In most insects, the endosymbiont Wolbachia induces cytoplasmic incompatibility (CI), an embryonic mortality observed when infected males mate either with uninfected females or with females infected by an incompatible Wolbachia strain. Although the molecular mechanism of CI remains elusive, it is classically viewed as a modification–rescue model, in which a Wolbachia mod function disables the reproductive success of the sperm of infected males, unless eggs are infected and express a compatible resc function. The extent to which the modification–rescue model can predict highly complex CI pattern remains a challenging issue. Here, we show the rapid evolution of the mod–resc system in the Culex pipiens mosquito. We have surveyed four incompatible laboratory isofemale lines over 50 generations and observed in two of them that CI has evolved from complete to partial incompatibility (i.e. the production of a mixture of compatible and incompatible clutches). Emergence of the new CI types depends only on Wolbachia determinants and can be simply explained by the gain of new resc functions. Evolution of CI types in Cx. pipiens thus appears as a gradual process, in which one or several resc functions can coexist in the same individual host in addition to the ones involved in the self-compatibility. Our data identified CI as a very dynamic process. We suggest that ancestral and mutant Wolbachia expressing distinct resc functions can co-infect individual hosts, opening the possibility for the mod functions to evolve subsequently. This gives a first clue towards the understanding of how Wolbachia reached highly complex CI pattern in host populations.
Cytoplasmically inherited bacteria have profoundly changed our view of insect sexuality within the last decade [1–3]. These bacteria are generally transmitted only by female hosts, through the egg cytoplasm, and have evolved complex interactions with insects. They are found to be widespread , and the most common of them, the alpha-proteobacterium Wolbachia, is suspected to infect more than 65 per cent of insect species . Wolbachia is also well known to have evolved a sperm–egg incompatibility known as cytoplasmic incompatibility (CI) promoting its own spread [3,6]. Among the most far-reaching effects of CI is the facilitation of speciation, making Wolbachia an important driver of insect evolutionary ecology [1,3].
The most spectacular effect of CI is that it specifically kills the embryos of uninfected females mated with infected males, so that infected females have a reproductive advantage. Cytological studies have shown that, in incompatible crosses, the paternal chromosomes are damaged and lost during the first mitotic divisions . Although the genetic basis of CI is still unknown, it can be interpreted as a ‘modification–rescue’ model: sperm is modified by a mod function in infected males, and the infected eggs rescue the modified sperm with a resc function, enabling the progeny to develop normally [8,9]. Apart from this simple case, CI can also occur between males and females carrying incompatible Wolbachia strains, a situation explained by assuming that each strain carries its own mod–resc pair, resulting in its particular CI type, as observed in the fruitfly Drosophila simulans .
Understanding the evolution of new CI types was mainly addressed using a combination of theoretical and empirical approaches. Mathematical investigations have produced a variety of predictions, with the assumption that CI can be explained with a single pair of mod and resc loci [10–14]. In such models, a mutation affecting either the mod or the resc locus will generate a new CI type incompatible with the ancestral type. Hence, the evolution of new CI types is classically viewed as the sudden transition of a Wolbachia strain from one CI type to another one that will be incompatible with the ancestral type. However, recent studies on Wolbachia strains from Drosophila spp.  and from mosquitoes of the Culex pipiens complex [16,17] suggested that a Wolbachia strain may use different mod–resc mechanisms, and therefore loci, to explain its compatibility pattern with other strains. This feature will condition the potential for evolution of CI types: increasing the number of mod and resc loci implies that CI types can change gradually rather than suddenly, with new CI types remaining compatible with ancestral ones.
Here, we report on a rapid evolution of CI types induced by Wolbachia strains, known as wPip strains, infecting Cx. pipiens. A wealth of CI types has been documented in Cx. pipiens populations for many years, with crossing relationships exhibiting frequently either unidirectional CI (one direction of a cross is sterile) or bidirectional CI (both directions of a cross are sterile) [16,18–20]. In this study, we showed that wPip can evolve new CI types in less than 50 host generations. The evolution of these new CI types can be explained by the accumulation of different resc functions by individual hosts, without apparent change in their mod functions. As a result, the new CI types remain compatible with their respective ancestral types.
2. Material and methods
(a) Mosquito lines
Four isofemale Cx. pipiens lines, each descended from a single founder female, were used in this study: Lv, Is, Ko and Tn (see the electronic supplementary material, table S1). They harboured different Cx. pipiens genetic background and were naturally infected by different wPip strains, namely wPip(Lv), wPip(Is), wPip(Ko) and wPip(Tn). In addition, one Wolbachia-uninfected line (Sl-TC), artificially created by antibiotic treatment as described by Duron et al. , was used (see the electronic supplementary material, table S1). Mosquito lines were kept in standard laboratory conditions in 65 dm3 cages at a constant temperature of 25°C, under a 12 L : 12 D cycle. Larvae were fed with a mixture of shrimp powder and rabbit pellets; adults were fed with honey solution.
The cytoplasms of the Lv, Is, Ko and Tn lines, including their respective wPip strains, were separately introduced into the Sl-TC nuclear background through eight generations of backcrosses, a procedure that should result in at least 99 per cent genome replacement of the original lines by the Sl nuclear genome. A first cross was made using 200 virgin females and 250 Sl-TC males. For the following generations, 200 hybrid females were backcrossed with 250 Sl-TC males. Using this protocol, we obtained the SlwLv, SlwIs, SlwKo and SlwTn lines that carry the Sl-TC nuclear genome and the wPip(Lv), wPip(Is), wPip(Ko) and wPip(Tn) strains, respectively.
(c) Crossing experiments
The CI types of Cx. pipiens lines were determined through crossing experiments. Mass crosses were carried out using an equal number of 2- to 5-day-old virgin males and females. All crosses were performed at least twice independently, and the results pooled for analysis. Note that CI was previously shown to be expressed at the same intensity throughout the Cx. pipiens male lifespan (i.e. without age effect) [22,23]. Females were allowed to blood feed 5 days after caging and their clutches were stored separately until hatching at 25°C ± 2°C. Hatching rates (HRs) were scored 72 h after egg collection to determine the CI phenotype. All unhatched clutches were checked for fertilization through observation of embryonic development following the procedure of Duron & Weill .
(d) Molecular typing
The wPip infections were characterized through the sequencing of four Wolbachia markers: the ankyrin domains encoding genes pk1 and ank2 , one putative guanylate kinase encoding gene (WPa_679; this study) and one phage-related gene (the putative secreted protein gene VrlC, also known as GP15; electronic supplementary material, table S2) . These markers are not amplified from the Wolbachia-uninfected line (Sl-TC) and discriminate the four wPip strains examined here (see the electronic supplementary material, table S3). Independent assays for infection by four other endosymbionts commonly found in insects (i.e. Cardinium, Rickettsia, Arsenophonus and Spiroplasma) were performed using specific polymerase chain reaction (PCR) amplifications as described by Duron et al.  (see the electronic supplementary material, table S2).
The Cx. pipiens mitochondrial haplotypes were determined through sequencing of an 852 bp fragment from the cytochrome b (cytb) gene (see the electronic supplementary material, table S2). The examination of the Cx. pipiens nuclear genome was assessed by PCR/restriction fragment length polymorphism (RFLP) tests based on ace-2 and Ester2 genes (see the electronic supplementary material, table S2). The ace-2 gene is located on chromosome I and encodes acetylcholinesterase 2 . The Ester2 gene is located on chromosome II and encodes a carboxylester hydrolase . A PCR/RFLP test on ace-2 using the restriction enzyme ScaI (37°C, 3 h ) allows the discrimination between the Lv, Is, Ko, Tn (two fragments: 230 and 470 bp) and the Sl-TC (three fragments: 120, 230 and 350 bp) nuclear genomes. We developed a PCR/RFLP test on Ester2 using the AvaII enzyme (37°C, 3 h), which generated different restriction fragments for the Lv (four fragments: 92, 176, 313 and 519 bp), Is (three fragments: 37, 519 and 544 bp), Ko (three fragments: 37, 519 and 544 bp), Tn (three fragments: 37, 519 and 544 bp) and Sl-TC (two fragments: 519 and 581 bp) nuclear genomes.
Mosquito DNA was extracted using a cetyltrimethylammonium bromide protocol . All PCRs were performed with approximately 20 ng of genomic DNA solution in a 40 μl final volume reaction for 35 cycles (94°C, 5 min; 94°C, 30 s; 52°C, 30 s; 72°C, 1 min). Primers are listed in electronic supplementary material, table S2. Direct sequencing of PCR products was performed on an ABI Prism 3130 sequencer using the BigDye Terminator Kit (Applied Biosystems) after purification with the QIAquick gel extraction kit (Qiagen, Valencia, CA). Sequence alignment and analyses were carried out using software mega . The sequences produced in this study are deposited in GenBank (accession numbers: JX188403–JX188404).
(a) New cytoplasmic incompatibility types evolve in laboratory Culex pipiens lines
We analysed experimentally the evolution of CI types over five years (2005–2009), which is roughly 50 mosquito generations. To this end, four isofemale lines of Cx. pipiens, designated as Lv, Is, Ko and Tn, and infected by the wPip(Lv), wPip(Is), wPip(Ko) and wPip(Tn) strains, respectively (electronic supplementary material, table S1), were crossed with each other and their CI types examined in 2005. The same crosses were repeated in 2009.
Although no self-incompatibility was observed between mosquitoes of the same line, there was clear evidence that some CI types had evolved. In 2005, the 16 reciprocal mass crosses between the four infected lines revealed a variety of crossing relationships with bidirectional CI (e.g. Lv × Is, Ko × Is and Tn × Is), unidirectional CI (e.g. Lv × Tn) and bidirectional compatibility (e.g. Lv × Ko and Ko × Tn; figure 1). In 2009, we repeated the same crosses, using the same methodology, but obtained different outcomes in two crosses, ♂Tn × ♀Lv and ♂Ko × ♀Is. Both crosses were incompatible and gave very few progeny in 2005, but they showed partial compatibility in 2009, producing a continuum from incompatible to compatible clutches (figure 1; electronic supplementary material, figure S1). We verified that no inadvertent contamination had occurred between 2005 and 2009 in the four strains by examining an array of genetic markers (electronic supplementary material, table S2). (i) The molecular typing of Cx. pipiens mitochondrial and nuclear backgrounds confirmed that the mosquito lines used in 2009 were derived from those initially used in 2005 (electronic supplementary material, table S3). (ii) Multi-locus Wolbachia sequencing showed that the wPip strains observed in 2009 descended from those characterized in 2005, and that no other Wolbachia strain was present (electronic supplementary material, table S3). In addition, we tested for the presence of a range of inherited bacteria also known to manipulate insect reproduction (i.e. Cardinium, Arsenophonus, Rickettsia and Spiroplasma), but did not find any of them.
Thus, the changes of CI types observed in 2009 when compared with 2005 were clearly due to the evolution of Cx. pipiens lines and/or to their infecting wPip strains. Two non-exclusive mechanisms could explain the observed variations of CI properties: the presence of a host restorer gene that prevents the expression of CI, and the emergence of new mutated wPip strain types.
(b) Cytoplasmic incompatibility properties are only due to Wolbachia
Each Cx. pipiens line induced unidirectional CI with an uninfected line, Sl-TC. In 2005 as well as in 2009, infected males always sterilized uninfected females, producing no progeny, whereas the reverse crosses were always fertile, with normal HRs (electronic supplementary material, figure S2). An antibiotic treatment of Lv, Is, Ko and Tn had restored full compatibility with Sl-TC females in 2006, indicating that the sterility was due to wPip infection rather than to nuclear incompatibility . Thus, it can be concluded that each wPip strain expressed a constitutive mod.
In 2009, we introduced separately each wPip infection into the Sl-TC nuclear background through eight generations of cytoplasmic introgression using Sl-TC males. The backcrossed lines, designated SlwLv, SlwIs, SlwKo and SlwTn, shared a homogenized Sl-TC nuclear background but harboured distinct cytoplasm, as confirmed by examining wPip, mitochondrial and nuclear Cx. pipiens genetic markers (electronic supplementary material, table S3). Through crossing experiments, we found that the four wPip strains exhibited the same CI types in the Sl-TC nuclear background as in their original nuclear backgrounds (figure 2; electronic supplementary material, figure S2). Thus, the variations of CI types were dependent only on wPip factors and not on Cx. pipiens nuclear factors.
(c) Mutations affecting resc loci explain the evolution of cytoplasmic incompatibility types
Having established that the 2005–2009 CI differences observed in the crosses ♂Tn × ♀Lv and ♂Ko × ♀Is were due to the evolution of their Wolbachia, we next examined which wPip functions had changed—either the mod functions of the wPip(Tn) and wPip(Ko) strains or the resc functions of wPip(Lv) and wPip(Is), or both. We sub-cloned each of the four Cx. pipiens lines in 10–14 new isofemales lines, and determined their respective CI types when crossed with original lines (see figure 3a–d). The crosses of females from 14 Lv sub-lines (named Lv-sub1 to Lv-sub14) with Tn males from the original line showed that two distinct CI types were coexisting among the females of the Lv original line (figure 3a; electronic supplementary material, figure S3a). Females from six Lv sub-lines exhibited the ancestral CI type (complete incompatibility) initially observed in 2005. Females of the eight other Lv sub-lines exhibited the new CI type (partial compatibility) observed in 2009; for instance, Lv-sub10 females mated with Tn males produced clutches with an HR continuum from low to high. Similarly, two CI types were also found among females of the Is line (figure 3b; electronic supplementary material, figure S3b). Females from five Is sub-lines exhibited the ancestral CI type, whereas females of the six other sub-lines exhibited the new CI type. By contrast, crosses between males from 12 Tn sub-lines and Lv females from the original line produced clutches with a HR continuum, suggesting that only one CI type occurred among males of the Tn line (figure 3c); an identical observation was made for the males of the Ko line using males from 10 Ko sub-lines crossed with Is females (figure 3d). That there is no variation among Tn and Ko males is also suggested by crosses with Lv and Is females exhibiting the ancestral CI types: all these crosses are incompatible, showing that all Tn and Ko males exhibit only the ancestral CI type (figure 3a,b).
The crossing patterns exhibited by isofemale sub-lines showed that the differences in CI patterns observed between 2005 and 2009 were due to changes in the resc function of the wPip(Lv) and wPip(Is) strains, rather than in the mod function of wPip(Tn) and wPip(Ko). These new mutated resc functions were found in approximately half of the examined isofemale sub-lines, and they only partially restored the compatibility, exhibiting an HR continuum from low to high. These data showed that the ancestral and the new mutated resc functions coexisted within females from whom sub-lines originated.
(d) Characterization of the new resc functions
We first examined if the new CI types (called N hereafter) remained compatible with their ancestral counterparts (called A hereafter) in Lv and Is mosquito lines. Reciprocal crosses between individuals from the Lv-sub12 (A) and Lv-sub10 (N) sub-lines showed full compatibility (electronic supplementary material, figure S4a). An identical observation was made in crosses with the Is-sub5 (A) and Is-sub1 (N) sub-lines (electronic supplementary material, figure S4b). These results indicated that two resc functions were present in the Lv and Is N CI types: (i) N CI types must have retained their ancestral resc function, able to restore the ancestral mod function (otherwise, the crosses between males with A CI types and females with N CI types would have been incompatible); (ii) in addition, N CI types have acquired new resc functions able to restore compatibility of Lv and Is females with Tn and Ko males, respectively. An alternative interpretation is that a single resc function can do all this in the N CI types. Under this hypothesis, the ancestral resc function has evolved towards a more generalist resc function that can restore distinct mod functions. In the case of Lv, the new resc function can then restore the Lv and Tn mod functions while the ancestral can restore only the Lv mod function. Similarly, a new and generalist resc function could also exist in the Is line. Note that this evolution of resc function in Lv and Is Wolbachia strains was independent of Lv and Is mod functions, which had not varied since 2005 (figure 1), confirming that the A and N CI types share the same mod function in each strain.
We then surveyed the new CI types in the four freshly isolated Lv and Is sub-lines. In general, we found that the A and N CI types of Lv and of Is were stably expressed over 10 host generations (electronic supplementary material, figure S5a–d). It is thus obvious that the new resc functions were permanently gained by the N Lv and by the N Is CI types: females from Lv-sub10 and Is-sub1 sub-lines crossed with Tn and Ko males, respectively, always produced clutches with an HR continuum from incompatible to compatible over 10 generations, while females from Lv-sub12 and Is-sub5 sub-lines remained incompatible (electronic supplementary material, figure S5a–d). However, there was also evidence that N CI types were still evolving from the A CI type: the incompatible cross ♂Ko × ♀Is-sub5 produced only incompatible clutches in G1 and G2, but few clutches with intermediate to high HR appeared in the next generations (electronic supplementary material, figure S5c).
Finally, new Lv and Is females that produce only fully compatible clutches when crossed with Tn and Ko males, respectively, were selected during two generations. As F2-selected females mated with incompatible males still produced clutches with low to high HR in the same proportions as at the F0 generation (Wilcoxon two-tailed tests; p = 0.26 for comparisons between F0 and F2 Lv females; p = 0.17 for comparisons between F0 and F2 Is females), the experiment was discontinued.
The evolution of Wolbachia CI types is classically viewed as a sharp transition from one CI type to a new one, incompatible with the former. In this study, we showed that changes in CI properties of Wolbachia infecting Cx. pipiens can be observed on a laboratory time scale, and support the notion that complexity of the mod–resc system allows a more gradual transition [32,33]. Within 5 years, new CI types emerged independently in two Cx. pipiens lines with a similar phenotypic scheme. Ancestral (A) and new (N) CI types remain mutually compatible, but they display different CI properties with specific strains, termed discriminating (D) strains. Appearance of the N CI types is independent of the host nuclear background and is driven only by variations of Wolbachia resc factors.
Shifts in CI properties had already been reported in pioneering work on laboratory lines [34–37]. In particular, Sasa et al.  described how unidirectional CI between two Cx. pipiens lines maintained in the laboratory was modified after an unspecified number of generations, producing a mixture of compatible and incompatible clutches as in our study. Rapid change of CI types could thus represent a general feature of Wolbachia, at least for wPip. Each new CI type observed here resulted from the gain of a new resc function with no apparent change in its mod function. This supports the notion that the mod and resc functions vary independently, in agreement with the fact that these functions are encoded by different genes of the Wolbachia genome [8,11,12,15,16].
How were the new resc functions acquired in the wPip strains? A first possibility could be that their resc factors have been modified either qualitatively (punctual mutation) or quantitatively (increased expression). Punctual mutations within resc sequences could reduce their specificity and thus allow the rescue of additional mod types. Additionally, mutant Wolbachia could carry two different resc loci inserted into their genome. Such genomic evolution could arise either through duplication of the existing resc locus, followed by divergence through point mutation, or through genetic exchange with another Wolbachia strain, resulting in the introduction of a new resc locus onto the genome. On the other hand, a higher level of resc expression might also lower their specificity. We do not favour this last hypothesis, because the new resc function in Lv and Is N lines each rescued only a single specific D strain, whereas one would have expected the rescue of more mod functions. A second possibility is the activation of a hitherto silent resc factor of distinct specificity. Expression of such additional functions was proposed to rely on the transposition activity of mobile genetic elements (MGEs), expected either to modify the expression of genes located close to their insertion sites, or to break or restore the function of genes in which they are inserted [26,38–41]. The potential implication of MGEs as modifiers of resc functions may thus also hold true for the present study, and the emergence of N CI types in Lv and Is Cx. pipiens lines may result from the activation of a new resc factor on the same Wolbachia genome. The fact that the two N CI types each rescue only one D strain also suggests that the number of genetic determinants of CI in a single Wolbachia genome could actually be larger than originally thought. This situation is reminiscent of our recent study of Cx. pipiens populations on Réunion Island, in which closely related Wolbachia were found to encode a variety of resc functions in addition to the one involved in their self-compatibility .
The thorough examination of compatibility pattern suggests the presence of single infection in A CI types and of co-infection in N CI types. Interestingly, A CI types represent around half of isofemale Lv and Is sub-lines; they were incompatible with the D males and remained so across generations, suggesting that they were mainly infected by the ancestral Wolbachia. It is likely that the other half of isofemale sub-lines (N CI types) were co-infected by a mixture of ancestral and mutant Wolbachia. Such a co-infection can easily explain that these females produced a continuum from incompatible to compatible clutches, depending on the abundance ratio of ancestral and mutant Wolbachia they carried. Thus, the compatibility with the D strain probably requires a threshold frequency of the mutant Wolbachia responsible for the N CI type. Following that logic, females harbouring a majority of mutant Wolbachia will be compatible with D males, but not female hosts harbouring a majority of ancestral Wolbachia. Worth noting is that no female host producing only compatible clutches with D males was observed, suggesting that none was mono-infected by the mutant Wolbachia. In other words, double infection seems thus maintained over several host generations, meaning that the efficiency of maternal transmission has the potential to impede segregation from double to single infection. Overall, it appears that the mutant Wolbachia have spread as neutral variants in Lv and Is lines where they coexist with the ancestral Wolbachia in about half of the females with a proportion sufficiently high to allow compatibility with their respective D strains.
Whether the co-infection of ancestral and mutant Wolbachia is a general feature in Cx. pipiens populations is still unknown. As long as the molecular mechanism of CI is not known and CI markers are not available, the only way to find out is to maintain strains with different CI types in laboratory conditions and detect variants by crossing with discriminating D strains. However, it is very likely that the presence of several resc functions within individual hosts is common, a feature that may facilitate the gradual evolution of the CI types. Indeed, expression of a new resc factor is neutral for the host, because the A and N CI types remain fully compatible, as in Lv and Is lines. The situation would be opposite for the mod factors, because a mutant Wolbachia expressing a new mod function is expected to be self-incompatible, and thus to be rapidly counter-selected at the population level until it evolves a new resc function. This probably explains why changes only in resc functions have been identified from our study. Co-infection should also allow for the emergence of mutant Wolbachia that no longer express the ancestral resc factor, because the production of the ancestral resc factor will be ensured by the co-infecting bacteria. This opens up the possibility for a gradual and neutral transition towards a greater complexity of CI types.
In conclusion, the present study provides experimental support for the rapid evolution of CI types through changes that separately affect the mod and resc functions of Wolbachia. It also demonstrates that multiple resc functions can be expressed by individual hosts, probably as a result of co-infection. With the increasing ability to genetically characterize Wolbachia [26,38–40], it is likely that infections currently defined as clonal will be shown to be a mixture of closely related Wolbachia strains, and deep genomic analysis should be valuable for the understanding of the genetic basis of CI.
We are very grateful to Nicole Pasteur, Philippe Fort and two anonymous reviewers for helpful comments and suggestions, and to A. Berthomieu, P. Makoundou and S. Unal for technical help. All sequence data were obtained on the Environmental Genomic Platform of the IFR Montpellier-Environnement-Biodiversité. Contribution 2012-095 of the Institut des Sciences de l'Evolution de Montpellier (UMR 5554 CNRS—Université Montpellier 2).
- Received June 14, 2012.
- Accepted August 16, 2012.
- This journal is © 2012 The Royal Society