusters corresponding to cytb lineage assignments (Figure 1). Seven out of 139 individuals had maximum cluster assignment probabilities lower than 96%. Subsequent sequencing of the ITS2 region for these individuals (Supplementary Figure S3) supported their initial cytb lineage assignment.

Table 2 Genetic differentiation among P. imperialis (cytb) species
Full size table
Figure 1
figure 1

Genetic cluster assignments of 139 P. imperialis individuals based on nuclear microsatellite genotypes. Each vertical column represents an individual, and the colours indicate its probability of assignment to each of the three genetic clusters. Samples are grouped by mtDNA assignment. Circles denote individuals whose highest cluster assignment probability was <0.96.

Full size image

Discussion

We used a population genetic approach to assess gene flow among three sympatric cryptic fig wasp species in north-eastern Australia. Our results revealed that >13% of syconia contained wasps from at least two mtDNA lineages. The co-occurrence of multiple species in the same syconium indicates that there is potential for hybridisation; however, we did not detect any hybrids or evidence of introgression. This study supports the reproductive isolation in local sympatry and species status of lineages delimited by Darwell et al. (2014), who used a few specimens from each of many sites over a wide geographic range.

Bayesian clustering analyses of female wasps indicated a clear concordance between mtDNA and multilocus microsatellite genotypes, despite seven individuals exhibiting admixed cluster assignment probabilities. It is possible that these individuals are F1 hybrids, which should suggest the potential for gene flow among mtDNA lineages. However, we found no evidence of mtDNA introgression, which could have one of three explanations. Firstly, the frequency of multiple heterospecific foundresses entering a single syconium may be too low to facilitate repeated backcrossing with the paternal nuclear background. Alternatively, F1 hybrids may be sterile or have reduced fitness. Molbo et al. (2004) found diploid F1 hybrid males in another genus of fig wasps from Panama, suggesting a breakdown in the haplo-diploid sex determination system. Finally, these fig wasps may simply avoid mating with non-conspecifics, even though these are closely related species.

There is a growing number of reports of cryptic ‘species’ in the fig—fig wasp mutualism, existing both in sympatry and allopatry (Molbo et al., 2003; Haine et al., 2006; Moe and Weiblen, 2010; Lin et al., 2011; Sun et al., 2011). Moe and Weiblen (2010) identified cryptic Ceratosolen fig wasp ‘species’ in association with several widespread Ficus species, suggesting an extensive role for vicariance in the generation of cryptic biodiversity. This notion is supported by population genetic analyses of fig wasps in Asia and Australia (Kobmoo et al., 2010; Sutton et al., 2016). Despite the local coexistence of cryptic fig wasp ‘species’ in at least four genera (Molbo et al., 2003, 2004; Haine et al., 2006; Lin et al., 2011; Sun et al., 2011; Darwell et al., 2014), true tests of hybridisation between such mtDNA lineages have only been undertaken in one of these. Molbo et al. (2003) identified two lineages of the fig wasp Pegoscapus hoffmeyeri that exhibited ~4% mtDNA divergence (they used COI, while here we use cytb). Subsequently, Molbo et al. (2004) used diagnostic microsatellite loci to identify hybrids between these lineages in only 1% of syconia sampled, suggesting that gene flow between them is rare. Our results involve wasps from a distantly related genus to those studied by Molbo et al. (2004), and the consistency between studies supports the idea that mtDNA-delimited fig wasp lineages are generally reproductively isolated species.

Our data highlight the need to address the question of how cryptic species that appear to be ecological equivalents can coexist on the same resources. Zhang et al. (2004) proposed a model to address this fundamental ecological issue by building on the idea that stable coexistence is possible if species discriminatively direct appropriate behaviours towards conspecifics and heterospecifics (Chesson, 1991). We have shown that sympatric P. imperialis lineages do not exchange genes and thus represent ‘true’ reproductively isolated species. Future work should aim to investigate potential divergence in the intricate ecology and morphology of these species and to test the model of cryptic fig wasp coexistence proposed by Zhang et al. (2004).

Data archiving

DNA sequences were deposited in GenBank with accession numbers KU316794–KU316933 (cytb) and KU353510–KU353516 (ITS2). Microsatellite genotype data are available from the Dryad Digital Repository: http://dx.doi.org/10.5061/dryad.54527.