Wolbachia pipientis Hertig and Wolbach (Rickettsiales: Rickettsiaceae) are intracellular α-proteobacteria that occur naturally in Aedes albopictus (Skuse) (Diptera: Culicidae) and numerous other invertebrates. These endosymbionts can invade host populations by manipulating host reproduction. Wolbachia infections have been shown to impart both costs and benefits to hosts in terms of development, survival, and fecundity. Here, we monitor intraspecific competition among independent cohorts of infected or uninfected larvae. Levels of competition are manipulated by varying initial larval densities and food levels. Although larval density is observed to have major impacts on immature survivorship, sex ratio of eclosing adults, and developmental rates, the Wolbachia infection status had minimal impact on male immatures and no effect on immature females under these experimental conditions. Female and male immatures were observed to respond differently to competitive pressure, with the functional relationships of females and males consistent with scramble and contest competition, respectively. The results are discussed in relation to the evolution of naturally occurring Wolbachia infections in Ae. albopictus (i.e., natural population replacement events) and public health strategies that propose the manipulation of Wolbachia infections in Ae. albopictus populations.
As a result of their high fecundity, short generation time, small size, and genetic diversity, insects lend themselves to testable models of animal fitness. Additionally, many species are important for their economic or health impacts. The Asian tiger mosquito, Aedes albopictus (Skuse) (Diptera: Culicidae), is a vector of dengue, West Nile, and chikungunya viruses to humans (Rai 1991, Gratz 2004, Reiter et al. 2006) and filarial worms to various mammals (Nayar and Knight 1999, Cancrini et al. 2003). After its accidental introduction into North America, Ae. albopictus has quickly extended its range (Knudsen 1995), resulting in declines of resident mosquito species (Hobbs et al. 1991, Hornby et al. 1994, O’Meara et al. 1995), often in directions unpredicted by species interactions in their native range (Black et al. 1989, Rai 1991). Such observations have prompted studies of the interspecific competitive abilities of this species (Black et al. 1989, Ho et al. 1989, Livdahl and Willey 1991, Teng and Apperson 2000, Lounibos et al. 2002, Juliano et al. 2004, Yee et al. 2004). Research involving intraspecific competition by Ae. albopictus is less common (Tsuda et al. 1994, 1997; Lord 1999, Blackmore and Lord 2000). As with other mosquito species that develop in natural and artificial containers, Ae. albopictus eggs are laid in habitats that generally provide limiting resources, resulting in competition during larval stages even among pure populations (Wynn and Paradise 2001, Gimnig et al. 2002). Thus, intraspecific larval competition is critical in determining population performance.
Insects have evolved numerous responses to intraspecific larval competition, but these can be simplified by Nicholson’s dichotomy (Nicholson 1954, Lomnicki 1988). During contest competition for limiting resources, dominant individuals monopolize resources and the number of survivors does not exceed a constant, maximum level independent of initial larval density. During scramble competition, each larva has similar access to resources. Therefore, individual resource uptake decreases with increasing larval density, resulting in decreased survivorship above a critical density because fewer larvae receive a minimum resource requirement. A plot of adult production as a function of increasing larval density would increase to a plateau (i.e., fit by an asymptotic curve) for contest competition but increase to a maximum followed by a decrease (i.e., fit by a parabolic curve) for scramble competition. The shape of the response curve by mosquito species undergoing intraspecific competition is important because of the potential response of a population to control strategies (Washburn et al. 1991). In particular, if the mosquito species undergoes scramble competition, reducing levels of competition can increase availability of resources, with the potential for more, larger or longer-lived adult mosquitoes.
Relatively little information is available regarding potential effects of symbiotic bacteria on immature mosquito development, including infections of the intracellular α-proteobacteria Wolbachia pipientis (Hertig and Wolbach 1924; Rickettsiales: Rickettsiaceae). Wolbachia are among the most widespread animal endosymbionts known, infecting >17% of arthropods (Werren 1997, Bourtzis and Miller 2003) and many filarial nematode species (Bandi et al. 1998). Aedes albopictus in natural populations are superinfected by two different Wolbachia strains named wAlbA and wAlbB (Sinkins et al. 1995a; Kittayapong et al. 2002a, b). In Ae. albopictus, Wolbachia reduces egg hatch in matings between males and females infected by different Wolbachia types, a phenotype known as cytoplasmic incompatibility (CI) (Sinkins et al. 1995b).
Wolbachia have been generally described as "reproductive parasites" (Werren 1997), affecting the fitness of females and males differently. Similar to other parasites, Wolbachia must use host cell resources to meet their metabolic needs, resulting in potential physiological and fitness costs to their host. However, extreme cost to the host can be maladaptive. Prior studies report Wolbachia effects ranging from relatively minor fitness costs (Hoffmann et al. 1990, Fleury et al. 2000, Olsen et al. 2001, Fry and Rand 2002) to benefits (Shoemaker et al. 2002, Dobson et al. 2004, Dean 2006). However, relatively few studies have focused on Wolbachia effects on hosts during immature stages. Harcombe and Hoffmann (2004) found no effect of Wolbachia infection on larval developmental times or size at adult emergence in Drosophila melanogaster (Diptera: Drosophilidae), whereas Islam and Dobson (2006) showed a cost to infected male Ae. albopictus larvae reared under food-limited conditions.
Larval developmental conditions are used to predict population growth of Ae. albopictus cohorts experiencing inter- (Livdahl and Willey 1991) or intraspecific (Lord 1999) competition. Relative costs and benefits of Wolbachia infection on competing Ae. albopictus immatures can have population-level impacts. Specifically, the host response to competition (i.e., scramble versus contest) can influence infection dynamics during a Wolbachia invasion event; likewise, the outcomes of the two types of competition predict different host population dynamics if per capita resource availability changes. Models predict that reduced egg hatch that occurs during an invasion event by a CI-inducing Wolbachia strain will have differing impacts on host populations that are regulated by contest versus scramble competition (Dobson et al. 2002b). It is only scramble competition that can result in an increased host population size during such an invasion, because of disproportionately higher immature survivorship that results from increased per capita resource availability as competing larvae die. Thus, an understanding of the host response to intraspecific competition at different resource levels is important to understanding potential impacts of Wolbachia invasion on the host population.
Experiments presented here were designed to observe for (1) effects of larval competition across sparse to dense larval densities on development time and adult emergence among Ae. albopictus with natural Wolbachia superinfections by manipulating resource availability and (2) an impact of Wolbachia infection on Ae. albopictus larval competition. For each comparison, impacts are described for the cohort (i.e., considering both sexes combined) and then for responses of females and males separately. Although male and female larvae compete with individuals of both sexes, it is useful to monitor competitive outcomes by sex because resources exploited by emerging males will impact female development and, thereby, opportunity for population growth. The results provide a better understanding of fitness costs associated with Wolbachia invasion into Ae. albopictus and could be used to improve models of Wolbachia invasion. Moreover, these data can be used to anticipate effects caused by the proposed artificial manipulation of medically important vector populations, using Wolbachia as a vehicle for population replacement/gene drive (Dobson 2003, Sinkins and Godfray 2004, Sinkins and Gould 2006) or as a tool for population suppression (Laven 1967, Dobson et al. 2002a, Zabalou et al. 2004, Brelsfoard et al. 2008).
Materials and Methods
The two mosquito strains used are believed to share a similar host genetic background and differ only by Wolbachia infection status (Dobson et al. 2004). UjuTet (UT) is an aposymbiotic (i.e., uninfected) strain generated by tetracycline treatment and has been free of Wolbachia >50 generations (Dobson and Rattanadechakul 2001). Individuals of the IH strain are superinfected by both the natural wAlbA and wAlbB Wolbachia types (Sinkins et al. 1995a) and were backcrossed with UT males to create a similar host genomic background (Dobson et al. 2004). Mosquitoes were maintained as previously described (Dobson and Rattanadechakul 2001).
For experiments, mosquitoes were reared at 28°C in identical cylindrical plastic containers (Mosquito Breeders; BioQuip, Rancho Dominguez, CA) holding 200 ml deionized water and food. Because larval competition is influenced not only by the quality of the larval diet but also its physical complexity (Barrera 1996), a standard artificial diet was used (liver powder; ICN Biomedical, Aurora, OH). For the IH strain, replicated cohorts of different initial densities (i.e., 10, 25, 50, 100, 200, and 400 per container) of newly hatched first-instar larvae were provided with one of two food regimens (12 or 18 mg/wk) by pipetting from a stock suspension. For the UT strain, different initial densities were tested using a single food regimen (12 mg/wk). Emerging adults were removed, sexed, and counted each week. The experiment continued until all individuals had eclosed or died.
Attributes including total adults produced, numbers of females and males, proportion mortality (i.e., 1 - [number of emerging adults/number of initial larvae]), proportion females, and times to mean female and male emergence were determined for each cohort. Mean responses were determined for each combination of larval density, food level, and Wolbachia infection status. Data were tested for normality (Kolmogorov-Smirnov test), and proportions were arcsine transformed. Multivariate analysis of variance (MANOVA) models were used to test for effects of treatment levels on the dependent variables proportion mortality and proportion females (Scheiner 1993). Density (as a class variable), food, and infection were used as main effects. Interaction terms included infection × density and food × density; other interaction terms were invalid because of the absence of the combination of UT and 18 mg food. Wilks’ Λ was used to estimate the amount of unexplained variance in the response terms. A second MANOVA model was generated to test responses in female emergence time × male emergence time. For both models, significant factors were subsequently tested for covariance and then for significant effects on response variables using univariate ANOVA and post hoc LSM pairwise comparisons after Bonferroni correction (Stata 10, 2008; StataCorp, College Station, TX).
Quadratic terms were fit to a linear model with least squares criterion (type III tests with initial larval density as fixed effects, SAS 9, 2002; SAS Institute, Cary, NC) and used to independently determine whether (1) sex and food or (2) sex and infection had significant effects on the shapes of the best-fit curves for male and female survivorship to adult stage.
Neither Wolbachia infection type nor food was observed to affect the total number of adults. The mean numbers of adults increased significantly to maxima across the density (ANOVA: F5,64 = 72.1, P < 0.0001) treatments but were unaffected by food (F1,64 = 1.26, P = 0.266) levels or infection (F1,64 = 1.31, P = 0.258) type (Fig. 1A). The effect of resource limitation (12 or 18 mg food) was determined for each initial larval density using comparisons of the IH strain, which harbors the natural Wolbachia infection type. The effect of Wolbachia infection was assessed using comparisons of the IH and UT (i.e., aposymbiotic) strains, while keeping food addition constant (12 mg liver powder). The density × infection interaction was significant (F5,64 = 3.34, P = 0.010), whereas the density × food interaction was not (F1,64 = 1.67, P = 0.155). The significant interaction is represented in Fig. 1A: equivalent numbers of mosquitoes resulted from low densities, despite infection type; in contrast, relatively fewer adults emerged from the UT than the IH cohorts at the two highest initial larval densities.
MANOVA on proportion mortality and proportion female (both arcsine transformed) was significant (Wilks’ Λ = 0.101, F34,126 = 7.96, P < 0.0001). Multivariate effects of density (Λ = 0.193, F10,126 = 16.06, P < 0.0001) and food (Λ = 0.886, F2,63 = 4.06, P = 0.0212) were significant, with differences in both proportion mortality (Fig. 1B) and proportion female adults (Fig. 1C). The multivariate effect of infection (Λ = 0.980, F2,63 = 0.64, P = 0.531) was not significant. Neither of the interaction terms was significant.
Post hoc multivariate multiple comparisons were used to further study density effects with respect to proportion mortality and proportion female (Scheiner 1993). Pairwise comparisons that were significantly different (after Bonferroni corrections) included density levels 10 and 200 (Λ = 0.690, F2,63 = 14.15, Padj = 0.0003); 10 and 400 (Λ = 0.524, F2,63 = 28.58, Padj < 0.0001); 25 and 200 (Λ = 0.733, F2,63 = 11.50, Padj = 0.0015); 25 and 400 (Λ = 0.575, F2,63 = 23.29, Padj < 0.0001); and 100 and 400 (Λ = 0.6782, F2,63 = 14.94, Padj = 0.0001). Standardized canonical coefficients of the first variate indicated that proportion female (0.471) and proportion mortality (2.043) behave similarly across density levels, but proportion female explained more of the variation across density levels. Proportion female and proportion mortality explained nearly equal amounts of variation for the two food levels based on standardized canonical coefficients for the first variate (1.150 and -1.01, respectively), with opposite signs indicating a change in pattern.
Estimating female and male survivorship to emergence separately requires the assumption of an equal sex ratio among the larvae used to initiate the cohort, which is supported by prior studies showing a 50:50 sex ratio in Ae. albopictus (Islam and Dobson 2006). Figure 2A shows mean numbers of females and males for each combination of density, food, and infection, whereas Fig. 2B represents the curves generated by fitting quadratic terms using a linear model to the data. The resulting equations are included in Table 1. The patterns for female and male production differed across density levels; the pattern for males was asymptotic, whereas that for females was parabolic.
Fitted curves differed significantly between females and males across density (P < 0.001) levels but not as a function of food (P = 0.371) provision. At both food levels, the number of eclosing females reached a peak at the 200 larvae density and decreased at the 400 larvae density to levels of emergence that were not different from cohorts initiated with the lowest larval density. In contrast, male survivorship at both food levels reached a plateau at higher densities. Survival patterns did not differ in comparisons of the two food levels for either sex. Similarly, patterns of female and male survival differed significantly across density (P < 0.0001) levels but not with infection (P = 0.409) type (Fig. 2B).
Adult emergence times were significantly different among treatments for female emergence time × male emergence time by MANOVA: overall Λ = 0.082, F34,126 = 9.22, P < 0.0001. Ignoring sex, the developmental times were significantly affected by food level (Λ = 0.628, F2,63 = 18.68, P < 0.0001), density (Λ = 0.174, F10,126 = 17.64, P < 0.0001), and infection (Λ = 0.858, F2,63 = 5.23, P = 0.0079). Post hoc multivariate multiple comparisons were again used to further study the main effect density with respect to female and male emergence times. Pairwise comparisons that were significantly different (after Bonferroni corrections) included density levels 10 and 100 (Λ = 0.214, F2,63 = 42.19, Padj < 0.0001); 10 and 200 (Λ = 0.604, F2,63 = 20.69, Padj < 0.0001); 10 and 400 (Λ = 0.501, F2,63 = 31.44, Padj < 0.0001); 25 and 100 (Λ = 0.769, F2,63 = 9.50, Padj = 0.0030); 25 and 200 (Λ = 0.522, F2,63 = 28.83, Padj < 0.0001); 25 and 400 (Λ = 0.434, F2,63 = 41.12, Padj < 0.0001); 50 and 200 (Λ = 0.593, F2,63 = 21.66, Padj < 0.0001); 50 and 400 (Λ = 0.492, F2,63 = 32.47, Padj < 0.0001); and 100 and 400 (Λ = 0.718, F2,63 = 12.38, Padj = 0.0009). Standardized canonical coefficients (males: 2.24, females: –0.090) indicated that male emergence times explained more of the variation in food levels. Coefficients for density (males: 1.353, females: 1.305) suggested that males and females were behaving similarly across density levels and explaining roughly equal amounts of variation for this factor.
To further assess the impact of sex on emergence times across density and food, we considered females and males separately; sex had a significant effect on time until mean adult emergence (analysis of covariance [ANCOVA]: F1,138 = 33.13, P < 0.0001). Sex also interacted significantly (F5,138 = 6.90, P < 0.0001) with density because mean female emergence (Fig. 3A) was delayed more than that for the mean male (Fig. 3B) for the higher density treatments.
Times for emergence were lengthened for both females and males reared at higher densities and lower food, but the sexes responded differently to endosymbiont infection. In the individual ANOVA, female development to adulthood (Fig. 3A) differed among treatments (overall ANOVA model, F12,81 = 25.08, P < 0.0001). Mean female emergence time was significantly affected by density (F5,69 = 46.53, P < 0.0001) and food (F1,69 = 4.53, P = 0.040) but not by infection (F1,69 = 0.04, P = 0.840). Female developmental time was significantly prolonged for cohorts with initial larval densities >100; mean time for female emergence did not differ between the 200 and 400 larval cohorts. The interaction between infection and density was not significant (F5,69 = 1.82, P = 0.121).
In contrast to females, male development time (Fig. 3B; overall ANOVA model F12,69 = 24.89, P < 0.0001) was significantly affected by infection status (F1,69 = 9.45, P = 0.003). Male developmental time was also affected by initial larval density (F1,69 = 40.64, P < 0.0001) and food provision (F1,69 = 37.72, P < 0.0001), similar to females. Male emergence was significantly delayed when density was ≥50 but did not differ among cohorts with initial densities of 100, 200, and 400 larvae. Similar to females, the interaction between infection and density was not significant (F5,69 = 1.47, P = 0.210).
As previously observed (Tsuda et al. 1997), larval density is a major determinant of competition among immature Ae. albopictus. Lord (1999) showed that larval density and food allocation influenced survival, rate of development, and adult (female) size during intraspecific competition within Ae. albopictus cohorts. Blackmore and Lord (2000) showed significant positive correlations between female size (wing length) and fecundity (size of initial egg batch) after emergence from cohorts at a variety of densities. Naturally superinfected Ae. albopictus, analogous to the IH strain, were presumably used for all these earlier experiments. Under the present developmental conditions, relative survival, female production, and developmental rates were all negatively impacted at higher competitive densities. Food provision also impacted sex ratio and development times for both females and males. In contrast, Wolbachia infection status had relatively little impact on immature Ae. albopictus performance under the tested conditions, with no observed effect on females and minor effects on males (increased developmental time in infected males).
Nicholson (1954) defined contest and scramble responses to intraspecific competition as phenomena characteristic of species. Overall, cohorts of Ae. albopictus showed contest competition when faced with intense competition, and the total number of adults reached an asymptote. However, when considered separately, females and males displayed differing patterns of competition: scramble and contest, respectively. As a possible explanation, it is hypothesized that males and precocious females monopolized resources, which were subsequently unavailable for lagging females, especially during later instars when biomass accumulation is most critical for survival and fecundity.
Although female and male larvae competed with individuals of both sexes, the outcomes of this competition tended to be unequal for the sexes. Such sexually asymmetric competition has been observed for other mosquito species including Aedes polynesiensis Marks (Mercer 1999) and Aedes aegypti L. (Bedhomme et al. 2003) and seems to be a widespread occurrence among mosquito species (Mercer et al. 2008). Whatever the starting conditions, males (with about half the adult biomass of females) consume fewer resources and emerge earlier than their sisters.
The low initial larval density treatments showed high variation in proportion females among cohorts (Fig. 1C) even while there was high survivorship within these cohorts. This high level of variation is probably stochastic rather than biological; the relatively few individuals used to establish the low-density cohorts probably misrepresented an initial 1:1 sex ratio by chance. However, this experimental effect was less important at higher densities for which sampling variation around the expected value of 50% female larvae at experimental setup is expected to be lower (Sokal and Rohlf 1981). Therefore, departures from 1:1 seen at higher larval densities are more likely to represent biological effects.
The simple laboratory model system used here was selected for initial experiments to reduce potential confounding effects that would complicate interpretation. However, the reliance on artificial conditions (including diet) during studies of container-developing mosquitoes has been criticized (Barrera 1996, Lord 1999, Dieng et al. 2002) because habitat complexity may mitigate negative effects of high density. For example, if endosymbiont infection were to reduce active movement by larvae, artificial rearing conditions may mask changes in susceptibility to predation or the impacts of physical contacts at high larval densities (Alto et al. 2005). Clearly, the potential for such effects should be studied in future experiments.
Observations of no difference between infected and uninfected cohorts in sex ratio, female number, or female developmental rate are consistent with the previous report of Islam and Dobson (2006). However, the prior report differs in that higher survivorship of males and total adults were observed in the UT strain relative to IH (Islam and Dobson 2006). Also in the prior report, no difference was observed between UT and IH in male development time. However direct comparison of results here with the prior results is complicated, because the prior study was conducted with conditions of low larval density and excess food, which resulted in ≈85–95% immature survivorship (Islam and Dobson 2006). Taken together, the results presented here and in the prior report suggest that Wolbachia infection does not negatively impact the survivorship or development of immature female Ae. albopictus but does negatively impact immature male fitness. As an extension of our studies, future experiments should be designed to examine direct competition between infected and uninfected immatures (i.e., competing for the same resources within the same larval pool).
The results presented here reinforce observations of minimal physiological costs associated with Wolbachia infection. Minimal fitness costs are consistent with theory, which predicts evolution toward mutualism in obligate, vertically inherited symbionts (Herre et al. 1999, Moran 2006, Weeks et al. 2007). Furthermore, under experimental conditions, the only observed fitness cost associated with Wolbachia infection was in males, which is consistent with predictions for maternally inherited symbionts (i.e., stronger selection toward commensalism or mutualism in females relative to males).
The sex-specific patterns of survivorship that result from increasing larval competition have important implications for public health interventions targeting egg hatch (e.g., sterile insect technique strategies based on CI or irradiation; Benedict and Robinson 2003, Brelsfoard et al. 2008). Under conditions of intense intraspecific competition, high larval mortality will serve to reduce immature competition levels. During contest competition experienced by males (asymptotic curve; Fig. 2B), reducing the number of larval competitors is expected to result in fewer adult males. In contrast, the number of adult females could increase because of a relaxation of intense scramble competition (essentially moving back up the parabolic curve, Fig. 2B, as individuals drop out of competition). Because female mosquitoes are responsible for biting and disease transmission, any impacts on female size or time of emergence caused by control interventions must be carefully evaluated.
We thank A. Eitam and M. S. Islam for experimental assistance. This research was supported by grants from the National Institutes of Health (AI-51533 and AI-67434). This is Publication 08–08–141 of the University of Kentucky Agricultural Experiment Station.