Wolbachia Transfer from Rhagoletis cerasi to Drosophila simulans: Investigating the Outcomes of Host-Symbiont Coevolution

ABSTRACT Wolbachia is an endosymbiont of diverse arthropod lineages that can induce various alterations of host reproduction for its own benefice. Cytoplasmic incompatibility (CI) is the most common phenomenon, which results in embryonic lethality when males that bear Wolbachia are mated with females that do not. In the cherry fruit fly, Rhagoletis cerasi, Wolbachia seems to be responsible for previously reported patterns of incompatibility between populations. Here we report on the artificial transfer of two Wolbachia variants (wCer1 and wCer2) from R. cerasi into Drosophila simulans, which was performed with two major goals in mind: first, to isolate wCer1 from wCer2 in order to individually test their respective abilities to induce CI in the new host; and, second, to test the theoretical prediction that recent Wolbachia-host associations should be characterized by high levels of CI, fitness costs to the new host, and inefficient transmission from mothers to offspring. wCer1 was unable to develop in the new host, resulting in its rapid loss after successful injection, while wCer2 was established in the new host. Transmission rates of wCer2 were low, and the infection showed negative fitness effects, consistent with our prediction, but CI levels were unexpectedly lower in the new host. Based on these parameter estimates, neither wCer1 nor wCer2 could be naturally maintained in D. simulans. The experiment thus suggests that natural Wolbachia transfer between species might be restricted by many factors, should the ecological barriers be bypassed.

Wolbachia is a maternally inherited ␣-proteobacterium and symbiont of arthropods (4,26,34,42). This bacterium has an intracellular lifestyle, and infections occur throughout host somatic and germ line tissues of insect species (15). As a reproductive parasite, it manipulates host reproduction and favors in this way its own dispersal in host populations. The most common Wolbachia effect described so far is cytoplasmic incompatibility (CI) (8,21). CI arises when infected males mate with uninfected females and results in embryonic lethality. Reciprocal crosses between infected females and uninfected males do not express CI. This pattern can be interpreted through a two-function model (29,42): Wolbachia would somehow modify the sperm of infected males during spermatogenesis (modification, or mod function), leading to embryo death unless Wolbachia is present in the egg and restores viability (rescue, or resc function). The mod and resc functions seem to interact in a specific manner, because CI can also be observed in crosses between males and females that are both infected, if the two partners bear different Wolbachia variants.
CI allows Wolbachia to invade host populations because it increases the fitness of infected females relative to that of uninfected ones. Both theoretical and empirical studies (6,16,19,36) have highlighted the key role of three parameters in the invasion dynamics: (i) CI level (the percentage of embryos killed by CI in incompatible crosses), (ii) the fitness effect of infection on female hosts (apart from CI), and (iii) the bacterial transmission efficiency from mothers to offspring. The studies described above showed that the frequency of infected individuals presents a stable equilibrium depending on these three parameters. The infection frequency reaches this stable equilibrium value only if it first passes a threshold frequency, the level of which also depends upon these three parameters.
CI is known for a variety of insect species, including the European cherry fruit fly, Rhagoletis cerasi (Diptera, Tephritidae). Early studies demonstrated high levels of incompatibility between populations of R. cerasi (1,2), the basis of which was recently shown to involve Wolbachia (32). Populations of R. cerasi are either infected by a single Wolbachia variant, wCer1, or superinfected by two variants, wCer1 and -2. Incompatibility occurs between males from doubly infected populations and females from singly infected populations, suggesting the wCer2 infection as the cause of CI (32). However, the picture is not perfectly clear. First, although it is likely that wCer1 once invaded the species through CI, the ability of this variant to induce CI cannot be tested, because populations lacking wCer1 have never been found. Second, the direct demonstration that wCer2 is responsible for CI has not yet been provided by a set of replicate crosses with individuals of known infection status. The establishment of standardized infected and uninfected laboratory lines is time-consuming and not straightforward, given the long generation time and specialized biology of R. cerasi.
In this paper, we report on the artificial transfer of Wolbachia between two different dipteran families, from the true fruit fly, R. cerasi, into the geneticist's fruit fly, Drosophila simulans (Diptera, Drosophilidae), an extensively studied Wolbachia host (24). These experiments were done with two major goals in mind. The first objective was to obtain lines singly infected by wCer1 and wCer2 in order to test their ability to induce CI. Cytoplasmic injections have indeed been proven to be an efficient technique for stimulating Wolbachia segregation (9). The second objective was to test the prediction regarding the consequences of Wolbachia-host coevolution on three key parameters: maternal transmission efficiency, fitness effects, and CI levels. Selection on host factors tends to increase the efficiency of maternal transmission and to decrease CI levels and fitness costs (35). Selection on bacterial factors tends to increase the efficiency of maternal transmission and to decrease fitness costs. Selection on Wolbachia factors for CI levels is neutral as long as population structure is not too pronounced (30,35). Coevolution is thus expected to lead to high transmission rates, low fitness costs, and low levels of CI. Reciprocally, low transmission efficiency, negative fitness effects, and high CI levels are expected after an injection of Wolbachia into a new host (11). The results presented are partially in agreement with these predictions. Indeed, a fitness cost to the host and low transmission efficiency are observed, as expected, but the level of CI is clearly reduced.

R. cerasi and D. simulans lines.
Larvae of R. cerasi were collected from a wCer1-and -2-infected population on honeysuckle (Lonicera xylosteum) in Vienna, Austria, in 1999. After pupation, puparia were stored under the optimal conditions (37). Emerging flies were kept in cages with water, adult diet, and artificial egg-laying devices (3). D. simulans STC was used as a recipient for the Wolbachia from R. cerasi. STC is an inbred stock from the Seychelles archipelago, originally infected by two Wolbachia strains, wHa and wNo, that was cured of infection following tetracycline treatment (28).
Wolbachia injection and line establishment. The transfer of wCer1 and wCer2 into the D. simulans STC strain was performed by cytoplasmic injection (33). Using a microneedle (Femtotips; Eppendorf), cytoplasm was taken from R. cerasi eggs and injected into the posterior part of recipient eggs. Donor eggs were obtained by dissection directly from ovaries, providing fresh and weakly differentiated embryos. Fresh receiver eggs were collected from the egg-laying plates every hour. Recipient eggs were dechorionated manually prior to injection.
D. simulans females developing from injected eggs represent the generation 0 (G 0 ). Each G 0 female was crossed with one G 0 male and was left for laying before its infection status was determined by PCR. The infection status of the offspring was determined by PCR on a mass extraction of three G 1 females. In lines in which infection was detected in G 1 , 10 G 1 sisters were mated to their brothers and left to lay separately before their infection status was determined.
During the experiment, all lines were maintained at 25°C at low larval densities in vials with axenic medium (14). Rates of transmission from mothers to offspring were low in transinfected lines, imposing stringent conditions for maintenance of infection. Thus, at every generation, and for every transinfected line, six females were left to lay independently before their infection status was determined. The next generation was then started by using offspring from infected females only.
CI tests. Individual crosses were done with 3-day-old virgin males and 4-to 5-day-old virgin females. Each cross was initiated by placing one male and one female in a vial with axenic medium. Copulation was monitored, allowing the discarding of pairs in which it lasted less than 15 min, to ensure that sperm was actually transferred. The male was then removed, and the female was supplied with an egg-laying plate for 48 h. Upon removal of the female, the eggs were placed at 25°C for 24 h before the egg hatch was measured by counting all eggs. Laying plates with less than 20 eggs were discarded. All individuals from infected strains were checked by PCR for the presence of Wolbachia.
Maternal transmission rates. Maternal transmission was first roughly estimated as the proportion of infected female daughters from infected mothers during the line establishment, up to G 10 . The proportion of infected males was similarly assessed in G 8 , G 9 , and G 10 . If CI occurs, this infection rate is an overestimate of the actual transmission rate: CI will increase the proportion of infected adults, because uninfected eggs tend to die. The actual maternal transmission rate of two lines was thus estimated after crossing infected females with uninfected males in G 20 .

Measurements of fitness effects.
Female fertility and fecundity were taken as parameters for the fitness effects of infections. These were investigated during CI assay experiments and therefore by using the same mating protocol. For fertility assays, uninfected males were crossed to infected and uninfected females, and hatching rates were compared. For fecundity assays, infected and uninfected males were crossed with infected and uninfected females. Fecundity was estimated by counting the eggs laid per female in 48 h.
PCR-RFLP and sequencing. DNA was extracted from flies according to the method described by O'Neill et al. (25). The PCR primers used were general primer 81F-691R of the Wolbachia surface protein gene wsp (44) as well as wCer1-and wCer2-specific wsp primer pairs (32), ftsZf1-ftsZr1 of the cell cycle gene ftsZ (41), and the 16S rRNA-specific primer for Wolbachia (25). PCRs were done in reaction volumes of 12.5 l for the infection screening or in 50 l for post-PCR procedures: 1 or 4 l of template DNA, 1ϫ reaction buffer, 0.2 mM deoxynucleoside triphosphates (dNTPs), 0.2 M forward and reverse primers, and 0.5 or 2 U of Taq DNA polymerase (Gibco), and sterile water was added to the final volume. PCR was run under conditions described by Zhou et al. (44). wsp, ftsZ, and 16S rRNA PCR products from wCer1-infected R. cerasi, wCer2infected D. simulans, and wAu infected D. simulans were cycle sequenced with Big Dye (Perkin-Elmer). wCer2 and wAu differ in their 81F-691R wsp sequence by one substitution (32). This mutation site proved to be a wCer2-specific restriction site for Fnu4H1. wCer2-infected lines were PCR-restriction fragment length polymorphism (RFLP) digested with Fnu4HI (New England Biolabs) under the standard conditions recommended by the restriction enzyme provider, in order to exclude any line or strain contamination with wAu.
Statistical analysis. CI and fertility data were analyzed with Wilcoxon's nonparametric tests. Fecundity data were analyzed by analysis of variance (ANOVA).
Nucleotide sequence accession number. The ftsZ sequences from wCer1, wCer2, and wAu have been deposited in the GenBank nucleotide sequence database under accession no. AY227737 to -39, respectively. The 16S rRNA gene sequences from wCer1, wCer2, and wAu have been deposited under accession no. AY227740 to -42, respectively.

RESULTS
Line establishment. A total of 1,036 embryos of the uninfected D. simulans STC line were injected with cytoplasm of wCer1-and -2-infected R. cerasi. From these, 82 embryos developed into adult females, 51 of which were infected. The different infection types were wCer1 and -2 (n ϭ 31), wCer2 (n ϭ 12), and wCer1 (n ϭ 8). Thus, segregation between wCer1 and wCer2 already occurred after injection into generation 0 (G 0 ). Transmission of wCer1 and/or wCer2 from G 0 to G 1 was found in 18 females. From these G 0 females, about 10 daughters were taken for line establishment. Only 3 out of 187 G 1 females were superinfected with wCer1 and -2, 38 were infected with wCer2, and 8 were infected with wCer1. wCer1 was lost from all lines between G 1 and G 2 , despite efforts to detect rare infected G 2 females. In G 6 , six isofemale lines remained infected by wCer2: RC20, RC21, RC33, RC45, RC50, and RC78. The six lines were from six different G 0 females injected with wCer1 and -2 cytoplasm. Uninfected lines RC20Ø, RC21Ø, RC33Ø, RC45Ø, RC50Ø, and RC78Ø were founded with uninfected G 1 females, sisters of the infected females used for the establishment of the infected lines.
CI assays. The expression of CI was tested by crossing uninfected females with infected and uninfected males. CI is observed if embryonic mortality is significantly higher when males are infected. This was investigated by using four infected lines (RC21, RC45, RC33, and RC50) and their uninfected counterparts (RC21Ø, RC45Ø, RC33Ø, and RC50Ø). As shown in Table 1, wCer2 was found to induce CI in 8 out of 10 experiments, although at a low level.
The ability of wCer2 to rescue its own CI expression was tested by crossing infected males with infected and uninfected females. Rescue is observed if embryonic mortality is significantly lower when females are infected. This was investigated by using two infected lines (RC21 and RC45) and their uninfected counterparts (RC21Ø and RC45Ø). As shown in Table  2, significant rescue was found in both experiments.
To test if this rescue was complete, infected females were crossed with infected and uninfected males. Rescue can be considered as complete if embryonic mortality is not significantly higher when males are infected. This was investigated by using two infected lines (RC21 and RC45) and their uninfected counterparts (RC21Ø and RC45Ø). As shown in Table 3, rescue was not found complete in the experiment involving the RC45 and RC45Ø lines, while P was found just above the 5%  threshold in the experiment involving RC21 and RC21Ø. Thus, the data suggest that wCer2 does not fully rescue its own CI. As discussed below, imperfect transmission is thought to be the likely explanation. Fitness effects. The effect of wCer2 on female fertility can be tested by crossing uninfected males with infected and uninfected females. A positive or negative effect on fertility is detected if hatching rates differ in the two crosses. This was investigated by using two infected lines (RC21 and RC45) and their uninfected counterparts (RC21Ø and RC45Ø). As shown in Table 4, wCer2 was not found to affect female fertility.
The effects of wCer2 on female fecundity were tested by crossing infected and uninfected females with both infected and uninfected males (lines RC21 and RC45 and RC21Ø and RC45Ø, respectively). The results, presented in Table 5, were analyzed by ANOVA (Table 6). In the experiment involving RC21 and RC21Ø, a surprising effect of male infection status was observed. Indeed, females appeared to lay significantly more eggs when mated with infected males. In this experiment, infected females were less fecund than uninfected ones, but this difference was not significant at the 0.05 threshold. In the experiment involving RC45 and RC45Ø, no effect of male infection was found. Again, infected females were less fecund than uninfected ones, and here the difference was significant. Thus, the data suggest that wCer2 reduces fecundity in infected females.
PCR-RFLP and sequencing. Sequenced wsp PCR products and PCR-RFLP from single flies of strains RC21 and RC45 confirmed the presence of wCer2 in these lines. Contamination with wAu did not occur. ftsZ PCR products of wCer2-infected D. simulans, wAu-infected D. simulans Coffs Harbor, and of wCer1-infected R. cerasi flies were sequenced. wCer2 and wAu shared the same ftsZ sequences, confirming their close genetic relationship. wCer1 was more distantly related, and sequence divergences in ftsZ (2.23% in 941 bp) and wsp (2.38 to 2.55% to wCer2 and wAu, respectively, in 588 bp) (32) were similar. Interestingly, substitutions were equally spread through ftsZ of wCer1, whereas they were restricted to the 3Ј region of wsp. Most substitutions in wsp of wCer1 were nonsynonymous. All three strains wCer1, wCer2, and wAu shared the same 16S rRNA sequences.

Injection, segregation, and infection loss.
After injection from superinfected R. cerasi into D. simulans, wCer1 and wCer2 segregated in G 0 . In their original host, segregation of wCer1 and wCer2 was observed at a rate of Ͻ1% in field populations, whereby in all cases, wCer1 was the leaking variant (32). High segregation rates during injection most probably result from the low number of bacterial cells that are injected within a single recipient egg and actually survive.
Both wCer1 and wCer2 were still detectable by PCR in G 1 following injection, suggesting that both variants reached the germ cells of G 0 females. However, wCer1 was lost from all lines between G 1 and G 2 , suggesting that it was unable to develop properly in this new host or to actively maintain itself in the germ line. This loss was unfortunate, because it prevented us from determining the phenotypic effects of wCer1, yet it also proved to be an informative result. The incapacity of wCer1 to develop in a new host might reflect a higher genetic divergence from wCer2 and a very tight and specific adaptation to the original host. This interpretation is consistent with the view that wCer1 is a more ancient infection in R. cerasi than is a To increase sample size, data were pooled from two experiments (performed in G 9 and G 10 ), after testing for homogencity. Crosses between infected males and infected females are the same as in Table 2.
b The infecting Wolbachia variant is given in parentheses. Ø, uninfected. c The Wilcoxon's tests were performed by comparing each pair of crosses. d P, associated ␣ probability. a To increase sample size, data were pooled from two experiments (performed in G 9 and G 10 ) after testing for homogeneity. Crosses between uninfected males and infected females are the same as in Table 3. Crosses between uninfected males and uninfected females are the same as in Table 1 wCer2, as suggested by infection patterns in natural populations (32). On the contrary, wCer2 was still present in G 2 .
Although the efficiency of maternal transmission is low in D. simulans, imposing a stringent protocol for infection maintenance, we still possess, at the time of writing, the six lines derived from six different G 0 females. CI levels, fitness effects, and transmission efficiency. We found that wCer2 can induce CI in D. simulans, although embryonic lethality is far from 100%. This confirms that wCer2 is able to induce CI and strengthens the view that it is responsible for the patterns of incompatibility observed between R. cerasi populations (2). We observed that wCer2 is able to rescue its own CI, but only partially so. This probably results from imperfect maternal transmission (i.e., not all eggs are infected and therefore protected from CI). The transmission rates that would be necessary to explain the imperfect rescue would be 55 to 65% for RC21 and RC45. Similar transmission rate values were observed for both lines at G 20 . Thus, it seems that wCer2 is not, strictly speaking, self-incompatible. Partial nonrescue is simply due to imperfect maternal transmission.
wCer2 does not affect female fertility, but seems to reduce female fecundity by at least 10%. Negative effects on host fitness have been reported previously in natural as well as artificial Wolbachia-host associations (19,21). Intriguingly, in one data set (involving lines RC21 and RC21Ø), females were found to lay more when mated with infected males-a result that we fail to interpret in adaptive terms.
wCer1 was not transmitted after G 1 , while wCer2 had a low transmission rate. This can be seen by the infection frequency observed during line maintenance, giving a mean value of 66% for the six transinfected lines. Transmission efficiency per se was estimated at G 20 in lines RC21 and RC45, giving a mean value of 65.5%, which is much lower than any maternal transmission rate reported so far for natural Wolbachia-host associations. We observed considerable variability within and between the transinfected lines in their infection rates with wCer2 over a long time, here represented by the data from generations 1 to 10 and from generation 20. This variability was not correlated to generation number or lines. We do not yet have an explanation for this finding.
Testing theory. Theory predicts that Wolbachia-host coevolution should lead to a decline of CI level and fitness costs and to an increase in maternal transmission (30,35). Inversely, strong CI, strong costs, and low transmission rates are expected in new associations (11). We tested this prediction by creating a new association and measuring the parameters. As expected, fitness costs to the host and low transmission rates were observed, but CI levels were very low. Wolbachia density in male testes has been recognized as a key factor for the expression of CI in Wolbachia associations (8,12,40). Whether the lower expression of CI of wCer2 in D. simulans is correlated with a reduced density still needs to be assessed. However, from an evolutionary perspective, there are two possible explanations why CI levels might be low in the novel wCer2 D. simulans association.
First, D. simulans might actively repress the expression of wCer2. This is plausible because wCer2 is very closely related to wAu, a natural Wolbachia variant of D. simulans, which does not appear to induce CI in this host (10,20,23,31). Although wAu might have lost its ability to induce CI, regardless of the host background, a possibility remains that D. simulans actively and specifically represses its expression. This being so, D. simulans might recognize wCer2 as wAu-like Wolbachia and therefore repress it.
Alternatively, the wCer2 infection might be maladapted to the new host and therefore not be able to induce high levels of CI in a new host background. Hence, the prediction that CI should be high in new associations might be incorrect. Levels of CI expressed in different host species have so far only been compared in experiments in which the original and novel host were closely related (5,11,27). High levels of CI were observed after the transfer of wRi from D. simulans into Drosophila serrata (11) and after the transfer of wMel-infected D. melanogaster into D. simulans (27). However, these results could reflect the evolutionary closeness of Drosophila species rather than the ability of Wolbachia to express high CI in any background. High CI levels might in fact not always be the sign of a recent Wolbachia-host association. Prout (30) and Turelli (35) demonstrated that within panmictic populations, bacterial variants inducing higher CI levels are not selected for, but Frank (17) showed that if the population is structured, bacte-   (18,22,25,38,41,44). Wolbachia in arthropods could be seen as a huge metapopulation with infected host species as habitats for various subpopulations (7). Within host species, extinction and colonization might regularly occur through loss or gain of infection, and the current distribution of Wolbachia could represent a global and dynamic equilibrium between these two processes (43).
Following the ideas of Combes (13,39), it can be generalized that Wolbachia must cross three filters (ecological, physiological, and population) before it is established in a new host species. The ecological filter is defined by the interaction between an existing and a potential new host species. It will condition the probability for Wolbachia of getting in contact with a new species within an individual's body. The physiological filter is defined by the ability of Wolbachia to colonize the germ line of an individual. Finally, the population filter conditions the ability of Wolbachia to invade and maintain itself in host populations, which depends on the values of the three main parameters: strength of CI, maternal transmission efficiency, and fitness effects on the host (19,36).
Here, the ecological filter was bypassed as Wolbachia was intentionally injected into the new host. wCer1 and wCer2 were both established in the germ line. However, wCer1 was lost after the first generations, whereas wCer2 was maintained. The three parameters influencing Wolbachia invasion dynamics (CI level, transmission efficiency, and fitness effects) were far from optimal. Based on formulas from the model of Hoffmann et al. (19), and using the estimated parameter values, the only possible infection frequency at equilibrium for wCer2 is 0. In other words, should wCer2 cross the ecological barriers by natural means, it would not be able to invade populations of D. simulans, nor would it be able to maintain itself starting from a high frequency. Our results thus suggest that the horizontal transfer between evolutionarily distant species was, in this case at least, very unlikely or impossible. Within the Wolbachia metapopulation, subpopulations (i.e., Wolbachia variants) seem to be adapted to local habitats (species or groups of closely related species). The population filter, the ability to invade host populations, might in fact be the most critical step, preventing Wolbachia from invading all arthropod species.