High genetic divergences indicate ancient separation of parthenogenetic lineages of the oribatid mite Platynothrus peltifer (Acari, Oribatida)

Authors


Michael Heethoff, Abteilung für Evolutionsbiologie der Invertebraten, Zoologisches Institut, Universität Tübingen, Auf der Morgenstelle 28E, D-72076 Tübingen, Germany.
Tel.: +49 7071 2974832; fax: +49 7071 294634;
e-mail: heethoff@gmx.de

Abstract

Theories on the evolution and maintenance of sex are challenged by the existence of ancient parthenogenetic lineages such as bdelloid rotifers and darwinulid ostracods. It has been proposed that several parthenogenetic and speciose taxa of oribatid mites (Acari) also have an ancient origin. We used nucleotide sequences of the mitochondrial gene cytochrome oxidase I to estimate the age of the parthenogenetic oribatid mite species Platynothrus peltifer. Sixty-five specimens from 16 sites in North America, Europe and Asia were analysed. Seven major clades were identified. Within-clade genetic distances were below 2 % similar to the total intraspecific genetic diversity of most organisms. However, distances between clades averaged 56 % with a maximum of 125 %. We conclude that P. peltifer, as it is currently conceived, has existed for perhaps 100 million years, has an extant distribution that results from continental drift rather than dispersal and was subject to several cryptic speciations.

Introduction

Sexual reproduction is the predominant form of reproduction in eukaryotes and both probably evolved together (Cavalier-Smith, 2002). The origin of eukaryotes probably occurred about 2.0–3.5 billion years ago (Miyamoto & Fitch, 1996). Recent calculations date the origin of eukaryotes and sex at 2.5 billion years (Gu, 1997). Extant eukaryotes without sexual reproduction therefore have abandoned sex at some time in their evolution due to hybridization, cytological dysfunction or bacterial infection (Lynch, 1984; Hurst et al., 1993). About 2000 parthenogenetic species are known (Milius, 2003), distributed in almost all groups of organisms. However, the existence and recognition of parthenogenetic species are controversial, due to the focus on the biological species concept which is axiomatically related to sexual reproduction (Mayr, 1940). Additionally, misunderstandings of basic population genetics of parthenogenetic organisms have led to the concept that parthenogens must form a continuum of genetic variation (Birky et al., 2005). This, however, is not necessarily true because parthenogenetic lineages can split into independently evolving lineages (e.g. by geographic isolation), and speciation of parthenogens can therefore be addressed empirically (Barraclough et al., 2003).

Despite being widespread, virtually all parthenogenetically reproducing taxa form singular lineages surrounded by sexually reproducing taxa, presumably due to long-term costs that offset short-term advantages of parthenogenesis (Maynard Smith, 1978; Bell, 1982; Kondrashov, 1993; Butlin et al., 1999; West et al., 1999; Butlin, 2002). Short-term advantages of parthenogenesis are manifold, the most prominent being the omission of producing males and diluting the genome, and it has therefore been called the ‘queen of problems in evolutionary biology’ to explain why most eukaryotic organisms reproduce sexually (Bell, 1982). Nevertheless, parthenogenetic lineages are assumed to be short-lived – evolutionary dead ends – due to the lack of recombination and the inability to either rapidly generate new genotypes or eliminate deleterious mutations in the long term (Muller, 1964; Maynard Smith, 1978; Kondrashov, 1988, 1993; West et al., 1999). The few exceptions that seem to contradict this dogma of evolutionary biology have been called ‘evolutionary scandals’ (Maynard Smith, 1978) or more generally ‘ancient asexuals’. Known ancient parthenogenetic groups are the bdelloid rotifers with 363 species (Mark Welch & Meselson, 2000), the darwinulid ostracods with 26 species (Martens et al., 2003) and probably several speciose groups of parthenogenetic oribatid mites (Norton & Palmer, 1991; Maraun et al., 2003, 2004; Schaefer et al., 2006). In general, there are few empirical data on the age of parthenogenetic lineages (Judson & Normark, 1996; Schön et al., 1996, 1998; Sandoval et al., 1998; Butlin et al., 1999; Mark Welch & Meselson, 2000; Butlin, 2002; Normark et al., 2003), but sex may have been abandoned in bdelloid rotifers tens of millions of years ago (Mark Welch & Meselson, 2000) and in darwinulid ostracods (Martens et al., 2003) and some oribatid mites (Hammer & Wallwork, 1979; Norton et al., 1988a) about 200 million years ago (Ma).

Oribatid mites (Acari, Oribatida) are a speciose group of mainly soil-living invertebrates with about 10 000 described species (Schatz, 2002) and an estimated total number of 50 000 (Travéet al., 1996) to 100 000 species (Schatz, 2002). The first indisputable fossil records of oribatid mites are from Devonian sediments deposited at least 380 Ma (Shear et al., 1984; Norton et al., 1988b), but the origin of the group presumably dates back 400–440 million years (Myr) (Lindquist, 1984). Due to the specific patterns of distribution and the low dispersal power of oribatid mites, it was concluded that the extant distribution of oribatid mite species is mostly the result of continental drift rather than dispersal; some species presumably predated the breakup of Pangea about 200 Ma and kept their distinct morphology (Hammer & Wallwork, 1979). Oribatid mites are important decomposers in forest ecosystems, fallows, fields and meadows with densities up to several hundred thousands per square metre in acidic soils of northern boreal forests (Lussenhop, 1992; Maraun & Scheu, 2000).

Parthenogenesis is widespread among different oribatid mite groups and there is both morphological and molecular evidence for radiations of several speciose taxa for which no sexual species is known (Norton & Palmer, 1991; Palmer & Norton, 1992; Maraun et al., 2003, 2004). The best known are members of the taxon Desmonomata, such as Trhypochthoniidea, Malaconothridae, Camisiidae, Nanhermanniidae and the genus Nothrus; according to Subias (2004), these comprise 51, 137, 78, 58 and 67 species, respectively. Less well studied, but also without known sexual species, are the enarthronote Brachychthoniidae (158 species) and Lohmanniidae (179 species).

Here we use sequence divergences of the mitochondrial gene for cytochrome oxidase I (COI) to estimate the genetic diversity of a well-studied member of Camisiidae –Platynothrus peltifer (C. Koch) – which is a proven parthenogen (Taberly, 1987, 1988). We show that this is indeed an ancient parthenogenetic species, one that has undergone high genetic diversification according to patterns that are consistent with continental drift. These genetically divergent but morphologically similar lineages could be considered separate cryptic phylogenetic species.

Materials and methods

Species studied and sampling of populations

Platynothrus peltifer is distributed across the Palaearctic and Nearctic regions (Karppinen & Krivolutsky, 1982; Golosova et al., 1983; Marshall et al., 1987) and occurs in a wide spectrum of habitats (Weigmann & Kratz, 1981; Siepel, 1990; Heethoff et al., 2000). Although it was earlier described as mycophagous (Luxton, 1972), recent studies suggest that the species directly feeds on plant litter material (Schneider et al., 2004). Reproduction is by automictic thelytoky (Taberly, 1987) with one to four eggs per clutch which are laid once a year between March and September (Grandjean, 1950). Generation time in nature is probably 2 years; adult longevity is unknown, but adults have been kept in laboratory for more than 200 days (Grandjean, 1950; Weigmann, 1975). Therefore, P. peltifer has a K-style reproductive biology despite its parthenogenetic reproduction. There is no fossil record of any Platynothrus species, but other species from the exclusively parthenogenetic taxon Camisiidae are represented in Cretaceous amber (85 Myr; Bulanova-Zachvatkina, 1974). It has been assumed that the distribution of extant populations reflects continental drift (Hammer & Wallwork, 1979) – i.e. that the species predates the breakup of Laurasia, the separation of Europe from North America.

The COI gene of 70 individual oribatid mites was analysed (Table 1). Sixty-five specimens of Platynothrus peltifer (Oribatida, Desmonomata, Camisiidae) came from 16 sites in North America, Europe and Asia. Four specimens of the parthenogenetic species Nothrus silvestris Nicolet (Oribatida, Desmonomata) originated from a single site in Germany and were included to estimate the genetic distances between different species from Desmonomata. As an outgroup, we sequenced one specimen of Oribatula tibialis (Nicolet) (Oribatida, Poronota). The latter species is sexual, but P. peltifer and N. silvestris reproduce exclusively via parthenogenesis (Palmer & Norton, 1991, 1992 and included references).

Table 1.   Origin of the analysed species of oribatid mites (n: number of analysed specimens).
SpeciesOriginnAbbreviation
Platynothrus peltiferUSA, West Virginia7PPUWa–g
USA, New York, Tully3PPUHa–c
Norway, Bergen3PPNBa–c
Germany, Schwedt1PPDBa
Germany, Solling6PPDSa–f
Germany, Darmstadt5PPKWa–e
Belgium, Rockroi4PPBAa–d
Belgium, Calestienne2PPBCa–b
Belgium, Ottignies2PPBOa–b
Austria, Graz8PPOGa–h
Italy, Siena3PPISa–c
Italy, Northern Tyrolia4PPINa–d
Italy, Monte Bodone1PPIMa
Russia, Toropetz1PPRUa
Japan, Yatsugatake10PPJYa–j
Japan, Fuji Yoshida5PPJFa–e
Nothrus silvestrisGermany, Darmstadt4NSKWa–d
Oribatula tibialisGermany, Darmstadt1Outgroup

Sequences

The mitochondrial genes coding for the cytochrome oxidase subunits I and II (COI and COII) are commonly used to address phylogenetic and evolutionary questions (Simon et al., 1994; Lunt et al., 1996; Schön et al., 1998; Salomone et al., 2002; Heethoff et al., 2004; Birky et al., 2005). The COI gene is an auspicious candidate for calculations of divergence times (Knowlton et al., 1993; Andersen et al., 2000; Salomone et al., 2002) and has been investigated in other groups of putative ancient and recent parthenogens including bdelloid rotifers (Birky et al., 2005), darwinulid ostracods (Schön et al., 1998), Schmidtea freshwater flatworms (Pongratz et al., 2003) and Timema walking sticks (Sandoval et al., 1998). Evolutionary rates of COI differ among taxonomic groups, but on average the gene evolves in a clock-like manner in arthropods (DeSalle et al., 1987; Brower, 1994) with a pairwise divergence rate of 2–2.3 % Myr−1 corresponding to an evolutionary rate of 1–1.15 % Myr−1. It was shown that a divergence rate of 2.15 % is also applicable for divergence time estimations in oribatid mites (Salomone et al., 2002).

Molecular techniques

Oribatid mite specimens were preserved in 70–100 % v/v ethanol until preparation. DNA was extracted using the DNeasy Tissue Kit (Qiagen, Hilden, Germany). PCR was performed with the HotStarTaq Master Mix Kit (Qiagen) with primers COIarch1 (5′-GGTCAACAAATCATAAAGAYATYG-3′) and COIarch2 (5′-TAAACTTCAGGGTGACCAAAAAATCA-3′). The total reaction volume of 50 μL contained 1.5 mm of MgCl2, 200 μm of each dNTP, 200 pmol of each primer and 2.5 units of Taq polymerase. PCR was specific; conditions were 15 min at 95 °C for polymerase activation, 30 s at 94 °C for denaturation, 60 s at 51 °C for primer annealing and 60 s at 72 °C for elongation. Thirty-six cycles were performed followed by a terminal 10-min elongation at 72 °C. Products were purified on a 1 % w/v agarose gel and stained with ethidium bromide; bands were excised, purified using chaotropic reagents, cloned with the Perfectly Blunt Cloning Kit (Novagen, Merck, Darmstadt, Germany) and transfected in Nova Blue Singles competent cells (Novagen) by heat shock. Positive clones were selected by blue/white screening. Plasmids were purified by alkaline lysis. Inserts were sequenced in both directions on an ABI capillary sequencer (SRD, Oberursel, Germany).

Data analysis

Sequences were verified to be of arthropod origin by comparisons with known sequences in GenBank using the blast search algorithm (Altschul et al., 1997) and aligned by hand in BioEdit 7 (Hall, 1999). Models for sequence evolution and corresponding parameters were estimated using likelihood ratio tests (hlrt) with modeltest 3.7 (Posada & Crandall, 1998) and corrected maximum likelihood distances were calculated in paup* (Swofford, 1999). A clock-like evolution of the sequences was tested in paup* by enforcing a molecular clock and comparing the resulting tree with an unconstrained one using a likelihood ratio test. In addition, relative rate tests (Tajima, 1993) were performed in mega 3 (Kumar et al., 2004) using O. tibialis as outgroup. A reduced dataset was generated containing only one sequence from each geographical lineage of P. peltifer, one sequence of N. silvestris and O. tibialis as outgroup.

Phylogenetic analyses were conducted in paup* using neighbour-joining (Saitou & Nei, 1987) algorithms based on the maximum likelihood distance matrix; maximum likelihood algorithms (tbr heuristics, random addition sequence with 10 replicates), maximum parsimony (tbr heuristics, random addition sequence with 10 replicates) for the whole dataset; and maximum likelihood (branch and bound) for the reduced dataset. We also used statistical parsimony (Templeton et al., 1992) with the tcs software (Clement et al., 2000) and a 200-step connection limit to estimate a phylogeographical network for the P. peltifer COI haplotypes. Four-fold degenerate sites (D4) were identified in mega 3. Genetic distances were calculated in paup*. Rates of synonymous and nonsynonymous substitutions were estimated in mega 3 using different algorithms (Li et al., 1985; Li, 1993; Pamilo & Bianchi, 1993; Nei & Kumar, 2000). Nucleotide sequences were translated into amino acids in mega 3 using the invertebrate mitochondrial genetic code (Clary & Wolstenholme, 1985).

Results

A 600-bp fragment of the COI gene corresponding to the positions 61–660 of the Drosophila yakuba COI gene and 21–220 of the D. yakuba protein (Clary & Wolstenholme, 1985) was analysed. All sequences were free of stop codons and highly conserved amino acids (Lunt et al., 1996) were equal in all sequences. The analysed part of the protein comprised 200 amino acids with two complete and one partial external loops, two complete internal loops and four complete and one partial transmembrane domains (see Lunt et al., 1996 for the structure of insect COI protein). The alignment was free of gaps and unambiguous. All sequences are available at GenBank (AN: DQ381157–DQ381226); an alignment is available from the corresponding author upon request.

The selected model for the whole dataset (containing 70 sequences of all Platynothrus, Nothrus and the outgroup) was HKY + I + Γ (Hasegawa et al., 1985; Yang, 1996; Table 2). The difference of synonymous and nonsynonymous substitutions averaged 1.12, indicating strong negative selection. Maximum corrected distances of 158 % were found between the outgroup and PPBOa (Table 2). All D4 positions were variable and the maximum p-distance was found between the outgroup and PPRUa. Maximum amino acid distances of 10 % were found between the outgroup and PPBOa. The likelihood ratio test between analyses with and without enforcing a molecular clock led to a rejection of the molecular clock hypothesis [χ2(68) = 150.56, P < 0.001]. However, relative rate tests indicated a clock-like behaviour of the sequences [no rate heterogeneity with χ2(1) < 3.85, P > 0.05].

Table 2.   Description of sequence variation, distances and estimated parameters.
 Whole dataset PlatynothrusReduced dataset
General
 Sequences (n)70659
 Variable sites (n)259227239
modeltest hlrt
 Selected modelHKY + I + ΓHKY + ΓHKY + I + Γ
 Frequence [A]0.29270.27630.2941
 Frequence [C]0.24740.24990.2576
 Frequence [G]0.14740.15410.138
 Frequence [T]0.31250.31960.3104
 Ti/Tv5.57166.49459.6994
 Invariable sites (I)0.507800.5796
 Gamma shape (α)1.06230.18661.0833
Distances
 Uncorrected (averaged)0.1620.150.211
 Uncorrected (maximum)0.2720.2450.268
 Corrected (averaged)0.5130.5591.5997
 Corrected (maximum)1.5771.2483.0725
D4 sites
 Number of sites (n)919696
 Variable sites (%)10096.88100
 Distance (averaged)0.4890.460.621
 Distance (maximum)0.7580.750.75
Amino acids
 Variable sites (n)411924
 Distance (averaged)0.0150.00380.036
 Distance (maximum)0.10.020.09

Genetic distances among P. peltifer sequences were calculated with the exclusion of N. silvestris and the outgroup (Table 2). The difference of synonymous and nonsynonymous substitutions averaged 1.03, indicating negative selection. Corrected distances reached 125 % and were maximal between PPJFb and PPUWe; D4 distances were at a maximum between PPJFb and PPUWa. The molecular clock hypothesis was rejected by the likelihood ratio test [χ2(63) = 1492.2, P < 0.001], but again was supported by relative rate tests [no rate heterogeneity with χ2(1) < 3.85, P > 0.05].

The reduced dataset (nine sequences: outgroup, N. silvestris, one P. peltifer from each geographical lineage) was constructed based on the phylogenetic analyses of the whole dataset (Fig. 1). The selected model was HKY + I + Γ (Table 2). The difference of synonymous and nonsynonymous substitutions averaged 1.48, again indicating strong negative selection. Maximum corrected distances reached 308 % (outgroup–PPDBa), indicating a high degree of saturation. The molecular clock hypothesis for the reduced dataset was supported by the likelihood ratio test [χ2(7) = 7.37, P = 0.39] and also by relative rate tests [no rate heterogeneity with χ2(1) < 3.85, P > 0.05].

Figure 1.

 Neighbour-joining tree of the whole dataset calculated in paup*. Distances are based on parameters estimated by a likelihood ratio test (see text for details). Numbers at nodes indicate bootstrap values from 1000 replicates. Seven clades of Platynothrus peltifer can be identified and are correlated to geographical origin of the specimens. Note the weak bootstrap support on nodes related to the sites from North America compared with other node supports. For abbreviations, see Table 1.

Phylogenetic analysis of the whole dataset indicated that the analysed specimens of P. peltifer constituted seven clades, which correlated with their geographical origin. The gross topology of phylogenetic trees using corrected distances with neighbour joining (Fig. 1), maximum likelihood and maximum parsimony (not shown) was similar. The seven geographical clades were: Northern/Central Europe (NCE), Eastern Europe (EE), Northern Tyrolia (NT), Southern Europe (SE), USA – New York State (UNY), USA – West Virginia (UWV) and Japan (J). Genetic distances (corrected and uncorrected) within the geographical clades were below 2 %. Distances between clades averaged 56 % with a maximum of 125 % (J–UWV).

The seven geographical lineages were also recovered using statistical parsimony as implemented in tcs (Fig. 2). To obtain a complete connection of the network, a connection limit of 200 steps was defined. The COI sequences of P. peltifer consisted of 35 haplotypes. The haplotypes within the geographical lineages were separated by one to six substitutions, whereas the geographical lineages were separated by 27–133 with the highest number of suggested substitutions being 374 (PPNBb–PPUWe,g). Maximum genetic distances apply to the deepest branching in the group and thus to its age (Avise, 1994). The maximum pairwise percentage distance of D4 sites for P. peltifer was 0.75 (PPUWa–PPFJb,c). Using this nonparametric distance estimation with a maximum conservative evolutionary rate of 5 % for neutrally evolving sites in mitochondria (Tautz et al., 2003), an age of 15 Myr was estimated for P. peltifer. This estimation can be refined by parametrical approaches. The maximum corrected distance within P. peltifer reached 1.248 corresponding to an age of 58 Myr [using 2.15 % divergence per Myr as estimated for oribatid mites by Salomone et al. (2002)]. Using the reduced dataset with a molecular clock assumption, the separation of Nothrus and Platynothrus (an estimate of the age of Desmonomata) occurred 110 Ma (Fig. 3) which also likely represents the time since parthenogenetic reproduction in this taxon existed. Two large clusters of P. peltifer could be separated (Fig. 3): one containing the genotypes from North America (UNY, UWV), the other containing the European and Asian genotypes (NCE, EE, NT, SE, J); these clusters separated 64 Ma. Within the European/Asian group, SE is separated from the rest 31 Ma, J separated from NT, NCE and EE 17 Ma, NT from NCE and EE 12 Ma and NCE and EE split 2.5 Ma (Fig. 3). Without the molecular clock assumption, the topology of the reduced dataset differed with regard to the split of North America and Europe (Fig. 4). The lineages from North America were not monophyletic, UNY appears as the sister group of a clade containing UWV and all European and Asian lineages. This topology is also supported by the analysis of the whole dataset (Fig. 1) without assuming a molecular clock. However, the bootstrap support is higher for a clade containing both sites in North America than for a separation of these lineages (Figs 1, 3 and 4) and a monophyletic origin of the North American lineages is supported by statistical parsimony (Fig. 2).

Figure 2.

 Reconstructed tcs plot of 35 COI haplotypes of Platynothrus peltifer. Connection limit: 200 steps. Seven geographical groups can be separated. Haplotypes within the lineages are separated by one to six substitutions and groups are separated by 27–133 substitutions. Length of branches is meaningless and thickness indicates the number of substitutions between nodes (also written on branches). The haplotype in the dotted box (PPBAb,c, PPNBa) has the highest outgroup weight. Sites from North America are monophyletic. For abbreviations, see Table 1.

Figure 3.

 Maximum likelihood (branch and bound) analysis of the reduced dataset with a molecular clock assumption. Parameters estimated by a likelihood ratio test (see text for details). Numbers at nodes indicate bootstrap values from 100 replicates. Bold numbers at nodes indicate the age (Myr) of the split as calculated by assuming a divergence rate of 2.15 % Myr−1. Sites from North America are monophyletic with a high bootstrap support. SE, Southern Europe; J, Japan; NT, Northern Tyrolia; NCE, Northern/Central Europe; EE, Eastern Europe; UWV, USA – West Virginia; UNY, USA – New York.

Figure 4.

 Maximum likelihood (branch and bound) analysis of the reduced dataset without a molecular clock assumption. Parameters estimated by a likelihood ratio test (see text for details). Numbers at nodes indicate bootstrap values from 100 replicates. Sites from North America are not monophyletic with rather weak bootstrap support. SE, Southern Europe; J, Japan; NT, Northern Tyrolia; NCE, Northern/Central Europe; EE, Eastern Europe; UWV, USA – West Virginia; UNY, USA – New York.

Discussion

Evidence for selection

Genes coding for proteins are usually under negative selection when synonymous substitution rates (SS) are higher than nonsynonymous substitution rates (NS) (SS > NS; Page & Holmes, 1998). The rate of SS and NS can therefore be used to identify mitochondrial pseudogenes in the nuclear genome that have no function and are under neutral selection (SS = NS). Different algorithms have been developed to measure SS and NS (Li et al., 1985; Li, 1993; Pamilo & Bianchi, 1993; Nei & Kumar, 2000), but how each estimates the numbers of potential and realized synonymous and nonsynonymous substitutions can be questioned (Nei & Kumar, 2000). Measurements of SS and NS in this study suggest that the predominant type of substitutions in the COI gene of the oribatid mites investigated was synonymous (SS − NS > 1) and that the gene is under strong negative selection. Together with the absence of stop codons and the conservation of functional amino acids in the COI protein, it is unlikely that the analysed sequences represent nuclear pseudogenes. The clustering of all sequences in phylogenetic analysis (Fig. 1) in comparison with the outgroup O. tibialis, as well as the identification of sequences in blast search, makes contaminations from gut contents or parasites unlikely. The fact that the gene is under strong negative selection and that the evolutionary rate is comparable with that in other arthropods (DeSalle et al., 1987; Brower, 1994; Salomone et al., 2002) contradicts the concept that parthenogenetic organisms should accumulate deleterious mutations (Birky et al., 2005).

Genetic distances and phylogeography

The admissibility of genetic distance for time estimation depends on the clock-like evolution of the analysed gene. A simple way to test the accuracy of the molecular clock is to estimate the difference in the number of substitutions between two closely related taxa in comparison with a third, more distantly related outgroup species using a relative rate test (Tajima, 1993; Page & Holmes, 1998). This test does not require any knowledge of the divergence times of the taxa in question. We compared all pairs of sequences with the outgroup O. tibialis and found no evidence for significant rate heterogeneity, which allows sequence distances to be used for time estimates. However, the likelihood ratio test indicated that there is no molecular clock except for the reduced dataset. This may be explained by the fact that the geographical lineages are each represented by several specimens with a very small divergence compared with the divergences between the geographical lineages. This can also be seen from the branch lengths (Fig. 1) and the numbers of substitutions in the tcs plot (Fig. 2). Using all sequences therefore results in an over-representation of identical (or nearly identical) genotypes that may not be independent. Therefore, it might be erroneous to use the whole dataset for the estimation of a molecular clock by a likelihood ratio test. After removing all sequences except one from each geographical lineage, the molecular clock hypothesis holds true just as in all relative rate tests. Time estimations were therefore performed using only the reduced dataset.

A nonparametric distance estimation for nucleotide sequences is the measurement of percentage distances in four-fold degenerate sites (D4). D4 sites do not affect the protein sequence and are assumed to have a slower saturation and less sensitivity to transition–transversion bias (Li, 1993). Almost all D4 sites were variable and pairwise distances reached 75 %; this indicates an ancient split of the lineages and high saturation of sequences. The observed distances likely underestimate the actual amount of evolutionary change due to high saturation. A number of parametric models have been developed to convert observed distances into measures of actual evolutionary distances (Jukes & Cantor, 1969; Kimura, 1980; Felsenstein, 1981; Hasegawa et al., 1985; Rodriguez et al., 1990; Yang et al., 1994; Yang, 1996). The measurement of genetic distances and phylogenetic analyses of nucleotide sequences depend strongly on the choice of evolutionary model and the estimated parameters (Goldman, 1993; Hoyle & Higgs, 2003). A likelihood ratio test uses log-likelihood scores to establish the model of DNA evolution that best fits the data (Posada & Crandall, 1998). We used fixed parameters established by a likelihood ratio test to increase the accuracy of likelihood-based distance estimates (Hoyle & Higgs, 2003). The corrected maximum pairwise distances within P. peltifer were 125 % and 307 % for the reduced dataset (O. tibialis– EE, NCE). Distances of more than 100 % are due to a substantial proportion of back-mutations and indicate an ancient split of the COI lineages. With such highly saturated sequences, it becomes difficult to estimate divergence times even if the sequences evolved clock-like. Time estimates therefore have to be interpreted with some caution, but that such large genetic distances evolved over at least tens of millions of years seems certain. Time estimates based on the evolutionary rate of 2.15 % for the COI sequences fit with data from continental drift: North America and Europe completely separated roughly 60–80 Ma (Palmer, 1999; Stanley, 1999) which is in accordance with our calculations for that split (64 Myr; Fig. 3). The distances from SE indicate that those populations have been separated from other European and Asian populations for 31 Myr, which might be explained by the uplifting of the Alps (Palmer, 1999). The samples from northern Tyrol (PPIN) originate from a more northern part of the Alps and represent a unique lineage (NT). Although a high degree of saturation renders time estimates equivocal, we conclude that the correlation of geological events with the genetic divergences of P. peltifer suggests that the current distribution is due to continental drift. This parthenogenetic species is therefore ancient, probably predating the separation of Europe and North America (see also Hammer & Wallwork, 1979).

Indication of speciation within P. peltifer

Nucleotide diversity within animal species is generally below 2 % (Avise, 1994), such that higher degrees of divergences indicate speciation events (Birky et al., 2005). Although the biological species concept is inapplicable to parthenogenetic organisms, parthenogens could evolve differentiated populations (Barraclough et al., 2003) after events like geographical isolation; these would constitute new phylogenetic species. This speciation might be seen only in nucleotide diversification, without morphological manifestations (Birky et al., 2005), or it could become obvious from both morphological and genetic traits, as seen in the radiation of several parthenogenetic oribatid mite lineages (Maraun et al., 2004). Genetic distances of the seven identified geographical lineages of P. peltifer are large enough to indicate parthenogenetic speciation. Due to the clonal structure of this species, plus the absence of closely related sexual species, it is unlikely that new genotypes in P. peltifer have evolved frequently from sexual ancestors (Norton & Palmer, 1991; Palmer & Norton, 1991, 1992).

The seven geographic lineages distinguished by COI have not yet been studied morphologically in detail, and it is premature to propose and name new species. It may not be too difficult to distinguish some of the populations. For example, our specimens from the eastern USA are lighter in colour, slightly flatter and usually (but not always) have more genital setae than those from Germany (where the type locality is). Also, individuals from Scandinavia seem to have shorter notogastral setae than those in central Europe (Sellnick & Forsslund, 1955; Olszanowski, 1996). However, commonly studied traits are known to be variable within populations, even within clones (e.g. Grandjean, 1974; Travé & Olszanowski, 1988), and this variation overlaps among examined populations. While the latter authors concluded that P. peltifer is a single widespread, variable species, they reported some geographic differences in average size and certain meristic setal characters. So, with discriminate analysis, specimens of the seven genetic lineages – and others not yet found – might prove to be identifiable by as-yet undefined sets of morphological characters.

Speciation without sexual reproduction?

The fact that sexual reproduction was completely absent during the evolution of P. peltifer is almost impossible to prove unambiguously. However, there is good evidence that this in fact was the case. Males that occur in parthenogenetic populations are sterile and have very low frequencies (Grandjean, 1941; Taberly, 1988). Sexual species of oribatid mites are known to have balanced sex ratios, and neither cyclical nor geographical parthenogenesis has been reported in any oribatid mite species (Norton & Palmer, 1991; Palmer & Norton, 1991; Norton et al., 1993). Natural populations show a small number of genotypes, each with fixed heterozygosity which is also fixed in mother–daughter lineages, so despite the presence of rare and sterile males, sexual reproduction and recombination presumably are absent (Lynch, 1984; Palmer & Norton, 1991, 1992). Additionally, it was shown that oribatid mites underwent substantial parthenogenetic radiations (Maraun et al., 2004). Contradictory to our results from the mitochondrial genome, the nuclear genome of P. peltifer was shown to be homogenized (Schaefer et al., 2006), although there is no indication of recombination. Such homogenized nuclear genomes were also reported for the ancient parthenogenetic darwinulid ostracods (Schön & Martens, 2003).

The mechanism responsible for the nuclear homogenization is not clear, although gene conversion has been suggested (Schaefer et al., 2006). In fact, a cellular mechanism, rather than a mechanism at the population level (i.e. sex), should be responsible for the nuclear homogenization in P. peltifer; otherwise, the contradictory COI diversity cannot be explained. This cellular mechanism corresponds to the mode of parthenogenesis. It was shown that P. peltifer reproduces by automictic thelytoky in combination with terminal fusion (Taberly, 1987). Such a mechanism should lead to homozygous genomes. However, isoenzyme electrophoretic studies indicated fixed heterozygosity within apparent clones (Palmer & Norton, 1992) and nuclear recombination seems to be absent (Schaefer et al., 2006). These findings could be explained if the sequence of meiotic divisions is inverted (‘inverted meiosis’; Wrensch et al., 1993). Further investigations are needed to unravel the cellular mechanism of parthenogenesis of oribatid mites and the results will be of great importance to explain the low degree of nuclear when compared with mitochondrial diversification (Heethoff et al., 2006).

Altogether, the available evidence indicates that oribatid mites represent more than a third ‘evolutionary scandal’; they contain several ancient and speciose parthenogenetic groups representing a number of evolutionary scandals. Even within a commonly recognized species like P. peltifer, it seems that substantial divergence, qualifying as cryptic speciation events, happened after the geographical isolation of populations. With a potential age of more than 100 Myr, they are among the oldest living parthenogenetic metazoan taxa. Even this time period likely is an underestimation for the time of parthenogenetic reproduction in the group as a whole due to the fact that the Camisiidae consist of 78 morphologically distinct species, all of which reproduce exclusively via parthenogenesis and radiated despite the absence of sex (Maraun et al., 2004). The combination of nuclear homogenization, morphological flexibility and mitochondrial diversification makes this group unique for further investigating the mechanisms responsible for long-term stability and speciation of parthenogenetic lineages.

Acknowledgments

We thank Reinhart Schuster, Heinz Schatz, Valerie Behan-Pelletier, Nobuhiro Kaneko, Torstein Solhøy, Gerd Weigmann, Philippe Lebrun, Fabio Bernini and Sonja Migge for providing specimens for analysis and Richard Thomas for providing primer sequences. Constructive criticisms on earlier drafts of this manuscript were offered by Joel Peck, Pekka Pamilo, Alex Kondrashov, Roger Butlin, Diethard Tautz, Andrew Weeks, Isa Schön, Koen Martens, Benjamin Normark and Matthew Meselson. The initiation of this work was supported by F. K. Zimmermann. The study was supported by the German Science Foundation (DFG).

Ancillary