The density and regulation of microbial populations are important factors in the success of symbiotic associations. High bacterial density may improve transmission to the next generation, but excessive replication could turn out to be costly to the host and result in higher virulence. Moreover, differences in virulence may also depend on the diversity of symbionts. Using the maternally transmitted symbiont Wolbachia, we investigated how bacterial density and diversity are regulated and influence virulence in host insects subject to multiple infection. The model we used was the wasp Asobara tabida that naturally harbors three different Wolbachia strains, of which two are facultative and induce cytoplasmic incompatibility, whereas the third is necessary for the host to achieve oogenesis. Using insect lines infected with different subsets of Wolbachia strains, we show that: (i) some traits of A. tabida are negatively affected by Wolbachia; (ii) the physiological cost increases with the number of co-infecting strains, which also corresponds to an increase in the total bacterial density; and (iii) the densities of the two facultative Wolbachia strains are independent of one another, whereas the obligatory strain is less abundant when it is alone, suggesting that there is some positive interaction with the other strains.

THE genetic diversity of parasites co-infecting individual hosts is often thought to be an important factor in the evolution of their virulence (Ewald 1994; Frank 1996a; Galvani 2003). Theoretical studies have shown that lower relatedness among the parasites within the host could lead to increased virulence (Frank 1994, 1996a; Van Baalen and Sabelis 1995). The classical idea, referred to as the “tragedy of the commons,” is that competition for a limiting resource puts faster exploiters at an advantage over more prudent ones (Hardin 1968). However, recent articles have pointed out that in commonly used models, selection among parasites affects only the host exploitation rate, but that different outcomes could be reached when other types of competition are also considered (Chao et al. 2000; Read and Taylor 2001; Brown et al. 2002). For example, interference among parasites could lead to underexploitation of hosts and hence to reduced virulence (Chao et al. 2000). The relationship beween virulence and multiple infection is thus still under debate and needs further empirical documentation.

So far these questions have received little study in vertically transmitted symbionts. There are two main reasons for this. First, the vertical transmission of symbionts tends to limit multiple infection, because the oocytes of the host are colonized by only a few individuals, and the resulting bottleneck greatly reduces the genetic diversity of the symbiotic population (Mira and Moran 2002). Second, many vertically transmitted symbionts have evolved a mutualistic relationship with their hosts, and the benefit they confer considerably outweighs their cost; this makes the cost difficult to assess or even detect (Thompson 1988; Bronstein 1994). However, these biological models could provide interesting data, because the closely linked evolutionary fates of host and symbionts should have led to the selection of mechanisms that reduce virulence to a minimum (Lipsitch et al. 1995).

A good model for studying this question is the maternally transmitted symbiotic bacterium Wolbachia, which is able to induce cytoplasmic incompatibility (CI), leading to postzygotic reproductive isolation between any male infected by a Wolbachia strain and a female lacking this strain (for review see Hoffmann and Turelli 1997). This puts females with multiple infection at an advantage and promotes the spread and maintenance of multiple infection (Frank 1998), which has proved rather common. Second, infected individuals benefit only indirectly from infection, as a result of the disadvantage suffered by uninfected females or females infected by only a subset of Wolbachia strains. Thus, the cost of infection (that we define here as the virulence of Wolbachia) and the advantage to the host are clearly distinguishable and can be measured separately in distinct individuals.

In this study, we used the association between Wolbachia and the parasitic wasp Asobara tabida, whose larvae develop as solitary endoparasites of Drosophila larvae. A. tabida individuals are co-infected by three different Wolbachia strains, wAtab1, wAtab2, and wAtab3 (Vavre et al. 1999). While both wAtab1 and wAtab2 are facultative parasites and induce CI (F. Dedeine, personal communication), wAtab3 is necessary for oogenesis to occur in this species: A. tabida females lacking wAtab3 do not produce oocytes (Dedeine et al. 2001). Using antibiotherapy, we created A. tabida lines that harbored different combinations of Wolbachia strains, but shared the same nuclear background.

This study addresses two main questions: (i) Is there any relationship between the diversity and virulence of Wolbachia and/or between Wolbachia density and virulence? and (ii) How are Wolbachia diversity and density regulated? The findings are discussed in the context of the evolution of virulence in multiple infections, by considering the peculiar selective pressures that act on this system.


Insect strains and rearing:

A. tabida (Hymenoptera: Braconidae) is a solitary endoparasitoid wasp of several Drosophila species. In the laboratory, parasitoids are reared on a Wolbachia-free strain of Drosophila melanogaster originating from Ste Foy-les-Lyon (France). Rearing and experiments were carried out without larval or adult competition at 20° under 12/12 light/dark (LD) cycle and 70% relative humidity.

A. tabida individuals are naturally infected with three Wolbachia strains, named wAtab1, wAtab2, and wAtab3 (Vavre et al. 1999). The triply infected line, named Pi(123), is an inbred line originating from Pierrefeu (Pi), France, which has been maintained by regular sib-mating for four generations. Derived A. tabida lines with different infection statuses were obtained using moderate antibiotic treatments (F. Dedeine, personal communication). Since wAtab3 is obligatory for reproduction in this species, only one singly infected and two doubly infected lines could be obtained and have proved stable in time: the singly infected line Pi(3) harboring wAtab3 and the doubly infected lines Pi(13) and Pi(23) harboring wAtab3 and wAtab1 and wAtab3 and wAtab2, respectively.

Components of the fitness cost of infection:

Several fitness traits have been measured in individuals of each line of A. tabida.

Offspring production and sex ratio:

After mating with 3- to 4-day-old males with the same infection status as themselves (checked by individual visual inspection), 1- to 2-day-old females were each provided with 150 Drosophila larvae (24 hr old) and allowed to parasitize hosts for 48 hr. The infested host larvae were then kept and allowed to develop. At emergence, the adult Drosophila and the wasps of both sexes were counted in each vial. We performed three series (i.e., blocks) to test these traits (at least seven replicates per line for each block).

Egg production:

A. tabida females are mainly proovogenic, and so most of their oocytes are already mature at emergence. To estimate the oocyte load, newly emerged females were kept for 5 days with water and honey to allow oocyte maturation to be completed. The ovaries were dissected in physiological saline, transferred into neutral red solution for 5 min, and then gently squashed between the slide and cover glass to disperse the eggs. The stained eggs were then counted under the microscope with the help of a video system.

Tibia length:

Left hind tibia was measured on the adult males and females of each line using a micrometer.

Dry weight:

Emerging parasitoids of both sexes were sampled in each line and dried at 65° for 48 hr before weighing.

Adult ability to survive starvation:

Two-day-old males and females were put in vials with moist cotton, but without food. Mortality was checked every day at the same time (±1 hr) until all the individuals had died. Six vials, each containing 10 parasitoids, were studied for each sex in each line.

Locomotor activity pattern:

Individual locomotor activity was monitored using a video-tracking and image-analysis system to provide automatic continuous measurements of the insects over several days (Allemand et al. 1994). Individual insects were isolated in experimental circular glass arenas with honey as food. The locomotor activity of each individual was measured every 3 min using binary data (1 if the wasp had moved during a 2-sec video recording and 0 if it had not), and the hourly activity was calculated as the percentage of active recordings obtained over 3 days with a 12/12 hr LD cycle. To evaluate the average daily pattern of activity, two independent parameters were estimated for each individual: the rate of locomotor activity, calculated as the average of active recordings over a 24-hr period, and the profile of the rhythm, which establishes the pattern of the total daily activity in terms of the hourly percentages. The rate of locomotor activity measures the locomotor performance of wasps, while the activity profile determines how this activity is organized throughout the day. Data are reported in terms of the Zeitgeber time (Zt, time within the environmental cycle); the light is turned off at Zt0 and turned on at Zt12.

Real-time quantitative PCR:

DNA extraction:

Insects or parts of insects were individually squashed in 150 μl 5% (w/vol) Chelex solution (Bio-Rad, Richmond, CA) and proteinase K (Eurobio, Les Ulis, France; final concentration 0.5 μg/μl) and kept at 56° for 6 hr. After 15 min at 95°, the samples were centrifuged at 16,000 × g for 4 min.


Quantification of Wolbachia bacteria was achieved by amplifying the Wolbachia surface protein gene wsp. To detect all the Wolbachia strains present we used the general forward primers 81F: 5′-TGG TCC AAT AAG TGA TGA AGA AAC-3′ (Braig et al. 1998). Specific PCR detection of each Wolbachia strain was conducted using three other forward primers: 165′F (5′-TGG TAT TAC AAA TGT AGC-3′) for wAtab1, 172F (5′-ACC TAT AAG AAA GAC AAG-3′) for wAtab2 (Zhou et al. 1998), and Aso3 (5′-AAA GGG GAC TGA TGA TGT-3′) for wAtab3. All these forward primers were used with the same reverse primer 691R: 5′-AAA AAT TAA ACG CTA CTC CA-3′ (Zhou et al. 1998).

Quantitative PCR:

Real-time quantitative PCR was performed using the LightCycler system (Roche). The 20-μl reaction mixture consisted of 10% (vol/vol) LightCycler DNA master SYBR Green I (Roche Diagnostics), 3 mm MgCl2, 500 nm each primer, and 2 μl of template DNA. The amplification consisted of 40 cycles of 15 sec at 95°, followed by 14 sec at 53° and 28 sec at 72° for 81F/691R and 11 sec at 52° and 22 sec at 72° for 165′F/691R, 172F/691R, and Aso3/691R.

Standard curves were drawn on clones of the three Wolbachia strains. Amplification with 81F/691R primers was performed on a triply infected female. PCR products were purified (GIBCO BRL, Gaithersburg, MD) and cloned into the pDrive cloning vector (QIAGEN, Valencia, CA). Specific PCR assays were used to identify the Wolbachia strain present in the clone. The DNA concentration of each sample was measured by OD absorbance at 260 nm. Standard curves were plotted using five dilutions of this vector (from 102 to 108 copies) containing one copy of a specific wsp sequence, which is a single-copy gene (Braig et al. 1998). The number of Wolbachia cells was calculated as described in Noda et al. (2001). These values must be considered as semiquantitative estimates of Wolbachia cell numbers.

Wolbachia abundance was measured individually on 5-day-old males and females, either in whole bodies or separately in head plus thorax and abdomen of the same individual. Density was obtained by correcting the number of Wolbachia by the mean fresh weight of insects of the line. The ratio of Wolbachia cells in the abdomen to the sum of Wolbachia in head plus thorax and abdomen was calculated individually.


Infection cost:

Among the seven fitness components studied here, three do not vary according to the infection status: productivity, fecundity, and tibia length. For sex ratio, a marginally significant effect was detected, with a potentially higher proportion of females in the triply infected line. However, this trait is highly variable and no such effect was detected in the third block. Clear conclusions were obtained only for dry weight, adult survival, and rate of locomotor activity (Table 1).

View this table:

Fitness traits of the Pi(123), Pi(13), Pi(23), and Pi(3) lines

Pi(3) individuals of both sexes were the heaviest, Pi(13) and Pi(23) had similar and intermediate weights, and Pi(123) were the lightest. Thus, the greater the diversity of Wolbachia lineages harbored by the insects, the less they weighed. More diversity in Wolbachia also led to a shorter life span of the host (Table 1; Figure 1). Life span of doubly infected wasps was intermediate for both sexes, but the difference from simply infected wasps was not significant in the males. The rate of locomotor activity also varied with the infection status of individuals (Table 1). As for dry weight and life span, singly infected wasps had greater locomotor activity than triply infected ones, but in this case the difference was not significant for the females. In contrast, the profile of the rhythm was the same regardless of infection status (ANOVA, d.f. = 3189, F = 1.763, P = 0.16, Figure 2).

Figure 1.—

Starvation survival curves of female and male wasps with differing Wolbachia infection statuses. (□) Pi(123), (▴) Pi(13), (♦) Pi(23), (○) Pi(3).

Figure 2.—

Mean curves of the locomotor activity rhythms of A. tabida lines Pi(123), Pi(13), Pi(23), and Pi(3). Males and females were measured for 3 days under LD 12/12 (Zeitgeber time) with light off at Zt0 and light on at Zt12. The black-and-white rectangles at the top of the figure represent night and day, respectively. (□) Pi(123), (▴) Pi(13), (♦) Pi(23), (○) Pi(3).

Total Wolbachia density:

Previous findings have shown that infection cost increases with bacterial diversity. To find out whether this is linked to density variations, we measured the number of Wolbachia cells in male and female Pi(3), Pi(13), Pi(23), and Pi(123) (Table 2). In all these lines, the numbers of Wolbachia and their relative densities (number of cells per milligram of fresh weight) were higher in females than in males (Mann-Whitney test, P < 0.001).

View this table:

Wolbachia density in A. tabida species

Despite the fact that the cell numbers obtained by real-time quantitative PCR were rather low and may have been underestimated, in both sexes the total Wolbachia density depended on the combination of the Wolbachia strains co-infecting the same individual host (Kruskal-Wallis test, P < 0.005). Singly infected individuals of the Pi(3) strain had the lowest density, which was less than one-quarter of that in Pi(123). In females, the density in the two doubly infected lines was intermediate between those of the singly and triply infected lines, but it was lower in Pi(13) than in Pi(23). In contrast, there was no significant difference between males of the two doubly and the one triply infected lines.

Overall, these findings demonstrate that increasing the mixture of Wolbachia strains results in a higher physiological cost to the host and also leads to higher Wolbachia density. This means that the cost of infection is also positively correlated with bacterial density, as shown in Figure 3.

Figure 3.—

Wolbachia density and dry weight. The Wolbachia density (histogram) and dry weight (curve) of Pi(123), Pi(13), Pi(23), and Pi(3) individuals are presented. Values correspond to the average of eight individuals for density and of 30 individuals for dry weight. Bars show the standard error.

Strain-specific Wolbachia density:

We have shown that total bacterial density increases with bacterial diversity, but we still do not know how each Wolbachia strain responds to the presence of other Wolbachia strains. We therefore measured the specific density of each Wolbachia strain co-infecting the same individual in all A. tabida lines (Figure 4).

Figure 4.—

Specific densities of wAtab1, wAtab2, and wAtab3. Specific densities of wAtab1, wAtab2, and wAtab3 in females and males of various infection statuses are presented. Values correspond to the average of eight individuals per sex and line. Bars show the standard error. P-values of Kruskal-Wallis tests are indicated. For wAtab3, means marked with the same letters are not significantly different (Mann-Whitney tests, P = 0.05).

First, the sums of the specific densities are equal to the estimations of total densities in all lines (Wilcoxon rank test, P = 0.11), thus indicating the reliability of the method and ruling out any concern about significant bias.

In both sexes, we found that the specific densities of wAtab1 and wAtab2 were the same in the Pi(123) and Pi(13) lines and the Pi(123) and Pi(23) lines (Mann-Whitney test, P > 0.05), respectively, and are thus independent of the infection status of the individuals. The density of wAtab3 was lower in Pi(3) than in multiply infected lines and, whereas the differences were not significant in the males (Kruskal-Wallis test, P = 0.11), they were significant in the females (P = 0.02), with the Pi(13) and Pi(123) lines harboring more wAtab3 than Pi(3) lines (Mann-Whitney test, P ≤ 0.02).

Considering the relative abundance of Wolbachia strains in triply infected individuals, wAtab1 was found to be the least represented in both sexes (16% in females and 22% in males), whereas the most abundant strains were wAtab2 in females (61%) and wAtab3 in males (near 41%). Despite the different abundances of wAtab1 and wAtab2 in the host, they induced similar infection costs.

Wolbachia distribution in the host body:

The localization of Wolbachia may have an influence on the infection cost (McGraw et al. 2002). We then studied the preferential localization of these bacteria in A. tabida species of each line by measuring the total number of Wolbachia cells in head plus thorax and in abdomen of the same host body. Comparison of the percentage of Wolbachia in abdomen compared to the entire body in the four lines demonstrates that Wolbachia are preferentially localized in abdomen in males as well as in females whatever the infection status (Table 3). However, the percentage of Wolbachia in abdomen is higher in females than in males (Student's t-test, P < 0.0001). This repartition does not differ between lines (ANOVA, P > 0.05 for both sexes), even though the percentage of Wolbachia in abdomen of males for the Pi(3) line seems to be lower.

View this table:

Percentage of Wolbachia cells in abdomen compared to the entire body in A. tabida females and males


Three main conclusions can be drawn from our findings:

  1. Despite the vertical transmission of Wolbachia and the high selective pressures that tend to reduce infection costs, some traits of A. tabida are still negatively affected by Wolbachia. However, the resulting costs are low, may have little influence on the fitness of adults in the wild, and probably do not affect the maintenance of multiple infection. This finding is consistent with numerous studies on the cost of Wolbachia where a high variability among species and traits exists (Hoffmann et al. 1990, 1994, 1998; Giordano et al. 1995; Girin and Boulétreau 1995; Turelli and Hoffmann 1995; Wade and Chang 1995; Bourtzis et al. 1996; Clancy and Hoffmann 1997; Min and Benzer 1997; Poinsot and Merçot 1997; Hoerauf et al. 1999; Vavre et al. 1999; Fleury et al. 2000; Dobson et al. 2002; Fry and Rand 2002). On the other hand, the nearly significant effect on sex ratio might promote multiple infections through increase in the proportion of females.

  2. The cost of infection depends directly on the diversity of Wolbachia and increases with the number of Wolbachia strains within the insect host. The same localization has been observed between individuals of all infection statuses; therefore the differences of infection cost observed in the various lines of A. tabida cannot be explained by a difference of localization in the host body. These results mean that even though variations in density are due to the differing diversity of the Wolbachia strains infecting individuals, higher bacterial densities also correlate with an increase in the cost of infection. A relationship between density and virulence has already been documented by McGraw et al. (2002), who demonstrated that the fitness cost associated with the popcorn strain was reduced in a new transfected Drosophila host harboring a lower Wolbachia density. Therefore, with results of McGraw et al. (2002), the data presented here strongly support the existence of a relationship between bacterial density and infection cost, which has rarely been demonstrated hitherto, but is generally accepted since more symbionts can be expected to require more energy (Thompson 1988). However other factors, such as the particular traits of each strain, may affect the impact of bacteria on their host. For example, the same cost is induced by wAtab1 and wAtab2, in spite of the lower density of wAtab1. Thus, all differences in the cost of infection cannot be attributed to bacterial density alone.

  3. The densities of the two facultative strains wAtab1 and wAtab2 are specifically regulated and do not depend on the presence of other strains. The situation is less clear for the obligatory bacterium wAtab3, suggesting that there may be some positive interaction with other strains. Such specific bacterial regulation seems to be rather common in multiple infections by Wolbachia, as suggested by studies in D. simulans, Ephestia kuehniella, and Leptopilina heterotoma (Rousset et al. 1999; Ikeda et al. 2003; Mouton et al. 2003).

The puzzling question that now arises is how the infection cost for A. tabida has persisted despite selective pressures toward reduced virulence that act on both the bacteria and the insects. Is multiple infection responsible for this persistence? The strain-specific regulation of bacterial density suggests that strains do not compete with one another for limited resources within the host, and this should prevent the increased virulence that is usually expected from multiple infection (Frank 1996a). The maintenance of infection cost in A. tabida probably expresses specific constraints, such as the classical trade-off between virulence and bacterial transmission (for review, see Frank 1996b). On the one hand, bacterial density should be kept as low as possible to reduce fitness costs to the host, but on the other hand, the intracellular density of microorganisms must be high enough to ensure transmission to the next generation. Moreover, convergent selective pressures also act on the host (Turelli 1994; Vavre et al. 2003), since strain loss can have dramatic indirect effects on individual fitness. In A. tabida, this is obvious for the obligatory wAtab3 strain, since loss of this strain results in female sterility. It is also true to some extent for the two facultative strains, both of which induce high CI levels (>70%; F. Dedeine, personal communication). Losing one or both of them can be expected to expose females to CI and counterselection, whereas it should have no effect on males. Consequently, selective pressures do promote the maintenance of diversity of infection, and this could explain how the specificity of density regulation has evolved in spite of other types of competition, such as interference between Wolbachia strains. In contrast, the density data for wAtab3 suggest that there may be some positive interaction between this obligatory strain and the two facultative ones, since the density of wAtab3 is lower when it is the only strain present.

Finally, the bacterial density regulating system exhibited by A. tabida and other multiply infected species may limit both competition for resources and interference between different Wolbachia strains, and this may reflect the peculiar selective pressures acting on the system. Multiple infection by Wolbachia can be viewed as a criminal conspiracy among bacteria, with each partner relying on the others for its own fitness. Clearly, the success of the plot relies on the reciprocal agreement among accomplices, and reciprocal damage would be unacceptable.


We thank the Développement Technologique et Analyse Moléculaire de la Biodiversité for use of the Light Cycler System. This study was partly supported by Centre National de la Recherche Scientifique (IFR 41-UMR5558).


  • Communicating editor: D. M. Rand

  • Received January 23, 2004.
  • Accepted May 10, 2004.


View Abstract