INAUGURAL ARTICLE by a Recently Elected Academy Member:Why Wolbachia-induced cytoplasmic incompatibility is so common

Cytoplasmic incompatibility (CI) is the most common reproductive manipulation produced by Wolbachia, obligately intracellular alphaproteobacteria that infect approximately half of all insect species. Once infection frequencies within host populations approach 10%, intense CI can drive Wolbachia to near fixation within 10 generations. However, natural selection among Wolbachia variants within individual host populations does not favor enhanced CI. Indeed, variants that do not cause CI but increase host fitness or are more reliably maternally transmitted are expected to spread if infected females remain protected from CI. Nevertheless, approximately half of analyzed Wolbachia infections cause detectable CI. Why? The frequency and persistence of CI are more plausibly explained by preferential spread to new host species (clade selection) rather than by natural selection among variants within host populations. CI-causing Wolbachia lineages preferentially spread into new host species because 1) CI increases equilibrium Wolbachia frequencies within host populations, and 2) CI-causing variants can remain at high frequencies within populations even when conditions change so that initially beneficial Wolbachia infections become harmful. An epidemiological model describing Wolbachia acquisition and loss by host species and the loss of CI-induction within Wolbachia lineages yields simple expressions for the incidence of Wolbachia infections and the fraction of those infections causing CI. Supporting a determinative role for differential interspecific spread in maintaining CI, many Wolbachia infections were recently acquired by their host species, many show evidence for contemporary spatial spread or retreat, and rapid evolution of CI-inducing loci, especially degradation, is common.

Wolbachia, maternally inherited alphaproteobacteria, may be the most common animal endosymbiont, occurring in about half of all insect species as well as other arthropods and nematodes (1). Relatively few Wolbachia infections of arthropods have been characterized for reproductive manipulation or any other effects, but among those tested, approximately half cause cytoplasmic incompatibility (CI) (e.g., see ref. 2 for Drosophila data). CI is defined by elevated embryo mortality when uninfected ova are fertilized by sperm from Wolbachia-infected males (3). CI intensity (i.e., the fraction of embryos killed) varies from a few percent to 100% and depends on Wolbachia genotype, host genotype, and various conditions, including temperature and host age (4–7). CI can also occur in matings of males and females carrying incompatible Wolbachia variants (8–10). CI was first described in the mosquito Culex pipiens and its close relatives (9, 11). The pioneering work of Beckmann and Fallon (12) on a Wolbachia protein found in Culex sperm initiated progress toward identifying pairs of loci that underlie CI in many taxa (reviewed in refs. 13–15). Our analyses address the evolutionary forces determining the prevalence of CI-causing Wolbachia. Although initially associated with Wolbachia (9, 16), other maternally inherited microbes also produce CI (17–21). Our analyses apply to all such microbes, but we focus on Wolbachia because its population biology, molecular biology, and patterns of acquisition are more completely characterized.The prevalence of CI-causing Wolbachia presents a puzzle. As noted by Prout (22) and Turelli (23), natural selection among mutually compatible Wolbachia variants in a host species does not favor CI. As first proposed by Hurst and McVean (24), the prevalence of CI may be more plausibly explained by a process of clade selection in which CI-causing Wolbachia lineages are more likely than non-CI-causing lineages to spread to new host species. Consistent with the data then available (e.g., see refs. 24–26), Hurst and McVean (24) assumed that Wolbachia infections generally decrease host fitness. This now seems doubtful, with increasing evidence, reviewed below, suggesting that many Wolbachia infections are mutualistic. We generalize the Hurst and McVean (24) clade-selection hypothesis, showing that both mutualistic and deleterious Wolbachia variants are more likely to spread to new host species if they induce CI. In support of this hypothesis, we review data indicating that many Wolbachia infections are relatively young (originating on the order of tens of thousands of years ago, long after speciation), that spatial and temporal Wolbachia frequencies within species often vary, and that Wolbachia regularly lose the ability to induce CI while retaining the ability to resist it. These observations suggest regular turnover of Wolbachia infections within and among host species.Hurst (27) proposed that natural selection would favor increased CI, but this conjecture was refuted by algebraic analyses of the fate of Wolbachia variants within individual host populations (22, 23) and metapopulations (28). Those analyses focused on mutually compatible variants that differ in the intensity of CI produced by matings of infected males to uninfected females (i.e., the average fraction of embryos that die because of incompatibility), the fidelity of Wolbachia maternal transmission, and the relative fitness (specifically viability and fecundity) of infected versus uninfected females. Within host populations, there is no selection among Wolbachia variants for increased CI. Specifically, among mutually compatible Wolbachia variants within a population (i.e., females carrying each variant are immune to the CI-inducing effects of the others), natural selection favors the variant whose female carriers produce the largest number of Wolbachia-infected progeny (i.e., product of relative fecundity times fraction of offspring that carry the infection). This is true irrespective of whether males carrying the favored variant produce CI when mated to uninfected females (23). Metapopulation structure, namely small local populations linked by migration, produces weak selection for CI, but very small positive effects on relative fitness (i.e., increases on the order of 10⁻³) generally suffice to overcome the intergroup selection advantage associated with even strong CI (28). Consistent with this prediction, several studies of Wolbachia infections in a wide range of hosts indicate relatively recent loss of function for the loci that cause CI [but typically not loss of functional loci that protect hosts from CI (15, 29, 30)].Because very closely related Wolbachia (separated by 1,000 to 10,000 y) infect distantly related, reproductively isolated host lineages (separated by 1 My to 10 My, e.g., refs. 31–33), processes both among and within host lineages can contribute to differential proliferation of Wolbachia variants across the tree of life (24). Recent data, reviewed below, indicate relatively rapid movement of Wolbachia lineages between host species by a combination of both introgression between closely related species and nonsexual horizontal transmission between more distantly related hosts. Nonsexual horizontal transmission can be mediated by both parasitoids (34) and host plants (35). The turnover of Wolbachia within host species often seems to occur much faster than the timescale of the origin and extinction of host species (32). Hence, to understand Wolbachia evolution, we must consider the frequency dynamics of variants both within individual host species and among host species, specifically the rate of spread to new host species, the duration of typical Wolbachia–host associations, and the persistence of CI within Wolbachia lineages. Debates concerning the relative importance of levels of selection often emphasize discordant selection at different levels (e.g., natural selection within groups may favor selfish behavior, but selection among groups may favor groups with more altruists) (36–38). Understanding CI evolution across Wolbachia lineages is simplified by the fact that there is essentially no selection for or against CI among Wolbachia lineages within individual host species (22, 23, 28). Hence, the maintenance and evolution of CI are plausibly determined by relative movement of Wolbachia lineages among host species and the persistence of Wolbachia infections and CI induction within host species.This interspecific versus intraspecific transmission perspective is explicit in the analyses of Wolbachia pervasiveness by Hurst and McVean (24) and Werren and Windsor (39). Building on the work of Turelli (23) and Prout (22), Hurst and McVean (24) proposed a “reversible evolution” model for CI in which CI-causing Wolbachia invade an uninfected host but are displaced by non-CI-causing variants (resistant to CI), which are then outcompeted by more fit Wolbachia-uninfected cytotypes. This cycle assumes that CI-causing variants impose a greater fitness cost on hosts than non-CI-causing variants, which are implicitly assumed to also reduce host fitness. Hurst and McVean (24) argued that the Wolbachia variants that persist among insect species are those best able to invade new host species through horizontal transmission. Their analyses suggest that deleterious CI-causing Wolbachia persist because CI facilitates invasion of new hosts. We generalize this framework to consider both mutualistic and deleterious Wolbachia, motivated by data suggesting that many, and plausibly most, natural Wolbachia infections are mutualistic, whether or not they induce CI (2, 3, 30, 40–42).Initial field and laboratory studies suggested that Wolbachia might generally reduce host fitness, specifically fecundity (25, 26). As illustrated by Eq. 1, direct fitness effects dominate the dynamics of rare Wolbachia infections, whether or not they cause CI, because CI is effectively nonexistent when Wolbachia-infected males are very rare. The deleterious-Wolbachia paradigm is demonstrably correct for Wolbachia transinfections (i.e., Wolbachia experimentally transferred from one host species to another) that are being used to control insect-vectored diseases of humans (43–45) and plants (46). For these systems, there is an unstable equilibrium frequency that CI-causing variants must exceed before their frequencies tend to increase deterministically through the frequency-dependent advantage associated with CI (Eq. 1). Once established locally, these infections with bistable dynamics can spread spatially (25, 47). But initial local establishment requires purposeful introduction (48, 49) or a genetic drift–like sampling process that gets local frequencies above the unstable equilibrium (50, 51).The Hurst and McVean (24) assumption that naturally occurring, CI-causing Wolbachia are generally deleterious no longer seems plausible. The paradigm shift is based on several observations concerning temporal and spatial variation of Wolbachia frequencies in nature. First, the rate of spatial spread of the CI-causing wRi Wolbachia in both California and Australian D. simulans populations was on the order of 100 km/y (25, 40). This makes sense only if long-distance, human-mediated dispersal can initiate local spread starting from very low frequencies. Bistability produced by deleterious Wolbachia effects precludes this. Indeed, for Wolbachia transinfections that are demonstrably deleterious, such as wMel introduced from D. melanogaster into Aedes aegypti, spatial spread is orders of magnitude slower (on the order of 100 m/y for Ae. aegypti rather than 100 km/y for D. simulans), despite comparable dispersal distances and generation times for D. simulans and Ae. aegypti (43). Second, the non-CI-causing Wolbachia variant wAu was observed to spread through Australian D. simulans; this makes sense only if wAu is mutualistic (40). Third, many Wolbachia that cause little or no CI, or other detectable reproductive manipulation, persist in natural populations [e.g., the variants wMel in D. melanogaster (52, 53); wSuz in D. suzukii (3); wMau in D. mauritiana (30); and the Wolbachia in the three-species D. yakuba clade (54, 55)]. Fourth, we now have several plausible examples of direct fitness benefits associated with Wolbachia, including protection from viruses (e.g., refs. 56–58), nutritional provisioning (e.g., refs. 59 and 60) and various life history effects (61). The temporal and spatial frequencies of Wolbachia infections that cause little or no CI seem most compatible with a balance between positive fitness effects (many of which remain to be identified) and imperfect maternal transmission (53, 62). Because of maternal transmission, we expect Wolbachia to evolve toward mutualism within host lineages (23), and this has been observed over a timescale of decades (63). Hence, it now seems likely that many Wolbachia invade new hosts through mutualism rather than reproductive manipulation. Although CI is not favored within individual host species, we argue that CI enhances spread among host species for both mutualistic and deleterious Wolbachia.The pervasiveness of CI-causing Wolbachia can be understood by analogy to the spread of disease microbes within and among conspecific individuals. This epidemiological perspective on the Wolbachia pandemic among insects was invoked by Werren and Windsor (39) to explain the relative constancy of the fraction of insect species infected (Wolbachia “incidence”) across continental regions. Their model considered only a transmission rate to new host species (T) and a loss rate for infections in host species (L). We extend their model by considering the relative transmission and loss rates for Wolbachia variants that do or do not cause CI, allowing for loss of functional CI loci within Wolbachia lineages (15, 29, 30). Simple models illustrate that CI both increases the transmission rate, T, and decreases the loss rate, L.Epidemiological models, which focus on disease-causing microbe density within host individuals and frequency among conspecific hosts, can be adapted to illuminate the incidence of alternative Wolbachia forms among host species. For instance, among disease microbes, if variants provide immunity to one another, competition favors the variant with the largest R₀, corresponding to “the average number of secondary cases arising from an average primary case in an entirely susceptible population” (64, p. 20). This corresponds to selection among mutually compatible Wolbachia variants favoring a higher T and longer persistence time within each host species. For disease microbes, a classical explanation for the evolution of intermediate virulence, as exemplified by myxoma in Australian rabbits (65), is that there is often a tradeoff between transmission rate and infectious duration (64, 66). For example, increased myxoma titer may increase transmission but accelerate host death. In contrast, no comparable tradeoff, now between Wolbachia frequencies within host species and the duration of Wolbachia infections within those host species, is expected for Wolbachia variants that cause CI. As discussed below, CI-causing Wolbachia variants are expected to be at higher frequencies within host species (producing a higher transmission rate between species) and also to persist longer in their host species than non-CI-causing variants. We illustrate both ideas with simple calculations and simulations. Because so much Wolbachia biology remains unknown, our goal is not to produce a fully parameterized model that predicts the frequency of alternative Wolbachia forms across all insects (or potential arthropod hosts) but simply to present a plausible hypothesis explaining why CI is so prevalent.