Introduction

The parasitic wasp Nasonia vitripennis is a model organism for studying developmental biology, sex-ratio studies, evolutionary genetics and host–microbe interaction (Werren, 1980; Breeuwer and Werren, 1990, 1995; Bordenstein et al., 2001; Beukeboom and Desplan, 2003; Pultz and Leaf, 2003; Shuker et al., 2003; Werren and Loehlin, 2009). Its genome has recently been sequenced, providing a wealth of information for population genetic studies (Werren et al., 2010). It has a cosmopolitan distribution and has been found almost everywhere it has been looked for (Whiting, 1967; see also http://www.nhm.ac.uk/research-curation/projects/chalcidoids/). Three additional species have been described in the genus: N. giraulti found in eastern North America (NA); N. longicornis found in western North America (Darling and Werren, 1990); and the recently identified N. oneida from New York state (Raychoudhury et al., 2010). Thus, there are four species found in North America, N. vitripennis, N. giraulti, N. longicornis and N. oneida, and N. vitripennis occurs microsympatrically with the other three. N. giraulti and N. longicornis are largely allopatric, whereas N. giraulti and N. oneida have overlapping distributions. In Europe (EU), however, only one species is currently found, N. vitripennis. An unresolved question is whether N. vitripennis originated in North America along with other known representatives of the genus, or from Eurasia. The divergence of the Nasonia genus has been placed in the Pleistocene era around 106 years ago (Campbell et al., 1993), with N. vitripennis diverging first and the other three around 0.4 × 106 years ago (Raychoudhury et al., 2010). All the species harbour the maternally inherited, reproductive endosymbiont Wolbachia, which causes sperm–egg incompatibilities and has a large role in reproductive incompatibility between several of the species, at least under laboratory conditions (Breeuwer and Werren, 1990; Bordenstein et al., 2001). However, elimination of Wolbachia by antibiotic treatment leads to the production of viable hybrid offspring (Breeuwer and Werren, 1990). Breeuwer and Werren (1995), Bordenstein et al. (2001) and Raychoudhury et al. (2010) have shown, using antibiotically cured strains from each species, that N. vitripennis has diverged from the other three species genetically to the point where some F2-hybrid breakdown occurs, whereas the three North American species show little or no hybrid breakdown (Bordenstein et al., 2001; Raychoudhury et al., 2010).

A comparison of the genetic diversity in the three Nasonia species in nuclear, mitochondrial and Wolbachia sequences by van Opijnen et al. (2005), revealed a remarkably reduced mitochondrial variation in N. vitripennis compared with the other species, whereas the nuclear and Wolbachia genomes did not show such a pattern. Oliveira et al. (2008) confirmed this pattern in the mitochondria with more strains from both Europe and North America. Moreover, the N. vitripennis mitochondria were found to be divided into two distinct clades, one on each side of the Atlantic, with the EU strains showing distinctly more variation than their NA counterparts. In addition, some of the NA haplotypes were nested within the EU mitochondrial haplotype variation. To explain this phenomenon of reduced mitochondrial variation in North America, van Opijnen et al. (2005) hypothesized a relatively recent selective sweep of a well-adapted mitochondrial haplotype. The low sequence variation in NA populations is especially noteworthy because Oliveira et al. (2008) have shown the Nasonia species have an exceptionally high mitochondrial mutation rate relative to the nuclear genes. As mitochondria and Wolbachia are co-inherited, a Wolbachia-induced sweep can produce very high frequency of the associated mitochondrial haplotypes (Turelli and Hoffmann, 1991; Werren, 1997; Ballard and Whitlock, 2004; Hurst and Jiggins, 2005) and such a sweep could also have produced a reduction in mitochondrial diversity in NA populations. However, an alternative explanation is that NA N. vitripennis have undergone a genetic bottleneck, possibly during a founding event from European populations.

In this study, we have collected and analysed additional data to resolve among these possibilities. By comparing genetic diversity indices among a larger sample of NA (N=52) and EU (N=27) Nasonia vitripennis in mitochondria, microsatellites, single-copy nuclear gene sequences and Wolbachia, we tested the two alternative hypotheses of mitochondrial–Wolbachia sweep versus a bottleneck event in North America. One of the key predictions of a mitochondrial sweep in NA populations would be the presence of significantly less mitochondrial variation compared with EU populations. Another prediction would be relatively similar diversities in microsatellites and nuclear sequences in the two populations. A mitochondrial sweep would asymmetrically affect variation in the cytoplasmic genome compared with the nuclear genome. However, in the event of a bottleneck in NA populations both nuclear and mitochondrial genomes would be expected to show a reduction in variation compared with the EU population. We provide analytical evidence for and against these two competing hypotheses and find that the data clearly support a cytoplasmic sweep in North America.

Materials and methods

Nasonia sampling and DNA extraction

We acquired 52 specimens from North America and 27 from Europe (Figure 1) from bird nests and artificial baits (see supplementary online Table 1 for a complete sample list). The NA specimens were obtained from previous samplings by members of the Werren laboratory (University of Rochester) and from field collections by the senior author in 2005. The EU specimens were obtained from wild derived stocks in the laboratory in Groningen, and from field collections in summer of 2006. All field materials were stored at −80°C or in 95% ethanol. DNA isolation followed a high-salt isolation protocol (Maniatis et al., 1982). As some of the laboratory lines had been in culture for a long time and may have lost some genetic variation, we used only one individual per line. This precluded any population genetic estimates that are based on the measured heterozygosity in the sampled population (for example, FIS). For the field samples, we again used only one individual per sampling point (mostly a bird nest box) to avoid sampling of highly related individuals. For the sequencing of nuclear markers, 18 different strains of N. vitripennis were chosen. These comprised of 10 strains from North America and 8 from Europe (summarized in Supplementary Table 2).

Figure 1
figure 1

Location of the sampling sites in (a) North America and (b) Europe.

Full size image

Mitochondrial DNA analysis

We designed primers for the N. vitripennis cytochrome oxidase I (COI) gene sequence obtained from GENBANK (LOC100113910). The combination of the primers: NL COI Fwd (5′-GTTATACCTKTWATAATWGGAGGATTTGG-3′) and NV COI Rev (5′-CTTTGAAAACCACGTTACCC-3′) amplified a ∼400-bp fragment in Nasonia using a standard PCR protocol at 52 °C annealing temperature.

The amplification of the fragment was checked on agarose gels (1%) and the PCR product was purified with the Nucleospin II kit (Machery Nagel, Düren, Germany). The sequencing reaction was performed on both strands using the same primers and the Big Dye Termination sequencing kit (Applied Biosystems, Foster City, CA, USA). Purified sequencing products were analysed using an ABI 377 automatic sequencer (Applied Biosystems). For the phylogenetic analysis, we used the ALIGNX program included in the VECTOR NTI software package (Invitrogen, Carlsbad, CA, USA) to align a 399-bp fragment without gaps. The software DNASP4 (Rozas et al., 2003) was used to calculate haplotype diversity and nucleotide diversity indices. To find the ancestral haplotype, we used a median joining haplotype network (Bandelt et al., 1999) constructed with NETWORK 4.5 (available at: http://www.fluxus-technology.com/sharenet.htm). The PHYLIP software package (available at: http://evolution.gs.washington.edu/phylip.html) was used to construct a neighbour-joining tree rooted with an out group formed by two Trichogramma species obtained from GENBANK (T. ostriniae: DQ177914, and T. achaeae: DQ177918). Two N. longicornis (GENBANK accession # EU935415 and EU935416) and two N. giraulti (GENBANK accession # EU935417 and EU935418) sequences obtained from this study were also added for comparison. All N. vitripennis COI sequences are deposited in GENBANK (accession # EU935326–EU935414; see Supplementary Table 1 for sample names and corresponding accession numbers).

Simulations were performed to estimate the time since the last mitochondrial sweep in NA populations using a Bayesian summary likelihood method (Garrigan et al., in press). The method is based on forward-time simulations of a Wright–Fisher model, starting from monomorphism. It estimates the time since the last complete mitochondrial sweep (Ts) for a population that has not yet achieved an equilibrium distribution of mutations. The approach uses aspects of the site frequency spectrum, particularly the proportion of single polymorphisms, to estimate the relaxation time (in generations) from monomorphism to drift–mutation equilibrium. On average, this relaxation time was found to be 4N generations, and time to equilibrium is dependent upon the population mutation rate, θ=Nμ, where N is the population effective number of chromosomes and μ is the per locus mutation rate per generation. The method accounts for uncertainty in the population size and mutation rate by randomly sampling values from a gamma prior distribution, and it also assumes a constant effective population size. The Wright–Fisher model simulation stops when both the number of segregating sites and number of singleton polymorphisms in the simulation match those observed from the data, and the generation is recorded and incorporated into the posterior probability distribution of Ts. The posterior distribution for the time since monomorphism can then be estimated from 1000 iterations of the model.

Two different approaches were used to estimate μ, the CO1 gene mutation rate. First, we took the Drosphila melanogaster empirically derived per base mutation rate for mitochondria of 6.2 × 10−8 (Haag-Liautard et al., 2008) and extrapolated it to Nasonia. However, Nasonia is known to have a much higher mitochondrial mutation rate. Oliveira et al. (2008) used synonymous divergence rates between Nasonia species in mitochondria and nuclear genes to estimate the mitochondrial mutation rate and observed it to be 35-fold higher than the nucleus. Extrapolating to predictions of the time of D. simulans–D. melanogaster split and N. giraulti–N. longicornis split, indicated that the mitochondrial mutation rate is 13 × greater in Nasonia than in Drosophila. As an alternative approach, the ratio of mitochondrial/nuclear mutation rates in D. melanogaster is 10 × (Haag-Liautard et al., 2008), compared with 35 × for Nasonia. Assuming that nuclear mutation rates are similar in the Drosophila and Nasonia lineages, this gives a 3.5 × higher mutation rate for Nasonia mitochondria. Adjusting the number of synonymous (neutral) sites in CO1 (92) gives an estimated μ of 7.4 × 10−5 and 2.0 × 10−5, respectively.

The EU populations have an empirically estimated θ=8 and much higher mitochondrial variation than the NA populations. Under the assumption that EU populations are at equilibrium (and have similar population sizes), we estimated the effective number of chromosomes (N) corresponding to the mutation rates to be 107 886 and 400 721, respectively. These parameters were then used in the simulation to estimate the time since monomorphism in NA populations, along with the empirically derived number of segregating sites (8), singleton mutations (4) and sample size (57).

Microsatellite analysis

A total of nine polymorphic microsatellites were examined (di- and trinucleotide repeats with variable repeat length; summarized in Supplementary online Table 3), spread over all five chromosomes to estimate the genetic diversity of the nuclear genome of the EU and the NA populations. The microsatellites Nv 22 and Nv 23 have originally been developed by Pietsch et al. (2004), but the primers were redesigned in our laboratory. All other loci were developed in the Groningen laboratory (Beukeboom and Desplan, 2003). As the samples were not distributed evenly over the areas, we grouped samples of neighbouring origin into arbitrary subpopulations to be able to calculate reliable genetic distance estimates. We grouped the samples from the following locations into corresponding subpopulations: Germany=GER; the Netherlands=NL; Switzerland, Italy and France=CH-IT-FR; Canada=CN; Idaho=ID; New York=NY; Utah=UT; Montana, Oregon, Wyoming, Nevada and South Dakota=MT-OR-WY-NV-SD; Pennsylvania and Virginia=PE-VI; Indiana, Michigan, Ohio=IN-MI-OH. The genetic diversity indices were calculated using the F-STAT software package (Goudet, 2001) or manually. Nei's standard genetic distance DS (Nei, 1987) and Goldstein's δμ2 value (Goldstein and Pollock, 1994) were used to construct neighbour-joining trees between subpopulations within Europe and North America, using the POPULATIONS 1.2.30 software (available at: http://bioinformatics.org/~tryphon/populations/) and 1000 bootstraps over loci.

Nuclear sequence analysis

A total of 11 different nuclear markers were sequenced across 17 different N. vitripennis strains. One EU isolate, AsymCX, is the source strain for the genome sequence (Werren et al., 2010) and relevant sequences were downloaded from NCBI. Primers for five of these markers were designed previously by Raychoudhury et al. (2009) and the rest were designed for this study. The primers were chosen specifically to provide coverage across the genome and markers across all five chromosomes have been incorporated in this study (summarized in Supplementary Table 4). The condition for all sets of primers were 94 °C for 2 min; 38 cycles at 94 °C for 45 s, 55 °C for 60 s, 72 °C for 1 min 30 s; 72 °C for 10 min; then held at 4 °C. Regular Taq (Invitrogen) polymerase was used for PCR reactions and the amplified products were treated with shrimp alkaline phosphatase and exonuclease I (Amersham, CA, USA) to clean the reactions before sequencing. Sequencing was performed at the Functional Genomics Centre, University of Rochester, using BigDye v2.0 or v3.0 terminator sequencing kit (Applied Biosystem) and an ABI 3700 or 3730xl automated sequencer (Applied Biosystem). The chromatograms were manually inspected and cleaned with Sequencher 4.7 software (Gene Codes, MI, USA) and the sequences were aligned using Bioedit vs 7.0.1 (Hall, 1999). The sequences have been deposited in gene bank with accession numbers GQ896384–GQ896485.

Wolbachia analysis

To identify the Wolbachia infection, we used the multi locus sequence typing (MLST) system developed specifically for this bacteria (Baldo et al., 2006b). Essentially, information from five genes (gatB, coxA, hcpA, ftsZ and fbpA) is used to characterize the Wolbachia infections by comparison with other sequences on the PubMLST website (http://pubmlst.org/wolbachia/), which archives the sequences. The combination of the five haplotypes forms a unique allelic profile for each Wolbachia infection. In addition, we also amplified a part of the wsp gene. Primers have been developed for both super group A (wNvitA)- and B (wNvitB)-specific Wolbachia for the five MLST genes and the sequences and conditions can be found on the website. As mitochondria and Wolbachia are maternally co-inherited, we screened 21 different field isolates, selected to cover the range of variation in mitochondrial DNA haplotypes, for the presence and characterization of Wolbachia with primers from both super groups for the MLST genes as well as wsp. The primers and their conditions for the A and B super group wsp gene were taken from Zhou et al. (1998).

Results

Mitochondrial DNA sequence data

We found a sevenfold higher mitochondrial diversity in EU compared with NA samples (π for EU: 2.36% and for NA: 0.31%; Supplementary Table 5). The level of variation in the NA N. vitripennis samples is in the same range as found by van Opijnen et al. (2005). It is striking that the diversity in the EU N. vitripennis samples found in this study is in the same range as that of N. longicornis and N. giraulti found by van Opijnen et al. (2005) in NA. Tajima's D (Tajima, 1989) is −0.88 but not significantly different from 0 for the NA samples and 0.53 (not significant) for the EU samples (Supplementary Table 5). The NA samples differ in their nucleotide sequence (using Nei's DXY) from the EU samples by, on average, 2.3% (compared with 0.3 and 2.4% within NA and EU, respectively), whereas N. vitripennis samples differ from the two sister species by, on average, 14.2% (N. vitripennis−N. longicornis=13.6%, N. vitripennis−N. giraulti=14.8%; Supplementary Table 6). However, there was considerably less COI variation in NA populations relative to EU populations, consistent with either the NA founder hypothesis or the NA mitochondrial–Wolbachia sweep hypothesis. The low variation in NA population is particularly noteworthy given the 35–40-fold higher synonymous substitution rate in Nasonia mitochondrial genes relative to nuclear genes (Oliveira et al., 2008; Raychoudhury et al., 2009).

In total, there were 24 unique mitochondrial haplotypes containing 34 synonymous and three non-synonymous mutations in the sequenced isolates (Table 1). The non-synonymous mutations lead to three haplotypes that were represented by one individual each. The only haplotype that was shared between EU and NA populations was the most common haplotype in North America, but it was not the most frequent in Europe (haplotype # 2; Figure 2 and Table 1). A nested clade analysis (Templeton et al., 1992) did not yield any significant geographic clustering of haplotypes within continents. The phylogenetic analysis (Supplementary online Figure 1) showed that all NA samples cluster together with a few interspersed EU samples. The whole NA group is nested within the EU group, and structure among the EU samples is much more pronounced compared with the NA samples.

Table 1 Mitochondrial haplotype frequencies in NA, EU and pooled (total) samples
Full size table
Figure 2
figure 2

Haplotype network based on the 399-bp mitochondrial COI sequence. Numbers in circles represent particular haplotypes, numbers along lines are the nucleotide positions in the sequence that changed. The sizes of the circles reflect the haplotype frequency. Nodes without a label represent intermediate haplotypes that were not sampled. Dashed lines indicate alternative mutation routes. The grey haplotypes have been found in North America, the white ones in Europe. Haplotye # 2 has been found on both continents (see Table 1 and text).

Full size image

The estimated divergence time of N. longicornis–N. giraulti split was used to calibrate divergence times of the two populations of N. vitripennis. Two estimates of the N. longicornis–N. giraulti split are available and are 0.4 × 106 years ago as estimated by Campbell et al. (1993) and 0.51 × 106 years ago as estimated by Raychoudhury et al. (2009). The total synonymous divergence (ks) between these two species has been estimated to be 38.9% (Oliveira et al., 2008), whereas the ks for the NA and EU populations are 1.32 and 10.55%, respectively. Thus, extrapolating this rate to N. vitripennis gives an age of the NA mitochondria of around 0.027–0.035 × 106 years and the EU mitochondria to be 0.217–0.277 × 106 years.

A Bayesian analysis method based on mutational spectral patterns (Garrigan et al., in press) was used to estimate the time since the mitochondrial sweep in NA populations (see Materials and Methods: Mitochondrial DNA analysis). Using EU populations as a reference and assuming them to be in mutational equilibrium, NA populations were estimated to have undergone a mitochondrial sweep 1945 generations ago (95% confidence limits of 1219–2671) and 3193 (3127–3259) generations ago, based on two estimates of the mitochondrial mutation rate in Nasonia. Given the potential error in estimating the parameters in these simulations, these times should be taken with caution. Nevertheless, the estimates support the contention that NA populations have undergone a selective sweep relatively recently and after the NA–EU split, thus accounting for their reduced variation relative to both the EU population and nuclear variation (see below).

Microsatellite data

Nei's gene diversity Ĥ (Nei, 1987) and allelic richness R are equal in the NA and EU samples, indicating similar levels of nucleotide variation in the two areas. This pattern is consistent over loci with various repeat lengths and motifs (Table 2). This is indicative of an evolutionary force affecting mitochondrial and nuclear genes differentially, as there is a sevenfold difference in the mitochondrial variation between NA and EU populations but not in the nuclear markers. There is low, but highly significant (P<0.001) differentiation between the two continents (FST=0.037±0.013 s.e.). As this measure is dependent on the variability of the marker used, we also calculated Hedrick's G′ST=0.23 (Hedrick, 2005), to compare the differentiation to other studies. This corrected differentiation index shows the same level of differentiation between the two continents as did a previous study between two populations within Europe (0.23, Grillenberger et al., 2008). Both neighbour-joining trees, based on the genetic distances (DS and δμ2) between sub-populations within Europe and North America, indicate no clear split between the two continents. We only present the tree based on Nei's DS, as the qualitative results for the tree based on δμ2 are identical (Supplementary online Figure 2). The mean genetic distance estimated with both parameters indicates no clear difference between the continents. The variation within EU population is slightly higher than that within NA population (DS±s.e.: EU: 0.36±0.10; NA: 0.35±0.02; between EU and NA: 0.43±0.04; δμ2±s.e.: EU: 22.05±8.06; NA: 13.92±2.30; between EU and NA: 17.18±2.54).

Table 2 Summary statistics of microsatellite data
Full size table

Nuclear sequence variation

As no distinct differences in microsatellites could be detected between the two populations, a survey of nuclear sequence diversity was initiated. The expectation was that the greater resolving power of these sequences would shed light on the level of nuclear diversity of the strains from these two areas. Furthermore, because of the lower mutation rate of single-copy nuclear DNA relative to microsatellites, these single-copy sequences are likely to clarify whether NA populations underwent a nuclear genetic bottleneck relative to EU populations.

The 11 different nuclear markers together yielded a data set of 7079 base pairs for each of the 18 different strains. Two measures of nucleotide diversity were enumerated: diversity in silent sites (synonymous and non-coding) and in non-synonymous sites (summarized in Table 3). Unlike the mitochondrial data, there was no greater variation in the EU strains and in fact the levels of silent site diversity were the same in both (0.16%), and thus, not significantly different (Mann–Whitney U-test: U=61, DF=11, P=0.5). Diversity in non-synonymous sites for NA and EU strains were 0.05 and 0.1%, respectively, and again there was no significant difference in level of variation between the two (Mann–Whitney U-test: U=44, DF=11, P=0.86). Although there are unique mutations within the two populations, they are mainly strain-specific and there are no fixed polymorphisms separating the two populations. The EU population showed six unique single nucleotide polymorphisms (SNPs), whereas the NA population had three. All the other SNPs were shared polymorphisms between the two populations. Thus, there is no significant difference between the levels of nuclear sequence diversity between the two regions. These shared polymorphisms could be due to retention of ancestral polymorphisms or due to gene flow between NA and EU populations, and we currently cannot distinguish between these alternatives. This is also true for the shared polymorphisms in microsatellites.

Table 3 Average nucleotide diversity of North American and European strains based on eleven markers sequenced totaling 7079 base pairs.
Full size table

Wolbachia variation

A key feature of a Wolbachia-induced sweep is the hitch-hiking of the associated cytotype with the invading infection (Turelli and Hoffmann, 1991; Werren, 1997; Ballard and Whitlock, 2004; Hurst and Jiggins, 2005). As the N. vitripennis strains clearly differ in the levels of mitochondrial diversity, an obvious hypothesis was that the causal force was a different Wolbachia infection sweeping through the NA population. Previously, Raychoudhury et al. (2009) showed, with a smaller data set of N. vitripennis samples, that there is no Wolbachia variation in the five housekeeping MLST genes in NA and EU populations. Using the larger data set in this study, we screened additional Wolbachia from representative samples from each of the major mitochondrial haplotypes. A total of five different genes from the MLST scheme and an additional marker, wsp, were sequenced from each of the 22 different strains for both the two Wolbachia strains found in N. vitripennis (wNvitA and wNvitB). We found three to be uninfected, three to be infected with type wNvitA only, and 15 to be doubly infected with wNvitA and wNvitB (Supplementary Table 1). One sample, HV-736, could not be sequenced completely because of poor DNA quality. In all the sequenced individuals, we did not find any variation within Wolbachia strains across NA and EU populations. It should be emphasized that failure to detect sequence variation in these six genes does not rule it out in other portions of the genome, particularly in hypervariable and rapidly evolving regions or regions under directional selection (Riegler et al., 2005). A previous study shows that Wolbachia have an estimated mutation rate 120-fold lower than mitochondria in Nasonia (Raychoudhury et al., 2009), and therefore the absence of variation in Wolbachia compared with mitochondria is not surprising.

Previous studies have shown that nearly 100% of N. vitripennis collected from nature are double-infected with both A and B super group Wolbachia (van Opijnen et al., 2005; Raychoudhury et al., 2009). All the single-infected and uninfected strains in this study are laboratory lines that have been in culture for at least a few years (Supplementary Table 1, indicated with *), and it has been shown that laboratory culturing can result in loss of Wolbachia infections (Perrot-Minnot et al., 1996), particularly during long-term refrigeration storage of diapausing larvae, routinely used for maintenance of strains. Thus the most parsimonious explanation of the infection status of these strains is the secondary loss of infections in the laboratory rather than actual infection status in the field.

Comparing mitochondrial, nuclear and Wolbachia variation

On the basis of nuclear sequence and microsatellite variation between NA and EU populations, it can be concluded that the two populations on either side of the Atlantic did not experience a reduction in the diversity of the nuclear genes relative to each other. Thus, these data are inconsistent with a founder effect in NA N. vitripennis. The other phenomenon that can explain the observed reduction in mitochondrial diversity in NA population is a mitochondrial sweep. This would show a reduction in the mitochondrial diversity of the NA strains but the nuclear variation would remain similar to that of the EU strains. We compared the number of polymorphic positions and number of mutations across the mitochondria and the nuclear markers from the strains of the two populations. As seen in Table 4, there are significantly lower numbers of both polymorphisms and polymorphic sites in NA mitochondria relative to nuclear variation. The NA population shows a reduced number of polymorphisms in its mitochondria compared with EU population (8 versus 33, Fisher's Exact test, P=18.28 × 10−4). This contrasts with the number of polymorphisms in the nuclear genes (28 in each). The second measure (Table 4b) enumerates the total polymorphic positions in the two populations. There is a significant reduction in the number of mitochondrial polymorphic positions in the NA samples compared with EU samples (7 versus 32 positions, Fisher's Exact test, P=6.51 × 10−4). This is in contrast to very similar levels of polymorphic nuclear positions (30 and 28 in NA and EU population, respectively). The mitochondria of the EU strains show 8% sites to be polymorphic, whereas, in contrast, the NA populations show only 1.8% polymorphic positions. Although Nasonia mitochondria are known to have an approximately 35–40-fold higher substitution rate than the nucleus (Oliveira et al., 2008; Raychoudhury et al., 2009), there is actually lower variation in mitochondrial genes relative to nuclear genes in NA population. This clearly shows that the NA population has undergone a significant reduction in its mitochondrial diversity but not in its nuclear genes.

Table 4 The number of polymorphic sites and mutations in the NA and EU populations
Full size table

No variation was found in either the A or B super group Wolbachia for the five MLST gene set, which are housekeeping genes generally under stabilizing selection. This reflects the lower estimated mutation rate for Wolbachia relative to that for the nuclear genome of approximately 1/3 (Raychoudhury et al., 2009), and is consistent with a Wolbachia sweep in the NA population and insufficient time for Wolbachia synonymous variation to accumulate.

Discussion

We found that the genetic diversity of N. vitripennis at the mitochondrial level is much higher in Europe than in North America, even though the latter specimens were spread over a larger geographical area (nucleotide diversity π: NA=0.31%, EU=2.36%, see Figure 1). As a consequence, the phylogenetic analysis shows a much weaker structuring in North America (Figure 2). There is one predominant haplotype (# 2) in North America that is also present in Europe but is infrequent. The negative Tajima's D value suggests that there has been a rapid expansion of N. vitripennis in North America, whereas there is no evidence for such an event in Europe.

In the case of a founder effect during the colonization of North America, one would expect a drastic reduction in genetic diversity. However, the differences detected in the microsatellite variation are only subtle, and therefore do not indicate a recent bottleneck. Moreover, the microsatellite data do not reveal a split between the EU and NA populations (Supplementary online Figure 2). Therefore, microsatellite, nuclear SNP and mitochondrial sequence data all show shared polymorphisms between NA and EU populations. Presently, we cannot distinguish between the retention of ancestral polymorphisms and gene flow to explain this pattern. Although microsatellites show high mutation rates that can lead to convergence (homoplasy) over time (Nauta and Weissing, 1996), this is unlikely for SNPs and mitochondrial sequences given the observed divergence levels. Moreover, although levels of SNP variation are low, there are both continent-specific unique SNPs and shared polymorphisms, but no fixed polymorphism between the two continents, making it difficult to resolve between gene flow and retention of ancestral polymorphisms. Although mitochondrial variation is severely reduced in NA population, nuclear sequence and microsatellite variation is similar in NA and EU populations, supporting the hypothesis that there has been a recent cytoplasmic sweep in NA populations, and ruling out a genetic bottleneck during a founder event of NA N. vitripennis derived from EU populations. The likely cause of a mitochondrial sweep is either an advantageous mitochondrial mutation or an advantageous Wolbachia mutation. A selective sweep of a Wolbachia type would result in hitchhiking of the associated mitochondrial haplotype due to their maternal coinheritance (Turelli and Hoffmann, 1991). As pointed out by Turelli (1994), advantageous mutations in Wolbachia, which makes it unidirectionally incompatible with the resident infection (that is, males are incompatible with resident bacteria-infected females) or reduce the fecundity cost of the infection, can sweep through the population and replace the resident Wolbachia strain. This is accompanied by the sweep of the linked mitochondrial haplotype in the process. There is some evidence that such Wolbachia sweeps have happened subsequent to establishment of an infection in a species, both in D. melanogaster (Riegler et al., 2005) and in D. simulans (Weeks et al., 2007). Furthermore, it would be expected to occur as Wolbachia endosymbionts adapt to a new host environment after horizontal transmission from a different insect, as we know has happened for several Wolbachia endosymbionts in the Nasonia system (Raychoudhury et al., 2009). The absence of variation in the five Wolbachia MLST loci in NA and EU populations is not problematic for the hypothesis of a Wolbachia-induced sweep, as beneficial mutation could occur anywhere in the ∼1 Mb genome and can result in a selective sweep. However, given that both the mitochondria-induced and the Wolbachia-induced sweep will create the same patterns of reduced mitochondrial variation, we cannot distinguish between these two causes with the current data. Fitness tests of the relevant Wolbachia and mitochondrial haplotypes together and separated from each other could help resolve this question.

The star-like pattern of the NA portion of the mitochondrial haplotype network can be interpreted as the result of a recently expanding population (Avise, 2000) or as incomplete mitochondrial sweep. Such historical patterns of mitochondrial variation have been interpreted to reflect genetic bottlenecks due to founder populations. Scheffer and Grissell (2003) found no mitochondrial genetic variation in the introduced populations of South American seed-feeding wasps (Megastigmus transvaalensis), but found a high variability in the African source population. Aoki et al. (2008) investigated the consequences of a glacial bottleneck on the mitochondrial genetic variation of a seed parasitic weevil (Cuculio hilgendorfi) in Japan and found a similar pattern. In contrast to these studies, in N. vitripennis the reduced variation is mirrored only in the mitochondria and not in the nuclear variation. The nuclear markers (both microsatellites and multi-gene sequences) show equal levels of variation. So, unlike the instances mentioned, there is no evidence that N. vitripennis in North America was introduced from Europe.

Our findings emphasize the importance of examining both mitochondrial and nuclear variation to infer genetic bottlenecks versus mitochondrial sweeps. The issue is particularly relevant to arthropods due to the high incidence of Wolbachia (Hilgenboecker et al., 2008), which can promote the incidences of mitochondrial sweeps.

Nasonia has an extremely high rate of synonymous mutation in its mitochondrial genes that are estimated to evolve at a rate nearly 35–40 times faster than the nuclear genes (Oliveira et al., 2008). However, the Wolbachia genome evolves at a rate that is approximately 1/3 that of the nuclear genes (Raychoudhury et al., 2009). Extrapolating, there are roughly 105–120 mutations in the Nasonia mitochondria for every mutation in Wolbachia. This can explain the mitochondrial diversity with respect to the lack of diversity in the Wolbachia infections of N. vitripennis. Furthermore, given the high mutation rate observed in Nasonia mitochondria, the high similarity between the common haplotype in North America and an uncommon variant in Europe implies that the sweep in North America would have been very recent. Thus, comparing nuclear, mitochondrial and Wolbachia variation in NA and EU populations, we conclude that NA populations have gone through a recent mitochondrial sweep, possibly due to a selective sweep of their resident Wolbachia infections.

Conflict of interest

The authors declare no conflict of interest.