Introduction

The endosymbiont Wolbachia is a widespread maternally-inherited bacterium found in numerous Arthropods (Werren and O'Neill, 1997) and also in Nematodes (Bandi et al, 1998). Its most common effects are feminisation of genetic males in isopods (Rigaud, 1997), parthenogenesis induction (PI) in haplodiploid Hymenoptera and one thrip (Stouthamer, 1997; Arakaki et al, 2001), male killing (Hurst et al, 1999; Fialho and Stevens, 2000) and cytoplasmic incompatibility (CI) in many species (Hoffmann and Turelli, 1997). The latter effect occurs in crosses between infected males and uninfected females (unidirectional incompatibility) or between males and females harbouring different Wolbachia variants (bidirectional incompatibility) (Hoffmann and Turelli, 1997). Wolbachia can also increase fecundity in Trichogramma (Girin and Boulétreau, 1995; Vavre et al, 1999b) or be obligatory for egg production in Asobara tabida (Dedeine et al, 2001). All these effects put infected individuals at a reproductive advantage over uninfected ones, and allow the spread of the infection in the host population.

Theoretical models of CI-Wolbachia predict that the frequency of infection within a population should correspond to a balance between the bacterial transmission efficiency, the CI intensity, and the cost suffered by infected individuals (review in Hoffmann and Turelli, 1997; Vavre et al, 2000). The actual infection cost is sometimes clear, but most often the cost is low or undetectable (Poinsot and Merçot, 1997; Stouthamer et al, 1999; Fleury et al, 2000); transmission rate ranges from 90 to 100% (Werren, 1997); and CI level shows high range of variation (Hoffmann and Turelli, 1997). Some parallel between the diversity of CI phenotypes and the natural infection frequency is thus expected.

Hymenoptera are of special interest for studying such relationships, since their haplodiploid reproduction allows variation in CI effects. In these species, haploid eggs develop into males, and incompatible eggs either die or develop into males if CI restores complete haploidy, with all possible intermediate cases (Vavre et al, 2001). Indeed, CI probably involves modification of male chromosomes by Wolbachia (mod function), such that when sperm from infected males fertilise oöcytes that are either uninfected, or infected by another bacterial variant, paternal chromosomes become improperly condensed and are lost (incompatible cross). If such sperm fertilise oöcytes bearing the same bacterial variant as their own, modified paternal chromosomes are rescued (resc function) and the cross is compatible (Breeuwer and Werren, 1993). In the pteromalid Nasonia, complete destruction of male chromosomes leads incompatible fertilised eggs to revert to haploidy and to develop into normal males (Breeuwer and Werren, 1990). No extra-mortality occurs and incompatible crosses produce all-male offspring without reduction in number. This CI type will be referred to as the ‘Male Development’ CI type, or MD type. In the Drosophila parasitoid Leptopilina heterotoma (Vavre et al, 2000), all incompatible fertilised eggs fail to develop and incompatible crosses produce reduced offspring consisting only in males issued from unfertilised eggs (‘female mortality’ CI type). This FM CI type is similar to that of the haplodiploid mite Tetranychus, where some of the incompatible fertilised eggs fail to develop (Breeuwer, 1997). In Cotesia sesamiae, Wolbachia also induces CI, probably of the MD type, even though some mortality can also be suspected (Ngi-Song et al, 1998).

Despite different CI phenotypes being expected in Hymenoptera, few data are available to estimate the extent of this diversity. The present study investigates the effect of Wolbachia in two haplodiploid, non-thelytokous Drosophila parasitoids: the diapriid Trichopria cf. drosophilae and the pteromalid Pachycrepoideus dubius (Vavre et al, 1999a). The results extend the diversity of Wolbachia-induced cytoplasmic incompatibility in Hymenoptera, and show parallel variation in the number of Wolbachia variants, level of incompatibility, and natural infection frequency.

Materials and methods

Strains

Trichopria (Hymenoptera: Diapriidae) females were collected in several places in south-eastern France, where they are uncommon. They belong to a species close to T. drosophilae (D Notton, Museum of Reading, UK, personal communication), and will be referred to as T. cf. drosophilae. This species develops as a solitary pupal endoparasitoid of various Drosophila species (Carton et al, 1986). It is infected by two Wolbachia variants (Vavre et al, 1999a) which differ from that described by Werren et al (1995) and Van Meer et al (1999) in T. drosophilae.

Crosses involved a naturally infected strain (Tw line) originating from Pierrefeu (French Riviera), and an uninfected one (To) derived from the former using rifampicin treatment following Vavre et al (2000) This cured strain proved stable and no restoration of infection has been observed after 20 generations. Experimental crosses were performed six generations after the end of antibiotic treatment to avoid any direct effect of antibiotics.

Pachycrepoideus dubius (Hymenoptera: Pteromalidae) can develop in pupae of numerous Diptera including Drosophila spp. where it behaves as a solitary ectoparasitoid (Carton et al, 1986). It is infected by a single Wolbachia variant closely related to that inducing thelytoky in Muscidifurax uniraptor (Vavre et al, 1999a). Infected (Pw) and uninfected (Po) lines were isolated from a single naturally polymorphic population originating from Lirac (southern France). Field-trapped females were used to establish isofemale lines and were checked for infection. Twenty infected and 20 uninfected lines were kept for five generations (five females at each generation), then tested again for infection, and finally pooled to establish two strains derived either from naturally infected, or uninfected wasps. The mixing of 20 different lines prevented genetic drift during early rearing generations.

In the laboratory, all parasitoids develop on a naturally Wolbachia-free Drosophila melanogaster strain originating from Lyon. Rearing and experiments were performed at 25°C, L.D. 12:12 and 70% R.H. In all experiments, Drosophila larvae were fed a standard diet (David and Clavel, 1965), adult wasps were fed honey.

Wolbachia detection

DNA extraction and PCR reaction were carried out according to Vavre et al (2000). Two sets of primers were used for PCR amplification: one specific for the Wolbachia FtsZ gene, and one specific for the insect ITS2 region to eliminate any false negative result due to amplification failure.

Crosses

In both T. cf. drosophilae and P. dubius, effects of Wolbachia were tested using all four possible crosses between infected and uninfected strains. For each cross, 15 pairs (48-h-old males and 12-h-old females) were isolated in Petri dishes and their behaviour observed to determine if they mated. After mating, females were transferred to Petri dishes containing either 60 (for Trichopria) or 30 (for Pachycrepoideus) Drosophila pupae, with the number differing considering the respective biology and daily fecundity of wasp species. For 3 days, females were daily transferred to Petri dishes with new hosts, and 10 control Petri dishes were kept unparasitised to estimate the natural mortality of the flies. Adult Drosophila and wasps of both sexes emerging from each dish were counted and used to estimate three parameters (defined in Vavre et al, 2000).

(1) Sex ratio (SR) is the proportion of males among adult offspring of each pair. SR is important to detect CI since both the male development and the female mortality CI types induce male bias in sex ratio, whereas the absolute number of males and females allow us to distinguish between the two CI types.

(2) Death by parasitism (DP) is the difference between the number of flies emerged from unparasitised and parasitised dishes, and measures the number of Drosophila killed by wasps. Death can be due either to the oviposition itself, or to the parasitic development of the wasp. Assuming that no host pupa recovers from parasitism (Carton et al, 1986), DP estimates the number of parasitised hosts. For each female, DP was summed over the 3 days.

(3) Parasite success (PS) is the ratio of emerging wasps to parasitised hosts as estimated by DP. PS is calculated for each female.

At the end of experiments, all individuals used in crosses were checked for infection. All proved infected in Tw and Pw strains, and uninfected in To and Po.

Results

Infection status and Wolbachia effects in Trichopria cf drosophilae

Out of 74 field-collected Trichopria females (five populations), 70 proved infected (Table 1) demonstrating a high natural infection rate in this species (94.5%). This is consistent with the complete infection of three laboratory strains established from French populations (25 individuals checked in each strain). However, it does demonstrate that some uninfected wasps are present in the wild.

Table 1 Infection status of field-collected individuals of Trichopria cf. drosophilae and Pachycrepoideus dubius (infected/checked individuals)
Full size table

Control crosses (Tw (infected) female × Tw male and To (uninfected) female × To male) produce offspring with identical sex ratios (Table 2), demonstrating that Wolbachia does not affect sex determination. Moreover, the offspring number is the same in both crosses, suggesting that Wolbachia does not affect the fertility of infected females, and has a low physiological cost.

Table 2 Results of crosses between infected (Tw) and uninfected (To) individuals (female × male) in Trichopria cf. drosophilae (mean and SD)
Full size table

In contrast, results of reciprocal crosses between infected and uninfected individuals depart strongly from each other (Table 2). While the cross between uninfected males and infected females does not differ from infected or uninfected control crosses, the reciprocal cross between infected males and uninfected females gives a highly male-biased sex ratio (87% males vs 52% in control), indicating CI. Death by parasitism (DP) is the same in the four crosses (Table 2), indicating that all females parasitised an equal number of hosts, whereas parasite success (PS) and total offspring production are far lower in the incompatible cross (Table 2), proving high mortality due to CI. The number of males is the same in the four crosses (Table 2), and thus the highly male-biased sex ratio (SR) in incompatible crosses is due to mortality among females (about 25 are lacking). Reduction in PS (about 35%) precisely corresponds to this decrease in female number. Altogether these data demonstrate that Wolbachia induces a female mortality CI type in T. cf. drosophilae, and from the difference between female numbers in compatible and incompatible crosses, mortality among fertilised eggs can be estimated to be about 82%.

All females issuing from incompatible crosses (76) proved Wolbachia-free. Therefore, their normal development resulted from escaping CI, and not from spontaneous restoration of infection. CI is incomplete, and about 18% of fertilised eggs develop normally.

Infection status and Wolbachia effects in Pachycrepoideus dubius

Most P. dubius populations are polymorphic for infection (Table 1), but no valid comparison can be drawn among populations due to the low number of individuals collected. Over all populations, infection frequency is only 35%, contrasting with the almost complete infection in T. cf. drosophilae and L. heterotoma (Vavre et al, 2000).

The effect of Wolbachia in P. dubius was studied using the same protocol as in T. cf. drosophilae. Neither death by parasitism (DP), nor parasite success (PS), nor total offspring production show significant variation between crosses (Table 3). Thus it seems that Wolbachia has no effect and is not able to induce CI in P. dubius. However, a slight difference appears in sex ratio (P = 0.06), due to a lower female offspring number in the cross between Po females and Pw males, while male production is unchanged. Together with a non significant decrease in DP, PS and total offspring in the ‘incompatible’ cross, this suggests some weak CI of the FM type.

Table 3 Results of crosses between infected (Pw) and uninfected (Po) individuals (female × male) in Pachycrepoideus dubius (mean and SD)
Full size table

Finally, we can conclude that Wolbachia has almost no effect on the reproduction of P. dubius, even if a very low CI of the FM type can be suspected.

Discussion

Up to now, most descriptions of Wolbachia effect in haplodiploid Hymenoptera have involved parthenogenesis induction, with only three cases of CI, resulting in various phenotypes, being described (Breeuwer and Werren, 1990; Ngi-Song et al, 1998; Vavre et al, 2000). The scarcity of reported CI cases is likely to be due partly to some sampling bias, since PI is far easier to detect than CI, as was pointed out by Cook and Butcher (1999). We describe here two other situations where Wolbachia either induces partial CI of the female mortality (FM) type, in T. cf drosophilae, or has almost no effect, in P. dubius. It is likely that many more examples of non-parthenogenetic Wolbachia will be forthcoming, and further studies, especially in arrhenotokous species, will help us to clearly appreciate the real distribution of Wolbachia effects in this group, and to better assess the variability of effects other than parthenogenesis induction.

The current hypothesis accounting for the difference between the male development (MD) and the female mortality (FM) CI types puts forward the destruction of paternal chromosomes, which would be complete in the MD type, thus restoring haploidy and allowing male development, or incomplete in the FM type, resulting in aneuploidy and the death of embryos (Breeuwer, 1997; Vavre et al, 2000). CI thus appears to be more severe in the MD type than in the FM type. This could be due to a weaker mod function in the FM type, achieved either through a lower bacterial density, or through a reduction in the mod function itself. This is consistent with the high frequency of partial CI in the FM type (three out of four cases, Breeuwer, 1997; Vavre et al, 2000; this study). Thus, variation in bacterial density could be responsible not only for variation in CI levels (Boyle et al, 1993; Breeuwer and Werren, 1993; Sinkins et al, 1995), but also for difference between the MD and FM CI types in Hymenoptera. Clearly, cytogenetic studies of early developmental events, in correlation with Wolbachia density, are needed to assess the basis of CI-Wolbachia diversity in Hymenoptera.

With respect to Wolbachia effects, comparison of the two species here studied with L. heterotoma is of special interest. The three species display cytoplasmic incompatibility of the FM type, but while in L. heterotoma all incompatible eggs fail to develop (Vavre et al, 2000), only 80% die in Trichopria, and none (or very few) in Pachycrepoideus. Considering the natural infection frequency in the three species, these observations closely fit theoretical predictions on the relationship between the intensity of CI and the natural infection frequency: no Wolbachia-free females can be found in L. heteroroma (Vavre et al, 2000), a few in Trichopria, and many in Pachycrepoideus.

Theoretical and empirical studies have predicted that evolution of CI could lead to a loss of infection in host-Wolbachia associations (Hurst and Mac Vean, 1996; Vavre, 2000; Werren and Windsor, 2000), provided Wolbachia reduces or loses its ability to manipulate host reproduction. The species here studied could thus reflect different evolutionary states of host-Wolbachia association.

In this evolutionary framework, the parallel between the multiple infections here observed (triple in L. heterotoma, double in T. cf drosophilae, single in P. dubius; Vavre et al, 1999a) and CI level could be explained by competition among Wolbachia variants within a host, that could lead to higher Wolbachia density (Van Baalen and Sabelis, 1995; Frank, 1996), and to selection towards higher CI level (Frank, 1998). It thus appears that multiple infection could be associated with stronger effects, making an association more stable, and reducing the risk of the loss of infection.

A similar pattern and scenario have been proposed in diploid D. simulans (Hoffmann et al, 1996; Bourtzis et al, 1998; Merçot and Poinsot, 1998; James and Ballard, 2000) and D. mauritiana (Giordano et al, 1995). It is true that, in the associations described here, both host and Wolbachia genotypes vary, and this makes it difficult to identify which partner is responsible for the CI difference. However, L. heteroroma, Trichopria and Pachycrepoideus could illustrate possible evolutionary steps of the CI effects of Wolbachia, showing a decrease in CI level, and the association of very weak CI with a low natural infection frequency, perhaps preceding infection loss.