The maternally inherited intracellular bacteria Wolbachia can manipulate host reproduction in various ways that foster frequency increases within and among host populations. Manipulations involving cytoplasmic incompatibility (CI), where matings between infected males and uninfected females produce non-viable embryos, are common in arthropods and produce a reproductive advantage for infected females. CI was associated with the spread of Wolbachia variant wRi in Californian populations of Drosophila simulans, which was interpreted as a bistable wave, in which local infection frequencies tend to increase only once the infection becomes sufficiently common to offset imperfect maternal transmission and infection costs. However, maternally inherited Wolbachia are expected to evolve towards mutualism, and they are known to increase host fitness by protecting against infectious microbes or increasing fecundity. We describe the sequential spread over approximately 20 years in natural populations of D. simulans on the east coast of Australia of two Wolbachia variants (wAu and wRi), only one of which causes significant CI, with wRi displacing wAu since 2004. Wolbachia and mtDNA frequency data and analyses suggest that these dynamics, as well as the earlier spread in California, are best understood as Fisherian waves of favourable variants, in which local spread tends to occur from arbitrarily low frequencies. We discuss implications for Wolbachia-host dynamics and coevolution and for applications of Wolbachia to disease control.
Wolbachia are bacteria that live within the cells of arthropod hosts and are widespread in many groups of insects. These bacteria can rapidly spread through a population through a process of cytoplasmic incompatibility whereby females uninfected by Wolbachia show embryo death when they mate with males carrying the bacteria. Because the infected females pass on Wolbachia to their offspring, this places them at a reproductive advantage, ensuring that the infection spreads through insect populations once it reaches a high enough frequency to overcome any negative fitness effects on its host. Yet while such a rapid spread has been predicted, it has rarely been observed in nature. Here we show that a Wolbachia infection of Drosophila simulans flies has spread very rapidly in eastern Australia, replacing another Wolbachia infection that has also spread in recent years. These invasions appear to have taken place from a very low frequency, implying that both infections are likely to have had a benefit to their hosts rather than a cost. These results have implications for the spread of Wolbachia infections currently being introduced into populations of mosquitoes and other insects for disease suppression.
Citation: Kriesner P, Hoffmann AA, Lee SF, Turelli M, Weeks AR (2013) Rapid Sequential Spread of Two Wolbachia Variants in Drosophila simulans. PLoS Pathog 9(9): e1003607. https://doi.org/10.1371/journal.ppat.1003607
Editor: Sylvain Charlat, CNRS - Université Lyon 1, France
Received: March 8, 2013; Accepted: July 24, 2013; Published: September 12, 2013
Copyright: © 2013 Kriesner et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This work was funded by program and project grants from the National Health and Medical Research Council of Australia, fellowships from the Australian Research Council and the National Science Foundation DEB 0815145. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
The vertically-transmitted intracellular bacterium Wolbachia may be the most widespread – and evolutionarily significant endosymbiont ,  of insects and other arthropods. Wolbachia induce many reproductive manipulations within hosts that increase their chance of spreading through females. Their ability to suppress other microbes in their hosts provides a novel method to control human vector-borne diseases such as dengue ,  and malaria , . Yet despite their ubiquity and potential importance in vector-borne disease control, there are few documented examples of Wolbachia infections spreading in natural host populations , .
The most commonly documented host reproductive manipulation by Wolbachia is cytoplasmic incompatibility (CI) , which reduces hatch rates for embryos produced when sperm from an infected male fertilizes uninfected ova or ova that carry an incompatible Wolbachia strain. CI provides a reproductive advantage to infected female hosts (whose infected eggs are protected from CI), and it can drive the spread of a Wolbachia infection within and among host populations , –. Several factors can affect the spread of CI-inducing Wolbachia infections, including the strength of incompatibility (quantified by H, the relative hatch rate of embryos from incompatible fertilizations), the maternal transmission frequency (1−μ, where μ is the frequency of uninfected ova produced by an infected female), and the fecundity of infected females relative to uninfected females (F, which can also approximate viability effects) .
Mathematical analyses – show that irrespective of the level of CI, as measured by H (0≤H≤1), if a CI-inducing infection satisfies F(1−μ)<1, it will tend to decrease in frequency when very rare. The reason is that the CI-induced reproductive advantage for infected females depends on the frequency of infected males, whereas disadvantages attributable to fitness costs, F<1, or imperfect maternal transmission, μ>0, are frequency independent. If the level of CI is sufficient that F(1−μ)>H (i.e., when the CI-inducing infection is very common, more infected offspring are produced by infected mothers than viable uninfected offspring are produced by uninfected mothers), the infection will tend to spread locally once it becomes sufficiently common. Hence for Wolbachia strains that satisfy H<F(1−μ)<1, we expect “bistable” dynamics –, where there is an unstable equilibrium infection frequency, denoted pu, that satisfies 0<pu<1; pu separates two stable equilibria, one at 0, the other a high frequency denoted ps (which always exceeds ½, ). If maternal transmission is imperfect (μ>0), then ps<1 because uninfected individuals are continuously introduced. The unstable equilibrium frequency, pu, plays a central role in both local dynamics and spatial spread. Local dynamics depend on the initial infection frequency, denoted p0. If p0>pu, the infection frequency tends to increase towards ps; conversely if p0<pu, the frequency tends to decrease to 0.
Spatial dynamics are more complex, and predictions depend on the initial frequencies over an extended area. Nevertheless spatial spread from a localized introduction cannot occur if pu is too large; roughly, bistable infections do not spread unless pu<½ (the exact condition is described in  and ). If a CI-inducing infection with pu<½ becomes established in a sufficiently large region (see Fig. 3 of ), it will tend to spread spatially at a rate determined by average dispersal distance of females and the intensity of CI. These bistable waves can be stopped by regional variation in population density or dispersal barriers .
In contrast to bistable infections, Wolbachia strains that provide a frequency-independent fitness advantage to infected hosts, such that F(1−μ)>1, would increase locally, even from low initial frequencies and regardless of whether they cause CI. Local spread would be followed by a “Fisherian” spatial wave ,  that, unlike a bistable wave, is unlikely to be halted. For such infections, 0 is the unstable equilibrium and the unique stable equilibrium satisfies ps<1 if maternal transmission is imperfect . As in , we use “Fisherian” to describe the spatial dynamics of variants that tend to increase even when locally very rare, unlike bistable variants that tend to increase only once they exceed a critical frequency pu>0. The mechanism that reduces the unstable equilibrium to zero may involve interactions between Wolbachia, D. simulans and various pathogenic microbes, as postulated by Fenton et al. . Alternatively it may involve nutritional provisioning ,  or other effects as yet unknown. In the absence of relevant field data, we approximate the effects of Wolbachia on D. simulans by a frequency-independent advantage that produces F(1−μ)>1. Nothing about our analyses or conclusions, which rest on the rapid sequential spread of two Wolbachia variants, requires a more detailed model. The key feature of Fisherian variants is that they spread much more rapidly than bistable variants, because small advance propagules can catalyze invasions of new areas [25, Ch. 5], in contrast to the steadily moving wave fronts expected with bistable dynamics. Unfortunately, the temporal and spatial resolutions of our data are insufficient for fitting mechanistic models of spatial spread.
Turelli & Hoffmann ,  first documented a Wolbachia infection spreading within and among natural populations. A strain of Wolbachia (wRi) was initially found infecting Drosophila simulans populations in southern California. Turelli and Hoffmann monitored the northward spread of wRi over ten years (1985–1994). Because they repeatedly found both imperfect maternal transmission in nature and reduced relative fecundity for infected females in the laboratory , , they assumed F(1−μ)<1 and inferred bistable dynamics (i.e., H<F(1−μ)<1). Using field-based estimates of μ, F and H, their mathematical analyses implied ps≈0.94, in close agreement with relatively constant infection frequencies observed in several natural populations for over 20 years . However, as emphasized by Jaenike  and elaborated by Carrington et al. , the CI-based prediction for ps is quite insensitive to variation in the relative fecundity of infected females, F. Based on field estimates of μ near 0.045, F(1−μ)>1 requires only that infected females have a fecundity advantage on the order of 5%. We present new mtDNA data from recent California samples of D. simulans and from a 1961 southern California collection which suggest that wRi invaded California less than 25 years before Hoffmann et al.  found it. This supports our new interpretation of Fisherian versus bistable spread in both California and Australia.
Other studies of Wolbachia in natural host populations have also assumed bistable dynamics , , in one case  based on observing very rare imperfect maternal transmission (μ≈0.014), in the other  based on an indication of spatial spread analogous to that seen in D. simulans. A definitive example of bistable dynamics comes from recent field releases of Wolbachia-infected Aedes aegypti mosquitoes. The wMel Wolbachia infection was transferred in the laboratory from its native host D. melanogaster to Ae. aegypti as part of a novel strategy for blocking transmission of the human dengue virus , . The observed rate of infection frequency increase within populations was consistent with significant fitness costs, and hence bistable dynamics; and low-frequency introductions into neighbouring populations have not led to Wolbachia establishment outside the release areas, as predicted with bistability (Eliminate Dengue Team, unpubl. data).
These mosquito data provide experimental support for bistable Wolbachia dynamics in nature. However, other infections including wMel in D. melanogaster cause minimal CI and persist despite incomplete maternal transmission. These infections may show Fisherian dynamics , , , consistent with the apparent global spread of alternative forms of wMel, none of which causes appreciable CI .
Here we present Wolbachia and mtDNA data that document the sequential recent spread of two Wolbachia infections in eastern Australian D. simulans populations. Previous research  indicated that D. simulans were polymorphic for a novel Wolbachia infection (wAu) present at a low frequency that induced no detectable CI. The persistence of this infection is most easily understood if it increases fitness consistent with F(1−μ)>1. Although various potential fitness advantages have been associated with Wolbachia infections in the laboratory, including virus protection ,  and fecundity increases , , they have not been demonstrated in nature for wAu or any other Wolbachia strain .
We show that the temporal and spatial data for wAu, and for wRi that has recently invaded Australia , are best explained by postulating that each variant increases host fitness in nature. We also present a new observation relevant to the history of wRi spread in Californian populations of D. simulans. These new data suggest that wRi invasion may be consistent with a Fisherian rather than a bistable wave, increasing our understanding of Wolbachia infection dynamics in nature and supporting theoretical analyses  suggesting that the global distribution of Wolbachia throughout insects and other arthropods is most easily understood if these infections tend to spread deterministically from low initial frequencies.
Wolbachia Infection Frequency Estimates
We assayed individual D. simulans samples for Wolbachia infection status and strain type using a real-time PCR/high resolution melt (RT/HTM) method designed to amplify a fragment of the wsp gene . Samples which successfully amplified using this method yielded amplicons with melting temperature peak rates (Tm) that clustered into two distinct groups which showed a consistent difference of ∼0.5°C. Wolbachia strain type for positively infected D. simulans individuals was assigned on this basis. Positive controls where DNA extracted from both a wAu-infected and a wRi-infected individual fly was combined and amplified in the same reaction produced intermediate and less distinct Tm peaks. This pattern was never observed for amplicons derived from individual specimens, suggesting that double infections are unlikely to occur amongst Australian field populations of D. simulans at any appreciable frequency. Sequencing of the amplicons derived from 13 of these samples using a standard PCR method with primers wsp_81F_Fwd: 5′-TGGTCCAATAAGTGATGAAGAAAC-3′ and wsp_691_Rev: 5′-AAAAATTAAACGCTACTCCA-3′  confirmed that 9 samples from the high-Tm cluster were the wRi haplotype (cf. GI: 225591853), whilst the remaining 4 from the low-Tm cluster were wAu (cf. GI: 2687519). For a further 5 high-Tm and 5 low-Tm cluster individuals we also obtained sequence data using standard PCR methods for each of the Wolbachia MLST genes gatB, coxA, hcpA, ftsZ and fbpA , and for the sucB gene . These samples were drawn from isofemale lines derived from diverse geographic locations including, for the high-Tm cluster, lines established in 2011 from Brisbane, Melbourne and Gosford, New South Wales; and for the low-Tm cluster, recently sourced lines from Perth and Geraldton (Western Australia), Y6 originally from Cameroon, West Africa (E.A. McGraw, pers. comm.), and Coff1 sourced from Coffs Harbour, New South Wales in the mid-1990s bearing the originally designated wAu infection . A single haplotype was obtained for the high-Tm samples which showed 100% sequence identity with the wRi strain of Wolbachia (cf. GI:225629872). A single distinct haplotype was also obtained for the low-Tm samples which was identical to previous data for the wAu strain at the sucB locus (cf. GI:84028372).
Further validation of these results was obtained using the RT/HTM method with two sets of strain-specific primers (wRi_wsp_Fwd: 5′-TGATGTTGAAGGGCTTTATTCACAG-3′; wRi_wsp_Rev: 5′-GTATCTGGGTTAAATGCTGCACCTG and wAu_wsp_Fwd: 5′-TGATGTTGAAGGAGTTTATTCATAC-3′; wAu_wsp_Rev: 5′-TTTGCTGGGTCAAATGTTACATCTT-3′). 34 of the original samples from the high-Tm cluster each amplified successfully using the wRi-specific primers but did not amplify with the wAu primers, whilst the reverse was true for 27 samples from the original low-Tm cluster.
Results for samples collected in 2004 (n = 162), 2008 (n = 499), and 2011/12 (n = 799) are summarised in Figure 1 (full details in Table S1). In contrast to the infection frequencies found for samples collected in 1993/94 [33; Figure 1], a second Wolbachia strain (wRi) was already prevalent amongst some east Australian D. simulans populations by 2004, both in the far north (Cairns) and the south (Melbourne); it occurred at a lower frequency further down the Queensland coast (Maryborough) and in northern Tasmania, but was not yet present in the central coastal area of northern New South Wales or in southern Tasmania. Additionally, for three central populations (Maryborough, Kingscliff and Red Rock), wAu infection frequencies increased significantly compared to geographically equivalent populations sampled previously (P<0.001, G-test). Ballard  also reports significantly higher wAu frequencies for samples collected in December 1999 from both Coffs Harbour and Brisbane than those previously found by Hoffmann et al. .
Blue shaded parts of the pie are wAu-infected D. simulans, red shaded are wRi-infected and unshaded are uninfected individuals. Pie size represents sample size (n).
The 2008 data show wRi spreading considerably since 2004, being at high frequency throughout Queensland, and now detected in both southern Tasmania and 6 of 7 populations from New South Wales. By 2011/12, wRi was at high frequency at every location we sampled in mainland eastern Australia and Tasmania (n = 794) with an overall frequency of 0.929 (0.909, 0.946); whilst the wAu infection was only detected from an island (Hayman Island) ∼29 kms offshore from Airlie Beach in Queensland. We did not detect wRi infection from two sites in Western Australia sampled in late 2011 and early 2012. Samples from Perth (n = 105) and Geraldton (n = 40), 370 km to the north, were polymorphic for the wAu infection, with infection frequencies of 0.648 (0.548, 0.738) and 0.500 (0.338, 0.662) respectively or 0.607 (0.522, 0.687) when pooled. Populations of D. simulans in Western Australia are separated from those in the east by ∼2000 km of dry habitat, preventing direct migration.
Laboratory crosses indicate that the wRi-infected Australian flies have a CI phenotype indistinguishable from California wRi. Hatch rates (Table 1) from crosses between infected females from four north Queensland lines established from females collected in 2011 and five-day-old males from a wRi-infected laboratory line provided no evidence for CI and were similar to the average hatch rate when females from these lines were crossed to males from an uninfected and a wAu-infected laboratory line. Females from these lines therefore behaved like wRi lines in crossing type. In addition, when males from these lines were crossed, they were compatible with the wRi line females, but generated incompatibility in crosses with females from both the uninfected and wAu-infected lines. All differences between reciprocal crosses were significant (Mann-Whitney U-tests) at the table wide α' = 0.05 level after corrections for multiple comparisons.
To test maternal transmission of wAu in D. simulans in Western Australia, we established 55 isofemale lines from field-caught females obtained from two locations near Perth and from Geraldton in Western Australia in 2011/12. We assayed the Wolbachia infection status of the mothers and several F1 progeny, including between 10 and 12 progeny for each of 34 lines infected with wAu. From these lines, 350 progeny were tested, and all but 8 were found to be infected. This yields a field estimate for μ of 0.023 with 95% bootstrapped confidence interval (bCI) of 0.003, 0.049 based on 10,000 replicates). The bCI is broader than the 95% binomial confidence interval based on total numbers (0.01, 0.045), but is more appropriate because of heterogeneity in transmission frequencies across females [cf. 26].
We also tested maternal transmission of wRi in eastern Australia using isofemale lines established from two locations near Melbourne in 2013. Ten F1 progeny were tested for each of 23 of these lines where the mother was found to be infected. All but 7 of the F1s were also infected, yielding a field estimate of μ of 0.026 (95% bCI 0.004, 0.057) which is consistent with previous field estimates of μ for wRi in California , .
The wRi infection has previously been found associated with significant fecundity increases in D. simulans lines from California . We performed a three-way comparison for fecundity using Australian wRi-infected, wAu-infected and uninfected laboratory lines which had been outcrossed to a common genetic background for 5 generations. Results (Figure 2) indicate that females from the wRi-infected line showed a marginally significant fecundity advantage relative to the uninfected line (p = 0.045, t-test) but strain differences were marginally non-significant in the three-way comparison (F2,37 = 2.7, p = 0.082).
To examine the association between Australian wAu and wRi infections across time, we obtained sequence data for a 1256 bp mtDNA fragment from 7 wAu-infected and 11 wRi-infected samples, and found no variation apart from a single G to A base substitution previously identified by Ballard  (position 457 on our fragment, position 8201 in , see Table 2) with the former sequence (here denoted “A haplotype”) found only in wAu-positive samples and the latter (denoted “R haplotype”) only in wRi-positive samples. Additional mtDNA haplotype sequence data obtained for 45 Wolbachia uninfected (w–) samples from 2004, 39 w– samples from 2008 and 6 w– samples from 2011 were similarly invariant, comprising only A or R haplotypes. The mtDNA haplotype for a further 9 w- samples from 2004, and 55 w- samples from 2008, was determined by treatment of PCR amplicons with the restriction enzyme HinfI. Full details of the mtDNA haplotypes determined for w- samples are summarised in Table S1. Both A and R haplotypes were detected amongst the 2004 and 2008 eastern Australian w- samples (simultaneously for some locations), whereas sequence data for w- samples collected in 2011 (n = 5) all confirmed the presence of the R haplotype. A single 2011 w– sample from Perth, Western Australia had the A haplotype.
To better understand the history of association of wRi with D. simulans, we also examined mtDNA variation in California D. simulans stocks by sequencing two regions which included those identified in Table 3 of Ballard (2004) as showing the greatest variation among siII mtDNA haplotypes, positions 1175 to 1626 and 7838 to 8287. We examined mtDNA from a stock founded in 1961 in Nueva, California, ∼25 km SE of Riverside, the type locality for wRi . This stock is not infected with Wolbachia, and its origin pre-dates the initial 1985 Wolbachia analyses. It yielded a novel haplotype, denoted CA61, different from those described in Ballard (2004) by two base pairs from both the canonical “R haplotype,” which Ballard calls DSR, and the canonical haplotype (DSW) associated with northern California D. simulans prior to the invasion of wRi. To better understand the novel CA61 haplotype, we examined mtDNA from two reference wRi stocks, Riv84 and Riv88, collected in Riverside in 1984 and 1988, respectively, and from seven stocks (denoted Y26, Y35, Y36, Y42, Y46, Y54 and Y62) collected from an orchard approximately 5 km east of Winters, California in August 2011. All stocks apart from Y35 and Y62 were wRi infected. The Y54 stock produced a novel haplotype which differs from the canonical DSR by one nucleotide (Table 1). As expected, the mtDNA from Riv84, Riv88, and all the other stocks matched DSR.
Given the paucity of sequence variation in siII D. simulans mtDNA sequences, and in the mtDNA associated with wRi , , there is little statistical power to infer phylogenetic relationships among these sequences, apart from the fact that the sequences from the Indian Ocean seem to be part of a distinct clade (see Fig. 3 of ). However, with only one exception, all wRi-infected stocks share the nucleotide A at position 8201, in contrast to the G found in DSW and CA61 (as well as AU23 and all other wAu-infected stocks examined by Ballard . This substitution corresponds to the HinfI polymorphism described by Hale and Hoffmann  differentiating mtDNA of wRi-infected stocks sampled worldwide from DSW. Nucleotide A was found at position 8201 in all wRi-infected stocks examined worldwide (Ballard , 92 wRi-infected stocks from California  and 29 of 30 additional wRi-infected stocks . The only exception may correspond to rare paternal transmission of mtDNA [cf. 47]. Given that uninfected flies from wRi-infected populations quickly become associated with the mtDNA from their infected maternal ancestors , it seems unlikely that wRi was present at an appreciable frequency near Riverside in 1961.
As noted by Hoffmann and Turelli , for a Wolbachia variant such as wAu to persist despite imperfect maternal transmission and little CI, it must increase fitness sufficiently to offset imperfect transmission, i.e., F(1−μ)>1. Let IA (IR) denote individuals infected with wAu (wRi), and let U denote uninfected individuals. We measure fitness of IA and IR females relative to uninfected females, and denote their relative fitnesses FA and FR. Let μA (μR) denote the frequency of U ova produced by IA (IR) females. The condition for wAu to increase when rare is FA(1−μA)>1. When this is satisfied, wAu will reach a stable equilibrium frequency of(1)The wAu frequencies observed in southern Queensland and northern New South Wales in 1999 and 2004, as well as the 2011/12 wAu frequencies in Western Australia, all suggest an equilibrium frequency for wAu of roughly 0.6. Using (1) and our estimate μA≈0.023 implies FA≈1.061. Given the wide confidence interval for μA (0.003, 0.049), (1) implies that plausible values for FA range from 1.008 to 1.140.
However, estimates of FA are further constrained by the observed wAu frequency changes. A pooled frequency estimate (n = 1641) for mainland eastern Australia in 1993/1994  is 0.054, with 95% confidence interval (0.044, 0.066); by December 1999, the estimated frequency (n = 92) had increased to 0.565 (0.458, 0.669). This increase occurred over about 120 generations, assuming roughly 20 generations per year for 6 years. With FA = 1.061 and μA = 0.023, pA would increase from 0.054 to 0.529 in 120 generations, consistent with our data. If μA were significantly smaller, say 0.01, (1) with = 0.6 implies FA = 1.026. Assuming μA = 0.01 and FA = 1.026, pA would increase in 120 generations from 0.054 to only 0.229, far below our 1999 estimate. However, if μA were appreciably larger, say 0.04, (1) with = 0.6 implies FA = 1.11. With those values, pA would increase from 0.054 to 0.597 in 120 generations. Thus, our frequency estimates and maternal transmission data are consistent with a positive fitness effect for wAu on the order of 5–10%, but inconsistent with appreciably smaller fitness advantages of 3% or less (Figure 3). Given the uncertainty of our estimates for both the apparent stable equilibrium frequency, , and the maternal transmission rate, μA, for wAu, our data documenting the frequency increase of wAu from 1993 to 1999 provide the most robust estimate for the fitness advantage it seems to induce.
As shown in Materials and Methods, once wAu is established, wRi will tend to increase when very rare only if(2)The roughly simultaneous spread of wRi from three geographically disparate foci in north Queensland, Victoria and Tasmania (Figure 1) indicates that condition (2) was met. Moreover, the observed temporal displacement of wAu by wRi is consistent with (2) and the levels of CI and maternal transmission for wRi observed in California , . For instance, the wRi and wAu infection frequencies estimated in March–May 2008 from seven coastal cities from New South Wales were statistically consistent (G test; G = 20.39, P = 0.06) with pooled frequencies (n = 253) of 0.09 and 0.54 for wRi and wAu respectively (and 37% uninfected). In 2011, seven New South Wales coastal populations spanning the same range of latitudes were sampled (n = 302); again their infection frequencies were statistically consistent (G-test; G = 3.74, P>0.68), but wAu had been eliminated and the pooled frequency estimate for wRi had risen to 0.91 (0.88, 0.94). We conjecture that there may be on the order of 15 generations per year on average in this area of coastal New South Wales (comparable to the Central Valley of California, versus 20 for sub-tropical coastal populations farther north).
To compare these data to the predictions of our simple model (see Eqs. 3 below), we need estimates of fitness effects, maternal transmission rates and CI. As above, we assume FA = 1.061 and μA = 0.023. Assuming that H, the relative hatch rate from incompatible fertilizations (i.e., sperm from wRi-infected males fertilizing either uninfected or wAu-infected ova), and the fidelity of maternal transmission were as observed in California, e.g., H = 0.55 and μR = 0.045 , , we predict that starting with pA = 0.54 and pR = 0.09, pA will fall below 0.01 within 40 generations only if FR is on the order of 1.08 (Figure 4). After 45 generations, these parameter values imply that pR should rise from 0.09 to very near its predicted stable equilibrium value of 0.94, consistent with our data.
FR (the fitness of wRi-infected hosts relative to uninfecteds) is predicted to be on the order of 1.08 or higher in a simple model (employing previously estimated parameter values of FA = 1.061 and μA = 0.023 for wAu, and μR = 0.045 and H = 0.55 for wRi).
The stable equilibrium frequency of wRi is very insensitive to FR . For instance, with H = 0.55 and μR = 0.045, as FR increases from 0.95 to 1.1, increases from 0.93 to only 0.94. In contrast, the unstable equilibrium for wRi (when entering an uninfected population) plummets from 0.22 to 0, converting the predicted spatial dynamics from bistable to Fisherian. Our temporal data are too coarse to approximate accurately the speed of wRi's spatial spread. However, between 2008 and 2011, wRi spread southward from Coffs Harbour and northward from Bega (Figure 1), filling in roughly 1000 km of the New South Wales coast in three years. The speed of this bidirectional spread is comparable to the northward spread of wRi observed in California, roughly 100 km/year . Such rapid spatial spread can be explained by human-mediated, long-distance dispersal and Fisherian local dynamics that allow rare long-distance migrants to greatly accelerate spatial advance [25, Ch. 5].
Despite the ubiquity of Wolbachia in natural populations of arthropods, there are very few documented examples of Wolbachia spread , . This has limited our understanding of Wolbachia spatial dynamics, with bistable spread for CI-causing Wolbachia suggested by imperfect maternal Wolbachia transmission by wild-caught females ,  and demonstrated fecundity costs for infected females in the laboratory , , . However, bistable dynamics cannot explain the persistence of Wolbachia that cause little or no reproductive manipulation . Moreover, as emphasized by Fenton et al. , bistable dynamics are difficult to reconcile with molecular data indicating that the widespread occurrence of Wolbachia is attributable to rare horizontal transmission events involving one or very few infected founders , , , . These phenomena are more easily explained by Fisherian dynamics , in which the frequency dynamics of rare Wolbachia infections are dominated by positive fitness effects. Our temporal and spatial data, describing the sequential spread of two Wolbachia strains (wAu and wRi) through natural populations of D. simulans along the east coast of Australia over the last two decades, support the view that both strains may have spread under Fisherian dynamics.
The recent spread of wRi is particularly compelling, with apparently independent introductions in both the north and south of the country between 1994 and 2004, leading to near-fixation of wRi in all populations sampled on the east coast mainland of Australia by late 2011/early 2012. The appearance of wRi in three geographically separated locations – Queensland, Victoria and Tasmania – in 2004 cannot be explained as a single bistable wave that originated and spread from elsewhere. In 1994, before the arrival of wRi, wAu was rare in several populations. Ballard's (2000) data from 1999 and our samples from 2004 and 2008 (Figure 1) document wAu frequency increases across the Australian east coast. Given that wAu does not induce CI , its spread from low frequency in multiple populations implies that it confers a fitness benefit. The rapid displacement of wAu by wRi suggests that wRi must enhance fitness even more than wAu.
Both wAu and wRi have been detected in D. simulans populations around the world , , , and these infections have co-occurred previously (e.g. Sangoqui and Rocafuerta, Ecuador in 2000; Brooksville, FL, USA in 2002 , and possibly also Japan ). The wAu infection in D. simulans may have a Neotropical origin as a result of a relatively recent horizontal transfer event from Drosophila willistoni . The origin of wRi in D. simulans is less certain, but very closely related Wolbachia have been found in various Drosophila that co-occur with this cosmopolitan species [e.g.], [ 54], [55,56]. It is not known how either strain was introduced into Australian D. simulans populations, but our data suggest that both introductions occurred within the last few decades. The pooled frequency of wRi (92.7%, Table S1) in our 2011/12 eastern Australian samples (including Tasmania) is consistent with the stable equilibrium frequency found in California over the past two decades [∼93%]; [ 17], [26,37]. This equilibrium frequency is determined primarily by significant CI and imperfect maternal transmission. Our data from crossing experiments, and mtDNA and Wolbachia sequencing suggest that the wRi strain now found throughout eastern Australia is essentially identical to that initially identified in California  and subsequently found in Africa, Asia, Europe and South America , . Genome comparisons support this (M. Turelli, unpublished data). Our mtDNA data from uninfected individuals in 2011/12 indicate that the R haplotype is being swept to fixation in eastern Australia as a consequence of the spread (with imperfect maternal transmission) of wRi [cf. 46]. A previous mitochondrial sweep of the A haplotype associated with wAu may have also occurred, but the ancestral haplotypes in eastern Australian D. simulans are not known.
Wolbachia and mtDNA are expected to be co-inherited maternally. Even though Wolbachia exhibit imperfect maternal transmission (μ>0), mtDNA haplotypes associated with a given infection will be driven to fixation, by CI or other Wolbachia-associated positive fitness effects, if all offspring of infected females inherit maternal mtDNA . Eventually, all individuals in the population have infected maternal ancestors who passed on their mtDNA even if not their Wolbachia. Although rare paternal transmission of Wolbachia has been detected in laboratory D. simulans , , Turelli et al.  and Turelli & Hoffmann  inferred that paternal transmission or horizontal transfer must be rare in nature. Similarly, there is evidence for rare paternal inheritance of mtDNA in D. simulans , , , but it is apparently too rare to create incongruity between Wolbachia and mtDNA lineages . Drosophila simulans populations worldwide carry three distinct mtDNA haplotype groups, with very little intra-haplogroup variation , , . Ballard  found both wAu and wRi only in flies carrying siII haplogroup mtDNA. He identified a single (synonymous) G to A base substitution (see Table 2) that was consistently different between wAu-infected versus wRi-infected flies respectively, and matched a restriction enzyme polymorphism identified by Hale and Hoffmann . In our analyses, all wRi flies had the same base substitution as identified by these researchers, consistent with the hypothesis of a unique origin for wRi in D. simulans.
Turelli and Hoffmann ,  had previously suggested bistable spatial dynamics of wRi in California populations of D. simulans. However, the rate at which wRi spread spatially, on the order of 100 km per year, is more consistent with Fisherian dynamics whose deterministic spread from very low frequencies allows rare long-distance dispersal to greatly accelerate spatial advance. Assuming that the mtDNA haplotype of the 1961 strain we characterized was widespread in southern California, our new analysis suggests that wRi was not yet at appreciable frequency at this time, and that the invasion detected by Turelli and Hoffmann  may reflect a rapid spatial spread from populations farther south.
The speed of wRi spread throughout eastern Australia is comparable to the spread of wRi in California, on the order of 100 km/year . The infection was common in two populations sampled in 2004 (Melbourne and Cairns), then swept to near fixation in all east Australian mainland populations and Tasmania by 2011/12. Although some genes and traits exhibit clinal variation over this range , , several molecular markers indicate that D. simulans is effectively a single panmictic unit throughout eastern Australia, including Tasmania . Thus, there are no apparent barriers to dispersal that would isolate populations from the spread of wRi. We did not, however, detect wRi in Western Australia, with the wAu infection persisting at relatively high frequency (≈60%). Western Australian populations of D. simulans may be geographically isolated from eastern Australian populations, although once wRi is introduced it is likely to sweep through this area in the coming decades.
The Drosophila-Wolbachia dynamics have been used to interpret and model the spread of Wolbachia-infected Aedes aegypti mosquitoes in Australia . In contrast to the apparent Fisherian spatial dynamics of wAu and wRi, these releases seem to follow bistable dynamics, with strong evidence for a non-trivial unstable equilibrium frequency, on the order of 10–20% , . After wMel was driven to a high frequency in two relatively isolated populations by 10 weeks of releases, it increased to near-fixation, and has remained at over 90% frequency for the past 2 years [7; unpublished data]. Although the infection was detected in disjunct residential areas near the release sites soon after the initial releases, it has not persisted or spread in these areas [7; unpublished data], as would be expected under Fisherian dynamics. Given the evolutionary pressure for Wolbachia to evolve towards mutualism with its hosts , , a decreasing unstable point may alter these dynamics in the future  particularly if fitness costs produced in novel hosts are overcome.
In summary, our results suggest that the wAu and wRi Wolbachia infections of D. simulans spread at least part of the time in a Fisherian manner. The wAu infection increased in frequency despite having imperfect transmission and not causing CI. Its spread and apparent equilibrium make sense only if it increased host fitness. Over the past decade, wAu was rapidly displaced by wRi throughout eastern Australia. The rapidity of this replacement starting from three locations in 2004 indicates that wRi also increases host fitness in nature, at least under some circumstances. These data plus evidence that wRi entered southern California after 1960 lead us to reinterpret that canonical example of Wolbachia spread in California as Fisherian rather than bistable. Depending on the rapidity with which Wolbachia adapt to novel hosts, its spread after artificial introductions to target species may be greatly accelerated.
Materials and Methods
Field-caught D. simulans samples were collected from various locations along the east coast of Australia including Tasmania during February–March 2004  and March–May 2008 [see 39]. Further samples were collected from mainland eastern Australia at similar coastal and some inland localities, and from the Perth region of Western Australia during 2011, and from both Geraldton (Western Australia) and Tasmania in early 2012. All samples were preserved in 100% ethanol and stored at −20°C. Specimens subsequently used in PCR analysis were males except where a few isofemale lines had been established prior to preservation of the mother. Species identity was established by male genital arch morphology and confirmed by molecular methods .
The new California samples were from isofemale lines (Y26, Y35, Y36, Y42, Y46, Y54 and Y62) established in August 2011. The collection site was a peach orchard approximately 5 km east of Winters, CA. The stocks were maintained as isofemale lines until analysed. The 1961 Nueva, California stock was obtained from the Drosophila Species Stock Center at University of California, San Diego (stock #14021-0251.006).
Wolbachia Infection Status
DNA extractions were performed using a standard Chelex based method, and assays for Wolbachia infection status and strain type were performed with a RT/HRM method using the Roche LightCycler® 480 system as previously described . Briefly, a conserved set of primers (wsp_validation_Fwd: 5′-TTGGTTACAAAATGGACGACATCAG-3′ and wsp_validation_Rev: 5′-CGAAATAACGAGCTCCAGCATAAAG-3′) were used to target a variable ∼340 bp region of the wsp gene. The wRi and wAu alleles are expected to differ by 22 SNPs and a single 3 bp indel over this wsp region, potentially yielding a significant difference in observed Tm for the respective PCR amplicons, with the wRi allele expected to have the higher average Tm .
Samples from D. simulans isofemale lines previously identified as wRi- or wAu-infected respectively, were included on each PCR plate as Wolbachia strain-type positive controls. Further positive controls entailed equal volumes of DNA extracted from an individual fly of each type being combined in a single reaction. Separate D. simulans specific primers (Dsim_RpS6_Fwd: 5′-CCAGATCGCTTCCAAGGAGGCTGCT-3′; Dsim_RpS6_Rev: 5′-GCCTCCTCGCGCTTGGCCTTAGAT-3′) were used as a host species-specific control for each sample.
The RT-PCR/HRM conditions were as follows: 10 minutes at 95°C; 45 cycles of 10 seconds at 95°C, 15 seconds at 58°C, 30 seconds at 72°C; 1 minute at 95°C; 1 minute at 40°C. High resolution melting; temperature ramped up to 95°C at 0.02°/second with continuous fluorescence acquisition; 30 seconds at 40°C.
Individual samples were scored as Wolbachia infected, uninfected or inconclusive based on machine reported wsp and RpS6 crossing point (Cp) relative and absolute value criteria and examination of individual Tm profiles for evidence of non-specific products. A relatively small number of samples, which initially yielded inconclusive results, were retested.
Additional standard PCR amplification of wsp, gatB, coxA, hcpA, ftsZ, fbpA and sucB gene fragments for selected individual samples followed  with annealing temperatures of 54°C for gatB, coxA, hcpA, ftsZ and sucB, and 59°C for wsp and fbpA. PCR products were checked using a 2% agarose gel for presence of an unambiguous single band of expected size. Amplified DNA for selected samples which yielded a clear and positive PCR result were sent to Macrogen (Korea) for purification and sequencing. Sequence data obtained was analyzed using Geneious v6.1 (Biomatters, Auckland, NZ).
Assays for mitochondrial DNA haplotype for a total of 172 individual Australian samples were performed using primers (Dsim_siII_7765_Fwd: 5′-ATTTAATATTCAAGCAATAGC-3′; Dsim_siII_8981_Rev: 5′-TTCTGGTTCTATAATTTTAGC-3′) designed to target a 1256 bp region of the D. simulans mitochondrial genome previously identified as containing at least one (synonymous) G to A base substitution at position 457 (outer strand) that was a characteristic difference between the haplotypes found to be associated with either the wAu or wRi strains of Wolbachia, respectively .
Amplification of mtDNA was performed using a standard PCR method with the following conditions: 5 minutes at 94°C; 38 cycles of 30 seconds at 94°C, 1 minute at 55°C, 1 minute at 72°C; 5 minutes at 72°C; hold at 4°C.
PCR products were checked using a 2% agarose gel for the presence of an unambiguous single band of expected size. Amplified DNA for 108 samples that yielded a clear and positive PCR result were sent to Macrogen (Korea) for purification and sequencing. Sequence data obtained was inspected using Sequencher v4.5 (Gene Codes, Ann Arbor, Mi). Subsequently PCR amplicons for a further 64 samples were checked on a 2% agarose gel after digestion with the restriction enzyme HinfI .
For each California line, DNA was extracted from 30 flies as described in . Following the position descriptions in , two sets of primers were used to amplify two mtDNA regions from positions 1034 to 1718 and from 7797 to 8425. The second region contains the G to A base substitution at position 8201 that distinguished the R haplotype from the A haplotype, associated with wRi and wAu infection, respectively. (The primers used were: region 1: DSRmt1033+: 5′-CCAAAATGACTTGTAATCCA-3′ and DSRmt1719−: 5′-GCACCTAATATTAAAGGCACT-3′, region 2: DSRmt7796+: 5′-GCTACATCTCCAATTCGATTA-3′ and DSRmt8426−: 5′-TTTATATTCTTTTAGACAACATGG-3′.)
The PCR conditions were: 3 minutes at 94°C; 34 cycles of 30 seconds at 94°C, 30 seconds at 55°C, 75 seconds at 75°C; 8 minutes at 72°C; hold at 10°C. The PCR products were checked on a 2% agarose gel for an unambiguous band of the correct size. The PCR products were purified using the QIAquick PCR purification kit. The amplified fragments were Sanger sequenced by the UC Davis College of Biological Sciences UCDNA Sequencing Facility, using the same primers used for the PCR reaction. The sequencing results were analysed using ApE alignment tool (A plasmid Editor v2.0.45, by M. Wayne Davis).
To confirm the sequence coordinates we realigned the mtDNA genomes described by Ballard in , ,  also including the D. yakuba mtDNA reference genome , and confirmed the positions of polymorphisms described in Table 3 of . We then aligned our new sequences against this reference.
Three separate laboratory lines previously confirmed to be wRi-infected (Riv88), wAu-infected (Coffs 1) and uninfected (W88) respectively were used in reciprocal crosses with each of four separate lines (CBQ40, CBQ46, CBQ72 and CBQ80) established from field caught females collected in 2011 from north Queensland to test for CI. The level of CI was determined by mating virgin 5 d-old males to virgin females (5–8 d old). Males were mated once, and females were placed after mating in a vial with a spoon containing 5 ml of agar-treacle-yeast medium and left for 24 h at 25°C. The number of unhatched eggs was counted 24–32 h later. CI data (egg hatch rates) were angular transformed prior to analysis. Mann-Whitney U-tests were used to compare CI levels between the different lines. Multiple comparisons were corrected at the table wide α' = 0.05 level using the Dunn-Sidak method .
Separate mass bred wAu-infected and uninfected fly lines were established from 20 to 30 Western Australian isofemales lines for which infection status had previously been determined. Virgin females from each of these mass bred lines were then mated with a similar number of field caught Brisbane males which had been aged in the laboratory for at least 5 d to reduce the impact of CI. An equivalent mass bred wRi-infected line was established from virgin female F1 progeny of field collected Brisbane females mated with males from the uninfected mass bred line. Each mass bred line was then outcrossed to field collected aged Brisbane males which were also aged for at least 5 d for a further 4 generations to homogenise the genetic backgrounds. Lines were then retested to confirm infection status. Flies were reared at low densities by transferring 25 eggs into vials on ∼20 ml of agar-cornmeal-yeast medium. Pairs of virgin females and males were transferred to fresh medium vials for 2 d, then females were transferred to vials with spoons containing 5 ml of medium with 20% fresh live yeast paste added to the medium surface. Spoons were replaced every 24 h for 5 d and eggs counted. Twelve to 13 females were assayed for each line. Model I ANOVA (analysis of variance) and t-tests were used to compare fecundity between lines.
We address two issues: first, the conditions for the initial spread and maintenance of wAu in an isolated population, then the conditions for its displacement by wRi. Let IA (IR) denote an individual infected with wAu (wRi), and let U denote an uninfected individual. Assuming discrete generations, we denote the frequency of these three types of adults in generation t by pA,t, pR,t and pØ,t. No individuals doubly infected for wAu and wRi have been found (and each infection is associated with distinct mtDNA haplotypes, ), hence we assume pA,t+pR,t+pØ,t = 1. We measure fitness of IA and IR females relative to uninfected females, and denote their relative fitnesses FA and FR. In principle, these Wolbachia may affect viability or fecundity. Given that our data indicate relatively small fitness effects (see Results), we can simplify the analysis by approximating fitness differences as female fecundity variation with no viability effects.
Field-collected IA and IR females have both been shown to produce uninfected offspring. Let μA (μR) denote the frequency of U ova produced by IA (IR) females. Before wRi arrives in the population, pA,t+pØ,t = 1; so we need only keep track of pA,t. Because wAu causes no CI,(3)Assuming FA(1−μA)>1, this produces the stable equilibrium (1) in Results. (Equilibrium (1) is equivalent to the standard mutation-selection equilibrium for haploids, in which the uninfected, less-fit type is maintained at frequency μ/s, where μ = μA is the fraction of uninfected ova produced by infected mothers and 1−s = 1/FA is the relative fitness of U females.) Equilibrium (1) implies that if and μA<<1 are known, FA≈1+μA/(1−).
To understand the dynamics of wRi entering a population with wAu, note that both Hoffmann et al. , and James and Ballard  demonstrated that wAu infection is not able to rescue CI caused by wRi. Indeed, wRi induces the same level of CI in matings with both uninfected (U) and wAu-infected (IA) females. The joint dynamics of pA,t, pR,t and pØ,t follow a simplified version of the recursions for double and single infections derived by Hoffmann and Turelli [12, Eqs. 2.5]. Let H = 1−sh<1 denote the relative hatch rate of embryos produced by IA or U ova fertilized by sperm from IR males. The relevant recursions are(4a)(4b)(4c)(4d)Note that in (4d) is just the sum of the right hand sides of (4a–c). Each of these three expressions has a simple interpretation as the product of two terms, the first (in square brackets) proportional to the fraction of ova of each type, the second proportional to the fraction of fertilized ova that hatch. (In (4b) this second term is one, because wRi-infected ova are compatible with all sperm.) Given that the three frequencies sum to 1, there are only two independent variables.
Now consider the conditions for wRi to increase when it is extremely rare and wAu is at equilibrium with U, as described by Eq. (1). When pR,t≈0, ≈pØ,t+FApA,t. When wAu and U are at equilibrium, Eq. (3) implies that pØ,t+FApA,t = FA(1−μA). Hence, from (4b), we see that the condition for wRi to increase when very rare is condition (2) provided in Results. This is just the condition for spread of a Wolbachia variant that is completely compatible with the existing type . Because essentially no CI occurs when wRi is very rare, the incompatibility of wRi with wAu does not enter condition (2).
When (2) is not met, pR,t increases only once it exceeds a frequency threshold so that the fitness advantage IR individuals receive from CI overcomes their frequency-independent disadvantage relative to variant A. This condition is a simple generalization of the unstable equilibrium that arises for a single Wolbachia infection that satisfies H<F(1−μ)<1.
Sequences for the MLST genes gatB, coxA, hcpA, ftsZ and fbpA, and the sucB gene obtained in this study have been deposited in GenBank (http://www.ncbi.nlm.nih.gov/Genbank) under accession numbers KF278668–KF278673.
We thank Jason Kennington for providing field-caught flies from Western Australia, Lea Rako for help with the CI assay, Chenling Xu for the molecular work and bioinformatics in California, Andrew Tremain for help with the California experiments, Aston Arthur, Carla Sgrò, Michele Schiffer, Belinda van Heerwaarden, Vanessa Kellerman and Jennifer Shirriffs for help with field collections of D. simulans, and Scott L. O'Neill for stimulating discussions in January 2011 about bistable versus Fisherian Wolbachia waves.
Conceived and designed the experiments: ARW PK AAH. Performed the experiments: PK SFL ARW. Analyzed the data: PK MT AAH ARW. Contributed reagents/materials/analysis tools: MT AAH. Wrote the paper: PK MT AAH ARW.
- 1. Werren JH, Windsor DM (2000) Wolbachia infection frequencies in insects: evidence of a global equilibrium? Proceedings of the Royal Society of London Series B-Biological Sciences 267: 1277–1285.
- 2. Hilgenboecker K, Hammerstein P, Schlattmann P, Telschow A, Werren JH (2008) How many species are infected with Wolbachia? - a statistical analysis of current data. Fems Microbiology Letters 281: 215–220.
- 3. Zug R, Hammerstein P (2012) Still a host of hosts for Wolbachia: Analysis of recent data suggests that 40% of terrestrial arthropod species are infected. Plos One 7(6): e38544.
- 4. Werren JH, Baldo L, Clark ME (2008) Wolbachia: master manipulators of invertebrate biology. Nature Reviews Microbiology 6: 741–751.
- 5. Engelstadter J, Hurst GDD (2009) The ecology and evolution of microbes that manipulate host reproduction. Annual Review of Ecology Evolution and Systematics 40: 127–149.
- 6. Walker T, Johnson PH, Moreira LA, Iturbe-Ormaetxe I, Frentiu FD, et al. (2011) The wMel Wolbachia strain blocks dengue and invades caged Aedes aegypti populations. Nature 476: 450–453.
- 7. Hoffmann AA, Montgomery BL, Popovici J, Iturbe-Ormaetxe I, Johnson PH, et al. (2011) Successful establishment of Wolbachia in Aedes populations to suppress dengue transmission. Nature 476: 454–457.
- 8. Hughes GL, Koga R, Xue P, Fukatsu T, Rasgon JL (2011) Wolbachia infections are virulent and inhibit the human malaria parasite Plasmodium falciparum in Anopheles gambiae. Plos Pathogens 7(5): e1002043.
- 9. Kambris Z, Blagborough AM, Pinto SB, Blagrove MSC, Godfray HCJ, et al. (2010) Wolbachia stimulates immune gene expression and inhibits Plasmodium development in Anopheles gambiae. Plos Pathogens 6(10): e1001143.
- 10. Hoshizaki S, Shimada T (1995) PCR-based detection of Wolbachia, cytoplasmic incompatibility microorganisms, infected in natural populations of Laodelphax striatellus (Homoptera: Delphacidae) in central Japan: Has the distribution of Wolbachia spread recently? Insect Molecular Biology 4: 237–243.
- 11. Turelli M, Hoffmann AA (1991) Rapid spread of an inherited incompatibility factor in California Drosophila. Nature 353: 440–442.
- 12. Hoffmann AA, Turelli M (1997) Cytoplasmic incompatibility in insects. In: O'Neill SL, Hoffmann AA, Werren JH, editors. Influential Passengers: Inherited Microorganisms and Arthropod Reproduction. New York: Oxford University Press. pp. 42–80.
- 13. Barton NH, Turelli M (2011) Spatial waves of advance with bistable dynamics: Cytoplasmic and genetic analogues of allee effects. American Naturalist 178: E48–E75.
- 14. Caspari E, Watson GS (1959) On the evolutionary importance of cytoplasmic sterility in mosquitoes. Evolution 13: 568–570.
- 15. Hoffmann AA, Turelli M, Harshman LG (1990) Factors affecting the distribution of cytoplasmic incompatibility in Drosophila simulans. Genetics 126: 933–948.
- 16. Fine PEM (1978) Dynamics of symbiote-dependent cytoplasmic incompatibility in culicine mosquitoes. Journal of Invertebrate Pathology 31: 10–18.
- 17. Turelli M, Hoffmann AA (1995) Cytoplasmic incompatibility in Drosophila simulans - dynamics and parameter estimates from natural populations. Genetics 140: 1319–1338.
- 18. Schraiber JG, Kaczmarczyk AN, Kwok R, Park M, Silverstein R, et al. (2012) Constraints on the use of lifespan-shortening Wolbachia to control dengue fever. Journal of Theoretical Biology 297: 26–32.
- 19. Barton NH (1979) Dynamics of hybrid zones. Heredity 43: 341–359.
- 20. Fisher RA (1937) The wave of advance of advantageous genes. Annals of Eugenics 7: 355–369.
- 21. Skellam JG (1951) Random dispersal in theoretical populations. Biometrika 38: 196–218.
- 22. Fenton A, Johnson KN, Brownlie JC, Hurst GDD (2011) Solving the Wolbachia paradox: Modeling the tripartite interaction between host, Wolbachia, and a natural enemy. American Naturalist 178: 333–342.
- 23. Kremer N, Voronin D, Charif D, Mavingui P, Mollereau B, et al. (2009) Wolbachia interferes with ferritin expression and iron metabolism in insects. Plos Pathogens 5: e1000630.
- 24. Brownlie JC, Cass BN, Riegler M, Witsenburg JJ, Iturbe-Ormaetxe I, et al. (2009) Evidence for metabolic provisioning by a common invertebrate endosymbiont, Wolbachia pipientis, during periods of nutritional stress. Plos Pathogens 5: e1000368.
- 25. Shigesada N, Kawasaki K (1997) Biological invasions : theory and practice New York: Oxford University Press. 205 p.
- 26. Carrington LB, Lipkowitz JR, Hoffmann AA, Turelli M (2011) A re-examination of Wolbachia-induced cytoplasmic incompatibility in California Drosophila simulans. Plos One 6(7): e22565.
- 27. Jaenike J (2009) Coupled population dynamics of endosymbionts within and between hosts. Oikos 118: 353–362.
- 28. Hoffmann AA, Turelli M, Simmons GM (1986) Unidirectional incompatibility between populations of Drosophila simulans. Evolution 40: 692–701.
- 29. Rasgon JL, Scott TW (2003) Wolbachia and cytoplasmic incompatibility in the california Culex pipiens mosquito species complex: Parameter estimates and infection dynamics in natural populations. Genetics 165: 2029–2038.
- 30. Hoffmann AA, Clancy DJ, Merton E (1994) Cytoplasmic incompatibility in Australian populations of Drosophila melanogaster. Genetics 136: 993–999.
- 31. Hoffmann AA, Hercus M, Dagher H (1998) Population dynamics of the Wolbachia infection causing cytoplasmic incompatibility in Drosophila melanogaster. Genetics 148: 221–231.
- 32. Riegler M, Sidhu M, Miller WJ, O'Neill SL (2005) Evidence for a global Wolbachia replacement in Drosophila melanogaster. Current Biology 15: 1428–1433.
- 33. Hoffmann AA, Clancy D, Duncan J (1996) Naturally-occurring Wolbachia infection in Drosophila simulans that does not cause cytoplasmic incompatibility. Heredity 76: 1–8.
- 34. Hedges LM, Brownlie JC, O'Neill SL, Johnson KN (2008) Wolbachia and virus protection in insects. Science 322: 702.
- 35. Teixeira L, Ferreira A, Ashburner M (2008) The bacterial symbiont Wolbachia induces resistance to RNA viral infections in Drosophila melanogaster. Plos Biology 6(12): e1000002.
- 36. Fast EM, Toomey ME, Panaram K, Desjardins D, Kolaczyk ED, et al. (2011) Wolbachia enhance Drosophila stem cell proliferation and target the germline stem cell niche. Science 334: 990–992.
- 37. Weeks AR, Turelli M, Harcombe WR, Reynolds KT, Hoffmann AA (2007) From parasite to mutualist: Rapid evolution of Wolbachia in natural populations of Drosophila. Plos Biology 5(5): e114.
- 38. Harcombe W, Hoffmann AA (2004) Wolbachia effects in Drosophila melanogaster: in search of fitness benefits. Journal of Invertebrate Pathology 87: 45–50.
- 39. Lee SF, White VL, Weeks AR, Hoffmann AA, Endersby NM (2012) High-throughput PCR assays to monitor Wolbachia infection in the dengue mosquito (Aedes aegypti) and Drosophila simulans. Applied and Environmental Microbiology 78: 4740–4743.
- 40. Zhou W, Rousset F, O'Neil S (1998) Phylogeny and PCR-based classification of Wolbachia strains using wsp gene sequences. Proc Biol Sci 265: 509–515.
- 41. Baldo L, Hotopp JCD, Jolley KA, Bordenstein SR, Biber SA, et al. (2006) Multilocus sequence typing system for the endosymbiont Wolbachia pipientis. Applied and Environmental Microbiology 72: 7098–7110.
- 42. Paraskevopoulos C, Bordenstein SR, Wernegreen JJ, Werren JH, Bourtzis K (2006) Toward a Wolbachia multilocus sequence typing system: Discrimination of Wolbachia strains present in Drosophila species. Current Microbiology 53: 388–395.
- 43. Ballard JWO (2004) Sequential evolution of a symbiont inferred from the host: Wolbachia and Drosophila simulans. Molecular Biology and Evolution 21: 428–442.
- 44. Ballard JWO (2000) Comparative genomics of mitochondrial DNA in Drosophila simulans. Journal of Molecular Evolution 51: 64–75.
- 45. Hale LR, Hoffmann AA (1990) Mitochondrial-DNA polymorphism and cytoplasmic incompatibility in natural populations of Drosophila simulans. Evolution 44: 1383–1386.
- 46. Turelli M, Hoffmann AA, McKechnie SW (1992) Dynamics of cytoplasmic incompatibility and mtDNA variation in natural Drosophila simulans populations. Genetics 132: 713–723.
- 47. Wolff JN, Nafisinia M, Sutovsky P, Ballard JWO (2012) Paternal transmission of mitochondrial DNA as an integral part of mitochondrial inheritance in metapopulations of Drosophila simulans. Heredity
- 48. Hoffmann AA, Turelli M (1988) Unidirectional incompatibility in Drosophila simulans: Inheritance, geographic variation and fitness effects. Genetics 119: 435–444.
- 49. Nigro L, Prout T (1990) Is there selection on RFLP differences in mitochondrial DNA. Genetics 125: 551–555.
- 50. O'Neill SL, Giordano R, Colbert AME, Karr TL, Robertson HM (1992) 16S Ribosomal-RNA phylogenetic analysis of the bacterial endosymbionts associated with cytoplasmic incompatibility in insects. Proceedings of the National Academy of Sciences of the United States of America 89: 2699–2702.
- 51. Richardson MF, Weinert LA, Welch JJ, Linheiro RS, Magwire MM, et al. (2012) Population genomics of the Wolbachia endosymbiont in Drosophila melanogaster. PLoS Genet 8(12): e1003129.
- 52. Charlat S, Le Chat L, Mercot H (2003) Characterization of non-cytoplasmic incompatibility inducing Wolbachia in two continental African populations of Drosophila simulans. Heredity 90: 49–55.
- 53. James AC, Ballard JWO (2000) Expression of cytoplasmic incompatibility in Drosophila simulans and its impact on infection frequencies and distribution of Wolbachia pipientis. Evolution 54: 1661–1672.
- 54. Miller WJ, Riegler M (2006) Evolutionary dynamics of wAu-Like Wolbachia variants in neotropical Drosophila spp. Applied and Environmental Microbiology 72: 826–835.
- 55. Bennett GM, Pantoja NA, O'Grady PM (2012) Diversity and phylogenetic relationships of Wolbachia in Drosophila and other native Hawaiian insects. Fly 6: 273–283.
- 56. Mateos M, Castrezana SJ, Nankivell BJ, Estes AM, Markow TA, et al. (2006) Heritable endosymbionts of Drosophila. Genetics 174: 363–376.
- 57. Sherengul W, Kondo R, Matsuura ET (2006) Analysis of paternal transmission of mitochondrial DNA in Drosophila. Genes & Genetic Systems 81: 399–404.
- 58. Kondo R, Satta Y, Matsuura ET, Ishiwa H, Takahata N, et al. (1990) Incomplete maternal transmission of mitochondrial-DNA in Drosophila. Genetics 126: 657–663.
- 59. Baba-Aïssa F, Solignac M, Dennebouy N, David JR (1988) Mitochondrial DNA variability in Drosophila simulans - Quasi absence of polymorphism within each of the 3 cytoplasmic races. Heredity 61: 419–426.
- 60. Solignac M, Monnerot M (1986) Race formation, speciation, and introgression within Drosophila simulans, Drosophila mauritiana, and Drosophila sechellia inferred from mitochondrial DNA analysis. Evolution 40: 531–539.
- 61. Arthur AL, Weeks AR, Sgro CM (2008) Investigating latitudinal clines for life history and stress resistance traits in Drosophila simulans from eastern Australia. Journal of Evolutionary Biology 21: 1470–1479.
- 62. Lee SF, Rako L, Hoffmann AA (2011) Genetic mapping of adaptive wing size variation in Drosophila simulans. Heredity 107: 22–29.
- 63. van Heerwaarden B, Lee RFH, Wegener B, Weeks AR, Sgro CM (2012) Complex patterns of local adaptation in heat tolerance in Drosophila simulans from eastern Australia. Journal of Evolutionary Biology 25: 1765–1778.
- 64. Turelli M (1994) Evolution of incompatibilty-inducing microbes and their hosts. Evolution 48: 1500–1513.
- 65. Huang AM, Rehm EJ, Rubin GM (2009) Quick preparation of genomic DNA from Drosophila. Cold Spring Harbor Protocols 2009: pdb.prot5198.
- 66. Ballard JWO (2000) Comparative genomics of mitochondrial DNA in members of the Drosophila melanogaster subgroup. Journal of Molecular Evolution 51: 48–63.
- 67. Clary DO, Wolstenholme DR (1985) The mitochondrial DNA molecule of Drosophila yakuba - Nucleotide sequence, gene organization, and genetic code. Journal of Molecular Evolution 22: 252–271.
- 68. Sokal RR, Rohlf FJ (1995) Biometry : the principles and practice of statistics in biological research. New York: W.H. Freeman. 887 p.