Effects of parasitic sex-ratio distorters on host genetic structure in the Armadillidium vulgareWolbachia association


F. Grandjean, Université de Poitiers, Ecologie, Evolution, Symbiose, UMR CNRS 6556, 40 avenue du Recteur Pineau, F-86022 Poitiers Cedex, France.
Tel.: +33 (0)5 49 45 42 76; fax: +33 (0)5 49 45 40 15;
e-mail: frederic.grandjean@univ-poitiers.fr


In the pill bug Armadillidium vulgare (Crustacea, Oniscidea), Wolbachia facilitates its spread through vertical transmission via the eggs by inducing feminization of genetic males. The spread of feminizing Wolbachia within and across populations is therefore expected to influence mitochondrial DNA (mtDNA) genetic structure by hitchhiking. To test this hypothesis, we analysed nuclear and mtDNA genetic structure, and Wolbachia prevalence in 13 populations of the pill bug host. Wolbachia prevalence (ranging from 0% to 100% of sampled females) was highly variable among populations. All three Wolbachia strains previously observed in A. vulgare were present (wVulC, wVulM and wVulP) with wVulC being the most prevalent (nine of 13 populations). The host showed a genetic structure on five microsatellite loci that is compatible with isolation by distance. The strong genetic structure observed on host mtDNA was correlated with Wolbachia prevalence: three mitotypes were in strong linkage disequilibrium with the three strains of Wolbachia. Neutrality tests showed that the mtDNA polymorphism is not neutral, and we thus suggest that this unusual pattern of mtDNA polymorphism found in A. vulgare was due to Wolbachia.


In population genetics, mitochondrial DNA (mtDNA) is widely used to study the phylogeographic history of a given species or for inferring female gene flow. mtDNA and nuclear genetic markers do not always lead to similar genetic structures because of their different patterns of inheritance (particularly in the case of sex-biased dispersal, e.g. Seielstad et al., 1998; Girman et al., 2001; Scribner et al., 2001) or because of the effects of natural selection (Bazin et al., 2006; Meiklejohn et al., 2007). Moreover, maternally inherited symbionts of arthropods (i.e. non-neutral cytoplasmic factors) can alter mtDNA genetic diversity and structure by hitchhiking (see Hurst & Jiggins, 2005 for review). Such a situation is found in Drosophila simulans (Ballard, 2000; Dean et al., 2003), in the woodlouse Porcellionides pruinosus (Marcadéet al., 1999; Michel-Salzat et al., 2001) and in several other arthropods (see Hurst & Jiggins, 2005 and references therein). In all cases, the rapid spread of the reproductive parasite Wolbachia dramatically reduces mtDNA polymorphism. Several theoretical studies also showed that Wolbachia has effects on nuclear genetic diversity (Telschow et al., 2002, 2006; Engelstädter et al., 2008; Kobayashi & Telschow, 2010; Kobayashi et al., 2011), although experimental validations are still missing. Particularly, feminizing Wolbachia are though to reduce the effective population size (and thus to increase genetic drift) and increase genetic influx in infected populations. Both effects cancel each other at neutral loci, and Wolbachia infection is thought to have only criptic effects on neutral allelic frequencies (Kobayashi et al., 2011). As Wolbachia is estimated to infect 20–73% of arthropods (Bouchon et al., 1998, 2008; Jeyaprakash & Hoy, 2000; Werren & Windsor, 2000), their possible effects must be considered when carrying out studies on host population genetics.

Many studies have focused on the effect of reproductive parasites such as Wolbachia on the population genetics of mtDNA of their hosts (e.g. Marcadéet al., 1999; Behura et al., 2001; Dewayne Shoemaker et al., 2003; Hurst & Jiggins, 2005). However, few of them clearly analysed the situation in which a polymorphism of infection occurs within populations. In such scenario, different parasites are associated with different mitotypes, all of which are expected to be hitchhiked by the parasites. As a consequence, the pattern of mtDNA polymorphism can be maintained and this pattern should differ significantly from the case of host species infected by a single parasite (Hurst & Jiggins, 2005). The terrestrial isopod Armadillidium vulgare exhibits a polymorphism of infection by feminizing Wolbachia (Cordaux et al., 2004; Verne et al., 2007), as well as a relatively high mtDNA polymorphism (Rigaud et al., 1999). Therefore, this species is a remarkable case for the study of the effects of Wolbachia on mtDNA polymorphism. In A. vulgare, Wolbachia inverts genetic males into functional females. Two feminizing Wolbachia strains (wVulC and wVulM), exhibiting different life history traits and consequently different evolutionary strategies, are well known in A. vulgare (Cordaux et al., 2004). A third strain, closely related to wVulC (wVulP, Verne et al., 2007), was recently discovered. Its phenotype is unknown but it is highly suspected to be feminizing as well. Another feminizing factor labelled f, a suspected Wolbachia genome fragment inserted into the host genome, has also been described in A. vulgare (Juchault & Mocquart, 1993; Cordaux et al., 2011). Although both f and Wolbachia can lead to female-biased sex ratios in the wild, they do not have identical patterns of transmission. Indeed, f is typically a non-Mendelian genetic element maternally transmitted. However, f undergoes a partial amount of paternal transmission when it is associated with a masculinizing gene (Rigaud, 1999). This masculinizing gene, named M, is able to restore male phenotype in the presence of the f element but not in the presence of Wolbachia (Rigaud & Juchault, 1993). In addition, a complex of resistance genes, named R genes, is able to reduce the vertical transmission rate of Wolbachia to the progeny without any effect on f (Rigaud & Juchault, 1992). As a result, true genetic females are rare in nature (Juchault et al., 1993) where populations are mainly composed of genetic males reversed into functional females when harbouring sex-ratio distorters (f or Wolbachia).

In A. vulgare, the theoretical models predict that feminizing Wolbachia should reach a stable prevalence equal to their transmission rates (i.e. around 70–80% of the whole population based on the transmission rates of Cordaux et al., 2004) and all the females should be infected by Wolbachia at equilibrium (Taylor, 1990; Hatcher, 2000). However, previous studies reveal that this prevalence is never reached (Juchault et al., 1993; Rigaud, 1999; Bouchon et al., 2008) and more complex models are therefore needed to explain the prevalence of Wolbachia in A. vulgare populations. Several nonexclusive hypotheses have been proposed to explain the low prevalence of Wolbachia. First, a high prevalence of Wolbachia could increase the risk of population extinction due to the rarity of males (Hatcher et al., 1999, 2000; Hatcher, 2000). Second, A. vulgare developed resistance mechanisms against Wolbachia that include a sexual selection against infected females (Moreau et al., 2001) and the control of the vertical transmission of Wolbachia by the host (R genes; Rigaud & Juchault, 1992). Third, the f element limits the spread of Wolbachia by competition (Grandjean et al., 1993; Juchault et al., 1993; Rigaud et al., 1999). Juchault et al. (1993) proposed a cyclic evolution of the system with successive selective sweep of the different female sex factors (Wolbachia, f, and W sex chromosome). If true, the cyclic raise and fall of the feminizing factors would lead by hitchhiking to recurrent decrease in the effective population size of mtDNA. In such a case, we can expect a particular pattern of mtDNA polymorphism, consisting of a few and highly divergent mitotypes (one per feminizing factor).

In this study, we propose to test the hypothesis that mtDNA is not a neutral marker in A. vulgare and reflects the dynamics of sexual factors (i.e. the different Wolbachia strains and f). To this end, we analysed the population genetic structure in 13 French populations of A. vulgare using one mtDNA locus and five microsatellite loci.

Materials and methods

Sample collection

Thirteen populations were sampled by hand along a transect of 97 km from Ensoulesse (north-east of Poitiers) to Villeneuve (south-west of Niort) in France (Fig. 1). Three of these (RBINord, RBISud and RBICentre) were sampled in a fully protected forest area, the ‘Réserve Biologique Intégrale de Chizé’, close to Chizé. A total of 400 individuals were collected (240 females and 160 males). Site sample sizes are summarized in Table 1. Departure from the 50% male percentage in the samples was tested using replicated goodness-of-fit tests as presented by Sokal & Rohlf (1995).

Figure 1.

 (a) Wolbachia prevalence in 13 populations of Armadillidium vulgare. One unifected intersex was observed in the Chizé population. Main city names are underlined. (b) Mitotype distribution in 13 populations of A. vulgare. Only females were genotyped.

Table 1.   Sex ratio and Wolbachia prevalence in studied populations. Sample size of collected individuals in the field (N), number of males and females, probability of goodness-of-fit test of biased percentage of males (G), and Wolbachia identification and prevalence in females are given.
LocalityN♂/♀P(G)Wolbachia prevalence (% among females)
wVulCwVulMwVulPAll strains
  1. *Significant results of the G test after Bonferroni correction for 13 simultaneous tests.

St Séverin3514/210.239066.7066.7
La Crèche3917/220.4264.59.1013.6
St Maixent416/35< 10−5*74.30074.3
Total400160/240 17.510.82.130.4

Molecular markers

Total genomic DNA was extracted as described in Michel-Salzat et al. (2001). Total genomic DNA concentrations were homogenized to 30 ng μL−1.

Wolbachia detection

Presence of Wolbachia was checked by PCR amplification of the wsp region (≈580 bp) in each individual using wsp 81F and wsp 691R primers, as described in Michel-Salzat et al. (2001). DNA concentrations were homogenized to 60 ng per PCR (final volume of 25 μL). The identification of Wolbachia strains was done by sequencing the wsp gene in all infected specimens as described in Cordaux et al. (2004). Sequences are deposited in GenBank under accession numbers DQ778095DQ778107.

Microsatellites markers

Five polymorphic microsatellite loci (Av1, Av2, Av5, Av6 and Av8;Verne et al., 2006) were used in this study to assess nuclear genetic variation in A. vulgare. All experimental procedures for microsatellite genotyping were strictly identical to those described in Verne et al. (2006). A total of 268 individuals (110 males and 158 females) were fully genotyped at the five loci with a minimum of 18 and a maximum of 24 individuals (depending on the number of specimens collected at each site) per population (Table 2).

Table 2.   Within population microsatellite and mtDNA polymorphism analysis. Microsatellites: sample size (Nmicro.), estimated mean number of alleles after rarefaction (NA), observed heterozygosity (Ho), FIS value. mtDNA, sample size (NmtDNA), mitotype number (M), corrected gene diversity (GNei);, nucleotide diversity (πn); statistic of Fu’s test (Fu, 1997).
PopulationMicrosatellite polymorphismmtDNA polymorphism
Nmicro.NAHoFISNmtDNAMGNeiπnFu’s Fs
  1. mtDNA, mitochondrial DNA.

  2. *,**Significant values (respectively at 5% and 1% probability level) after sequential Bonferroni corrections for 13 simultaneous tests.

Saint Séverin238.580.720.0861430.380.01228.44*
La Crèche207.370.650.0841530.530.01269.10*
Saint Maixent247.850.69−0.0031520.130.00466.19*

mtDNA marker

In all, 169 females were genotyped. Five to 16 females (mean = 13.2 females per population) were randomly chosen in each population, and part of their Cytochrome c oxidase subunit II (COX2) mtDNA gene was sequenced. After PCR amplification with primers 5f4 (5′-GCACCACATCACGCCATCATT) and 5r3 (5′-GCTTTACCCTCACTTCGGCTT) (Marcadéet al., 2007), a 610–615-bp fragment of mtDNA comprising the 3′-part of COX2 gene (354 bp) and its 3′ adjacent region (256–261 bp) was sequenced using the internal primer CO2dir (5′-GAGAGYCCYATCCTTGAAGAAG) and the 5r3 reverse primer. DNA sequences were manually aligned with Bioedit (Hall, 1999). All sequences were deposited in GenBank under accession numbers EF519916EF519952.

Within population diversity


The allelic richness, which corrects the observed number of alleles for differences in sample size, was computed by the rarefaction method using HP-Rare (Kalinowski, 2004, 2005). The observed heterozygosity (HO) and unbiased expected heterozygosity (HE) (Nei, 1978) were computed with GENETIX 4.05 (Belkhir et al., 2004). Differences between populations in allelic richness and HE were assessed using Friedman’s test. Deviations from Hardy–Weinberg equilibrium were tested using the exact probability test of Guo & Thompson (1992) available in genepop 3.2a (Raymond & Rousset, 1995). For each population, significance levels were calculated at each locus and over all loci. Genotypic linkage disequilibrium between each pair of loci was estimated by Fisher’s exact tests. Tests for both deviations from Hardy–Weinberg equilibrium and linkage disequilibrium were performed using Markov chains (10 000 dememorization steps, 100 batches, 5000 iterations per batch).

Mitochondrial DNA

Mitotype number and Nei’s gene diversity corrected for sample size (which is equivalent to expected heterozygosity of microsatellite loci; Nei, 1987) were calculated with genodive software (Meirmans & Van Tienderen, 2004). Nucleotide diversity (πn), defined as the average number of pairwise nucleotide differences per site (Nei & Li, 1979), was calculated using the program DnaSP (Rozas & Rozas, 1999). Construction of the phylogenetic network was performed by TCS version 1.21 (Clement et al., 2000), as described in Templeton et al. (1992). Linkage disequilibrium between mtDNA and Wolbachia infection status was measured using D′ statistic (Lewontin, 1964). Exact test of linkage disequilibrium was performed across the entire sample (n = 171 genotypes) using Arlequin software version 3.11 (Excoffier et al., 2005) with default parameters (50 000 dememorization steps, Markov chain of 100 000 steps).

Neutrality tests

Genetic diversity can be influenced by several factors; among them, selection pressure should not affect microsatellites diversity and Wolbachia is thought to have only criptic effects on microsatellite allelic frequencies (Kobayashi et al., 2011), whereas drastic reductions in population size should affect both mitochondrial and nuclear genetic diversities. To test the effects of these factors, neutrality tests were performed both on mtDNA and microsatellites.


Population bottlenecks were tested using two methods implemented in Bottleneck software v1.2.02 (Piry et al., 1999): the ‘sign test’ (Cornuet & Luikart, 1996) and the ‘Wilcoxon sign-ranks test’ (Luikart & Cornuet, 1998). Three mutation models were tested: Infinite Allele Model (IAM), one-step Stepwise Mutation Model and Two Phase Model with 95% single-step mutations and 5% multiple-step mutations.

Mitochondrial DNA

For each sampled population, deviations from neutrality were tested according to Fu’s Fs test (Fu, 1997) using NeutralityTest software kindly provided by Dr Yun-Xin Fu. Although it was primarily designed to detect an excess of haplotypes given the number of mutations (Fu, 1997), we used it here to detect a deficit of haplotypes.

Genetic structure

Genetic differentiation

Genetic differentiation between populations was estimated using pairwise FST estimates according to Weir & Cockerham (1984) as implemented in the genetix program 4.05 (Belkhir et al., 2004). Due to a limited number of both loci and individuals, FST was preferred over RST as a measure of genetic differentiation (Gaggiotti et al., 1999; Rossiter et al., 2000). Significance of pairwise FST was tested with GENETIX’s permutation procedure using 10 000 permutations. Estimates of mtDNA differentiation between pairs of populations were calculated by taking mitotype divergence into account (ϕST) or not (FST) and tested with a permutation procedure (10 000 permutations) using Arlequin software (Excoffier et al., 2005). Wolbachia prevalence distribution was analysed and tested in the same way as mtDNA, using FST based on the infection status of the individuals. amova was successively performed with Arlequin software (Excoffier et al., 2005) on microsatellite data, mtDNA and Wolbachia genetic structures. As results were highly similar whether mitotype divergence was taken into account or not, only the results that take mitotype divergence into account are shown.

Isolation by distance

Mantel tests and partial Mantel tests were performed to discriminate between the effects of isolation by distance and the effects of Wolbachia on microsatellite and mtDNA genetic structures. FST and ϕST were respectively transformed into FST/(1−FST) and ϕST/(1−ϕST) according to Rousset (1997). The program fstat (Goudet, 1995, 2001) was used to estimate the correlation among the matrices of genetic distances based on mitochondrial markers, microsatellite loci, Wolbachia prevalence and the matrix of log-transformed geographic distances.

Bonferroni corrections for multiple tests were applied whereever needed during the statistical analyses.


Within locality diversity

Observed percentage of males ranged from 15% to 74% in the samples (Table 1 and Fig. 1a). Replicated goodness-of-fit tests indicated a percentage of males significantly different from 50% for the whole sample (N = 400; GT, 1 = 16.11; P = 6 × 10−5). A significant heterogeneity is found among the 13 samples (GH, 12 = 35.25; P = 0.0004). After Bonferroni corrections, only one female-biased sex ratio (15% of males; St Maixent population) was observed in the samples. However, when the St Maixent population is removed, the sex ratio of the whole sample (N = 359) remains significantly female biased (42.9% of males; GT, 1 = 7.27; P = 0.007). One uninfected female intersex was found in the population of Chizé. As Wolbachia was not found in this population (Table 1), such intersex individuals can be the result of incomplete feminization of a genetic male by the f element (Legrand & Juchault, 1969).

Wolbachia diversity

Wolbachia were found to be present in 11 of the 13 sampled populations. Prevalence was highly variable ranging from 0% to 100% of sampled females (Table 1 and Fig. 1a). Only four of the Wolbachia infected populations exhibited a prevalence that was higher than 50% in sampled females, and only one of these four populations showed a significant female-biased sex ratio (St Maixent). The three Wolbachia strains previously observed in A. vulgare (wVulC, wVulM and wVulP; Cordaux et al., 2004; Verne et al., 2007) were present, and wsp gene analysis revealed that sequences were identical to those previously published. wVulC was the most prevalent (nine of 13 populations), followed by wVulM (six of 13 populations) and finally wVulP (two populations). One population hosted all three strains, five hosted two strains and another five hosted only one of the three strains. No multiple infection was found within a single individual.

Microsatellite diversity

Microsatellite genetic diversity was high in all populations and exhibited low variability among populations (Table 2). Allelic richness and expected heterozygosity ranged respectively from 6.88 to 8.77 and from 0.68 to 0.80. No significant difference in expected heterozygosity and allelic richness was found between populations (Friedman’s Test, respectively P = 0.0668 and P = 0.1553). No significant departure from Hardy–Weinberg equilibrium was detected. However, Fis values were globally positive (only one negative value in the Coulombiers population) and may suggest the existence of null alleles at low frequencies in the majority of the populations. No significant linkage disequilibrium between loci was detected.

mtDNA diversity

Based on COX2 sequences, eight mitotypes were found in the sampled populations of A. vulgare (Figs 1b and 2) with a divergence ranging from 0.16% to 6.15% (mean 3.78%). Network analysis showed the existence of four groups of highly divergent mitotypes: (AvA-AvB-AvD); (AvE); (AvF-AvG) and (AvH-AvI). Mitotype diversity, however, was low and not equally distributed among populations (Table 2). Nei’s gene diversity was highly variable, ranging from 0.13 to 0.76. The nucleotide diversity (πn) was also variable, ranging from 0.005 (St Maixent) to 0.029 (Coulombiers).

Figure 2.

 Phylogenetic network of eight mitotypes of Armadillidium vulgare, inferred from the cytochrome oxidase II region (610–615 bp). Pie size is proportionnal to the number of individuals. Frequencies of associated infection status are shown on bar plots.

Wolbachia–mtDNA linkage disequilibrium

Significant linkage disequilibrium was found between Wolbachia infection status and mtDNA (exact test of linkage disequilibrium, P < 10−5). Distribution of mitotypes according to the sampling location and the infection status of females are shown in Fig. 2. wVulP is associated with the AvB mitotype only (D′ = 1), whereas wVulC and wVulM are both associated with several mitotypes. However, both strains are mainly associated with one mitotype: 18 of 21 wVulM-infected females had an AvA mitotype (D′ = 0.82) and 30 of 32 wVulC-infected females had an AvD mitotype (D′ = 0.91). All mitotypes except AvB were found in uninfected females.

Neutrality tests and bottleneck detection


Results of the Wilcoxon tests show a significant excess of gene diversity under the IAM model in four populations only (RBINord, RBICentre, RBISud and Chizé). Sign tests were never significant. As a consequence, there is no clear evidence of demographic bottlenecks in the sampled populations.


According to Fu’s Fs test after Bonferroni corrections, all the studied populations showed a significant mitotype number deficit given the observed number of mutations (Table 2, Fig. 2). This means that the observed mitotypes diverged a long time ago and are not closely related.

Population differentiation

Wolbachia prevalence structure

F-statistics revealed a highly significant geographic structure of Wolbachia strains prevalence. Pairwise FST values ranged from −0.041 to 0.642. Eighteen of the 78 pairwise FST values were significant after sequential Bonferroni corrections at the 5% level (38 significant values before Bonferroni corrections). Ten and 5 of the 18 significant pairwise FST values concern the samples of St Maixent and St Séverin, respectively, both populations exhibiting the highest prevalence of one Wolbachia strain (wVulM and wVulC with a prevalence of 0.74 and of 0.67, respectively). amova performed on Wolbachia prevalence showed that the infection status explained 30.85% of genetic variance among populations (Table 3).

Table 3.   Analysis of molecular variance results. Four amova was conducted on allele frequencies for microsatellite data, infection status, mitotype frequencies and the combination of infection status and mitotype frequencies. One amova was conducted on mitochondrial DNA with pairwise distances between mitotypes. ϕST and FST were calculated respectively with and without taking into account allele/mitotype distances. P indicates the probability that estimates of ϕST and FST do not differ from zero, 10 000 permutations.
 Source of variationd.f.Sum of squaresVariance componentsPercentage of variation 
MicrosatellitesAmong populations1247.670.0532.85FST = 0.028
P < 0.0005
Within populations519938.071.80797.15
Infection statusAmong populations1214.970.08130.85FST = 0.308
P < 0.0005
Within populations15828.840.18369.15
ADN mitochondrial (haplotype frequencies)Among populations1228.870.16642.19FST = 0.421
P < 0.0005
Within populations15836.040.22957.81


A significant but low microsatellite genetic structure was found. Pairwise FST values ranged from −0.004 to 0.083. Thirty-five of 78 FST values were significant after Bonferroni correction. amova based on microsatellite loci supports this low but significant structure with a global FST = 0.029 (P < 0.0005; Table 3).


Mitochondrial DNA genetic structure is high, based on both mitotype frequencies and mitotype divergence. Pairwise ϕST ranged respectively from −0.108 to 0.830. Thirty-five of the 78 pairwise ϕST are significant after Bonferroni corrections. amova based on mitotype frequencies and mitotype divergence gave similar results, with FST = 0.421 (P < 0.0005) and ϕST = 0.453 (P < 0.0005), respectively (Table 3). The three most abundant mitotypes (AvA, AvD and AvF) are present in most of the populations, although four of the five remaining mitotypes (AvB, AvE, AvG, AvH and AvI) are present in only one to two populations. AvE is present in four populations.

Isolation by distance

Simple and partial Mantel tests were used to identify factors that can influence genetic structure of the analysed markers (Wolbachia, mtDNA, nuclear DNA). Results are summarized in Table 4 and Fig. 3. Simple Mantel tests showed a significant correlation between microsatellite structure and geographic distance (r² = 0.308) compatible with an isolation by distance. Kobayashi et al. (2011) showed that the effects of Wolbachia on genetic drift and genetic influx cancel each other and thus remain criptic. As a consequence, pairwise FST based on microsatellite loci are assumed to represent gene flows between populations and are used to test the gene flow impact on mtDNA and Wolbachia prevalence structure. No clear evidence of either gene flow or geographic distance effect on mtDNA structure was found. However, a strong correlation between Wolbachia prevalence structure and mtDNA structure (either based on FST or ϕST) was shown. Therefore, isolation by distance was observed for the microsatellite data but not for the mtDNA.

Table 4.   Simple and partial Mantel test results. (Partial) regression coefficient β, associated probability P(β) and explained variances (r²) are shown for each test. Gene flow corresponds to the FST/(1−FST) distance based on microsatellite loci. Wolbachia corresponds to the FST/(1−FST) distance based on Wolbachia prevalence.
Dependent variableExplanatory variableControlled variableβP(β)r²
  1. mtDNA, mitochondrial DNA.

Microsatellites (gene flow)Geographic distance 0.024< 5 × 10−50.308
Wolbachia 0.0090.5000.100
Geographic distanceWolbachia0.0220.0100.346
WolbachiaGeographic distance0.0060.084 
mtDNA (only mitotype frequencies)Geographic distance 0.4920.0340.058
Wolbachia 1.275< 5 × 10−50.859
Gene flow 10.1870.0520.048
Geographic distanceWolbachia0.0650.7950.860
WolbachiaGeographic distance1.265< 5 × 10−5 
Gene flowWolbachia−3.8810.4860.865
WolbachiaGene flow1.311< 5 × 10−5 
WolbachiaGeographic distance 0.3370.0430.051
Figure 3.

 Plots of genetic structure indices based on Wolbachia prevalence, cytochrome oxidase II and microsatellite loci against log-transformed geographic distances. Open circles and closed circles represent nonsignificant and significant pairwise comparisons after sequencial Bonferroni corrections for 78 comparisons, respectively. Lines represent regression of genetic differenciation on logarithm of geographic distance through all pairwise comparisons.


Low prevalence of Wolbachia and polymorphism of infection

In terms of Wolbachia incidence, 11 populations of 13 harboured Wolbachia with a prevalence ranging from 0% to 100% in females. This incidence is the highest recorded to date in A. vulgare populations: Juchault et al. (1993) showed only six Wolbachia-positive populations of 31, whereas Rigaud et al. (1999) showed three of 10 and Cordaux et al. (2004) showed 16 of 20. The increasing Wolbachia incidence over the years in these publications probably does not indicate that Wolbachia is still spreading in A. vulgare. The sampled populations are not the same in the different studies, leading comparisons difficult. Moreover, Wolbachia detection methods evolved across time (physiological tests in the first study, PCR based tests using different genetic markers in the latest ones) and as a consequence, it more probably reflects the increased efficiency of detection methods.

wVulC was the most prevalent strain, being present in nine of 13 populations and representing 58% of the infected individuals. wVulM was found in six of the 13 populations (36% of the infected individuals). Finally, wVulP was found in only two populations (Poitiers and Ensoulesse, 7% of infected individuals). These results are similar to those of Cordaux et al. (2004) who observed that wVulC was the dominant strain (12 of 20 populations) whereas wVulM was less prevalent (four of 20 populations). To date, wVulP has only been observed in the Poitiers area (Verne et al., 2007), suggesting that this variant has a much more restricted distribution than wVulC and wVulM. All five individuals harbouring the AvB mitotypes are also harbouring wVulP. As no male was found harbouring this mitotype, it might suggest a high transmission rate.

A strong linkage disequilibrium was found between Wolbachia strains and some mitotypes. wVulM, wVulP and wVulC are predominantly associated with the AvA, AvB and AvD mitotypes, respectively. The observed linkage disequilibrium is expected due to the maternal inheritance of both Wolbachia and mitochondria. However, a few other associations were found, such as wVulC-AvA and wVulM-AvG. Previous studies have shown that haemolymph contacts and parasitism are possible routes for horizontal transmission of Wolbachia (Rigaud & Juchault, 1995; Cordaux et al., 2001). The association of some mitotypes (e.g. AvA) each with several infection status suggests that horizontal transfers occured recently and that associations are lost either by genetic drift or by selection (i.e. because of a lower fitness of the association). Unfortunately, no information is available on the fitness of the different associations.

According to minimum models of dynamics of feminization, the coexistence of several feminizers within a single population is unstable at equilibrium (Ironside et al., 2003). The most competitive strain should reach a prevalence equal to its transmission rate within the whole population and all the females should be infected at equilibrium (Taylor, 1990; Hatcher, 2000; Ironside et al., 2003). However, the presence of two Wolbachia strains in the same sample was relatively common (five of 13 populations) and the simultaneous presence of three strains was observed in the population of Poitiers. The prevalence in females within any given population was low compared to the observed transmission rates of wVulC and wVulM (Cordaux et al., 2004), except in the population of Poitiers, where the three Wolbachia strains were found. All these prevalence and incidence data suggest that the Wolbachia transmission rates alone cannot explain the observed prevalence of Wolbachia in A. vulgare (Taylor, 1990; Hatcher, 2000). A number of nonexclusive hypotheses have already been proposed in the literature to explain the low prevalence of Wolbachia: sexual selection against infected females (Moreau et al., 2001), competition with the f element (Grandjean et al., 1993; Juchault et al., 1993; Rigaud et al., 1999; Bouchon et al., 2008) and control of the vertical transmission of Wolbachia by the host (Rigaud & Juchault, 1992).

Dannowski et al. (2009) showed that sibmating can reduce male-killing bacteria prevalence. Their computer simulations also showed that, depending on male mating capacity, a stable coexistence of two strains of male-killing bacteria is possible if sibmating occurs but is below a threshold. In A. vulgare, Moreau & Rigaud (2003) demonstrated a high male mating capacity. Moreau et al. (2001) also found that males prefer copulate and interact with uninfected genetic females. In their experiments, they used males whose mothers were uninfected genetic females (WZ). The preferences of the sons of infected females are not known. However, in the wild populations, we can expect that a large proportion of males are produced by uninfected females and these males prefer to copulate with genetic females. Because of this male preference, a low level of inbreeding might be present in the populations. The low but positive values of FIS found in the sampled populations might reflect this low level of inbreeding.

Discordance between mtDNA and nuclear genetic markers

Microsatellites and mtDNA markers show very different patterns of genetic structure in A. vulgare. These differences concern the isolation by distance, the level of polymorphism and the levels of population differenciation. Indeed, under neutrality, mtDNA FST should reflect female gene flow whereas microsatellite FST should reflect both male and female gene flow. Assuming neutrality, a balanced sex ratio and no sex-biased dispersal, both nuclear and mtDNA structures are linked to Nem parameter by Wright’s formula (Wright, 1943): FST ∼ 1/(1 + 4Nem) for nuclear diploid markers and FST ∼ 1/(1 + Nem) for mtDNA (with Ne = effective population size and m = migration rate). Under these assumptions, we can therefore estimate Nem from microsatellite loci and use this estimation to calculate the expected level of mtDNA genetic structure. A difference between the expected and observed mtDNA structure would therefore either indicate a sex-biased dispersal, non-neutrality of mtDNA or unbalanced sex ratio. The expected global mtDNA FST is 0.105, which is four times lower than the observed value (FST = 0.421). Moreover, the genetic structure observed on microsatellite markers is compatible with a model of isolation by distance, whereas the genetic structure observed on mtDNA is not. The latter is indeed significantly linked to Wolbachia prevalence structure. Our study also shows significant departures from neutrality of mtDNA markers. Therefore, a low female dispersal cannot explain these departures from neutrality. The observed female-biased sex ratios would tend to decrease the population mtDNA genetic differenciation because they are correlated with an increased female population size and, as a consequence, with a decreased genetic drift. It thus clearly appears that mtDNA population structure is not only the consequence of gene flow but is also strongly influenced by another evolutionary force. Looking to its prevalence, Wolbachia is a potential candidate to explain this pattern of mtDNA genetic structure.

The effects of Wolbachia on nuclear genetic structure are expected to increase in both genetic drift and genetic influx in infected populations. However, these effects are also expected to cancel each other (Kobayashi et al., 2011), and as a consequence, they are not easy to test. Our data set suggests there is no strong difference in the microsatellite polymorphism (in terms of heterozygosity and allele number) between infected and noninfected populations. Unfortunately, our data set is not enough to estimate separately the effective migration rate and effective population size of the populations.

Wolbachia and the mtDNA polymorphism of A. vulgare?

A clear demonstration that Wolbachia is responsible for the mtDNA genetic structure found in A. vulgare requires that infected and uninfected populations be compared. Unfortunately, uninfected populations are rare. Even in the event of uninfected populations, the f element is usually present and may also disturb mtDNA genetic structure in the same manner of Wolbachia as it is mostly transmitted from mother to daughter. Juchault et al. (1993) analysed 31 European populations and found only one population composed exclusively of genetic females. All the other populations comprised females harbouring either Wolbachia or the f element. Intraspecific comparisons are therefore impossible for now. Also, the population genetic structure based on mtDNA genetic markers has been studied in very few other terrestrial Oniscid species. Interspecific comparisons are therefore very limited and might reflect differences in ecology and life history traits (notably then comparing cave and surface species) as well as the effects of Wolbachia. However, comparison of mtDNA genetic structure found in A. vulgare and P. pruinosus is interesting.

As in A. vulgare, the P. pruinosus species complex host three different Wolbachia strains (Rigaud et al., 1997; Marcadéet al., 1999; Michel-Salzat et al., 2001). However, all populations examined so far have hosted only one Wolbachia strain. In French populations, no mtDNA polymorphism has been found, whereas the populations of Greece, Tunisia and Réunion island host two mitotypes. Wolbachia exhibits a very different pattern of infection in the latter populations (only females are infected; prevalence around 60%) as compared to in the French populations (both males and females can be infected; prevalence > 90%) (Michel-Salzat et al., 2001). A selective sweep of Wolbachia accross the French populations of Porcellionides is probably responsible for the observed strong reduction in mtDNA polymorphism. Such a dynamic does not seem to occur in A. vulgare. On the contrary, the observed polymorphism of infection suggests that evolutionary mechanisms maintain the within-population coexistence of several Wolbachia strains, and thus the linked mitotypes by hitchhiking. The strong divergence observed between the four groups of mitotypes also suggests that the maintenance of several Wolbachia–mitotype associations is ancient within and across populations. Several hypotheses could explain this maintenance: competition between Wolbachia strains, variation in Wolbachia strains response to selection and variation in bacterial phenotype. The model of cyclic evolution of the system proposed by Juchault et al. (1993) can also explain the observed pattern of mtDNA polymorphism and Wolbachia prevalence, if we consider that the Wolbachia selective sweeps are not always due to the same strain. The observed pattern of mtDNA polymorphism is in accord with the prediction of Hurst & Jiggins (2005) when several Wolbachia strains coexist within populations.

Male-killing- and cytoplasmic incompatibility-inducing Wolbachia can alter the mtDNA polymorphism of uninfected individuals. When the vertical transmission rate is not 100%, the infected females produce uninfected daughters at each generation. These uninfected daughters possess the same mitotype as their mothers. As a consequence, the mitotype associated with Wolbachia is repeatedly introduced in the pool of uninfected females at each generation and will invade the pool of uninfected females. At equilibrium, it will replace all the mitotypes not associated with Wolbachia. Such a situation does not seem to occur in A. vulgare because several haplotypes not associated with Wolbachia are well present in the sampled populations. Indeed, all infected females tested now in the laboratory are genetic males (ZZ), and their progenies are exclusively composed of infected daughters and uninfected sons. The only case still not well understood would be the production of f-harbouring daughters by infected females. This situation, creating a stable female-biased lineage without Wolbachia, was described only once in Legrand et al. (1984). The paternal inheritence of f, possible in presence of the masculinizing gene, will create new mitotype-f associations that might be maintained in the populations. As a consequence, the Wolbachia-associated mitotypes have a lower probability of replacing the other mitotypes.

In this study, the genetic structure of microstaellite loci and a mtDNA locus in A. vulgare exhibited a distinct pattern. MtDNA genetic structure is correlated with Wolbachia prevalence genetic structure and not with geographic distances. Polymorphism analysis suggests mtDNA is not neutral. Wolbachia could easily explain the coexistence of several highly divergent mitotypes within populations due to the existence of three Wolbachia strains within the analysed populations.


The PhD work of Sébastien Verne was supported by a doctoral fellowship from the Région Poitou-Charentes. We thank Nicolas Puillandre, Sylvie Patri and Daniel Guyonnet for technical assistance in genotyping and sequencing and Robert Comte for assistance in sampling. We are grateful to Richard Cordaux, Thierry Rigaud and Laurent Keller and two anonymous reviewers for comments on the manuscript.

Data deposited at Dryad: doi: 10.5061/dryad.p508s296