Introduction

The cosmopolitan terrestrial isopod Porcellionides (=Metoponorthus) pruinosus has a confused taxonomic history (Vandel, 1962). This species is highly synanthropic and is considered to be the most widely distributed species of terrestrial isopods (Vandel, 1962; Garthwaite & Sassaman, 1985). Originally native to Asia Minor (Vandel, 1962) and carried extensively elsewhere by humans, P. pruinosus has colonized the entire world except the polar regions. It presents important geographical variation and is considered to be a polytypic species. Based on morphological criteria about 20 subspecies were recognized throughout the world (Vandel, 1962) although the validity of these subspecies remains questionable (Böhme, 1978). More recently several observations supported the presence of distinct species (Garthwaite & Sassaman, 1985; Juchault et al., 1985; Marcadé et al., 1999).

As far as is known, all the studied populations of this species complex harbour Wolbachia bacteria. Wolbachia are intracytoplasmic alpha-proteobacteria which are maternally inherited and can alter the reproduction of their hosts. In the case of P. pruinosus, they are believed to induce the feminization of genetic males (Rigaud et al., 1997; Marcadé et al., 1999). Two different patterns of infection seem to be present in P. pruinosus populations, named type 1 and type 2 in the following text. Type 1 was found in all the French host populations examined: Wolbachia were present both in males and females, with a 90% prevalence (Rigaud et al., 1997; Marcadé et al., 1999). Type 2 was found in Greece, Tunisia and Réunion Island host populations: Wolbachia were detected in females only, and the prevalence was only 60% (Marcadé et al., 1999). These two groups of host populations and their associated patterns of infection could be well separated by mitochondrial RFLP analysis. On the other hand, two Wolbachia strains belonging to the B super group of Wolbachiae were found in P. pruinosus (Marcadé et al., 1999; Cordaux et al., 2001). These results have led the authors to hypothesize a possible correlation between Wolbachia lineages inducing different phenotypes, and differentiated populations in the P. pruinosus group (Marcadé et al., 1999). However, the high genetic divergence recorded between the two Wolbachia strains involved suggested possible horizontal transmission rather than common origin.

The aims of the present study were (1) to check the genetic differentiation of both hosts and parasites on a larger scale; (2) to determine the infection patterns in the host populations; and (3) to test the hypothesis of coevolution of P. pruinosus and Wolbachia as previously suggested by Marcadé et al. (1999). This was accomplished by analysing the sequences of mitochondrial LSU rDNA host genes, shown to be an adequate marker for this taxonomic level (Michel-Salzat & Bouchon, 2000) and to be more reliable than RFLP analysis. At the same time, a molecular analysis of the Wolbachia strains found in the host populations was carried out by sequencing the wsp gene, a rapidly evolving gene encoding a Wolbachia surface protein (Braig et al., 1998). On the basis of these two analyses, the correlation between the infection pattern and the Wolbachia strains is then examined and discussed.

Materials and methods

Sampling and Wolbachia detection

Nine populations from Tunisia (Tunis), Spain (Manzanares), Réunion Island (Saint Paul), Israel (Haifa), Greece (Athens) and France (Nevers, Montamisé, Celles-sur-Belle, Camarade) were studied (see Table 1). Camarade, Manzanares and Haifa were new collecting sites, whereas additional sampling was studied in populations previously investigated (i.e. Nevers and Celles-sur-Belle, Rigaud et al., 1997; Athens, Tunis and Saint-Paul, Marcadé et al., 1999). Total DNA was obtained by standard phenol extraction and ethanol precipitation (Kocher et al., 1989) from muscles, nerve tissue and gonads of each individual. PCR assays were performed to test for the presence of Wolbachia, using two specific primer sets for the 16S rDNA and wsp genes (O’Neill et al., 1992; Zhou et al., 1998). Positive and negative controls using known infected and uninfected isopods (Armadillidium vulgare, laboratory strains) were performed with each set of PCR reactions.

Table 1 Origin and number of Porcellionides pruinosus sampled, presence of Wolbachia (the number testing positive is in parentheses) and accession numbers in EMBL, GenBank and DDBJ databases for host and Wolbachia genes
Full size table

Sequencing

Individuals testing positive for Wolbachia were used for both host mitochondrial and Wolbachia gene sequencing. The 3′-half end of the LSU rDNA region of the host’s mitochondrial DNA was amplified using the specific primers 16Sar and 16Sbr (Xiong & Kocher, 1991). PCR was conducted as previously described (Michel-Salzat & Bouchon, 2000). The wsp region of the Wolbachia DNA was amplified using the specific primers wsp81F and wsp691R (Braig et al., 1998).

After purification, the double-stranded PCR products were sequenced on both strands with the amplification primer set, using the ABI Prism BigDye terminator cycle sequencing kit (PE Applied Biosystems, Foster City). Sequences were analysed on a ABI 310 automated sequencer. For each specimen, sequences were obtained independently at least twice from different PCR products.

Mitochondrial sequences were deposited in EMBL, GenBank and DDBJ databases under accession numbers AJ275207 to AJ275212 (Table 1). Two sequences used in analyses were previously obtained from the Celles-sur-Belle and Saint Paul populations (AJ388107 and AJ388108, Michel-Salzat & Bouchon, 2000). Wsp sequences were deposited in EMBL, GenBank and DDBJ databases under accession numbers AJ300579 to AJ300585 (Table 1). Two sequences were previously obtained from the Montamisé and Saint Paul populations (AJ276604 and AJ276605, Cordaux et al., 2001).

Alignment and phylogenetic inferences

LSU rDNA sequences from P. pruinosus were manually aligned with the sequence of Porcellionides cingendus (no. AJ388106), previously shown to be an appropriate outgroup (Michel-Salzat & Bouchon, 2000). Wsp sequences were manually aligned, taking into account the coding structure of the gene. The sequence of Dysdera erythrina symbionts (no. AJ276615) belonging to the A super group of Wolbachiae was added as outgroup (Cordaux et al., 2001). The corresponding data matrices (10 host sequences of 430 sites including gaps, 120 variable sites; 10 parasite sequences of 588 sites including gaps, 135 variable sites) were separately tested using the MODELTEST 3.0 software (Posada & Crandall, 1998) for a model of DNA substitution.

Phylogenetic trees were inferred using PAUP* 4.02B software (Swofford, 1998). Tree reconstructions were performed using maximum parsimony, maximum likelihood and distance methods using the General Time Reversible GTR model (Rodriguez et al., 1990). Bootstrap analyses were performed with 1000 replicates (Felsenstein, 1985). A Mantel test (Mantel, 1967) was performed with GENETIX 4.01 software (Belkhir, 2000) in order to estimate the correlation between the two distance matrices from parasite and host phylogenies (10 000 permutations).

Results

Infection status

All the populations were carrying Wolbachia. Infected males were only found in French and Spanish populations (i.e. Nevers, Celles-sur-Belle, Camarade and Manzanares) whereas no male was detected positive for Wolbachia in southern populations (i.e. Athens, Tunis and Saint-Paul; Table 1). The difference between these two infection patterns in males was significant (χ2=17.4, d.f.=1, P < 0.001). The presence of infected males in French and Spanish populations indicated that they corresponded to the type 1 infection. There is no difference in infection rates in males in these four populations (χ2=5.62, d.f.=3, P > 0.05). Southern populations with uninfected males corresponded to the type 2 infection. Assuming a 95% confidence limit, the maximum likelihood frequency of a positive male being present in these samples is therefore 10.8% (Post & Millest, 1991). Comparing the type of infection (type 1 vs. type 2) there was no difference in infection rate in females (χ2=1.48, d.f.=1, P > 0.05). In the same way, when males and females were pooled, there was no significant difference in prevalence among populations (χ2=11.99, d.f.=6, P > 0.05).

Host diversity

The alignment of LSU rDNA sequences from P. pruinosus revealed a first group of identical sequences that included French (Nevers, Montamisé, Celles-sur-Belle, Camarade) and Spanish (Manzanares) populations (Table 2). The second group of sequences corresponded to populations from Réunion Island (Saint Paul), Greece (Athens), and Tunisia (Tunis), with only one substitution between the Tunisian sequence and the others. The sequence obtained from the Israeli population (Haifa) exhibited a divergence of 22.5% with those of the second group and a divergence of 16.5% with those of the first group. The first and second groups differed by a mean divergence of 20.8% (Table 2).

Table 2 Matrix of total nucleotide differences of mitochondrial sequences (below diagonal) and Wolbachia sequences (above diagonal) for populations of Porcellionides pruinosus
Full size table

All the procedures used for phylogenetic reconstruction resulted in the same inferred tree (Fig. 1b). Two groups of populations were clearly separated, the first one comprising the Montamisé, Celles-sur-Belle, Manzanares, Camarade, Nevers and Haifa populations (group Pp A, for P. pruinosus haplotype A), the second one comprising the Tunis, Athens and Saint Paul populations (group PpB, for P. pruinosus haplotype B). Nodes were supported by high bootstrap values (>95%). The nucleotide divergence observed between the two mitochondrial lineages (20.8%) was similar to the distance separating all populations of P. pruinosus from the outgroup P. cingendus (from 22 to 24%). Assuming a molecular clock of about 2% of divergence per Myr (Simon et al., 1994; Ballard & Kreitman, 1995), the two haplotypes could have diverged 10 Myr ago. However, the variation in the molecular clock across taxa suggests taking this result with caution. A calibration of the molecular clock in Oniscidea, with biogeographical and palaeontological data, would be necessary to obtain a fully reliable estimation.

Fig. 1
figure 1

Phylogenetic relationships among Wolbachia strains and Porcellionides pruinosus haplotypes, showing distribution of infection patterns. Infection patterns: 1, type 1, Wolbachia in males; 2, type 2, no Wolbachia in males; *, type of infection unknown.Fig. 1(a) Maximum likelihood tree (−log likelihood=1674.84) obtained from the wsp sequences. wPruI, wPruII, wPruIII are Wolbachia strains of P. pruinosus (see text). Bootstrap values (1000 replicates) are shown above nodes.(b) Maximum likelihood tree (−log likelihood = 1382.77) obtained from partial sequences of mt LSU rDNA. PpA and Pp B are mitochondrial lineages of P. pruinosus (see text). Bootstrap values (1000 replicates) are shown above nodes.

Full size image

Wolbachia diversity

The alignment of the wsp DNA sequences revealed four different sequences corresponding to three groups (Table 2). The first group corresponded to Wolbachia isolated from Athens, Tunis, Haifa, St-Paul and Nevers P. pruinosus populations. Wolbachia from Tunis and Nevers exhibited the same wsp sequence and showed a divergence of 0.3% with the other strains of this group. Wolbachia from Camarade and Manzanares exhibited the same wsp sequence and showed a divergence of 4.8% with the first group (Table 2). The third group included Wolbachia from the Celles-sur-Belle and Montamisé populations which exhibited the same wsp sequence. This last group was highly divergent from the others (pairwise distance = 20.6%).

All the procedures used for phylogenetic reconstruction gave the same inferred tree (Fig. 1a). Three strains were clearly identified and supported by high bootstrap values. The first strain included the symbionts found in southern populations (Athens, Tunis and Saint-Paul) as well as the symbionts from Haifa and Nevers. This strain was called wPruI (for Wolbachia P. pruinosus I) according to the nomenclature of Zhou et al. (1998). The second strain, called wPruII, grouped the bacteria found in the Camarade and Manzanares populations. The last strain comprising the Montamisé and Celles-sur-Belle symbionts was called wPruIII.

Comparison between host mitochondrial lineages and bacterial strains

Comparison of P. pruinosus and Wolbachia relationships (Fig. 1) revealed that the PpB group of populations (St-Paul, Athens, and Tunis) was infected only by the wPruI strain of Wolbachia. The Pp A group of populations was found to carry the three different strains of Wolbachia. The Mantel test confirmed that genetic distances of hosts and symbionts were not correlated (r2=0.125, P=0.761), thus that the observed patterns of distribution can not be explained by co-speciation.

Discussion

We clearly identified three Wolbachia strains in the P. pruinosus complex. The comparison of the relationships of hosts and bacteria showed that the Wolbachia strains could not be strictly attributed to one or the other host lineages. The wPruI strain observed in the PpB group of host populations was also found in two samples of the PpA group (Haifa, Nevers) and host populations of the PpA group harboured the three different strains of Wolbachia. Furthermore, the distribution of Wolbachia strains can not be explained only by coevolution and probably involved several infestation events. The most parsimonious hypothesis requires two infestation events: an ancestral infection by the wPruI strain might have occurred, followed by a co-speciation leading to the wPruII strain. More recently, a second infection might have occurred by host switching of the distantly related wPruIII strain. This scenario is strengthened by previous analyses of Wolbachia phylogenetic relationships, which revealed that the wPruIII strain is closely related to insect Wolbachia strains and could result from recent interspecific transfers (Bouchon et al., 1998; Cordaux et al., 2001). On the other hand, the wPruI and wPruII strains were closely related, belonging to a woodlice-specialized Wolbachia clade (Cordaux et al., 2001; Bouchon et al., 1998).

Two different infection patterns were recorded which can not be related to variation in prevalence, but instead only in presence or absence in males. Infection patterns were not correlated with Wolbachia distribution (Fig. 1), since the same pattern was associated with the two very distantly related strains wPruII and wPruIII, whereas the two most related strains wPruI and wPruII were associated with two different infection patterns. This suggests that the pattern of infection is mainly mediated by the host and not by the Wolbachia strain. The best evidence appears in the wPruI clade even if we can not totally preclude a slight Wolbachia effect because of the low differentiation of wsp sequences between Nevers strain (pattern 1) and the southern ones (pattern 2). As the wPruII strain belongs to the feminizer clade of Wolbachia (Bouchon et al., 1998; Cordaux et al., 2001), host mediated pattern is also probable in the Manzanares and Camarade populations. The presence of the wPruIII strain in both sexes in the Celles-sur-Belle population could either be because of a host effect or because this Wolbachia strain has never been a feminizer. A host mediated pattern of infection has already been suggested in woodlice (Rigaud et al., 1999) and in insects (Fialho & Stevens, 2000). Because the feminizing process involves competition between a bacterial product and the male hormone (Juchault & Legrand, 1985), the two infection patterns could be due to differences in hormonal target. A high variability of the androgenic hormone was recently reported in the Oniscidea (Martin & Juchault, 1999). The host mediated patterns of infection shown in this paper could therefore be the result of a differentiation of the androgenic hormone in the two P. pruinosus sibling species. Moreover, host resistance against feminizing Wolbachia has been hypothesized in some cases and might have occurred in the P. pruinosus complex (Rigaud et al., 1997).

The molecular analysis of a mitochondrial gene clearly revealed the existence of two distinct groups of P. pruinosus populations (PpA vs. PpB) separated by the high genetic distance of 20.8%. When comparing the same portion of the mitochondrial LSU rDNA in crustaceans, a divergence of 0.5% is observed between two branchiopod species (Palmero et al., 1988; Perez et al., 1994), whereas a divergence of 11% is observed between two decapod species (Machado et al., 1993). Within the Oniscidea, the pairwise distances were 3% between two populations of the same species (Armadillidium vulgare) and 7.4% between two subspecies (Porcellio dilatatus dilatatus and P. d. petiti) (Michel-Salzat & Bouchon, 2000). The higher divergence value found here in P. pruinosus corresponds to those observed between species within a given genus (e.g. from 22 to 24% between P. cingendus and P. pruinosus populations, this study; 22% between two Armadillidium species, Michel-Salzat & Bouchon, 2000). This result supports the existence of two distinct species in the studied populations of P. pruinosus, which is congruent with Marcadé et al. (1999). As no clear morphological features have been found to separate the two species, they correspond to the definition of sibling species (Dobzhansky et al., 1968). Our molecular data suggest that the divergence between the two species may have occurred during the Miocene (10 Myr ago). As the genus Porcellionides originates from Asia Minor (Vandel, 1962) the interfertility of populations from Tunis and Réunion Island (Marcadé et al., 1999) suggests a recent introduction of specimens from the Mediterranean zone to Réunion Island. This may have occurred by means of natural expansion due to the facility of woodlice for rafting on floating vegetation (Taiti et al., 1992), or more probably by way of human migration (Schmalfuss & Ferrara, 1978). According to our data, the sibling species seems to be present in Israel and Western Europe. However, mainly due to the synanthropy of these animals, phylogeographic considerations on their actual distribution are difficult to clarify.

There is some empirical evidence for the role of Wolbachia in speciation in the Drosophila simulans complex (Hoffmann & Turelli, 1988), between Nasonia species (Breeuwer & Werren, 1990) and in the terrestrial isopod Porcellio dilatatus (Legrand & Juchault, 1986). Host speciation by Wolbachia-induced incompatibility has been recently demonstrated (Bordenstein et al., 2001) between two closely related species of Nasonia. As all the studied samples were carrying Wolbachia, the reproductive isolation in P. pruinosus populations (Marcadé et al., 1999) could therefore be related to Wolbachia effects on their hosts. However, by hitchhiking the mitochondria because they are associated cytoplasmic genomes, feminizing Wolbachia could promote the fixation of mitochondrial mutations (Werren, 1997). The only phenotype described in P. pruinosus up to date is feminization (Rigaud et al., 1997; Marcadé et al., 1999). Considering the distribution of Wolbachia strains in the two host species as well as the incongruent phylogenetic relationships of Wolbachia and their hosts (maybe due to recent Wolbachia horizontal transfer), we conclude that it is very unlikely that feminizing Wolbachia were involved in the differentiation of the two sibling species. Alternatively, we can suggest an ancestral CI-inducing Wolbachia acting in P. pruinosus speciation which could secondarily evolve to a feminizing phenotype. However this latter scenario seems highly hypothetical.