Dr M.-J. Perrot-Minnot, Equipe Ecologie-Evolution (UMR CNRS 5561 Biogéosciences), Université de bourgogne, 6 Bvd Gabriel, 21000 Dijon, France. Tel.: 33 380 396340; fax: 33 380 396231; e-mail: email@example.com
Recent studies on Wolbachia-induced incompatibility in haplodiploid insects and mites have revealed a diversity of cytoplasmic incompatibility (CI) patterns among host species. Here, we report intraspecific diversity in CI expression among four strains of the arrhenotokous mite Tetranychus urticae and in T. turkestani. Variability of CI expression within T. urticae ranged from no CI to complete CI, and included either female embryonic mortality or male conversion types of CI. A fecundity cost attributed to the infection with the high-CI Wolbachia strain was the highest ever recorded for Wolbachia (−80 to −100% decrease). Sequence polymorphism at a 550-bp-portion of Wolbachia wsp gene revealed two clusters distant by 21%, one of which included three Wolbachia strains infecting mite populations sampled from the same host-plant species, but showing distinct CI patterns. These data are discussed in the light of theoretical predictions on the evolutionary pathways followed in this symbiotic interaction.
As the first identification of the rickettsia-like Wolbachia as a factor inducing reproductive incompatibility, a diversity of effects exerted by these obligate intracellular micro-organisms on their hosts has been found in such distant groups as insects, mites, nematodes and crustaceans (see Werren, 1997; Bourtzis & Braig, 1999; Stouthamer et al., 1999; for reviews). Wolbachia have been shown to enhance their propagation through maternal transmission by manipulating host reproduction in various ways, such as inducing reproductive incompatibility, thelytokous parthenogenesis, feminization of genetic males, or male-killing. Wolbachia have also been shown to have both detrimental and beneficial effects on components of host fitness, including female fecundity or male fertility (see Stouthamer et al., 1999 and Bordenstein & Werren, 2000 for reviews), and behavioural traits (Fleury et al., 2000).
Variability in the expression of host–Wolbachia interactions is of great interest to evolutionary biologists as it may help to understand how different strategies can be selectively favoured that either accelerate the spread of infection (often to near fixation), or result in the maintenance of polymorphism at equilibrium frequency. The most common effect of Wolbachia on reproduction of their arthropod hosts is cytoplasmic incompatibility (CI). However, this effect is markedly variable in intensity and pattern (Stouthamer et al., 1999). CI occurs when uninfected females are fertilized by infected males (unidirectional CI), or when infected females are fertilized by males harbouring a different Wolbachia strain (bidirectional CI). Paternal genome fragmentation in fertilized eggs then results in partial to complete haploidization (Reed & Werren, 1995). Incompatibility is thus expressed as either an increased or complete embryonic mortality in diploids (O'Neill & Karr, 1990; Hoffmann & Turelli, 1997). Several factors have been identified that influence the expression of CI phenotype, such as bacterial density, environmental factors and host or bacteria genes (seeWerren, 1997; Stouthamer et al., 1999 for reviews). Such a multitude of interacting factors could therefore potentially translate into complex incompatibility relationships between hosts strains or populations (Hoffmann & Turelli, 1997; Werren, 1997; Bourtzis & Braig, 1999; Stouthamer et al., 1999).
Haplodiploid host species are particularly interesting in studies of population biology and evolution of host–Wolbachia interactions because the development of haploid embryos into males widens the array of CI patterns. As in diplodiploid species, paternal genome fragmentation in incompatible crosses results in partial to complete haploidization of fertilized eggs (O'Neill & Karr, 1990; Reed & Werren, 1995). However, in arrhenotokous species, haploid embryos resulting from paternal genome elimination in fertilized eggs can develop into viable males (Breeuwer & Werren, 1990). Consequently, in incompatible crosses, fertilized eggs can either develop as haploid males or abort. This leads to two CI types in single infections: (1) increased male production of an amount equal to the reduction in the number of females without reduction in offspring production, referred to as ‘male conversion’ (Bordenstein et al., in preparation) or ‘male development’ (Vavre et al., 2001); (2) reduction in female production through increased embryonic mortality (‘female mortality’) (Table 1). Thus, the diagnostic feature of these two CI types is whether or not an increased proportion of males in the brood of incompatible crosses is associated with a reduction in offspring viability. Only some haplodiploid species have been previously documented for patterns of Wolbachia-induced CI (see Table 1). It is still unclear whether the two patterns of CI in haplodiploids – male development (MD) or female mortality (FM) – are the results of intrinsic Wolbachia strain differences, or specific host–Wolbachia interactions. These CI types may even represent the extremes of a continuous variation in CI phenotype, as suggested recently for the parasitoid Leptopilina heterotoma, where FM and a mixed pattern of FM and MD co-occur among inbred host sublines with various experimentally manipulated infection status (Vavre et al., 2001). Other studies are needed in various haplodiploid species, to assess the extent of intraspecific variability of CI types, its underlying determinants, and the physiological costs potentially associated with them. In mites, it has been suggested that the holokinetic structure of the chromosomes could affect the type of CI encountered, by enhancing aneuploidy in incompatiblecrosses(Breeuwer, 1997; Stouthamer et al., 1999). Whether such features could constrain the evolution of CI expression is an important issue to resolve, particularly if fitness costs are associated with CI types in a pleiotropic manner. Specifically, theoretical models predict the evolution of CI as a correlated response to direct selection on female fecundity and maternal transmission rate on bacteria variants, and on infected male compatibility on host variants (Turelli, 1994; Vavre et al., 2000).
Table 1. Patterns of Wolbachia -induced cytoplasmic incompatibility in haplodiploid species reviewed from the literature.
Embryonic mortality (rate)
Offspring sex ratio (percentage of females produced)
CI pattern: MD (‘Male Development’), high-male to all-male progeny without reduction of offspring production, due to complete haploidization of fertilized eggs that developed into males; FM (‘Female Mortality’), male-biased sex-ratio and reduction of offspring production caused by embryonic mortality resulting from egg aneuploidy: N (‘neutral’), causing no reduction in neither female proportion nor offspring production. Embryonic mortality rate in parasitoid wasps (*) are indirect estimates based on the relative number of adult offspring emerging from crosses of uninfected females mated to infected males (incompatible) and to uninfected males (compatible). Standard sex ratio in these studies range from 70 to 93% in Nasonia species, 54–81% in Leptopilina heterotoma, and from 50% (Breeuwer, 1997) to 70% (Gotoh et al., 1999a; Gotoh et al., 1999b; Vala et al. 2000) in tetranychid mites.
In this study, we investigated the occurrence of various types of CI and fecundity effects induced by Wolbachia infection in several strains of the phytophagous mite species Tetranychus urticae, and one strain of Tetranychus turkestani. The aim was to document intraspecific variability of CI patterns in these arrhenotokous species and to determine whether CI levels correlate with fitness effects. Furthermore, genetic distances between the corresponding Wolbachia strains in T. urticae were estimated from partial sequences of the wsp gene (Zhou et al., 1998). The data are discussed in the light of current hypotheses on the evolution of Wolbachia–host interactions, with respect to CI type and level of infection, and associated effects on the host fitness.
Materials and methods
Strains and curing treatments
Three strains of T. urticae were established from 30 to 100 mites collected from Rose-bay (Nerium orleander) in Montpellier, France (strains L and N) and Athens, Greece (G). They were mass reared for 3 months prior to the experiment. A third T. urticae strain (U) originating from the Netherlands and collected on Sambucus sp., was mass-reared on kidney bean in a greenhouse for several years in Montpellier. A strain of the closely related species T.turkestani collected on sugar beet in Champagne,France (B), was also established and maintained in the same manner as the T. urticae strains L, N and G. Rearing and experiments were performed in a climate room at 21 °C, 16 : 8 L : D, and 60–70% relative humidity. All strains were reared on detached leaves of kidney bean Phaseolus vulgaris (var. Containder) in closed and aerated individual containers. Contamination between strains was prevented by surrounding each strain with tap water.
The effect of Wolbachia on reproduction was investigated using infected and uninfected individuals from thesame strain. Five cured strains were established from each original strain using antibiotic treatment. As tetranychid mites are phytophagous, the uptake of antibiotic was achieved by soaking the petiole of the leaf upon which mites were feeding, into a solution of 0.02% tetracycline in phosphate buffer (0.005 m KH2PO4). Thirty females were initially allowed to lay eggs for 2–3 days on a bean leaf, before moving the leaf to the antibiotic solution. Leaves were continuously replaced by freshly treated leaves in order to continuously expose the mites to the antibiotic. Twenty to thirty females were regularly transferred to a new rearing arena to avoid over-exploitation of the leaves by the growing colonies. Three replicates of treated colonies were performed per strain. After 2 months of antibiotic treatment (approximately four generations), three samples of five females pooled were collected per replicate to assess their infection status by a PCR assay (see below). All colonies were found to be uninfected, and the per-strain treatment replicates were pooled to start the five cured strains used in the crossing experiments (Ut, Gt, Nt, Lt, Bt). These strains were maintained in mass-rearing without antibiotic for about two generations (4 weeks) before use, to avoid potential side-effect of the antibiotic treatment. Another PCR test was performed on each strain at the time of the crossing experiments.
Female teliochrysalids, the last developmental stage before adult emergence, were individually isolated from stock cultures on a 2-cm-diameter disk of fresh bean leaf, placed on a water-soaked cotton layer. Upon emergence, females were presented with two virgin males. The duration of copulation was recorded for the crosses with the U and L strains only. We used young virgin males produced as a cohort by groups of females isolated as teliochrysalids. This procedure was designed to avoid a potential decrease of Wolbachia effect due to male ageing or to repeated consecutive matings. Males were discarded 2 days after and mated females were allowed to oviposit for 7 days. Multiple mating does not occur in T. urticae, and the quantity of sperm transferred in a single mating is enough to fertilize the eggs (Helle, 1967). Around 50% of life-time fecundity and 60% of female offspring production are achieved during the first 7 days of oviposition in the U strain (M.-J. Perrot-Minnot & A. Migeon, unpublished data), and oviposition rate decreases afterwards (Young et al., 1986; M.-J. Perrot-Minnot & A. Migeon, unpublished data). Consequently, the first 7 days provide good estimates because they include the oviposition peak of spider mite females. Eggs were scored and adult emergence was followed from the eighth to the twentieth day after last oviposition day. Offspring were sexed upon sexual differenciation (deutonymphal stage) or later, and removed to minimize over-exploitation of the leaf disk. Fecundity was therefore measured as the number of eggs laid in the first 7 days of the oviposition period. We estimated mortality rate as the difference between the number of eggs recorded and the number of adults emerged, weighted by the number of eggs recorded. Preliminary observations showed that mortality occurred mostly at the embryonic stage, but shriveling of aborted eggs makes them difficult to score. Because our rearing conditions prevented escapes from larva eclosion to adult emergence, recording eggs and adults ensured a reliable estimate of mortality. Sex ratio was calculated as the proportion of females among adult offspring, i.e. secondary sex ratio. In order to check the stability of CI pattern over time, the L strain was analysed further by setting a second crossing experiment 9 months after the first one. This strain was lab-reared as described above.
Reproductive parameters were analysed using parametric tests (anova, and the Tukey test for a posteriori multiple pairwise comparisons, Zar, 1996). Sex ratio and mortality rate were Arcsin-tranformed prior to analysis. Nonparametric correlation test was performed on sex-ratio and mortality rate (Siegel & Castellan, 1988).
PCR assay of Wolbachia infection, sequencing and phylogenetic analysis
Deoxyribonucleic acid (DNA) was extracted from single mites using a CTAB protocol (Navajas et al., 1998). A fifh T. urticae strain (F), collected on eggplants in a greenhouse near Montpellier (France), was included in the molecular analysis. Wolbachia were detected in a polymerase chain reaction (PCR)-based assay using the general wsp primers 81F (5′TGG TCC AAT AAG TGA TGA AGA AAC) and wsp 691R (5′AAA AAT TAA ACG CTA CTC CA) (Braig et al., 1998). PCR amplification was carried out in a final volume of 50 µL with 0.2 nmol of each primer, 20 nmol of each DNTP, 2 units of Taq polymerase (Eurogentech) with 1× enzyme buffer supplied by the manufacturer and 75 nmol of MgCl2. One-tenth of the total volume of extraction was used as template. The amplification conditions were: 1 min at 94 °C, 1 min at 55 °C and 1 min at 72 °C per cycle for 35 cycles (see Zhou et al., 1998). A volume of 10 μL of the PCR product was run on a 1% agarose gel to determine the presence and size of the amplified DNA. The tetracycline-treated U strain was used as negative control. PCR products were purified using the Promega Wizard kit. Direct sequencing with the amplification primers was performed in an ABI prism 377 automatic sequencer using the BigDye Terminator method (Perkin Elmer, Foster City, CA, USA). Both strands of each individual were sequenced. Sequences were manipulated using Bioedit software (Hall, 1999) and their alignment was performed using Clustal W1.5 (Thompson et al., 1994). The data set was analysed usingseveral programs included in the phylip package (Felsenstein, 1993). The distance matrix was generated using the program DNADIST, with Kimura two-parameter distance option. A phylogenetic tree was built using the neighbour-joining method (program NEIGHBOT). Several sequences of Wolbachia detected in other arthropod hosts were included in the analysis (GENBANK accession numbers: Culex pipiensAF020061; Drosophila melanogasterAF020072; Oribatida unid. sp. AJ276617).
The results of the intrastrain crosses between infected and cured individuals are presented in three separate tables according to the pattern of CI observed (Tables 2–4). Crosses between infected males and uninfected females of U and G strains of T. urticae and of the B strain of T. turkestani showed no sign of incompatibility, neither as increased offspring mortality nor as increased male production (Table 2). This result therefore suggests that the Wolbachia harboured by these strains are not capable of inducing incompatibility, because the sperm from infected males remains fertile in uninfected egg cytoplasm. In respect to the view of CI mechanism as a two step process – a modification of the sperm chromosomes (mod) and a rescue of modified sperm in the egg cytoplasm (resc) (Werren, 1997), these strains could accordingly be defined as nonmodifier (mod–) strains.
Table 2. Reproductive parameters in intrastrain crosses (female × male) between infected and uninfected (tetracycline-cured, t) individuals oftetranychid mites. See text for informations on U and G strains of T. urticae and B strain of T. turkestani . Sex ratio is expressed as the percentage of females in the brood. For each type of cross, parameters are given as mean (standard deviation) or median (first quartiles) for proportions; n is the number of replicate crosses. One-way anova and a posteriori pairwise comparisons (Tukey test) were performed for eachreproductive trait, after Arcsin transformation of proportions. Letters refer to significant differences at P < 0.05.
Number of eggs
Sex ratio (quartile)
U × U
U × Ut
Ut × U
Ut × Ut
G × G
G × Gt
Gt × G
Gt × Gt
B × B
B × Bt
Bt × B
Bt × Bt
Table 3. Reproductive parameters in intrastrain crosses (female × male) between infected and uninfected (tetracycline-cured, t) individuals of L T. urticae strain . Sex ratio is expressed as the percentage of females in the brood. For each type of cross, parameters are given as mean (standard deviation) or median (first quartiles) for proportions; n is the number of replicate crosses. Series 1 and 2 were performed 1 month and 10 months after antibiotic treatment, respectively. One-way anova and a posteriori pairwise comparisons (Tukey test) were performed for each reproductive trait, after Arcsin transformation of proportions. Letters refer to significant differences at P < 0.05. A two-sample t -test was also carried out for each type of cross and each reproductive parameter between the two series ( * P < 0.05).
Table 4. Reproductive parameters in intrastrain crosses (female × male) between infected and uninfected (tetracycline-cured, t) individuals of N T. urticae strain. Sex ratio is expressed as the percentage of females in the brood. For each type of cross, parameters are given as mean (standard deviation) or median (first quartiles) for proportions; n is the number of replicate crosses. One-way anova and a posteriori pairwise comparisons (Tukey test) were performed for each reproductive trait, after Arcsin transformation of proportions. Letters refer to significant differences ( P < 0.05).
Number of eggs
Sex ratio (% females)
N × N
N × Nt
Nt × N
Nt × Nt
In contrast with the above pattern, crosses between uninfected females and infected males of T. urticae L strain (Lt × L crosses) produced all-male progenies, suggesting complete incompatiblity (Table 3). The results of two series of crosses performed at a 9-month interval showed that this CI pattern was stable over time. In both series, the lack of any significant differences in fecundity and offspring mortality between Lt × L and Lt × Lt crosses suggests that in incompatible crosses, haploidization of fertilized eggs resulted in male development instead of egg abortion, leading to all-male progenies (Table 3). Indeed, haploid eggs develop into adult males in arrhenotokous species, including those resulting from paternal genome loss in fertilized eggs. Mating duration, recorded as the time spent in copula, did not reveal any significant differences between L and Lt males mated to Lt females (296 ± 88 s, n=13, and 305 ± 72 s, n=9, respectively; n.s.). This suggests that sperm transfer was similar in both types of crosses, indicating that poor fertilization could not be responsible for all-male production in Lt × L crosses.
A second pattern of CI was found in the N strain of T.urticae. Sex ratio was significantly different among crosses, as a result of decreased sex ratio in Nt × N crosses compared with others (Table 4). On average, more males and significantly fewer females were produced in Nt × N crosses (Table 4). A significantly negative correlation between mortality rate and sex ratio was also found in the Nt × N crosses (Spearman correlation test, d.f.=15, r=−0.576, P < 0.05), whereas such correlation was not significant in the compatible crosses Nt × Nt, N × Nt, N × N. This result suggests that the apparent increase in male production was actually the consequence of increased mortality of fertilized female eggs. However, mortality was not significantly increased in the incompatible crosses on average, as would have been expected if CI was expressed exclusively as FM. Failure to evidence small effect (such as low incompatibility level caused by embryonic mortality) might be because of low sample size. Alternatively, it could result from the fact that in some Nt × N crosses with a decreased sex ratio, no embryonic mortality was observed: two out of six crosses producing less than 50% females in their brood exhibited no embryonic mortality. These crosses contribute to overall significant differences in sex ratio but not in mortality rate (Table 4). Thus, CI expression in Nt × N incompatible crosses could not be unambiguously assigned to one or the other type of CI, e.g. FM or MD, but seemed rather to involve both male development and embryonic mortality, although a larger sample size would be necessary to conclude a mixed effect.
Fitness effects were investigated using the crosses performed to establish CI patterns. Thus, only changes in female fecundity and, more indirectly, male fertility, were addressed. When comparing these traits between infected and antibiotically cured individuals, fitness effects due to Wolbachia infection were evidenced in the L strain only. First, fecundity of treated females was two times higher than that of infected females (Table 3). This increased fecundity following the antibiotic treatment which was even more pronounced after several generations of laboratory rearing: the fecundity of cured females was increased to 28–50% 9 months later (Table 3). In the same time, the variance in fecundity was reduced: the phenotypic coefficient of variation (CV) decreased from 0.36 to 0.17 in Lt × Lt crosses [Z-test for bilateral comparison of CV in Zar (1996), Z=1.79, P=0.07], whereas it was not affected in the other compatible crosses (L × L and L × Lt crosses, from 0.25 to 0.17, Z=0.93, P=0.34 and from 0.20 to 0.29, Z=−0.93, P=0.34, respectively). Secondly, there was amale fertility effect: L females mated to infected males produced a significantly higher sex ratio than those mated touninfected males, in both series (Mann–Whitney bilateral test, n=9 and 10, U=14, P < 0.05, and U=1.5, P < 0.01, respectively) (Table 3). This increase was due to a reduction in the number of males produced in both series (Newman–Keuls test for paired comparisons, t=3.73, P < 0.01 and t=3.93, P < 0.01, respectively). An additional experiment was designed to test the hypothesis of a shortage of sperm during the oviposition period in L × Lt crosses, as a plausible explanation for a decrease in sex ratio in a haplodiploid species. A closer analysis of the oviposition sequence during the first 7 days showed that the decreased sex ratio in crosses with cured males was not due to a high male production later in the oviposition period, but rather to a consistently lower proportion of females produced (Fig. 1). The former pattern would have suggested sperm depletion consecutive to a lower production/insemination of sperm from cured males, or a lower sperm longevity during storage, compared to infected males. On the contrary, the females mated to cured males continuously produced a lower proportion of females in their brood (Fig. 1). As the proportion of fertilized eggs cannot be estimated in incompatible crosses, such comparison could only be done with infected females mated to infected males and uninfected males.
Molecular analysis of the bacterial strains
By using specific primers, we amplified an approx. 550 base pairs (bp) fragment of the wsp gene from the five Wolbachia strains infecting T. urticae. None of these strains were identical. Wsp sequences were deposited in the EMBL data library under the following accession numbers: F strain, G strain, L strain, N strain, U strain (AJ437286). Multiple infection (the presence of more than one distinct type of Wolbachia in individual host) was unlikely because the sequences obtained by direct sequencing of the PCR products were unambiguous. Thus, the phylogenetic pattern observed is showing interstrain polymorphism (Fig. 2), with genetic distances ranging from 0.015 to 0.214 (Table 5). Wolbachia strains were clustered in two groups: one including the strains infecting mites collected on rose-bay (L, N and G), the other those infecting mites from other host-plants (Fig. 2). Both clusters were highly supported by bootstrap values. This pattern of genetic divergence between Wolbachia strains was independent of CI type and of the geographical origin of their host (Table 5). According to the classification proposed by Zhou et al. (1998), the cluster (U, F) corresponds to the same Wolbachia strain (wsp sequence 97.5% identical or more, same phenotype), but not the cluster (G, N, L), which, despite low genetic divergence among the strains (below 4%), shows contrasted phenotypic effects. These sequences were aligned together with previously determined sequences of Wolbachia from arthropods (three insects and one mite). Interestingly, the U, F group was closer to a species from a distant mite order, Oribatida, than to Wolbachia infecting T. urticae from rose-bay (Fig. 2). All these Wolbachia strains belong to the Bgroup.
Table 5. Patterns of Wolbachia -induced cytoplasmic incompatibility (CI) in crosses between mite strains collected in different geographical areas. The genetic distances between pairs of Wolbachia calculated on the basis of the wsp sequences, is also indicated. Mites were collected either on rose-bay ( Nerium orleander ), or on standard host plants within the T. urticae host range. The CI pattern is defined as in Table 1.
Collection site refer to three geographical areas: A, Montpellier and its vicinity (50 km max); B, the Netherlands (reared in a greenhouse in Montpellier for more than 10 years); C, Greece.
Diversity of CI patterns and fitness effects
Wolbachia -induced incompatibility was variably expressed among the four strains infecting T. urticae . It included examples of no incompatibility (U and G strains), complete incompatibility with male development (L strain), and very low incompatibility expressed as an ambiguous mixed pattern of diploid egg mortality and male development (N strain). In addition, the absence of CI in the strain of T. turkestani tested contrasts with the FM type of CI found in another strain by Breeuwer (1997 ), thus showing that the outcome of Wolbachia –host interaction on CI is also variable in that species.
The L strain of T. urticae is the first record of the ‘male development’ type of CI and of complete incompatibility in Wolbachia-infected mites. So far, all the studies on Wolbachia-induced CI in arrhenotokous mites, including other strains of T. urticae and T. turkestani, only found either neutrality or partial incompatibility expressed as egg abortion (Table 1). The other strains studied here contribute to a general predominance of either neutrality (U, G, B) or intermediate to very low incompatibility (N) among mite–Wolbachia associations. The CI pattern expressed in the N strain is still ambiguous, because sample sizes were too small compared with the low CI level found, to definitely establish the co-occurrence of both embryonic mortality and male development in incompatible crosses. The heterogeneity of CI intensity and type could be the result of infection polymorphism in this strain or to imperfect curing of some females used in the crosses (PCR tests were not done on these individuals). The existence of this CI type intermediate between MD and FM is of great interest from an evolutionary point of view (Vavre et al., 2000) and has been only recently reported in another haplodiploid species (Vavre et al., 2001). Generating several N substrains cured from Wolbachia by independent antibiotic treatments is thus necessary to confirm the existence of a mixed FM–MD pattern of incompatibility.
The diversity of CI patterns in mites revealed in this study and previous ones compared withthatinhymenopterans (Table 1). It allows one to reject the hypothesis of the exclusive expression of CI in mites as ‘female mortality’, caused by a taxon-specific chromosomal structure. Embryonic mortality in Wolbachia-induced incompatible crosses results from the fragmentation of paternal chromosomes (Reed & Werren, 1995). Because chromosomes are holokinetics in most mites, they may be more prone to produce fragments and enhance aneuploidy instead of being lost, thus increasing mortality and perhaps survival of aneuploids (Breeuwer, 1997; Stouthamer et al., 1999). Our finding of the MDtype of CI in a T. urticae strain shows that, even with holokinetic chromosomes, paternal genome fragmentation and loss can be high enough and aneuploidy low enough to induce male development. To establish whether egg haploidization is complete or not would require cytological observations. Further experiments with the L strain are also necessary to assess the survival of F1 female aneuploids, a feature that could be found to be more specifically associated with holokinetic chromosomes (Breeuwer, 1997), and could contribute to the infection by increasing F2 mortality in incompatible crosses.
In addition to its high CI level, the Wolbachia–host association in the L strain of T. urticae also imposed the highest fecundity cost ever reported. Fecundity of infected females on the first 7 days of oviposition was decreased by 80–100% compared with cured ones, and this pattern was stable over several generations (more than 15). By contrast, no fecundity change was observed in the other strains studied exhibiting no CI (U, G, B) or very weak CI (N). Fitness effects attributed to Wolbachia infection in insect hosts ranges from positive to negative (reviewed in Hoffmann & Turelli, 1997; Bordenstein & Werren, 2000). Significant fecundity costs have previously been reported in hosts harbouring CI-inducing Wolbachia, but in a much lower proportion. For example, Vala et al. (2000) reported a 10% decrease in clutch size in T. urticae strain, and a 10–20% reduction in fecundity was also observed in Drosophila simulans (Hoffmann et al., 1996), both under laboratory conditions. In other cases of strong CI, no fecundity reduction was found (Bordenstein & Werren, 2000; Bordenstein et al., submitted). The underlying process leading to increased fecundity of treated L females is unknown, and two nonexclusive hypotheses can be advanced. Infection could impose physiological costs limiting egg production in infected females. Antibiotic curing would have allowed some re-allocation to oviposition in Lt females. It is not clear however, why fecundity still increased during the 15 generations following curing. Another hypothesis considers the role of Wolbachia as an epigenetic dominant factor controlling oviposition rate: curing could have allowed the phenotypic expression of host nuclear genes as an increase in fecundity. It could also have released some heritable variation in fecundity, maintained sheltered from the effects of natural selection as a consequence of Wolbachia infection. This could account for the increased fecundity (and decreased CV) observed after 9 months of mass-rearing of the cured L strain. This hypothesis is speculative and counter-intuitive, because Wolbachia strains decreasing their own vertical transmission rate are not expected to be maintained. However, it demonstrates the need for studies linking Wolbachia to the maintenance of genetic variation on its host's life history traits. Another component of fitness in the L strain, male fertility, also seems to be affected by the infection, but in a reverse manner: the proportion of females in the offspring of L females was higher when mated to L males than to Lt males. This effect of infection on male fertility was however, marginally significant. Because no treatment replicates were done on the L strain, these fitness effects should be interpreted cautiously. Further experiments are certainly needed to assess the range and consistency of the decreased fecundity of infected L females and the increased fertility of infected L males, and to identify the underlying mechanisms.
The finding that both MD and FM CI types occur in the mite T. urticae and are associated to variable CI levels (from none to complete) and fitness effects, raises the following questions: How has such diversity evolved and what are the factors inducing differences in CI type and level? Is there any pleiotropic effects between CI pattern and fitness costs?
How have these different Wolbachia–host interactions evolved?
Several not mutually exclusive hypotheses have been proposed to account for the diversity of CI patterns in haplodiploid insects and mites: Wolbachia strains, host genes or an interaction of both (Breeuwer, 1997; Stouthamer et al., 1999). The role of intrinsic differencesbetween Wolbachia strains or the host's genomic influence on the intensity and type of CI expressed can only be tested by introgression or trans-infection experiments (reviewed in Werren, 1997; Stouthamer et al., 1999). Some indications can however, be gained from the genetic divergence among the bacterial strains and among the host's populations studied here. Despite the high variability of the wsp gene (Zhou et al., 1998), no genetic divergence among the Wolbachia infecting mites from L, N and G populations was detected, a group of strains showing three distinct CI patterns. On the other hand, partial sequencing of the host nuclear gene ITS2 in the L and G strains revealed a 0.42% divergence rate, whereas no genetic difference was found between G and U strains [L and G strains correspond to FM and GA in Navajas et al. (2000), and U to NLs in Navajas et al. (1998)]. Thus, host genome could have influenced CI pattern despite a common Wolbachia strain in the L and G strains. The lack of CI in G and U strains could also be a host, rather than a symbiont, property, and the possibility remains that these Wolbachia strains may be able to modify sperm from other mite strains (as the L strain in the ‘rose-bay’Wolbachia cluster).
The independent or interactive effects of Wolbachia and host genetic variants could also be mediated by their control upon bacterial density. Bacterial density is an important factor determining the intensity of CI, in particular in haplodiploids (Breeuwer & Werren, 1993; Perrot-Minnot & Werren, 1999). It has been proposed that bacterial density could also affect CI type in haplodiploids, with low bacterial density associated with embryonic mortality, whereas high bacterial density would induce egg haploidization (Breeuwer & Werren, 1993; Breeuwer, 1997; Vavre et al., 2001). The rationale is that the extent of sperm modification covaries with bacterial density, resulting in variable levels of chromosome fragmentation and thus of aneuploidy. The patterns observed here in T. urticae– high CI level and MD in the L strain, and very low CI level and a mixed FM–MD pattern in the N strain – fits this hypothesis. However, counter-examples to pleiotropic effects of bacterial density on CI level and CI type are also suspected (Vala et al., 2000; Vavre et al., 2000; Bordenstein et al., submitted). To test whether variation in bacterial density is responsible for a phenotypic continuum from MD to FM CI type, we could directly assess the relationship between bacterial density (quantitative PCR) and egg mortality, or indirectly through experimental manipulation of bacterial density by partial curing.
Apart from these proximate causes, is it possible to provide an evolutionary scenario for the diversity of CI found among these mite strains? When mapping CI type to our phylogenetic data, the most parsimonious interpretation is to consider CI as the ancestral state, retained in the L and to a lesser extent in the N strains, but lost in the U strain in one cluster, and in the G strain in the other. The evolution of Wolbachia virulence towards neutrality would thus have occurred twice among the T.urticae populations studied. This interpretation relies on the theoretical prediction that in random mating populations infected with a single Wolbachia incompatibility type, selection favours parasite variants that maximize their transmission rate (e.g. maternal transmission rate and female fecundity), even if their virulence is correlatively decreased (Turelli, 1994). The findings of a mite population infected with a CI-inducing strain clustering with (U, F) would be necessary to reinforce the hypothesis of CI as the ancestral state. Interestingly, the absence of fecundity cost in the strains of T. urticae and T. turkestani exhibiting no CI (U, G, B) or very weak CI (N) is consistent with the prediction that the evolutionary coupling of Wolbachia transmission rate and host's reproduction in random-mating populations should eventually lead to the absence of fecundity costs and low to nul CI level (Turelli, 1994).
In addition, the clustering of Wolbachia strains infecting T. urticae seems mainly determined by the mite's host-plant. Bacteria infecting mites collected on rose-bay were more closely related to each other than to those infecting strains collected on other host plant species, whatever their geographical origin and CI pattern were. On the other hand, mites collected on rose-bay in the mediterranean area are phylogenetically split into two distinct groups based on ITS2 sequencing, with the L and G strains each belonging to one and the other (Navajas et al., 2000). Thus, at least two independent colonization events by T. urticae have occurred on rose-bay. The genetic proximity of Wolbachia infecting these G, N, and L strains could thus result from horizontal transfer, possibly mediated by a common ecological environment. Care should be taken however, in inferring the evolutionary history of host–Wolbachia interactions because recombination may occur in Wolbachia, in which case sequence homology does not necessarily indicate phylogenetic proximity of the symbiont strains (Jiggins et al., 2001; Werren & Bartos, 2001).
Overall, our data have demonstrated the existence of a high diversity of CI patterns within tetranychid mites, including polymorphism at the intraspecific level on both CI type, CI intensity and associated fitness effects. Specifically, higher levels of CI were associated with higher fecundity costs for females. In a gene (wsp) phylogeny of the Wolbachia present in these T. urticae strains, Wolbachia from mite populations collected from rose-bay clustered together and independently of the type of CI induced, questioning the role of ecological factor, e.g. a common host plant, in promoting horizontal transfer. The finding of elevated fecundity costs in the high-CI L strain and of low or nul CI and fecundity cost in the other strains of T. urticae fits theoretical prediction of evolutionary coupling of incompatibility pattern with fitness effects. Our data and other recent studies on other haplodiploids highlight the most important parameters to investigate if we are to understand the evolutionary history of these diverse host–Wolbachia associations: the strength and factors of covariation between CI type (FM and MD), CI intensity, fitness costs of infection and bacteria transmission rate.
We are greatful to two anonymous reviewers for their critical comments on an earlier draft. Hakim Hasnaoui helped with crossing experiments, and Frank Cézilly, Thierry Rigaud, and Jack Werren gave useful comments on the manuscript.