6 Corresponding author.


The genetic incompatibility avoidance hypothesis as an explanation for the polyandrous mating strategies (mating with more than one male) of females of many species has received significant attention in recent years. It has received support from both empirical studies and a meta-analysis, which concludes that polyandrous females enjoy increased reproductive success through improved offspring viability relative to monandrous females. In this study we investigate whether polyandrous female Drosophila simulans improve their fitness relative to monandrous females in the face of severe Wolbachia-associated reproductive incompatibilities. We use the results of this study to develop models that test the predictions that Wolbachia should promote polyandry, and that polyandry itself may constrain the spread of Wolbachia. Uniquely, our models allow biologically relevant rates of incompatibility to coevolve with a polyandry modifier allele, which allows us to evaluate the fate of the modifier and that of Wolbachia. Our empirical results reveal that polyandrous females significantly reduce the reproductive costs of Wolbachia, owing to infected males being poor sperm competitors. The models show that this disadvantage in sperm competition can inhibit or prevent the invasion of Wolbachia. However, despite the increased reproductive success obtained by polyandrous females, the spread of a polyandry modifier allele is constrained by any costs that might be associated with polyandry and the low frequency of incompatible matings when Wolbachia has reached a stable equilibrium. Therefore, although incompatibility avoidance may be a benefit of polyandry, our findings do not support the hypothesis that genetic incompatibilities caused by Wolbachia promote the evolution of polyandry.

Despite decades of research, one of the most widely studied topics in evolutionary and behavioral ecology focuses on the adaptive significance of the high incidence of polyandry among females of almost all taxa (e.g., Birkhead and Møller 1998). This is largely because it remains unclear why some females mate with more than one male when they do not appear to receive direct benefits from mating, such as nuptial gifts, paternal care, or access to territories. This is perplexing because mating is generally considered costly to females and many females do not need more than one ejaculate to guarantee successful lifetime fertility. Indirect genetic benefits to females are often evoked as an explanation for polyandry because, by mating with multiple males, female fitness could be increased by producing offspring of superior genetic quality. Evidence supporting this hypothesis is accumulating (e.g., Fisher et al. 2006) although it is often difficult to show that patterns of paternity directly reflect increased offspring genetic quality (Edvardsson et al. 2007). Polyandry may also enable females to avoid direct costs associated with being genetically or reproductively incompatible with their mates (genetic incompatibility avoidance hypothesis) (Zeh and Zeh 1996, 1997, 2001, 2003; Jennions and Petrie 2000; Tregenza and Wedell 2000). This hypothesis is supported by a growing body of work examining incompatibilities in the context of inbreeding avoidance and those that occur randomly in certain male/female combinations (Tregenza and Wedell 1998, 2002; Newcomer et al. 1999; Stockley 1999; Foerster et al. 2003; Bretman et al. 2004; Marshall and Evans 2005; Simmons et al. 2006) but see Brown et al. (2004) and Jennions et al. (2004). However, although a recent meta-analysis revealed a small but significant positive effect of polyandry on embryo viability, there is little clear-cut evidence that this arises because fertilizations are biased toward genetically compatible sperm (Simmons 2005). Additionally, with the exception of studies of inbreeding avoidance, little is known about the origins, mechanisms, or actual risk of “incompatibility” to females. Further studies of polyandry are therefore required to elucidate the benefits received by females and also the selective power of genetic incompatibilities in promoting polyandry.

Polyandry is hypothesized to reduce the fitness costs of genetic incompatibilities by improving the likelihood that a female's eggs are fertilized by the sperm of compatible males. However, this depends on females having at least partial control over paternity or possessing a mechanism that enables them to bias paternity toward compatible males (Zeh and Zeh 1997; Colegrave et al. 2002; Lorch and Chao 2003). This could be achieved by exploiting the outcome of sperm competition, providing the sperm of compatible males fare better in competition with the sperm of incompatible males, or by utilizing some form of postcopulatory choice. In the latter scenario, females may be able to detect differences in the “compatibility” or suitability of sperm using interactions at the molecular and cellular level between sperm/ejaculates and eggs/female reproductive tract (Zeh and Zeh 1996, 1997; Jennions and Petrie 2000; Tregenza and Wedell 2000; Simmons 2005). These sorts of interactions commonly prevent self fertilization in plants (Cowan et al. 2000; Bernasconi 2003) but have also been found in animals, for example, the compound ascidian Diplosoma listerianum, where fertilizations appear to be biased toward genetically less similar mates (Bishop 1996; Bishop et al. 1996).

Organisms manipulated by the maternally inherited, intracellular bacterium Wolbachia pipientis, which causes predictable and strong reproductive incompatibilities, provide excellent model systems for examining the benefits of polyandry in terms of incompatibility avoidance. Wolbachia causes reproductive failure known as cytoplasmic incompatibility (CI) (O'Neill et al. 1997) when the sperm of infected males fertilize the eggs of uninfected females. Although Wolbachia is not found in or transmitted by mature sperm, a large proportion of uninfected eggs fertilized by sperm from infected males undergo abnormal mitosis and die (Lassy and Karr 1996; Tram and Sullivan 2002). Crosses between infected females and either infected or uninfected males are viable. In the fruit fly Drosophila simulans infected with the Riverside variant of Wolbachia, close to all (≥95%) of an uninfected female's offspring die as a result of CI in crosses involving virgin-infected males. This makes it easy to determine the proportion of offspring that die as a result of genetic incompatibilities between parents in this species because, although embryos die as a result of CI, these eggs are laid. In addition to manipulating host reproduction, Wolbachia infections are often associated with costs to hosts. For instance, in some Drosophila species, Wolbachia is known to decrease the fecundity (Hoffmann et al. 1990) and fertility of its host (Snook et al. 2000), and has variable effects on host longevity (Fry et al. 2004). Importantly for investigations of the benefits of polyandry, Wolbachia infection is associated with decreased sperm competitive ability in D. simulans (Champion de Crespigny and Wedell 2006).

The population dynamics of Wolbachia have been modeled extensively and models typically predict Wolbachia to spread to, or close to, fixation within populations within a relatively short period of time (Caspari and Watson 1959; Hoffmann and Turelli 1997) owing to the large reproductive advantage of infected females over uninfected females. The critical parameters of Wolbachia infection dynamics are thought to be the maternal transmission fidelity of Wolbachia and cost of infection to infected females in terms of reduced fecundity. Variation in these parameters affect the minimum threshold frequency of Wolbachia required for successful invasion and also the stable equilibria of Wolbachia (the maximum frequency it attains). In nature, Wolbachia infection frequencies are often polymorphic, with the stable equilibria at intermediate frequencies within populations (Turelli and Hoffmann 1991, 1995; Vala et al. 2004). The persistence of these intermediate frequencies may be explained by various factors, including maternal transmission fidelity (Turelli and Hoffmann 1995), natural curing events (Clancy and Hoffmann 1998), and survival of some uninfected eggs in matings with infected males.

Uninfected individuals may also remain in populations because the reproductive cost of Wolbachia generates selection on the host to modify the action of Wolbachia. In terms of behavioral adaptations, this possibility has been modeled in the context of the evolution of precopulatory mate preferences (Champion de Crespigny et al. 2005). Similarly, the impact of male-killing Wolbachia on the evolution of mate choice has also been considered (Randerson et al. 2000) and both models predict mate preferences can evolve (note, however, that the dynamics of male killing and CI are so very different that conclusions stemming from models relating male-killing Wolbachia to the evolution of mate choice [Randerson et al. 2000] need have no generality). Despite models suggesting that Wolbachia can create conditions for the evolution of mate choice, precopulatory mate preferences based on Wolbachia infection status seem rare in nature (Hoffmann and Turelli 1988; Hoffmann et al. 1990; Wade and Chang 1995; Jiggins et al. 2002; Champion de Crespigny and Wedell 2007), although they have been demonstrated in spider mites (CI-inducing Wolbachia) (Vala et al. 2004) and woodlouse (feminizing Wolbachia) (Moreau et al. 2001). The potential for Wolbachia to promote evolutionary changes in host reproductive behavior (e.g., mating frequency) or postcopulatory adaptations, (e.g., biased paternity), in hosts has received little attention in the form of theoretical models. In addition, host behaviors are not usually incorporated into Wolbachia population dynamics models (but see Hoffmann and Turelli (1997) and Champion de Crespigny et al. (2005)). However, if polyandry undermines the transmission advantage of Wolbachia it has the potential to prevent or slow the spread of Wolbachia within populations.

Although the advantages of female multiple mating have not been modeled in the context of Wolbachia-associated incompatibilities, there have been three previous attempts to determine whether genetic incompatibilities in general promote polyandry (Haig and Bergstrom 1995; Colegrave et al. 2002; Lorch and Chao 2003) but see also Hosken and Blanckenhorn (1999). These models predict that, provided sperm competition or some other process favors “compatible” males, polyandry has a selective advantage even if multiple mating imposes some cost on females. However, none of the models allow levels of incompatibility to vary, but rather assume a fixed rate of incompatibility that is not permitted to evolve under the influence of polyandry. Furthermore, the models do not consider the fate of a modifier of female mating behavior over time but rather examine the effect of being polyandrous on mean fitness in a given scenario. Because the relationship between polyandry and levels of incompatibility will be dynamic, these previous models are unable to evaluate whether genetic incompatibilities generate evolutionary change in female reproductive strategies, that is, change from monandry to polyandry, and are restricted to describing which strategy (monandry or polyandry) has a selective advantage under a given set of circumstances.

Here, we explore the benefits of polyandry in the face of genetic incompatibilities caused by Wolbachia, on both empirical and theoretical fronts. First, we test experimentally whether polyandry increases uninfected female D. simulans reproductive success. If females benefit from polyandry when they are susceptible to CI, polyandrous females should suffer reduced fitness costs attributable to Wolbachia compared to monandrous females. Polyandrous uninfected females should therefore have increased offspring production relative to infected females than singly mated uninfected females. Second, we compare the remating rates of infected and uninfected females. Uninfected females are predicted to have shorter remating intervals than infected females, as this promotes sperm competition and may increase the likelihood of fertilizing eggs with compatible sperm. We use our empirical findings to develop models that enable us to test the following theoretical predictions. First, provided polyandry decreases the transmission advantage of Wolbachia, polyandry may inhibit or prevent the spread of Wolbachia within populations. Second, Wolbachia may promote the spread of polyandry in the host.

Materials and Methods


The D. simulans used in these experiments derive from iso-female lines infected with Wolbachia that were originally collected in Riverside, California but which have since been maintained in laboratory populations for many years. Uninfected flies were obtained by rearing the offspring of infected flies (>80 females) on food treated with antibiotics (tetracycline hydrochloride) over two consecutive generations approximately one year prior to the experiment (see Champion de Crespigny and Wedell [2006] for detailed methodology). The infection status of the stock populations was confirmed by PCR. Virgin flies were acquired by collecting eggs from stock populations and raising larvae in density controlled vials (five larvae mL−1Drosophila food medium) at 25°C on a 12:12 h light/dark cycle. Each vial contained 25 larvae and the vials were inspected at 6-h intervals during the light cycle for newly eclosed adults. New adults were chilled on ice, sexed, and the sexes placed in separate density controlled vials containing food.


To assess the effect of polyandry versus monandry on female offspring production, infected and uninfected female D. simulans were randomly allocated to one of two experimental treatments in which females were mated to either one (monandry) or two different (polyandry) males. Within these treatments, females were further randomly allocated to treatments according to the infection status of their mate(s) (Table 1) so that all 12 possible mating combinations based on male and female infection status were performed.

Table 1.  Experiment treatments and final sample sizes (N) for experiments. (A) Monandry versus polyandry; and (B) female remating interval. In experiment A, treatments affected by cytoplasmic incompatibility are marked with “CI.” All other crosses are entirely compatible matings. N1 is the final sample size of the original experiment. N2 is the final sample size of the repeated section of the original experiment.
A. Mating strategyMonandry versus polyandry experiment
Female infection statusMale 1 infection statusMale 2 infection statusCIN1N2
  Uninfected  26 
 UninfectedInfected CI31 
  Uninfected  18 
PolyandryInfectedInfectedInfected 18 
  InfectedUninfected 18 
UninfectedInfected 16 
  UninfectedUninfected 19 
  UninfectedUninfected 1618
B.Female remating interval experiment
Female infection statusMale 1 infection statusMale 2 infection status N 
 InfectedInfectedUninfected 22 
  UninfectedUninfected 27 
 UninfectedInfectedUninfected 27 
  UninfectedUninfected 31 

Females were placed in individual vials containing food approximately 15 h prior to their first mating. All females were virgins and three-day old when they mated for the first time. The experiment commenced within 2 h of “lights on” when a single, virgin, two-day-old male (anesthetized briefly on ice) was transferred into each females' vial. The males typically recovered within a minute of transfer and soon afterwards began courtship. All copulations were observed and the duration recorded. Males were removed from the vials shortly after mating was completed. Females were discarded if mating did not take place within 3 h.

Following mating and male removal, all females were maintained at 25°C for 24 h. Monandrous females were then transferred to new vials containing food for a further 24 h (Fig. 1). Polyandrous females, on the other hand, were provided with a new virgin two-day-old male and copulations were observed as above (Fig. 1). Males were removed after mating, or if 3 h elapsed without mating, and the females were immediately transferred to fresh vials containing food for a further 24 h. After the second 24-h interval was completed all females (monandry and polyandry) were transferred to new vials containing food for 48 h. This was repeated for a further 48 h so that females laid eggs for a total of 144 h (6 days) following their first mating. At the end of the final 48-h interval females were frozen for body size measurements. Female body size was estimated by measuring wing length: the distance between the intersection of the anterior cross vein and the longitudinal vein 3 (L3), and the intersection of the L3 with the distal wing margin (Partridge et al. 1987a, b; Champion de Crespigny and Wedell 2006).

Figure 1.

Experiment timeline: schedule of mating(s) and vial exchange. The dotted line indicates the period of time eggs were counted in the vials of uninfected females with incompatible matings. The dashed line indicates the period of time for which total offspring production is calculated for all females.

This experimental design generates four vials containing eggs and offspring per female (Fig. 1). Two vials containing eggs were collected in the 48 h after the first mating. This enabled us to compare female offspring production immediately after the first and second (where applicable) mating over identical periods of time (24 h). These two vials were inspected for eggs, and if the female was in a treatment that involved an incompatible cross (i.e., uninfected female mated to at least one infected male), the eggs were counted allowing estimation of the degree of CI induced by males (number of offspring produced/number of eggs laid). However, to reduce the time spent counting eggs, if both mates were compatible (i.e., uninfected females mated only to uninfected males or infected females mated to any male) we simply confirmed that females were laying eggs. Females were discarded from the analyses if they produced a total of < 10 offspring in treatments only involving matings with compatible males, or laid < 10 eggs in treatments where uninfected females mated to at least one incompatible male. The vials were maintained at 25°C with a 12:12 h light/dark cycle for 15 days following the commencement of oviposition. At this point all surviving offspring had eclosed within the vial and the number of offspring in each vial was counted as a measure of female reproductive success.

To confirm our findings with respect to CI induction, we repeated part of this experiment (hereafter referred to as the “replicate” experiment). Specifically, we mated uninfected females at 24-h intervals to two males in all combinations of infection status (exactly as in the original experiment: Table 1). For 24 h immediately following each mating, females laid eggs in individual vials containing food that had been dyed with red food coloring to better visualize eggs. We counted the number of hatched and unhatched eggs 48 h after the start of the laying period. Eggs hatch 24 h after being laid so all eggs should have hatched prior to being scored. The vials were maintained exactly as in the original experiment and the number of offspring that survived until adulthood was counted. The number of offspring that survived until adulthood was compared to the number of hatched eggs and was highly correlated (Spearman: r= 0.97, n= 80, P < 0.0001). This indicates both that the counting method is reliable and that posthatching larval mortality is low. When calculating CI induction in the replicate experiment, we first took into account natural hatching failure. This was achieved by scoring the hatching success of uninfected females mated to two compatible males. Approximately 12% of eggs produced by females mated to two compatible males failed to hatch. This figure was used to correct the hatching success of the remaining treatments. CI induction was then calculated by dividing the number of hatched eggs by the total number of eggs laid.

In both the original and replicate experiments, CI was used to assign and compare the paternity of competing infected and uninfected males in a manner analogous to the sterile male technique commonly employed in studies of sperm competition, for example, Schneider et al. (2000). Following the application of the correction factor described above, unhatched eggs were assumed to be fertilized by the incompatible, infected male, whereas hatched eggs were assumed to be fertilized by the compatible, uninfected male.


We investigated whether female remating interval (the time between the termination of the first mating and start of the second mating) is affected by whether females or their mates are infected with Wolbachia. Three-day-old infected and uninfected virgin females were randomly allocated to one of four possible experimental treatments in which they mated with either an infected or an uninfected male (Table 1). All males were virgin and two-day old. The females were enclosed in individual vials 1 h prior to the experiment commencing when a single male (anesthetized on ice) was transferred into each female's vial. The copulations were observed and the time each mating started and finished was recorded. Males were removed immediately after mating had finished, or if 2 h elapsed without mating. Approximately 18 h after the first copulation, all females were offered a single two-day-old uninfected male and the time of remating and copulation duration was recorded. Females were allowed 2 h in which to remate. Although equal numbers of replicates (n= 32 per treatment) were set up for each treatment, some females did not mate at the first opportunity and hence were discarded from the analyses. The final sample sizes are reported in Table 1.


Polyandry and female offspring production

Although females in the polyandry treatment were provided with two mating opportunities, approximately 20% of females did not remate. These generated an additional group (to the monandry treatments) of females with only one mating. Females that “choose” to mate once may differ from females with enforced monogamy in a variety of reproductive characteristics. However, we found no significant differences in terms of offspring production (all P > 0.09), copulation duration (ANOVA: F7, 98= 1.259, P= 0.279) or female size (ANOVA: F7, 96= 1.268, P= 0.275) between the matching treatments in these two groups. Note, the offspring production data were analyzed in separate treatment pairs because incompatible crosses make the data strongly nonnormal and resistant to transformation. Due to the lack of differences between the two groups, the data were pooled and all analyses involving “monandry” include both datasets. All data were tested for normality using Kolmogorov–Smirnov tests and transformed where appropriate. The total offspring production data were log transformed (log10(x + 1)). If data transformations did not improve the normality of the data nonparametric tests were used in the analyses. In analyses utilizing general linear models (GLMs), all two-way and three-way interactions were included in the model and nonsignificant terms were removed following stepwise deletion of the least significant terms. Interactions were removed before main effects. The models presented are the minimal models retaining all significant effects. The data were analyzed in SPSS 14.0. All data presented in the text are means ± standard error of the means.

Wolbachia and female remating interval

The female remating data were analyzed in two ways. First we analyzed the data with respect to whether females remated using χ2 analyses. However, this method is limited in its sensitivity and does not accommodate covariates. Therefore, female remating intervals were also analyzed using Survival Analysis in GLIM4 (Royal Statistical Society) assuming a Weibull distribution (Crawley 1993). This method allows us to take into account the speed of remating and includes data from females that did not remate. This is therefore a more sensitive analysis of female willingness to remate. The remating data were adjusted so that the remating interval of each female was expressed as the number of minutes greater than the quickest remating interval. Following this methodology, the fastest female to remate did so at time equals 1 min. The data were adjusted in this way to improve the fit with the assumed distribution. The terms included in the model were as follows: female infection status and the infection status of the first male to mate were fixed factors and the duration of the first copulation was a covariate. All two-way interactions were modeled. Terms were removed following stepwise deletion of the least significant terms as above. The validity of the error structure was verified using the Kaplan macro in GLIM4 (Crawley 1993).



Monandry versus polyandry: when all mating combinations are compatible (no CI)

We compared the offspring production of polyandrous and monandrous females that were solely mated to compatible males (uninfected females mated to uninfected males or infected females mated to either infected or uninfected males) to determine whether polyandry is generally advantageous to D. simulans females. When only compatible pairings were considered, polyandrous females produced significantly more offspring than monandrous females (General linear model (GLM): F1, 152= 26.67, P < 0.001) and offspring production was neither affected by female size (GLM: F1, 152= 0.17, P= 0.681) nor female infection status (GLM: F1, 152= 0.05, P= 0.822).

Monandry versus polyandry: incompatible versus compatible matings

As expected, CI dramatically reduced the reproductive success of uninfected females. Mating with a single incompatible male reduced uninfected female offspring production by 95% compared to females mated once to a compatible male (Fig. 2A). Likewise, polyandrous uninfected females suffered reduced reproductive success when they mated to incompatible males. However, there was a strong effect of mating order on total offspring production (Fig. 2B). Uninfected females mated to two incompatible males produced 95% fewer offspring, and uninfected females mated first to a compatible male followed by an incompatible male produced 75% fewer offspring, than uninfected females mated to two compatible males. However, females that mated first to an incompatible male followed by compatible male were able to recoup their fitness entirely and their offspring production was not significantly different from uninfected females mated to two compatible males (Mann–Whitney U: Z =−0.406, n= 102, P= 0.685). It is important to note that although offspring production of uninfected females mated first to a compatible (uninfected) followed by an incompatible (infected) male was substantially reduced, this reduction is attributable both to CI but also the number of eggs laid. Females in this treatment laid significantly fewer eggs than uninfected females mated to either two infected males or an infected male followed by an uninfected male (ANOVA: F2, 40= 8.311, P= 0.001). The reasons for this are not clear and this pattern was not repeated in the replicate experiment (ANOVA: F3,79= 1.002, P= 0.397). Therefore, our estimate of the impact of CI on the offspring production of uninfected females mated first to a compatible and second to an incompatible male is probably an overestimation because these females did not invest in reproduction (through egg laying) to the same extent as females in the other treatments.

Figure 2.

The mean ± SE offspring production of (A) monandrous females and (B) polyandrous females. Dark gray bars indicate crosses where all males are incompatible with females. Light gray bars indicate crosses where one of two males is incompatible. White bars indicate entirely compatible crosses. Female infection status is described in the lower divisions of the x-axes. Male infection status is given in the higher divisions of the x-axes. In B, male infection status and order of mating is described as follows I = infected male, U = uninfected male. The first letter indicates the infection status of the first male to mate and the second letter the infection status of the second male.

Monandry versus polyandry: the impact of CI (offspring production of uninfected females relative to offspring production of infected females)

To establish whether mating with more than one male enables females to reduce the impact of CI-inducing Wolbachia, we compared the impact of CI on the offspring production of uninfected females, between monandrous and polyandrous females. To do this we separately expressed the impact of CI for monandrous and polyandrous females as the mean proportion of offspring produced by uninfected females relative to infected females (see below). This calculation is essential because it takes in to account any differences in offspring production between monandrous and polyandrous females that occur because polyandrous females received two ejaculates whereas monandrous females received only one. The impact of CI was calculated as follows with the results based on our data (mean ± SE) presented to the right


where, A is the mean number of offspring produced by uninfected females mated to either one (monandry) or two (polyandry) infected i or uninfected u males; and B is the mean number of offspring produced by infected females mated to either one (monandry) or two (polyandry) infected i or uninfected u males. The order of males in the polyandry equation corresponds to the order of mating. Polyandry is advantageous if,


This calculation assumes that females are equally likely to mate with either an infected or uninfected male, as is the case with our experimental design. Under this scenario, we found the above to be true and therefore that, by mating with more than one male, on average polyandrous uninfected females produced almost twice as many offspring as monandrous uninfected females.

Polyandry: CI induction and patterns of paternity in polyandrous uninfected females

Although polyandry on average reduces the reproductive burden of CI to uninfected females, for individual females there is a strong effect of order of mating in terms of male infection status. In other words, the reproductive success of uninfected polyandrous females depends on when they mate with infected, incompatible mates (Fig. 2B). If polyandry is advantageous to females in terms of reducing the incidence of CI, then the mean hatching success, (H), of uninfected females that mate with both infected and uninfected males, (Hiu, Hui) must be greater than the mean hatching success of uninfected females that mate solely with either infected (Hii) or uninfected males (Huu) (see equation below where the order of mating and male infection status (i/u) is stated). The results of our experiments are presented beneath the equation.


In both experiments we found the above to be true and therefore that polyandry enables females to reduce the incidence of CI. We therefore expect that polyandrous uninfected females that mate to both an infected and uninfected male are able to exploit the outcome of sperm competition to bias the paternity of their offspring toward the uninfected compatible male. In both the original and the replicate experiment infected, incompatible males sired significantly fewer offspring than uninfected, compatible males when their sperm competed for fertilizations (Original experiment: t-test: t21= 2.11, P < 0.025; Replicate experiment: t-test: t41= 3.89, P < 0.005). When in the second male role in the original experiment, infected second males sired 33.4 ± 26.8% of the offspring whereas uninfected second males sired 95.2 ± 16.1%. In the replicate experiment infected males sired 42.8 ± 8.7% of the offspring compared to 78.8 ± 4.3% sired by uninfected males when second to mate. It is therefore clear that polyandry reduces the cost of CI because the ejaculates of infected males are competitively inferior to the ejaculates of uninfected males in sperm competition.


We investigated whether infected and uninfected female D. simulans vary in their propensity to remate based on both their own infection status and that of their first mate. We found no effect of female infection status (Pearson: χ2= 1.11, df = 1, P= 0.292) nor the infection status of the first male to mate (Pearson: χ2= 0.42, df = 1, P= 0.517) on the total numbers of females that remated. However, when we examined the speed in which these females remated we found that uninfected females had shorter remating intervals than infected females (χ2= 5.58, df = 1, P= 0.018; Fig. 3). Female remating was also influenced by male infection status. Contrary to expectation, females that mated to infected males were in general slower to remate (χ2= 9.72, df = 1, P < 0.002; Fig. 3). There was no effect of copulation duration on female remating interval (χ2= 0.54, df = 1, P= 0.462) and no significant interactions.

Figure 3.

The mean ± SE remating interval of female D. simulans based on female infection status and the infection status of the first male to mate. Gray bars are infected males and white bars are uninfected males. There is a significant effect of male infection status (P < 0.002) and of female infection status (P < 0.018) on female remating interval.


CI under monandry

Let us consider adult females, a proportion p of which are infected with Wolbachia and q= 1 –p that are not. Because sons obtain Wolbachia exclusively from their mothers, by knowing the frequency of Wolbachia in females we can assume we know the frequency in males. If we suppose mating is random, that a proportion a of the progeny of an infected female inherit Wolbachia (maternal transmission fidelity) and infected females suffer a fitness cost s owing to reduced fecundity (Hoffmann et al. 1990), then the frequency of Wolbachia in the next generation is given by


where w is the sum of the right-hand sides and b is the proportion of uninfected eggs that survive when a male has Wolbachia. The equilibria can be solved using p′=p. This resolves to two equilibria other than p= 0. These are


where A=a(b− 1), B= (1 +as− b), C=a(1 − s) − 1.

For a cost s= 0.05 and maternal transmission fidelity rates 0.9, 0.95, 0.98, and 1, these equilibria, for varying b, are given in Figure 4. Increasing b and thereby reducing the ability of Wolbachia-infected sperm to induce CI, which occurs as males age and after mating (Karr et al. 1998; Reynolds and Hoffmann 2002), has only a marginal effect on the stable equilibria (maximum infection frequency) achieved by Wolbachia when the maternal transmission rate is high (none when a= 1.0). Moreover, when maternal transmission rates are high (0.98 – 1.0), great variation in CI induction (b, 0 − 0.4), has practically no effect on the invasion conditions of Wolbachia. It is only when CI induction is really inefficient (i.e., high b) that Wolbachia invasion becomes much harder and the minimum threshold frequency for successful invasion of Wolbachia is substantially raised. Note that p*= 1, if a= 1 or b= 0. However, reducing the ability of infected sperm to induce CI lengthens the time taken for Wolbachia infections to reach their stable equilibria.

Figure 4.

The effect of variation in strength of CI induction (proportion of offspring that survive CI) (b) on the equilibrium frequency of Wolbachia (P*), for four values of maternal transmission fidelity (a= 1, 0.98, 0.95, 0.9). For each value of a, at any given value of b, there exist two equilibria. The lower value equilibrium represents the minimum frequency that Wolbachia must attain to be able to invade the population. For example, for a= 0.9, b= 0.2, approximately 20% (p*= 0.2) of the population must have Wolbachia for Wolbachia's frequency to deterministically increase. For any given value of a, there exists an upper value for the inefficiency of CI induction (b) consistent with the presence of Wolbachia in the population. For a= 0.9, this value is just under b= 0.6, for a= 0.95, this value is just over b= 0.7. Should the minimum threshold be reached then the population evolves to the upper stable equilibrium level (the upper part of each curve). From this plot we conclude that the lower the maternal transmission fidelity the higher the minimum threshold frequency of Wolbachia required for invasion and the lower the upper stable equilibrium.

Can polyandry in conjunction with sperm competition inhibit the spread of Wolbachia?

Let us now consider what might happen in a polyandrous population. Here we must make assumptions about mating order effects on the fertilization success of males in two scenarios; first when females mate to two males of the same infection status and second when females mate to two males of different infection status, where we expect sperm competition to play an important role in determining uninfected female reproductive success. In scenario one (both males are infected or both uninfected), let us assume that the first male to mate will on average fertilize a proportion e of the eggs with the second male fertilizing 1 −e. Now consider an uninfected female that mates first to an infected male and second to an uninfected male. We may now assume that e (1 –g1) of the first males' sperm are used and 1 –e(1 –g1) of the second males' sperm are used. For g1= 0 (no sperm competition), this resolves to e and 1 –e as before and at the limit, g1= 1, only the uninfected males' sperm are employed. Similarly, if the uninfected male inseminates first, a proportion (1 –e)(1 –g2) =h are fertilized by the second male and 1 –h of eggs are fertilized by sperm from the first, Wolbachia-uninfected, male. We can then write:




and W is the sum of the right-hand sides.

Solving for p′=p reveals the positions of the equilibria. Notice that if g1=g2= 0 then z resolves to p2b + q2+ pq(1 + b). Hence, in the absence of sperm competition, mating order effects are irrelevant and the parameter e disappears. Moreover, g1=g2= 0, then the equilibrium solutions under polyandry are identical to those under monandry and therefore polyandry per se has no effect on CI induction. In other words, the average rate of CI induction for polyandrous females (irrespective of mating order) is the same as for monandrous females.

However, if we permit differential sperm competitive ability between infected and uninfected males (g1 and g2 > 0), then the equilibrium solution changes. The resulting cubic equations can be solved analytically. The results are best viewed graphically. First we can ask about the effect of sperm competition on the equilibrium conditions for given cost (s), mating order effects (e), and maternal transmission fidelity (a). As a is probably around 0.98 (Hoffmann et al. 1990), e approximately 0.25 (i.e., 75% of sperm used in crosses in which both males are uninfected come from the second male (Price 1997)), we can examine the effects of varying g1 and g2 with varying b. For simplicity we will assume g1=g2. This can be seen below for b= 0.01, 0.2, and 0.4 (Fig. 5A).

Figure 5.

The effect of sperm competition (g) on the equilibrium frequency of CI-inducing Wolbachia (P*), for: (A) three values of CI induction strength (proportion of offspring that survive CI) (b= 0.01, b= 0.2, b= 0.4). The lowest of the lower equilibrium lines is for the most potent CI-inducing Wolbachia, b= 0.01. We assume maternal transmission fidelity, a= 0.98, mating order effects, e= 0.25 (75% of eggs are fertilized by the second male when females mate to two males of the same infection status), and cost of infection s= 0.05. Note that at g= 0, the conditions are identical to those found under monogamy; and (B) three values of maternal transmission fidelity (a= 0.98, 0.95 and 0.9). The lowest and highest lines are the two equilibrium frequencies (unstable and stable, respectively, with the highest vertical transmission rate). We assume b= 0.4, e= 0.25, and s= 0.05. At g= 0 the equilibrium is as under no sperm competition.

One result is most notable. Namely that, increasing competitive ability of uninfected sperm over infected sperm (g1, g2 tends to 1), the invasion conditions (minimum threshold frequency that must be achieved for successful invasion) for Wolbachia become ever more restrictive, although the upper stable equilibrium of Wolbachia is largely unaffected. From this we conclude that with a competitive ability of uninfected sperm over sperm from Wolbachia-infected males, polyandry can be a defense against invasion by CI-inducing Wolbachia. The logic here is fairly transparent. On invasion of Wolbachia, most polyandrous matings involving CI-inducing males will be with one male infected, one not. With sperm competition, the CI-inducing males have less killing effect because they are poor sperm competitors and polyandry hence acts to in effect increase the value of b, the proportion of eggs surviving. This is one of very few examples demonstrating that the spread of what is essentially a sexually transmitted disease can be inhibited by mating frequently with many different males. However, if Wolbachia successfully invades, at the upper stable equilibrium infection frequency, uninfected females will be unlikely to mate with both infected and uninfected males (most will be pairs of infected males). Hence the upper stable equilibrium of Wolbachia is largely unaffected by polyandry. Similarly, when we consider the effect of sperm competition on the invasion and stable equilibrium level for different levels of maternal transmission, the invasion conditions are profoundly worsened but the stable equilibrium is largely unchanged (Fig. 5B).

Note here, however, a further corollary of the above results. We have assumed that sperm competition favors the sperm of the uninfected males. Should it by contrast enable preferential fertilization by the infected male (g1=g2 < 0) then the conditions for invasion are relaxed. This finding supports the conclusions reached by Hoffmann and Turelli (1997) in their models of the effect of infected males having an advantage in sperm competition on Wolbachia population dynamics. This is one of the few forces identified to date that can aid invasion of Wolbachia that induce CI but, given no evidence for this in this study (but see Wade and Chang (1995)), this must be considered a hypothetical possibility at most.

Can Wolbachia promote the evolution of polyandry?

While the above results suggest that polyandry, coupled with sperm competition favoring the sperm of uninfected males, might be a bar to invasion of CI-inducing Wolbachia, we can also ask the inverse question: does the presence of Wolbachia enable the spread of polyandry? To investigate this we consider a modifier allele that increases the rate of polyandry. Intuitively this might be likely to invade due to the improvement in the fitness of polyandrous females resulting from sperm competition favoring the sperm of uninfected, compatible males, thereby reducing mortality rates of uninfected eggs. If we imagine that polyandry does not impose any costs on females, then an infected female might benefit by having fewer of her uninfected eggs being killed by CI. Likewise an uninfected female will see more of her progeny survive for the same reason. For these reasons we may also expect to see the modifier in linkage disequilibrium with the uninfected individuals. As males without Wolbachia do not induce CI, the modifier may also be expected to be associated with larger broods on average.

To investigate this possibility we considered a modifier, M, in a haploid species which when in a female ensures mating with two males chosen at random from the population. Females with the m allele are monandrous. The modifier allele we assume to be nuclear and not sex linked. In a male the modifier is associated with no unusual activity. Let us also suppose that there exists a cost, t, to polyandry. We can then consider the fate of the four haploid types (MI, MU, mI, and mU, where I and U indicates infected and uninfected and M and m are the two alleles at the modifier locus). These four types exist at frequencies x1, x2, x3, and x4 respectively. The recursions for such a model are extensive and the tcl script to simulate them is provided as online Supplementary material. To analyze the dynamics of the situation we consider the fate of the modifier by simulation. The modifier is introduced at a sum frequency of 0.001 in linkage equilibrium with the cytotypes, these being assumed to exist at the stable equilibrium frequency determined under monandry (see above). By inspection we find that if invasion of the modifier is possible, fixation is also possible.

The results of extensive simulations (8400 in total), as a function of t, the cost of the modifier, are presented in online Supplementary Figure S1. From the simulations several results are clear. First, for the modifier to spread there must be (1) both polymorphism of Wolbachia and (2) sperm competition favoring the wild-type sperm (g > 0). To fulfill the first condition the vertical transmission rate (a) must be less than 1 and b > 0, that is, Wolbachia must both have imperfect vertical transmission and imperfect CI induction. Under these conditions spread to fixation of the cost-free modifier occurs readily.

It would however be a mistake to suppose that this provides a realistic model for the evolution of polyandry. The most important weakness of this model is that invasion of the modifier is easily stopped if the cost of modification is even slight. The reason for this is straightforward. The modifier gains in matings when one male is infected and one uninfected and when it too is associated with uninfected eggs. However, although this is not necessarily the case for all species infected with CI-inducing Wolbachia (Vala et al. 2004), the stable equilibria of Wolbachia in D. simulans, typically occurs when 90% or more of the individuals are infected (Turelli and Hoffmann 1991). Hence the modifier is rarely associated with uninfected individuals and females usually mate with infected males. Therefore, under this scenario the modifier suffers the cost of polyandry with very little benefit, most of the time. If the equilibrium frequency of Wolbachia is of the order of 98%, then even with an extreme sperm competition advantage for uninfected males (g = 1), the modifier benefits from polyandry in fewer than 4% of matings. This benefit is maximally realized when the female is herself uninfected, which in turn is only 2% of the time. So a sizeable benefit is only realized in less than 0.1% of matings and polyandrous females suffer the costs of polyandry regardless. Therefore, the domain of cost of polyandry space that is consistent with invasion is limited in the extreme. Moreover, when a cost of polyandry of 0.01 (i.e., 1% cost) was consistent with invasion of the modifier allele the minimum threshold frequency that Wolbachia must assume to permit its invasion is unrealistically high (e.g., 40% or more). These conditions are associated with low maternal transmission fidelity (a) and low CI induction rates (high b). More generally, for any given parameter value, the cost of a modifier consistent with invasion is proportional to the frequency of uninfected individuals at monandrous equilibrium (Fig. 6). For relatively realistic parameter space, the maximum cost the modifier can suffer is very limited (Fig. 6). For this reason we consider this model unrealistic as this parameter space is too confined.

Figure 6.

The minimum equilibrium frequency of uninfected individuals in the population under the most permissive conditions of the alternative variables; maternal transmission fidelity (a), sperm competition (g), CI induction strength (b) and cost of infection (s) as a function of the cost of polyandry (t). Note that for realistic parameter space (a∼ 0.98, b∼ 0.1–0.4 and s∼ 0.05), around 0.01 of the population (1%) is uninfected, suggesting that even if sperm competition were absolute (g = 1), the modifier of polyandry could tolerate a cost not higher than about 0.1% (t= 0.0001).

In sum, if the sperm of uninfected males win in competition against sperm from infected males, polyandry may well act to inhibit the invasion and spread of Wolbachia, and although in principle sperm competition may promote polyandry when Wolbachia is established, the advantage is typically too weak to resist even very minor costs of polyandry.


Here we explore the benefits of polyandry in D. simulans females where offspring production is strongly affected by the maternally inherited bacterium Wolbachia. When females are equally likely to mate with infected or uninfected males, uninfected polyandrous females produce more offspring than monandrous uninfected females because they are able to exploit the outcome of sperm competition, in which infected, incompatible males are strongly disadvantaged. Our finding that virgin-infected males produce competitively inferior ejaculates is consistent with a previous study in this species, in which the ejaculates of nonvirgin males competed (Champion de Crespigny and Wedell 2006). The sperm competition process enables uninfected polyandrous females to bias the paternity of their offspring toward compatible, uninfected males, and in some circumstances entirely recoup the fitness costs associated with CI. Sperm competition thus undermines the reproductive advantage of infected females and therefore the transmission of Wolbachia. Our findings are analogous to other studies of incompatibility avoidance (Tregenza and Wedell 1998; Newcomer et al. 1999; Stockley 1999; Foerster et al. 2003; Bretman et al. 2004; Marshall and Evans 2005) where polyandrous females have greater offspring hatching success than monandrous females. Unlike some of these studies, in D. simulans the source of the reproductive incompatibility (Wolbachia) is known in terms of its impact on female reproduction and in which female/male combinations it will occur. As a result, variation in female reproductive success can be confidently attributed to incompatibility avoidance (i.e., avoidance of Wolbachia-induced CI).

In addition to demonstrating that polyandrous females can avoid investing in inviable offspring by exploiting sperm competition, our results indicate that uninfected females may actively promote sperm competition by remating more quickly than infected females. Short remating intervals are likely to be advantageous as they increase the level of competition between ejaculates, leading to an increase in the probability of fertilizing eggs with compatible sperm. However, further investigation is required to reveal the generality of these findings. Ideally, this would encompass field-based studies of female remating intervals in populations polymorphic for Wolbachia infection and/or replicate studies using different laboratory strains of D. simulans. Caution in drawing strong conclusions is motivated by the considerable variation in female remating intervals reported in different studies of D. simulans and the potential for strain-specific effects. For example, one study of D. simulans (of unknown Wolbachia infection status) found only 32% of females remated within a 12-day period (Singh and Singh 2004), whereas 64% of females in our study remated approximately 18 h after their first copulation. Female remating interval was also affected by the infection status of the first mate. Both infected and uninfected females that mated to infected males had longer remating intervals than females mated to uninfected males. This finding is unexpected because one would predict females mated to incompatible males to remate sooner to reduce the cost of CI. However, females may be unable to determine the infection status of their mates and longer remating intervals associated with Wolbachia infection in males could be male rather than female driven. Nevertheless, the findings of this study provide tantalizing evidence that female reproductive strategies could be responsive to their Wolbachia infection status.

At first sight, the results of these experiments appear to provide strong support for the genetic incompatibility avoidance hypothesis as an explanation for multiple mating in female D. simulans. However, the design of these experiments makes assumptions about the frequency with which females encounter incompatible males. In our experiment, females were equally likely to mate with an incompatible male as a compatible male. The same is true of other studies of genetic incompatibility avoidance where the paternity achieved by either related or nonrelated males is compared (Stockley 1999; Tregenza and Wedell 2002; Jennions et al. 2004). Although this design enables researchers to quantify the advantage to polyandrous females, it is unlikely to represent the actual probability of encountering incompatible mates in nature. The rate at which incompatible matings occur has crucial implications for reproductive benefits obtained by polyandrous females. To fully evaluate the impact of polyandry on incompatibilities, or conversely, the selective power of genetic incompatibilities in promoting change in host reproductive strategies, it is necessary to consider the dynamics of genetic incompatibilities between mates.

Based on the data generated by our experiments, we developed models that explored the consequences of differential sperm competitive ability on the fate of newly invading Wolbachia infections. When uninfected females mated to both infected and uninfected males, the most conservative estimate of the paternity achieved by infected, incompatible males (from the two experiments) was 32%. In the models, this estimate of paternity is approximately equivalent to g= 0.4, and providing that females mate multiply prior to the initial invasion of Wolbachia, polyandry raises the minimum threshold frequency required for successful spread of Wolbachia to fixation. In other words, when sperm competition favors uninfected males, polyandry increases the likelihood that Wolbachia fails to invade a population. It is also likely to slow the rate of spread of Wolbachia within a population, owing to the reduced transmission advantage of Wolbachia associated with the decreased incidence of CI induction. When polyandry and differential sperm competitive ability act in conjunction with reduced maternal transmission fidelity and incomplete CI induction by infected males, the conditions for invasion and spread of Wolbachia are worsened. This is likely to be exacerbated by stochastic curing events such as temperature extremes or naturally occurring antibiotics (Clancy and Hoffmann 1998) that increase the proportion of uninfected individuals within populations (Champion de Crespigny et al. 2005). However, if Wolbachia manages to invade, reduced sperm competitive ability of infected males does not greatly alter the stable equilibrium frequency achieved by Wolbachia. Therefore, in the long term polyandry and sperm competition alone are unlikely to generate stable intermediate infection polymorphisms.

The converse hypothesis examined by the models developed in this study is that Wolbachia-associated incompatibilities promote or drive the spread of polyandry. To investigate this possibility we developed a model that allows biologically relevant rates of incompatibility to coevolve with polyandry. Furthermore, we examined the fate of a modifier rather than mean fitness effects, which have been modeled previously (e.g., Colegrave et al. 2002). Our models indicate that the spread of a “polyandry” allele is theoretically possible, so long as uninfected males' sperm are competitively superior to infected males' sperm and Wolbachia is polymorphic. The spread of a modifying “polyandry” allele, however, was greatly constrained by costs associated with multiple mating. Even a very minor cost of, for example, a 1% reduction in fitness, caused the modifying allele to fail to spread within reasonably realistic parameter space. This finding is in line with findings from a previous model that compared the selective force of “good genes” and “mate compatibility” on the evolution of polyandry, where costs associated with multiple mating reduced the parameter space in which polyandry was advantageous (Colegrave et al. 2002). Costs associated with polyandry are thought to be significant for many species because female promiscuity can lead to, for example, an increased risk of acquiring sexually transmitted diseases or predation (Jennions 1997) and increased selfish manipulation by males (Chapman et al. 1995). However, these costs are rarely quantified in studies of polyandry despite being required for evaluating the potential benefits of multiple mating to females. Although costs of polyandry were not evident in our study, in the sense that females that received two ejaculates produced more offspring than females that were restricted to mating once, regardless of Wolbachia infection status, we did not measure female life time reproductive success or other female life-history parameters, such as longevity.

The probability of mating with an incompatible mate in conjunction with potential costs associated with polyandry imposes severe restrictions on the conditions in which incompatibilities caused by Wolbachia generate a selective advantage for polyandry. Our findings therefore do not support the hypothesis that genetic incompatibilities drive or promote the transition from monandry to polyandry. However, this is context dependent. The potential for other forms/causes of genetic incompatibility to promote polyandry requires investigation on a case-by-case basis. What is clear is that there must be a high probability of encountering and mating with incompatible mates for polyandry to have a selective advantage, assuming it is associated with costs. Otherwise, many polyandrous females will bear costs associated with multiple mating without deriving benefits through incompatibility avoidance. This situation is likely to be true for several other forms of selfish elements, such as maternal effect lethals, for example, the Medea factor in Tribolium castaneum (Beeman et al. 1992) and Scat in mice (Hurst 1993), which are likely to cause incompatibilities infrequently because, like Wolbachia, they typically sweep to fixation (Smith 1998). It is therefore necessary to determine the likelihood of the occurrence of genetic incompatibilities between mates to evaluate the importance of polyandry for the genetic incompatibility avoidance hypothesis.

In summary, cellular endosymbionts such as the bacterium Wolbachia are widespread reproductive parasites of arthropods. Our study has shown that when they cause reproductive incompatibilities in D. simulans, polyandry can relieve the fitness cost to uninfected females because sperm competition increases the number of fertilizations obtained by compatible males. Furthermore, there is some evidence that uninfected females may actively promote sperm competition by remating more rapidly than infected females. In species/populations that are already polyandrous, female multiple mating and the ensuing sperm competition can prevent the invasion and spread of CI-inducing Wolbachia. However, despite the benefits of avoiding incompatible mates, in populations harboring stable Wolbachia infections and where incompatible matings are infrequent, it is unlikely that Wolbachia and associated CI will promote the evolution of polyandry.

Associate Editor: M. Travisano


We wish to thank M. Edvardsson and A.Wright for laboratory assistance and D. Hodgson for performing the GLIM4 analyses of female remating interval. This research was supported by an ORS award and European Social Fund (ESF) funding to FEC and the Leverhulme Trust and a Royal Society Fellowship to NW.