We sampled Bryobia mites from 111 different locations (populations) between May 2000 and September 2006 (see Additional file 1 and 2 for details on Bryobia samples). This collection comprised samples from a wide range of host plant species (at least 12 different host plant genera in six families) in 14 different countries in Europe. We investigated intraspecific variation in one species (B. kissophila) by analyzing 61 populations across 12 different European countries. Additionally, we included samples from South Africa (one population) and the United States (two populations). As an outgroup reference, we sampled ten European and two Chinese Petrobia spp. (Acari: Tetranychidae) populations. Bryobia and Petrobia both belong to the subfamily Bryobiinae of the Tetranychidae 21. Mites were either directly used or stored in 96% ethanol until DNA extraction. Mites were morphologically identified by Dr. F. Faraji (Mitox, Amsterdam) and Dr. E. Ueckerman (PPRI, Pretoria, South Africa). However, not all samples could be identified morphologically. Currently, few morphological keys for distinguishing Bryobia species exist, but none of these includes all described species and all are developed for identifying locally occurring mites only (1926 (South Africa); 27 (Greece); 28 (New Zealand); 29 (United States)). Voucher specimens will be stored at the Zoological Museum of the University of Amsterdam (ZMA). DNA was extracted from single mites using the CTAB extraction method as previously described 30. Adult females as well as males, if present, were used.

Samples were identified to the species level using morphological keys. Bryobia sarothamni Geijskes, B. rubrioculus (Scheuten), B. berlesei van Eyndhoven, B. praetiosa Koch, and B. kissophila van Eyndhoven were identified based on morphology. Samples that could not be morphologically identified were a posteriori named B. spec. I-VII, based on molecular phylogenetic analysis. The latter designations will be maintained throughout the article, although the exact species status remains to be determined (see discussion). In addition, we sampled mites from the related genus Petrobia. These were identified as Petrobia tunisea and P. harti. Other Petrobia samples were a posteriori named P. spec. I and P. spec. II.

Where possible, field-collected females were reared as isofemale lines in the laboratory. In this way the reproductive mode could be assessed. Strains, in which males were neither observed in the cultures, nor encountered in the field, were classified as asexual. Males are easily recognized by their smaller body size and extremely long front legs (up to two times their body size) compared to females, and by their mating behavior 22. Additionally, the lower part of the abdomen is V-shaped in males, while it is circular in females. For those species that were not reared in the lab, the occurrence of males in the collected field samples was assessed (Table 1). If males were encountered, the clade was considered to reproduce sexually; if not, the clade was classified as asexual.

Part of the mitochondrial COI gene was amplified using various primer combinations (as individual primer sets did not work for all species) (Table 2). Two primers were adjusted based on sequences form other tetranychid taxa available from GenBank (Table 2). Depending on the primer combination, a fragment of 410–867 basepairs (bp) was obtained. A fragment of 410 bp, amplified by the primers COI_F1 and COI_R1 (Table 2; 31), was used in subsequent analyses. PCR conditions for COI were as described in Ros and Breeuwer 30. For one individual of each COI haplotype, the 5' end of the nuclear 28S rDNA (the D1 region) was amplified (see Additional file 1 and 2). This region is generally less variable than the mitochondrial COI region 32 and therefore, samples with identical COI haplotypes were assumed to have an identical 28S haplotype. However, several processes might violate this assumption (see discussion). The validity of the assumption was therefore investigated by sequencing more than one individual per COI haplotype in several cases (see Additional file 1 and 2). PCR reaction mix was the same as for COI 30, except that no additional MgCl₂ was added. PCR cycling conditions for 28S were 2 min. at 94°C, followed by 35 cycles of 40 sec. at 94°C, 40 sec. at 48°C, and 90 sec. at 72°C, and a final extension at 72°C for 5 min. A negative control in which water was added instead of DNA was included in all PCR reactions. Products were visualized on a 1% agarose gel stained with ethidium bromide in 0.5× TBE buffer (45 mM Tris base, 45 mM boric acid, and 1 mM EDTA; pH 8.0).

PCR products were purified using a DNA extraction kit (Fermentas, St. Leon-Rot, Germany). The purified products were directly sequenced using the ABI PRISM BigDye Terminator Sequence Kit (Applied Biosystems, Nieuwerkerk a/d IJssel, The Netherlands) according to the manufacturer's instructions but diluted 16 times. Both strands of the products were sequenced using the same primers as in the PCR amplification. Sequences were run on an ABI 3700 automated DNA sequencer. Sequences were checked visually for ambiguous nucleotides (double peaks) and the presence of stop codons. Sequences of representatives of all haplotypes were submitted to GenBank (see Additional file 1 and 2 for GenBank accession numbers).

Sequences were aligned using ClustalX version 1.8.0 with default settings 33. Analyses were performed for the 28S and COI datasets separately. PAUP* version 4.0b10 34 and DAMBE version 4.1.15 35 were used to calculate numbers of variable sites, uncorrected pairwise divergences, nucleotide composition, and transition and transversion ratios. PAUP was used to perform a chi-square test of base frequency homogeneity across all taxa.

Phylogenetic analyses were performed using Neighbour-Joining (NJ), Maximum Likelihood (ML), and Bayesian methods. Both PAUP and Modeltest 3.6 36 were used to select the optimal evolution model. The selected model was further optimized by critically evaluating the selected parameters 37 using the Akaike Information Criterion (AIC; 38). For 28S, a submodel of the GTR+G (General Time Reversible model with gamma distributed rate heterogeneity among sites) with a rateclass 'a b c a b d' had the highest likelihood score (i.e., the lowest -ln likelihood score) (AIC). Because COI is a protein coding gene, we tested if the likelihood of models could be further improved by incorporating specific rates for each codon position 39. Using this approach, the Transition model (TIM) with site specific rates for the three codon position was selected for COI. Under the selected models, parameters and tree topology were optimized using the successive approximations approach 40. NJ analyses (p-distances) and ML analyses (heuristic search, random addition of sequences with five replicates, TBR branch swapping, and a reconnection limit of 10 for COI analysis) were performed in PAUP. Robustness of nodes was assessed with 1,000 NJ- resp ML-bootstrap replicates. However, as PAUP does not allow for site-specific rates in bootstrap analysis, ML bootstrapping for COI was performed with gamma distributed rates, with 100 bootstrap replicates. Bootstrap values were then plotted on the phylogeny obtained with the TIM model with site specific rates. Bayesian analyses were performed as implemented in MrBayes 3.1.2 41. For 28S we used a GTR+G model; for COI we used a GTR model with separate rates for each codon position. Analyses were initiated form random starting trees. Two separate MCMC (Markov chain Monte Carlo) runs, each composed of four chains (one cold and three heated), were run for 2 million generations. The cold chain was sampled every 100 generations, the first 5,000 generations were discarded afterwards (burn-in of 25%). Posterior probabilities were computed from the remaining trees. We checked whether the MCMC analyses ran long enough using the program AWTY (Are We There Yet?) 42. Stationarity was assumed when there was convergence between the two MCMC runs and when the cumulative posterior probabilities of splits stabilized; in both analyses 2 million generations proofed sufficient. The final trees were rooted using four species of the genus Petrobia as outgroup.

The COI phylogeny depicts B. kissophila as a paraphyletic species (see results). We tested if a phylogeny with B. kissophila as a monophyletic group is significantly less likely than the presented phylogeny by performing a Kishino-Hasegawa test (KH-test; 43) and a Shimodaira-Hasegawa test (SH-test; 44) as implemented in PAUP. First, we performed a constrained heuristic search (B. kissophila as a monophyletic clade; search settings same as without constraint). The KH- and SH-test were used to test for significant differences between the likelihood scores of the trees (unconstrained and constrained). We used a one-sided KH-test to correct for comparing an a priori-specified phylogeny (the constrained tree) with an a posteriori-specified phylogeny (the ML tree) 45. We also tested for monophyly of all asexual respectively sexual species using the same approach (constrained search with asexuals resp. sexuals as a monophyletic clade).

Figure 3 shows the ML phylogeny reconstructed from the 28S D1 fragment, with ML bootstrap values and Bayesian posterior probabilities. Identical topologies were obtained from the different analyses (NJ, ML and Bayesian). The resulting phylogeny is well-resolved. Populations belonging to a single lineage have different geographic origins and are invariably found on the same host plant species or group of host plant species. This indicates that each species has a strong host plant association. Certain species group together, and this grouping is again linked to related host plant species. Bryobia berlesei, B. sarothamni, B. spec. III, and B. spec. VI form a monophyletic group (bootstrap value 77%) and are all found on host plant species of the Fabaceae, tribe Genisteae. Bryobia species found on host plant species of the tribe Genisteae have been named the Berlesei group 18. Closely related to this group are B. kissophila and B. praetiosa, although exact relationships remain unresolved (low bootstrap values). Bryobia kissophila is restricted to Hedera helix (Ivy), whereas B. praetiosa is found on grasses and herbaceous plant species growing along road sides. Samples from rosaceous fruit tree species also form a monophyletic group (bootstrap value 100%). Here, three haplotypes are distinguished. The p-distances between these haplotypes are low (0.6–1.0%) and these haplotypes most likely all concern B. rubrioculus. All aforementioned species group together (bootstrap value 64%) and are clearly separated from another group of species collected from grasses and herbs (B. spec. I, B. spec. V, and B. spec. VII) and Malva spec. (B. spec. IV). Average distances between Bryobia species range from 1.7 to 11.5%.

Figure 4 shows the ML phylogeny reconstructed from the COI fragment, with ML bootstrap values and Bayesian posterior probabilities. NJ, ML, and Bayesian analyses show identical topologies. The COI region is more variable than the 28S D1 region, with interspecific distances ranging from 7.5 to 16.8%. The ML phylogeny reconstructed from COI sequences shows clustering of haplotypes into distinct clades (Figure 4), which in most cases correspond to the 28S lineages found. Clades found for COI are well supported; however, the phylogenetic relationships between these clades remain largely unresolved. Therefore, higher-level groupings (e.g., the Berlesei group) are less pronounced than for 28S. The wide sampling of B. kissophila revealed a large amount of intraspecific variation. Samples from 61 populations cluster into four different clades (named A-D in Figure 4). Average pairwise distances between these clades range from 5.5 to 8.8%. Clade B is the largest, covering samples from all over Europe. Samples of clade A fall within the geographical range of clade B (clade A contains one French and two Dutch populations). Clade C is comprised of populations from the United States and clade D of populations from Spain and Portugal (note however that the sampling intensity in clade C is low).

Based on the COI sequences, B. kissophila is paraphyletic; B. praetiosa forms a monophyletic group, which is nested within the B. kissophila clades. The same topology is supported by the NJ and Bayesian analysis (data not shown). However, the KH-test and SH-test do not reject a topology with a monophyletic B. kissophila clade (p = 0.51 for both tests), so paraphyly of B. kissophila is not well supported. All B. kissophila COI clades share the same 28S haplotype, which differs 1.7% from B. praetiosa. Nucleotide differences between B. praetiosa and the four B. kissophila clades range from 5.1 to 5.9% and are thus smaller than those among B. kissophila clades (5.5 to 8.8%).

For all species except B. rubrioculus, all COI haplotypes found within the various species clades share an identical 28S haplotype. The assumption that samples with identical COI haplotypes have an identical 28S haplotype proved also to be true in cases where this was tested. In B. rubrioculus, however, this assumption was violated, as two samples sharing the same COI haplotype (NL16 and GR5) have different 28S haplotypes (distances between 0.6 and 1.0%).

Despite extensive sampling, from different host plant species in various countries, only seven out of 111 sampled populations contained males (6%). These samples concern two species: B. spec. IV and B. sarothamni. All other populations reproduced asexually, indicating that asexuality is widespread in the genus Bryobia (Table 1). In the 28S phylogeny, the asexual species do not form a monophyletic clade. This is supported by the KH- and SH-test: a phylogeny enforcing asexuals as a monophyletic clade is significantly different from the maximum likelihood phylogeny (p = 0.002 for both tests). The two sexual species form two apical branches within different asexual clades. This pattern does not support the general observation that each asexual species forms a single apical branch with a sexual sister species. The observed patterns indicate either a single origin of asexuality with subsequent reversals to sexuality or multiple origins of asexuality. Unfortunately, the observed patterns in the 28S phylogeny could not be confirmed by the COI phylogeny, which is unresolved at deeper branching patterns due to saturation.

Extensive sampling of B. kissophila shows that intraspecific diversity at COI is very large. Four clades (A-D) are distinguished, with interclade distances ranging from 5.5 to 8.8%. Clade (B) is the largest clade and comprises populations from all over Europe. Two other clades (C and D) are linked to geographically different areas (United States and Iberian Peninsula, respectively). The fourth clade (A) contains three populations from areas within clade B.

Intraspecific COI differences between some B. kissophila clades are larger than interspecific distances between B. kissophila and B. praetiosa. In the phylogenetic tree (NJ, ML, and Bayesian analyses) B. kissophila, which is only found on Hedera helix, forms a paraphyletic species: B. praetiosa falls within the B. kissophila clade. However, bootstrap support is not very high and a KH- and SH-tests do not exclude monophyly of B. kissophila. Still, the overlapping range of inter- and intraspecific divergence is remarkable. Intraspecific differences are much higher than values found in for example asexual Brevipalpus mites (Acari: Tenuipalpidae). In this mite genus intraspecific ML distances range between 1.9 to 4.3% for the same COI region 51. Extremely high intraspecific differences (up to 20.9%) were found between sexual and asexual lineages of the ostracod Eucypris virens 52.

There are several hypotheses that could explain this high level of intraspecific diversity in asexual species. Clonal diversity can originate in at least three different ways 53: 1) through separate and recurrent origin of clones from a sexual ancestor, 2) through hybridization between asexual females with males (either from the same or from other species), or 3) through mutation. Separate origins of clones from a sexual ancestor is often found in species with mixed reproduction (sexual as well as asexual lineages), where asexual clones are continuously formed over time 52. Mixed reproduction is absent in B. kissophila, because the whole species is asexual. It is still possible that the different asexual clones independently originated from a highly variable sexual ancestor 54 and that ancestral mitochondrial polymorphism is maintained. This would, however, result in a correlation between the nuclear and mitochondrial phylogeny. The fact that the distinction between B. kissophila and B. praetiosa (1.7% difference) at the nuclear 28S locus is not supported by the mitochondrial COI phylogeny is not consistent with the above expectation.

Hybridization as a current source of clonal diversity, is not very likely because males are completely absent in B. kissophila as well as in B. praetiosa. Males have never been observed in cultures or in the field. Moreover, Weeks and Breeuwer 23 showed that males obtained after antibiotic treatment do not successfully mate with females. Furthermore, although males do exist in other species, these species are restricted to very different host plant species. On the other hand, hybridization might have occurred in the past before the origin or fixation of asexuality. Then, the phylogenetic pattern we currently observe is simply a reflection of the relatively recent fixation of asexual reproduction in different lineages.

Finally, mutations could explain part of the clonal diversity. We observed high levels of intraspecific diversity at the mitochondrial level, but almost none at the nuclear level. High levels of intraspecific mitochondrial COI diversity either signify a long history of asexuality or an elevated rate of mutations at the mitochondrial DNA. In the case of a long asexual history, divergence at nuclear genes is also expected, and alleles within an individual would eventually diverge over time (Meselson effect; 13). However, if gene conversion occurs during meiosis, divergence at nuclear genes can be reduced or absent. Although heterozygosity is maintained in Bryobia, indicating that parthenogenesis is functionally apomictic 23, the mechanism of parthenogenesis is unknown. Maintenance of heterozygosity can be achieved by strictly apomictic parthenogenesis (no meiosis) or by premeiotic doubling. Only in the latter case, meiosis is present, and gene conversion can take place. It is unclear whether parthenogenesis in Bryobia involves meiosis, and at this moment we can not distinguish between the alternative explanations.

Reproductive parasites like Wolbachia and Cardinium can influence processes like hybridization and fixation of mutations and can therefore have a large impact on clonal diversity. They can cause selective sweeps of mitochondrial DNA, which increase the rate of fixation of mutations. Selective sweeps can either decrease (homogenization of mitochondrial variation in a population) or increase (when a species is infected with different bacterial strains each linked to a mitochondrial haplotype) genetic variation 55. The pattern we found for B. kissophila, with distinct mitochondrial clades with high interclade divergence, was also found in Drosophila simulans 5657. In D. simulans these mitochondrial clades were most likely associated with different Wolbachia strains. Selective sweeps combined with hybridization between different species can also cause homogenization of species after a hybridization event, when the parasites spread and drag along the associated mitochondrial haplotype. This can result in paraphyletic patterns at the mitochondrial DNA. Such mitochondrial introgression patterns have indeed been found in several closely related species of Drosophila 5859 and in species of the blowfly genus Protocalliphora 60. Wolbachia are present in both B. kissophila and B. praetiosa 23. It is possible that the different haplotypes observed within B. kissophila are a consequence of selective sweeps caused by different Wolbachia strains. Additionally, interspecific transfer of mitochondrial DNA from B. kissophila to B. praetiosa could also explain the paraphyly observed for B. kissophila and B. praetiosa.

In conclusion, both mutations and hybridization can explain the clonal diversity and paraphyletic patterns in B. kissophila and B. praetiosa and both processes are possibly driven by reproductive parasites. These processes might cause similar divergence patterns in other Bryobia species and explain incongruencies between 28S and COI phylogenies for B. rubrioculus as well.

The observed high levels of intraspecific variation at the COI gene have serious consequences for the use of COI as a tool for identifying species (DNA barcoding; 6162). Although the region we investigated is a different COI fragment than is usually used in barcoding studies (the Folmer fragment; 63), a similar pattern may be expected for the total COI gene 30. DNA barcoding assumes that intraspecific variation is lower than interspecific variation. A standard threshold of 2% divergence for identifying species has been proposed 61. With intensive intraspecific sampling over a wide geographic range we demonstrated that intraspecific variation can be extensive and easily exceeds interspecific differences, thus undermining current barcoding assumptions. Similar high levels of intraspecific variation are not restricted to Bryobia mites but were also observed in other tetranychid mites 30.

The 28S phylogeny shows that the asexual Bryobia species do not form a monophyletic clade (Figure 5). This implies two possible scenarios for the origin of asexuality in the genus: 1) asexuality originated once, with subsequent radiation of asexuals and at least two independent reversals to sexuality or 2) asexuality originated multiple times (at least seven times; Figure 5). The first scenario is the most parsimonious, because only three transitions (one from sexual to asexual, and two from asexual to sexual) are needed instead of seven (sexual to asexual) to explain the observed phylogenetic pattern. However, there are several arguments against this scenario.

First, reversals to sexuality are less likely than origins of asexuality. Once a species becomes strictly asexual, mutations are expected to accumulate in traits involved in sexual reproduction 26465 and the capacity to produce sexually is lost very quickly. This process will also operate in species where asexuality is caused by bacteria. In species with bacterial-induced asexuality, including B. praetiosa, experiments have shown that females fail to reproduce sexually after removal of the bacteria 2351666768. Moreover, Groot and Breeuwer 69 showed that in Brevipalpus mites where Cardinium causes asexual reproduction, some strains lost the Cardinium bacteria but still reproduced asexually. Apparently, loss of Cardinium did not result in a reversal to sexuality. Although we can not exclude the possibility of reversal to sexuality 25, the aforementioned studies suggest that it is not likely.

Second, although we performed an extensive geographic sampling of species, sampling of the genus is not complete. For example, van Eyndhoven 18 described thirteen species as members of the 'Berlesei' group, three of which are sexual. We have only sampled four species of this group, one of which reproduces sexually. If more sexual species belong to this group, we need to invoke more reversals to sexuality in order to maintain the hypothesis that asexuality has a single origin in Bryobia (Figure 5).

Third, several independent infections with Wolbachia could explain independent origins of asexuality. Asexuality in B. praetiosa is caused by Wolbachia 23. It is possible that Wolbachia or other reproductive parasites are widespread in this genus, causing asexuality in all asexual species. Wolbachia strain diversity and abundance in Bryobia has not been studied so far.

The above arguments suggest that multiple origins of asexuality are likely. At least seven independent origins of asexuality are required based on our current phylogenetic information (Figure 5). Radiation of asexuals within particular clades remains a possibility, especially because sexual species seem rare within the genus. The general thought that asexuals are always single lineages branching off from closely related sexual relatives is certainly not valid within Bryobia. Asexuality seems evolutionary successful, at least in the short term. Asexuality in Bryobia is functionally apomictic, resulting in the maintenance of heterozygosity 23. This may contribute to the success of asexual Bryobia species, as high levels of heterozygosity are assumed to be advantageous (heterosis or overdominance; 70). In other haplodiploid systems parthenogenesis leads to complete homozygosity 717273.

Additionally, bacterial parasites can play a role in the adaptive success of asexuals. Wolbachia has been found in six Bryobia species so far 23. Generally, asexual clones are considered genetically identical if they are identical at their own genomic DNA. However, such genetically identical clones may harbor different bacterial strains. Differences in bacterial composition can influence the fitness of clones, as bacterial symbionts may play a role in protection against parasitoid attack 74 or against fungal pathogens 75. These differences, and also changes in bacterial composition through occasional horizontal transfers, could play a role in the adaptive success of asexual Bryobia species, and of asexuals in general. Furthermore, bacterial genes can be transferred to the host DNA. Recently, Dunning Hotopp et al. 76 showed that an almost complete Wolbachia genome was transferred to genome of its host, Drosophila ananassae. Other arthropod and nematode species also contained fragments of Wolbachia DNA in their genome, indicating that lateral gene transfers occur more often 76. If occurring in asexuals, such transfers undermine the strict clonality of these asexuals and may play a crucial role in their long term evolutionary success.