No evidence for the ‘Meselson effect’ in parthenogenetic oribatid mites (Oribatida, Acari)

Authors


Mark Maraun, Technische Universität Darmstadt, Institut für Zoologie, Schnittspahnstrasse 3, 64287 Darmstadt, Germany.
Tel.: +49-6151-165219; fax: +49-6151-166111;
e-mail: maraun@bio.tu-darmstadt.de

Abstract

It has been hypothesized that in ancient apomictic, nonrecombining lineages the two alleles of a single copy gene will become highly divergent as a result of the independent accumulation of mutations (Meselson effect). We used a partial sequence of the elongation factor-1α (ef-1α) and the heat shock protein 82 (hsp82) genes to test this hypothesis for putative ancient parthenogenetic oribatid mite lineages. In addition, we tested if the hsp82 gene is fully transcribed by sequencing the cDNA and we also tested if there is evidence for recombination and gene conversion in sexual and parthenogenetic oribatid mite species. The average maximum intra-specific divergence in the ef-1α was 2.7% in three parthenogenetic species and 8.6% in three sexual species; the average maximum intra-individual genetic divergence was 0.9% in the parthenogenetic and 6.0% in the sexual species. In the hsp82 gene the average maximum intra-individual genetic divergence in the sexual species Steganacarus magnus and in the parthenogenetic species Platynothrus peltifer was 1.1% and 1.2%, respectively. None of the differences were statistically significant. The cDNA data indicated that the hsp82 sequence is transcribed and intron-free. Likelihood permutation tests indicate that ef-1α has undergone recombination in all three studied sexual species and gene conversion in two of the sexual species, but neither process has occurred in any of the parthenogenetic species. No evidence for recombination or gene conversion was found for sexual or parthenogenetic oribatid mite species in the hsp 82 gene. There appears to be no Meselson effect in parthenogenetic oribatid mite species. Presumably, their low genetic divergence is due to automixis, other homogenizing mechanisms or strong selection to keep both the ef-1α and the hsp82 gene functioning.

Introduction

Putative ancient parthenogenetic taxa, such as bdelloid rotifers, darwinulid ostracods and some clusters within oribatid mites (Mark Welch & Meselson, 2000; Schön & Martens, 2003; Maraun et al., 2004), are challenging theories of evolutionary biology. They are species rich (darwinulid ostracods: 26 species; bdelloid rotifers: 363 species; oribatid mites: between 54 and 156 species in different clusters) and may have radiated while being parthenogenetic. These taxa have been termed ‘evolutionary scandals’ (Maynard Smith, 1978) because theoretical considerations suggest that long-term advantages of sexual reproduction should eventually overcome the short-term advantages of parthenogenesis. Consequently, parthenogenetic lineages are doomed to extinction in the long term. Understanding how species managed to persist over long periods of time without sex and recombination will contribute significantly to our understanding of the prevalence of sex in the living world.

No single evolutionary theory is able to explain why sex is the prevalent reproductive mode in eukaryotes (Bell, 1982; Kondrashov, 1988,1993; Roughgarden, 1991; Browne, 1992; Lynch et al., 1993; Crow, 1994; West et al., 1999; Butlin, 2002). From the selfish DNA point of view parthenogenesis is much easier to explain than sexual reproduction, yet only a minority – about 1%– of the known eukaryotic species reproduce via parthenogenesis (White, 1978). It is still not clear if the main function of sex is the homogenisation of genomes (Roughgarden, 1991; Otto & Nuismer, 2004) or the production of different genotypes that allow adaptation to changing environments (Ghiselin, 1974; Maynard Smith, 1978; Hamilton, 1980; Perlman et al., 2003). Ancient asexual taxa provide the opportunity to test the hypothesis that sexual reproduction evolved not to produce genetic variation in the short term but to homogenize genomes over long periods of time. Contradicting the latter hypothesis, ancient asexual taxa have more genetic variation than sexual taxa of similar age (Barraclough & Herniou, 2003; Barraclough et al., 2003).

A precondition for understanding the survival of ancient asexuals is to know (a) whether these taxa really are ancient and (b) whether they are apomicts or automicts.

Of the three putative ancient asexual animal taxa, bdelloid rotifers are apomicts (Mark Welch & Meselson, 2000; Birky, 2004). In asexual, nonmarine ostracods, only apomixis has been reported so far (Butlin et al., 1998) although the Darwinulidae have not yet been investigated extensively. The exceptionally low genetic divergence of ITS1 in the ancient asexual species of the family, Darwinula stevensoni, could be attributed to gene conversion (Schön & Martens, 2003) whereas homogenising mechanisms such as highly efficient DNA repair (Schön & Martens, 1998) could keep other genomic regions intact (Schön & Martens, 2003). In contrast, there is evidence for automixis in parthenogenetic taxa of oribatid mites (e.g. Platynothrus peltifer) (Taberly, 1987a,b; Wrensch et al., 1994; Butlin et al., 1998; Gorelick, 2003).

Definitions of sexual and parthenogenetic reproduction are sometimes contradicting, which is why we provide our definition below. We use the term sex for the fusion of meiotically and independently produced gametes of two individuals. Parthenogenetic reproduction as used here relates specifically to thelytoky, the production of females from unfertilised eggs. Further, we distinguish between apomictic and automictic thelytoky. Apomixis describes a process in which oocytes are produced by mitotic cell division in the germ line. Daughters inherit a complete unrecombined maternal genome, which results in increased heterozygosity at any given locus over time as mutations accumulate. Automixis refers to the reconstitution of the diploid state from meiotically reduced oocytes within one organism. Several mechanisms for reconstituting the numbers of chromosomes in the presumptive egg are known, each either resulting in increased heterozygosity or homozygosity among the offspring. Doubling the chromosomes before meiosis and pairing of homologous chromosomes increases homozygosity as does post-meiotic fusion of meiotically produced ova (eggs with polar bodies), if nonhomologous chromosomes pair, crossing over can generate genetically diverse offspring. The degree of heterozygosity at a given locus depends on whether the first or the second meiotic division is suppressed. The same holds for the fusion of the first (central fusion) or the second (terminal fusion) polar body with the oocyte. In general, central fusion automixis results in heterozygosity (except in cross-over regions) whereas terminal fusion automixis increases homozygosity (except in cross-over regions) (Maynard Smith, 1978; Bell, 1982; Suomalainen et al., 1987). The patterns of these hypothesized mechanisms can be overwritten by nonmeiotic homogenising processes, such as gene conversion, mitotic recombination, or DNA repair.

One way to test for long-term absence of recombination is the so-called ‘Meselson effect’, as proposed by Birky (1996) and Mark Welch & Meselson (2000). The Meselson effect assumes that the two allelic copies at a given locus (in a diploid) become divergent over time as a result of the independent accumulation of mutations. Over millions of years of evolution the intra-individual allelic divergence of ancient apomicts should be higher than that of sexual species. However, the effect should not exist in ancient automicts, in which homogenising mechanisms such as recombination, gene conversion (Butlin, 2000), or DNA repair (Schön & Martens, 2000) should lower allelic diversity (Gorelick, 2003; Gandolfi et al., 2003). Therefore, the power of testing the Meselson effect is asymmetrical. Its presence indicates long-term lack of recombination but its absence does not distinguish sexual reproduction from any other sort of homogenising mechanism (Butlin, 2000). So far, the Meselson effect has been found in bdelloid rotifers (Mark Welch & Meselson, 2000) but not in D.stevensoni (Schön & Martens, 2003) or any other putative ancient parthenogenetic group (Birky, 2004). Even the presence of the Meselson effect does not prove long-term absence of recombination; Ceplitis (2003) showed that the effect is also compatible with low rates of sexual reproduction.

In contrast to darwinulid ostracods and bdelloid rotifers, the various putative ancient asexual taxa of oribatid mites have so far hardly been studied in an evolutionary context with molecular techniques. These mites are decomposer animals that reach high densities in virtually all soils of the world, ranging from a few hundreds in agricultural sites up to 500 000 m−2 in northern boreal forests (Maraun & Scheu, 2000). Fossils are known from Devonian sediments indicating an age of at least 380 My (Shear et al., 1984, Norton et al., 1988). There is increasing evidence that several species-rich clusters of oribatid mites represent ancient asexual lineages that radiated while being parthenogenetic (Norton & Palmer, 1991; Norton et al., 1993; Maraun et al., 2003,2004).

This study tests for the presence of the Meselson effect in parthenogenetic species of oribatid mites. The intra-individual genetic divergence in parthenogenetic taxa is compared with that of sexually reproducing oribatid mite species. We used the elongation factor-1α (ef-1α) and the heat shock protein 82 (hsp82) genes; both are conserved and most likely single copy genes (Klompen, 2000; Mark Welch & Meselson, 2000). The hsp82 gene has been used for the analysis of the Meselson effect in bdelloid rotifers (Mark Welch & Meselson, 2000) and darwinulid ostracods (Schön & Martens, 2003) facilitating the comparison of results from the different groups of ancient asexuals. The ef-1α gene has been successfully used for phylogenetic studies (e.g. Klompen, 2000). Additionally, the functionality of hsp82 was verified in the current study by sequencing its cDNA. We also applied likelihood permutation tests (McVean et al., 2002) to test for recombination and gene conversion in ef-1α and hsp82 from both sexual and parthenogenetic taxa.

Materials and methods

We sequenced between one and eleven clones of a fragment of the ef-1α gene (573 bp) of three sexual (Eupelops plicatus, Achipteria coleoptrata, Steganacarus magnus) and three parthenogenetic species of oribatid mites (Nanhermannia coronata, Nothrus silvestris, P. peltifer). We also sequenced two clones from one individual of the parthenogenetic oribatid species Hypochthonius rufulus (Enarthronota) as outgroup for the phylogenetic analysis (for details of the sample locations see Table S1). Additionally, we sequenced between eight and 17 clones of a fragment of the hsp82 gene (525–537 bp) from the sexual species S. magnus and the parthenogenetic species P. peltifer (genomic DNA and cDNA). We also sequenced one clone of the parthenogenetic species Tectocepheus minor (Oribatida) to root the tree. Mites were collected from litter in different forests in Germany (for details of the sample locations see Table S2).

Sample preparation, PCR and sequencing

Single animals were placed in an Eppendorf tube, frozen in liquid nitrogen and crushed against the tube wall with a steel rod. Genomic DNA was obtained using Qiagen DNeasy® kit for animal tissues following the manufacturer's protocol for animal tissues (Qiagen, Germany).

The ef-1α fragment was amplified using the forward primer 40.71 F (5′-TCN TTY AAR TAY GCN TGG GGT-3′) and the reverse primer 52.RC (5′-CCD ATY TTR TAN ACR TCY TG-3′; Klompen, 2000). Amplifications of the ef-1α fragment were performed in 50 μL volumes containing 2 μL of each primer (100 pm μL−1), 15 μL DNA and 25 μL of HotStarTaq Mastermix (2.5 units HotStarTaq® Polymerase, 200 μm of each dNTP and 15 mM MgCl2 buffer solution; Qiagen, Germany). Amplification conditions were: an initial denaturation step at 95 °C for 15 min, followed by 35 cycles of 94 °C (30 s) denaturation, 50 °C (70 s) annealing and 72 °C (90 s) extension and terminated with 10 min at 72 °C as final extension.

The hsp82 gene fragment of 525–537 bp length was amplified using the forward primer hsp1.2 (5′-TGC TCT AGA GCA CAR TTY GGT GTN GGT TTY TA-3′) and the reverse primer hsp8.x (5′-ACG TTC TAG ART GRT CYT CCC ART CRT TNG T-3′) (Schön & Martens, 2003). PCR compositions for hsp82 were identical to ef-1α but cycling conditions differed: an initial denaturation step at 95 °C for 15 min was followed by nine cycles of 95 °C (50 s) denaturation, 50 °C (50 s) annealing and 72 °C (2 min) extension. Then 34 cycles of 95 °C (50 s), 55 °C (50 s) and 72 °C (2 min) followed. The PCR was again concluded with a final extension of 10 min at 72 °C.

PCR products were purified on 2% agarose gels; bands were excised from gels, recovered and purified using QIAquick® Gel Extraction Kit (Qiagen, Germany). All PCR products were cloned using the Perfectly BluntTM Cloning Kit (Novagen, Germany) and transformed into E. coli Nova Blue SinglesTM competent cells (Novagen, Germany) by heat shock using the manufacturer's protocol. The plasmids were purified using QIAprep® Spin Miniprep System (Qiagen, Germany). The DNA was sequenced by Scientific Research and Development GmbH, Oberursel (Germany), using an ABI sequencer (Applied Biosystems, USA).

mRNA isolation, cDNA synthesis and sequencing

Total RNA was extracted from single and half individuals of S. magnus and P. peltifer. Animals were placed in reaction tubes, frozen in liquid nitrogen and disrupted using a steel rod. The solution was homogenized using a syringe with a small needle. The procedure followed the protocol for isolation of total RNA from animal tissue as given in the RNeasy® Mini Handbook (Qiagen, Germany). Complementary DNA (cDNA) was synthesized using the Qiagen One Step® RT-PCR Kit (Qiagen, Germany). Amplification conditions were: 30 min at 50 °C for reverse transcription of the RNA template to obtain cDNA, followed by PCR for hsp82 (see above). Cloning was done as described above and the cDNA was sequenced from Scientific Research and Development GmbH, Oberursel (Germany). All ef-1α and hsp82 sequences were deposited in GenBank (see Tables S1 and S2 for accession numbers).

Alignment and phylogenetic analysis

DNA and cDNA sequences were aligned with the multiple alignment algorithms in the program Clustal X (Thompson et al., 1994,1997) using 10 and 0.1 as parameters for gap opening and gap extension penalty, respectively. The alignment for ef-1α was unambiguous because all sequences had the same length. All sequences except three could be translated into amino acids. Those DNA sequences contained stop-codons; which may be due to PCR errors, or, alternatively, they may be pseudogenes. The sequences were included in the phylogenetic tree and clearly marked with asterisks but were not used for the analysis of the genetic distances within and between the oribatid mite species. The alignment of the amino acids and the nucleic acids for the hsp82 gene was also unambiguous since only a few gaps occurred and these were always 3 bp long, indicating the loss of single amino acids. No stop-codons were found in the hsp82 fragment when the sequences were translated (data not shown). All alignments are available from the corresponding author on request.

Phylogenetic trees were constructed using neighbour-joining (NJ), maximum parsimony (MP) and likelihood-based (ML) algorithms in PAUP* (Version 4b10; Swofford, 1999) treating gaps as missing data. Tree length and statistical indices are given for informative sites only (Posada & Crandall, 1998). Our settings for the ef-1α data corresponded to the TrNef model (Tamura and Nei, equal base frequencies) with gamma correction (α = 0.4667; Yang, 1996) and with substitution parameters A–C, A–T, C–G, G–T = 1.0; A–G = 2.9637; C–T = 3.1584. For the hsp82 data our settings corresponded to the TrN model (Tamura and Nei; base frequencies: A = 0.3788; C = 0.1772; G = 0.2743; T = 0.1696) with gamma correction (α = 0.8411) and with substitution parameters A–C, A–T, C–G, G–T = 1.0; A–G = 3.3943 and C–T = 6.979.

Reliability of the nodes was ascertained from 1000 bootstrap replicates. MP trees were constructed with heuristic search of 100 random additions, and the tree bisection-reconnection (TBR) branch-swapping algorithm with the option of collapsing zero branch length. A strict consensus tree was constructed for both ef-1α and the hsp82 sequence data.

Sequence divergence in sexual and parthenogenetic taxa

The average maximum intra-specific and intra-individual molecular divergence (=distances based on the evolutionary model obtained by MODELTEST; Posada & Crandall, 1998) in the ef-1α of the three sexual species was compared with those of the three parthenogenetic species by two separate one-way analyses of variance (anova) with the fixed factor ‘reproductive mode’ using SAS 8e (SAS Insitute Inc., Cary, USA).

McVean test for recombination and gene conversion

Likelihood permutation tests using the program LDhat Version 2.0 (http://www.stats.ox.ac.uk/mcvean/LDhat/) for recombination and gene conversion (McVean et al., 2002) were conducted with the two-allele model, no frequency cut-off of rare alleles, Watterson's theta estimates (θW) of population-scaled mutation rates per site and 10 000 permutations as parameters. Likelihood permutation tests were run with 101 grid points and different values of 4Ner until the value with ML as suggested by the program were obtained. Nonparametric permutation tests on measures of linkage disequilibrium and physical distance are calculated to provide values of PLKmax for recombination and PLKmax for gene conversion, which are new statistical parameters of this particular likelihood permutation test (McVean et al., 2002).

Results

The aligned ef-1α sequences were 573 bp and those of the hsp82 gene were 525 or 537 bp long (see Tables S1 and S2). Of the 573 aligned sites of ef-1α 283 were constant, 48 were variable but parsimony-uninformative and 242 characters were parsimony-informative. Of the 537 aligned sites of the hsp82 gene 360 were constant, 72 were variable but parsimony-uninformative and 105 were parsimony-informative.

Phylogeny and Meselson effect

For ef-1α (Fig. 1) the bootstrap replications of the NJ analysis clearly support the separation of the species but rarely separated individuals within the sexual or parthenogenetic species. The topologies of the MP and the ML tree for ef-1α were very similar to the NJ tree and are therefore not shown (both trees are available from the corresponding author on request). All species investigated were separated in the NJ tree. For hsp82 (Fig. 2) the bootstrapping supports the separation of the two species P. peltifer and S. magnus but within these two species few splits were supported by bootstrap values higher than 50. The MP and ML trees were nearly identical to the NJ tree (both trees are available from the corresponding author on request). Intra-specific and intra-individual molecular variation was generally low in each of the species. However, the average maximum intra-specific genetic divergence of ef-1α in the three sexual species (8.6%) was larger than that of the three parthenogenetic species (2.7%). Additionally, the average maximum intra-individual genetic divergence of the sexual species (6.0%) was larger than in the parthenogenetic species (0.9%). However, neither difference was statistically significant (anova: F1,4 = 3.28, P = 0.14 and F1,4 = 3.81, P = 0.12, respectively). The average maximum intra-individual genetic divergence of the hsp82 gene was 1.1% in the sexual species S. magnus and 1.2% in the parthenogenetic species P. peltifer.

Figure 1.

Neighbour joining tree of several clones of three parthenogenetic species and three sexual species of oribatid mites and the parthenogenetic out-group species H. rufulus. The tree was constructed using the 573 bp alignment of the partial sequence of the elongation factor 1α gene. Numbers at the nodes represent supporting percentages of 1000 bootstrap replicates (only values above 50% are reported). For abbreviations see Table S1.

Figure 2.

Neighbour joining tree of several clones of the parthenogenetic oribatid mite species P. peltifer, the sexual species S. magnus and the parthenogenetic species T. minor. The tree was constructed using the 537 bp alignment of the partial sequence of the heat shock protein 82 gene. Numbers at the nodes represent supporting percentages of 1000 bootstrap replicates (only values above 50% are reported). For abbreviations see Table S2.

Results of the likelihood permutation tests indicate that recombination occurs in ef-1α of all three sexual species (Table S3). In addition, there is also evidence for two of the three sexual species (E. plicatus and S. magnus) that gene conversion takes place in ef-1α. None of the parthenogenetic species showed any evidence for recombination or gene conversion. For hsp 82, likelihood permutation tests provided neither for the sexual species S. magnus nor the parthenogenetic species P. peltifer any evidence for recombination or gene conversion.

Discussion

Absence of the Meselson effect

Results of this study provide no evidence for the Meselson effect in parthenogenetic oribatid mites. Intra-individual molecular variation in both the ef-1α and the hsp82 genes in parthenogenetic species was very low and similar to that of sexual species. In the phylogenetic tree the DNA sequences of the parthenogenetic (and also of the sexual) species formed distinct clusters suggesting that neither the ef-1α nor the hsp82 alleles evolved independently, i.e. without homogenising mechanisms, over a long period of time. Additionally, the results of the RT-PCR supported the hypothesis that the hsp82 gene is transcribed; indicating that at least one of the two alleles did not deteriorate during evolution. Therefore, the studied parthenogenetic oribatid mite species do not appear to be ancient apomicts, which is consistent with the cytological evidence of Taberly (1987a,b), who found P. peltifer and another parthenogenetic oribatid species to be automictic. Due to the asymmetric power of the Meselson effect (Butlin, 2000) it is difficult to explain its absence in parthenogenetic oribatid mites, but a number of mechanisms that may have contributed to the homogenisation of the genome should be considered.

Several mechanisms can be dismissed as unlikely. Rare (spanandric) males are known from many parthenogenetic oribatid mite species, including those we studied (Palmer & Norton, 1991); if these males are functional, then rare sexual reproduction could explain the similarity of intra-individual genetic divergence in sexual and putative parthenogenetic species. However, our results show no evidence for recombination in the parthenogenetic species studied, so spanandric males appear to be non-functional, as was suggested by several previous lines of evidence. For example, spanandric males that have been studied cytologically were functionally incompetent and were ignored by females (Taberly, 1988). Furthermore, Palmer & Norton (1991) found the population sex ratios of 136 species of Desmonomata (including our studied parthenogenetic species) to be either approximately equal (50 : 50) or strongly female biased (about 95% females). Such a bimodal sex ratio distribution also eliminates frequent hybridisation as a possible contributor to the low intra-individual genetic variation. An allozyme study of a number of parthenogenetic Desmonomata (Palmer & Norton, 1992) suggested clonal population structure and examples of fixed heterozygosity despite the presence of rare males, and the population with the most males (6%; Mucronothrus nasalis) had nearly the lowest genotypic diversity. In a general context, spanandric males are found in many obligate parthenogenetic animals (Lynch, 1984) and, under realistic assumptions of heterozygote superiority or epistatic interactions between loci, rare sexual reproduction in an otherwise asexual population may actually be disadvantageous (Peck & Waxman, 2000).

Two possible explanations relate to the age of taxa. One is that very limited differences between alleles might result from a parthenogenetic oribatid species being recently split from a sexual ancestor. This idea is unlikely to have general explanatory power, since some taxa of oribatid mites (e.g. various clades in Desmonomata) are species-rich, yet include no sexual species. A contention that these taxa radiated while being parthenogenetic is much more parsimonious than one in which parthenogenesis evolved frequently and independently in these lineages while all sexual relatives have become extinct is highly unlikely (Norton et al., 1993; Maraun et al., 2004). A second idea is that if oribatid mites, as a group, radiated relatively recently, low intra-individual genetic divergence could still characterize both sexual and parthenogenetic taxa. This also is unlikely since oribatid mites existed at least since the Devonian; about 380 My ago (Shear et al., 1984; Norton et al., 1988), and rich fossil evidence shows that the main radiations are quite old as well (Labandeira et al., 1997). Biogeographic evidence even suggests that one parthenogenetic species (M. nasalis) predates the breakup of Pangea (Hammer & Wallwork, 1979). Estimates of the age of sexual and parthenogenetic oribatid mites using molecular markers that are thought to have clock-like divergence (e.g. COI, 18S rDNA) are needed to better refute this idea.

It has been proposed that parthenogenetic taxa might possess efficient mechanisms of DNA repair (Schön & Martens, 1998), which contribute to keeping alleles homogeneous. Together with strong selection pressure this might explain the limited genetic variability of the ef-1α and hsp82 genes. However, neither DNA repair nor strong selection pressure explain why even the third codon position changed little.

Other explanations for low allelic divergence might also be relevant here but await confirmation. (1) Those parthenogenetic oribatid mites that have been studied cytologically are automicts. Automixis can homogenize the genome by terminal fusion of (meiotically produced) oocytes with the second polar body. While oogenesis has been well studied in only two parthenogenetic oribatid mite species, our study species P. peltifer is one of them (Taberly, 1987a). We assume that automictic reproduction in our studied parthenogenetic oribatid mites contributed to their low genetic divergence. (2) Gene conversion also is an effective homogenising mechanism that can occur on nonhomologous pairs of chromosomes (Carpenter, 1994). The strongest evidence for gene conversion comes from gene families or multi-copy regions, such as rRNA clusters and the ITS region (Benevolenskaya et al., 1997; Butlin et al., 1998; Fuertes Aguilar et al., 1999), but it probably occurs everywhere in the genome (e.g. Bertran et al., 1997; Haubold et al., 2002). Gene conversion in the ancient asexual ostracod D. stevensoni has been restricted to the ITS1 region only (Schön & Martens, 2003) and might also act in bdelloid rotifers (David Mark Welch, personal communication). In the current study, the likelihood permutation tests provided only evidence for gene conversion in the ef-1α gene of sexually reproducing but not of parthenogenetic oribatid mites. Thus, it seems unlikely that gene conversion, at least in the screened genomic regions, caused the observed, low genetic divergence of parthenogenetic taxa. (3) Mitotic recombination and aneuploidy are other mechanisms that can potentially contribute to the homogenisation of genomes (Birky, 2004). The importance of these mechanisms for limiting intra-individual genetic diversity should be explored in more detail.

A puzzling result of this study is that parthenogenetic oribatid mites might rather be ancient automicts instead of apomicts but no evidence for recombination could be found. We therefore assume that the homogenisation of the genome by automixis has been ongoing for long periods of time. Consequently, any effects of recombination have been superimposed and cannot be detected anymore in parthenogenetic oribatid mites. However, this explanation does not solve the enigma why different lineages of parthenogenetic species have similar DNA sequences. Over evolutionary long periods of time, at least some mutations should have accumulated.

In conclusion, it seems that the majority of ancient asexual animal taxa might posses some kind of homogenizing mechanisms detaining their genomes from infinitely accumulating mutations (Schön & Martens, 2002). The main mechanisms suggested to date are automixis, gene conversion and highly efficient DNA repair. This imposes that the Meselson effect might not exist at all since it has not been confirmed for any putative ancient asexual taxon (Tramini: Normark, 1999; D. stevensoni: Schön & Martens, 2003; Giardia lamblia: Baruch et al., 1996; oribatid mites: this study), and even the presumed high allelic diversity in bdelloid rotifers may be the result of an ancient hybridisation effect between two taxa (Mark Welch et al., 2004, M. Meselson, personal communication). Then the very interesting question remains why not more taxa are ancient asexuals if the long-term disadvantages of parthenogenetic reproduction can be defeated.

Acknowledgments

We thank an anonymous referee and Koen Martens for providing valuable comments on this manuscript. We also thank Michael Laumann and Sonja Migge for commenting on earlier drafts. The first and the second author equally contributed to this work. The study was supported by a European Community Marie Curie Fellowship and by the German Research Foundation (DFG, SPP 1127, Ma 2461/1-2).

Ancillary