A Wolbachia-associated fitness benefit depends on genetic background in Drosophila simulans

Abstract

The α-proteobacteria Wolbachia infect a number of insect species and influence host reproduction to favour the spread of infected females through a population. The fitness effect of this infection is important in understanding the spread and maintenance of Wolbachia within and among host populations. However, a full elucidation of fitness effect requires careful control of host genetic background. Here, I transferred a single clone of Wolbachia (the wHa strain) into three genetically distinct isofemale lines of the fly Drosophila simulans using microinjection methodology. These lines carried one of the three described mitochondrial haplogroups (siI, siII or siIII) and differ in nuclear genome as well. Population cage assays showed that wHa-infected siIII flies enjoyed a dramatic fitness benefit compared to uninfected siIII. In contrast, wHa did not affect the fitness of siI or siII flies. This study points to the importance of host-by-symbiont interaction terms that may play an important role in organismal-fitness.

1. Introduction

It is well appreciated that symbionts can influence the fitness of their hosts. For vertically transmitted symbionts, the fitness of their host is an important determinant of their spread and maintenance within host populations (Lipsitch et al. 1995). Teasing apart fitness contributions of host and symbiont has proven to be a challenge in some groups. Furthermore, fitness effects may depend not on the particular host or symbiont type, but on an interaction between the two. Wolbachia, a group of vertically transmitted α-proteobacteria, are a highly successful group of symbionts that infect 17–76% of all insect species (Werren et al. 1995; Jeyaprakash & Hoy 2000). These bacteria manipulate host reproduction in ways that favour their spread through populations (Hoffmann & Turelli 1997). Fitness costs or benefits conferred by the symbiont also hinder or promote spread of the infection (Caspari & Watson 1959; Turelli & Hoffmann 1991; Turelli 1994). Fitness-effect experiments have shown that Wolbachia run the gamut from parasite to commensal to mutualist. However, it is not well understood how fitness effects are manifested differently on alternative host genetic backgrounds within the same species.

In some cases, Wolbachia cannot be associated with statistically significant fitness effects (Nigro & Prout 1990; Turelli & Hoffmann 1995; Hoffmann et al. 1996; Charlat et al. 2004). However, both fitness costs and benefits have been found in a number of different insect species. Infected Tribolium beetles show a 37% reduction in overall fitness relative to uninfected beetles (Stevens & Wade 1990). Infected Drosophila females experienced a 29% reduction in the number of progeny (Hoffmann & Turelli 1988), and a follow-up experiment showed a 10–20% reduction in fecundity (Hoffmann et al. 1990). Infected male D. melanogaster produced significantly fewer sperm and suffered a reduction in fertility (Snook et al. 2000). Artificially infected D. serrata, a species not normally infected with Wolbachia, experienced significant reduction in egg-to-pupa viability and egg-to-adult viability (Clancy & Hoffmann 1997).

Wolbachia may also be associated with a fitness benefit to their hosts. In fact, a fitness benefit was indirectly inferred to explain the maintenance of Wolbachia in wild D. melanogaster populations (Hoffmann et al. 1998). This prediction was later supported in studies where infected flies had higher fecundity (Olsen et al. 2001) and showed significantly longer survival (Fry & Rand 2002). In D. simulans, infected females showed an increase in overall productivity when compared to their antibiotic-cured counterparts, although this fitness advantage appeared to be transient (Poinsot & Merçot 1997). In mosquitoes, infected females lived longer, produced more eggs and had higher hatching rates than uninfected females (Dobson et al. 2002). If the wasp Asobara tabida is cured of its Wolbachia infection, oogenesis fails (Dedeine et al. 2001). Wolbachia may also affect male reproductive success; sperm from infected Tribolium beetles out-competed sperm from uninfected males (Wade & Chang 1995).

All of these studies compared infected to uninfected individuals. There are two common ways that uninfected individuals are derived. First, one can collect naturally uninfected individuals; however, this strategy prevents the standardization of host genetic background compared to the infected individuals. Furthermore, some species are fixed for infection. A second strategy is to cure infected lines with antibiotics to create an uninfected subline with the same genetic background. However, it is possible that antibiotic treatment itself affects fitness. An alternative method to these two methods is to artificially infect a subset of individuals from a naturally uninfected line. This third strategy was employed here, in an attempt to control host genetic background without the use of antibiotics.

The fly D. simulans offers an ideal model system to ask how Wolbachia-associated fitness effects are manifested, while controlling host genetic background. D. simulans serves as a host to four genetically distinct Wolbachia strains, called wRi, wAu, wHa and wMa (Hoffmann & Turelli 1988; O'Neill & Karr 1990; Nigro 1991; Turelli & Hoffmann 1995; Ballard 2004). These four strains differ by over 12% at the nucleotide level at the wsp locus (Zhou et al. 1998) but some are invariant at the 16S locus (O'Neill et al. 1992). Interestingly, these four genetically distinct strains of Wolbachia are non-randomly associated with three genetically distinct mitochondrial haplogroups in the host D. simulans (Ballard 2004). The three mitochondrial haplogroups, called siI, siII and siIII, differ by ∼2% at the nucleotide level, while there is ∼0.1% nucleotide polymorphism within each of these haplogroups (Solignac et al. 1986; Ballard & Kreitman 1994; Ballard 2000a,b). In contrast, nuclear genetic variation is random with respect to the structure seen in mitochondria and/or Wolbachia (Ballard 2000a; Ballard et al. 2002; Dean et al. 2003; Dean & Ballard 2004).

To directly test how host genetic background affects Wolbachia-associated fitness within a species, I transferred the wHa strain into three uninfected lines that differ in their nuclear genomes and carry one of the three distinct mitochondrial haplogroups (Ballard 2000a; Ballard et al. 2002; Dean & Ballard 2004). Of the hundreds of D. simulans flies collected from over 30 countries, the wHa strain has only been found infecting siI-carrying flies (Ballard 2004), so this experiment also offers a glimpse into the effects of disrupting naturally occurring Wolbachia/mitochondria complexes. Effects on fitness were identified using a population cage assay. Compared to its uninfected counterpart, the infected siIII-carrying line showed a dramatic fitness increase. No effect of infection was seen in siI- or siII-carrying lines. These results underscore the importance of a host×symbiont interaction term in the description of fitness costs and benefits of infection within a single species.

2. Material and methods

(a) Lines used

I transferred the wHa strain, which in D. simulans has only been found in hosts with siI mitochondria (Ballard 2004), into three genetically distinct host backgrounds. The donor siI/wHa isofemale line was originally collected from Hawaii (line DSH from O'Neill & Karr 1990). DSH was chosen as the donor line because the siI mitochondrial type is apparently unable to replace or exist in a heteroplasmic state with incumbent siII or siIII (de Stordeur et al. 1989; de Stordeur 1997). It should be noted that other cytoplasmic elements may be transferred during this process.

The uninfected siI recipient line was derived from DSH through stochastic loss of infection in the laboratory, which occurs rarely (Poinsot et al. 2000). Two additional, naturally uninfected recipient isofemale lines carried the siII (line DSW from Hoffmann & Turelli 1988) or siIII mitochondrial type (line MD111 from Ballard 2000a). Nuclear heterozygosity within each host line is unlikely to contribute detectable variance. Before commencing this study, all lines were full-sib-mated for five generations to reduce nuclear variation. Furthermore, the DSH and DSW lines were maintained in the laboratory for over 10 years, while the MD111 line was maintained for over 5 years. There were many differences in both mitochondrial and nuclear sequences among these three lines (Ballard 2000a; Ballard et al. 2002; Dean et al. 2003; Dean & Ballard 2004). Although nuclear genomes among these three fly lines differ, I refer to them as siI, siII and siIII throughout the remaining part of this manuscript for simplicity.

Two D. melanogaster lines, CyO/Sp and w118, were used as common competitors in population cage experiments, as described below. CyO/Sp displays a curly winged mutation. On an average, 50% of CyO/Sp eggs do not hatch. The w118 line is marked with a recessive white-eyed mutation.

(b) Microinjections and fly husbandry

Microinjection methodology followed Santamaria (1987) with modifications. Embryos less than 1 h old were retrieved from donor and recipient lines and dechorionated in 50% bleach for 1 min. They were then washed in distilled water, followed by washing in 0.005% Triton X solution. Embryos were placed on 3 M double stick tape (Whiteley & Kassis 1993) and covered in Halocarbon 700 oil.

Under a dissecting microscope, cytoplasm was drawn from infected DSH embryos into a needle, pulled from 1 mm diameter glass capillaries. A small amount of this cytoplasm was injected into uninfected recipient embryos with a pneumatic pump. Injected embryos were placed in a sealed chamber with high-humidity and oxygen. After 24–36 h, surviving larvae were removed from the oil and placed in a vial containing standard medium. Any surviving adult females were placed individually in fresh vials with 3–6 males from her same genetic background.

Several generations after establishing an isofemale line from injected females, infection status was checked with polymerase chain reaction (PCR; following Dean et al. 2003). In most cases, lines deemed uninfected were rechecked a few generations later. In no case was an uninfected line later found to be infected. A region of mtDNA was amplified and sequenced (following James et al. 2002) from transfected lines 5–10 generations later to confirm the absence of contaminating siI mitochondria. The siI, siII and siIII mitochondria differ by 26 single nucleotide polymorphisms and three length variants in this region, allowing unambiguous identification of each mitochondrial type (Ballard 2000a). Transfected lines were also sequenced for the Wolbachia wsp locus to further confirm the presence of wHa in a homoplasmic state (following Zhou et al. 1998).

(c) Population cage experiments

Following transfections, six unique genetic combinations existed—three host types (carrying siI, siII or siIII mitochondria, as well as nuclear differences) and two infection states (infected with wHa or not). It is possible that other unknown symbionts or genetic factors were transferred during microinjections.

Each genetic combination of D. simulans was competed against a common D. melanogaster mutant in three replicate perspex cages approximately 35×20×20 cm. The D. melanogaster mutant CyO/Sp was chosen since there is no meaningful gene flow between D. simulans and D. melanogaster (reviewed in Davis et al. 1996). More specifically, hybridization results in unisexual sterile hybrids. Differences in cuticular hydrocarbons form intense premating barriers between these two species (Coyne et al. 1994). The CyO/Sp mutant is easily distinguished from D. simulans by the presence of curly wings.

Under two different temperature regimes, 20 and 25 °C, three replicate cages were run for each D. simulans+CyO/Sp combination, for a total of 36 cages. A number of developmental, behavioural, and phenotypic characters differ among flies raised at different temperatures (Neat et al. 1995; Partridge & French 1996; Gilchrist et al. 1997; James et al. 1997). To seed each cage, 250 first instar larvae from each D. simulans and CyO/Sp line were placed in each of two bottles containing standard medium, for a total of 500 larvae per bottle and 1000 larvae per cage. Because egg-to-adult viability declines daily with the age of the mother (Hercus & Hoffmann 2000), mothers of these first instar larvae were the same age across all lines.

Cages were given 23 days of development at 20 °C and 13.5 days of development at 25 °C. After this period, adults were emptied into a perspex cage containing four fresh bottles and given 48 h to lay eggs. Adults were then frozen for later processing. Each cage was maintained in this manner until it became fixed for either D. simulans or CyO/Sp, or until the tenth generation.

All the adults were sorted and counted from each cage every second generation. The proportion of D. simulans was calculated and arcsin-transformed (Sokal & Rohlf 1995). This number was taken as a proxy for the competitive ability of each D. simulans line. The data were then analysed using a mixed model ANOVA, using host genetic background, infection status, temperature and their four possible interaction terms as the main factors. Since the adults of one generation seeded the next, data may not be independent among generations. Therefore, generation number was considered as a first-order autoregressive source of covariation. Since the very first generation (generation 0 in figure 1) measured larval competitive ability only, it was excluded from statistical analyses. All the statistical analyses were performed using SAS v. 9.0.

Figure 1.

Proportion of D. simulans versus CyO/Sp through 10 generations at two temperatures. The first set of population cages involved competition between CyO/Sp and (a) siI at 20 °C or (b) 25 °C; (c) siII at 20 °C or (d) 25 °C; siIII at (e) 20 °C or (f) 25 °. The second set of population cages involved competition between w118 and (g) siIII at 20 °C or (h) 25 °C. Straight lines indicate an uninfected host line; dashed lines indicate transfected counterparts.

(d) Identification accuracy

To assess the accuracy in scoring flies, an allele-specific PCR was designed that selectively amplified either a 1139 bp fragment from D. melanogaster or a 437 bp fragment from D. simulans mtDNA (Ballard 2000a). Flies were sampled from three stages of the experiment. First, three randomly chosen adult flies from each of the 36 cages, originally identified as D. simulans were assayed from generation 0. Second, three flies originally identified as D. melanogaster were assayed from generation 0. Third, three flies originally identified as D. simulans were assayed from either generation 10 or from the generation in which D. simulans reached fixation.

(e) Potential loss of infection

After approximately four generations of the population cage experiment, flies from each cage were assayed to confirm that they were still infected or uninfected as expected. Loss of infection may occur at low frequency in the laboratory (Poinsot et al. 2000), but should be selected against in a cage experiment.

3. Results

(a) Microinjections

5984 embryos were injected, from which eight transfected lines were successfully established. Among the 5984 embryos injected, 770 survived injections. Of these, 202 reached adulthood. Among these, 102 were males and not useful in establishing infected lines since Wolbachia is a maternally transmitted symbiont. Of the 100 females, 13 successfully established new isofemale lines. Of these, eight lines tested positive for the presence of Wolbachia, three of which were used in this study. These numbers are probably an overestimate of failure rate, as techniques were continually optimized throughout the experiment.

The three transfected lines used in the population cage experiments yielded ‘clean’ sequence from their incumbent mitochondria, confirming the absence of contamination by siI mitochondria. All transfected lines also yielded ‘clean’ wHa sequence. Therefore, the transfected siII and siIII lines represented a decoupling of natural host/symbiont associations.

(b) Population cage experiments

For this part of the experiment, 191 367 flies were counted, for an average (standard deviation, s.d.) of 1031 (498) flies per counted cage per generation. When in competition with CyO/Sp, the proportion of D. simulans increased over time in all cages except those containing uninfected siIII (figure 1a–f). Competitive ability differed significantly based on host genetic background (F_1,168=221.89, p<0.0001, table 1). The siII line increased most rapidly, reaching fixation by the fourth generation. At 25 °C, the siII line reached fixation by the second generation. At 25 °C, the siI and siIII lines showed more fluctuation in the proportion of D. simulans compared to a more steady increase seen at 20 °C.

Table 1.

Mixed model ANOVA results of population cage assays.

competitor	effect	num d.f.	den d.f.	F value	p
CyO/Sp	temp	1	168	4.12	0.04
	host	2	168	221.89	<0.0001
	infection	1	168	298.23	<0.0001
	temp×host	2	168	1.92	0.15
	temp×infection	1	168	2.45	0.12
	host×infection	2	168	246.07	<0.0001
	temp×host×infection	2	168	0.86	0.42
w118	temp	1	18	1.68	0.21
	infection	1	18	69.91	<0.0001
	temp×infection	1	18	2.87	0.11

The presence of the wHa strain of Wolbachia played a significant role in the competitive ability of these flies (F_1,168=298.23, p<0.0001, table 1), and interacted significantly with host genetic background (F_2,168=246.07, p<0.0001, table 1). Specifically, the presence of Wolbachia dramatically increased competitive ability in the siIII host compared to uninfected siIII flies, although it did not affect competitive ability in siI or siII hosts (figure 1e and f). Temperature significantly affected competitive ability (F_1,168=4.12, p≈0.0.04), but this was probably due to the greater fluctuation seen at 25 °C (figure 1b and f). Temperature did not interact significantly with either host or symbiont (table 1).

The dramatic difference between infected and uninfected siIII lines warranted further corroboration. The above experiment was repeated for the siIII lines only using the alternative D. melanogaster mutant, w118. This mutant was marked by white eyes and, in contrast to CyO/Sp, was not 50% lethal. Therefore, w118 was expected to be a more intense competitor. For this experiment, 49 599 flies were counted for an average (s.d.) of 1459 (728) flies per counted cage. The wHa strain significantly increased competitive ability of siIII flies (F_1,18=69.91, p<0.0001; figure 1g and h, table 1). Temperature did not contribute to variation (F_1,18=1.68, p≈0.21). Thus, in two independent experiments, the infected siIII host showed an increased competitive ability against two independent D. melanogaster competitors.

(c) Identification accuracy

Of the 232 PCR reactions, 229 confirmed the original identification. This constitutes an identification accuracy of nearly 99%. Among the three misidentifications, two were flies incorrectly identified as D. melanogaster and one was a fly incorrectly identified as D. simulans. Thus, misidentification is unlikely to contribute substantially to the analyses presented above.

(d) Potential loss of infection

In the fourth generation of the population cages, all transfected lines were confirmed to be infected and all uninfected lines were confirmed to be uninfected. The wHa strain of Wolbachia induces strong incompatibility in its natural siI host (O'Neill & Karr 1990; James & Ballard 2000), suggesting loss of infection would be selected against in a cage environment.

4. Discussion

Competitive ability is probably a complex interaction between life-history characteristics such as fecundity, development time and longevity. The genome of any individual will obviously influence these characteristics, but interaction terms between symbiont and host may also be important. While this experiment showed a clear difference in competitive ability among infected and uninfected siIII, it cannot necessarily reject differences between infected and uninfected siI or siII lines. Differences in competitive ability might be subtle and difficult to detect given the rapid increase in frequency of siI and siII lines during the population cage assay.

In terms of species interactions, Wolbachia run the gamut from parasite (Hoffmann & Turelli 1988; Hoffmann et al. 1990; Stevens & Wade 1990; Clancy & Hoffmann 1997; Snook et al. 2000) to commensal (Nigro & Prout 1990; Turelli & Hoffmann 1995) to mutualist (Wade & Chang 1995; Poinsot & Merçot 1997; Hoffmann et al. 1998; Dedeine et al. 2001; Olsen et al. 2001; Dobson et al. 2002; Fry & Rand 2002). By controlling host genetic background, this study points to the importance of host genetic background when describing fitness effects and the rate of spread within a single species. In addition, it is the first explicit disruption of natural Wolbachia/mitochondrial complexes. A fitness increase was observed in an unnatural host (siIII versus siI), supporting the hypothesis that association between particular Wolbachia strains and mitochondrial groups is not maintained directly by selection (Dean & Ballard 2005).

Unpublished data further support the conclusion that Wolbachia increases fitness of siIII-carrying flies (A. C. James 2003, personal communication). After artificially curing the naturally wMa-infected, siIII-carrying line MD199 with tetracycline, significantly fewer progeny developed after 3 days of egg-laying compared to the original infected line. One interpretation is that tetracycline had a detrimental effect on the cured line. However, curing several other infected hosts that carried the siI and siII mitochondrial types showed no effect. Therefore, a fitness advantage associated with Wolbachia was observed independently in a different host and symbiont than those studied here. The significant increase in fitness of siIII hosts due to Wolbachia infection may be a general phenomenon, at least in the laboratory.

How do these results apply to the distribution of Wolbachia in nature? In nature, siIII hosts may be infected with the wMa strain, which does not induce strong incompatibility (James & Ballard 2000). Theory predicts that without fitness benefits or improved transmission dynamics, the wMa symbiont should be stochastically lost from populations (Turelli 1994). The interaction term observed here may help to explain why Wolbachia has not been lost from siIII populations in nature. On the other hand, with such a dramatic increase in competitive ability, infection frequency might be expected to reach fixation. Instead, the wMa strain attains a relatively low-equilibrium infection frequency of 0.17 (95% C.I.=0.13–0.22, n=245 isofemale lines; combined data of Merçot & Poinsot 1998; James & Ballard 2000; Dean et al. 2003). Future field study should determine if and how competitive ability is affected in nature. It is also possible that some nuclear backgrounds of siIII-carrying flies do not enjoy such a fitness advantage.

Clearly, one weakness of this study is that each transfected line was created once in the laboratory. Without multiple transfected lines, the generality of the results observed here cannot be further assessed. The low success rate of microinjections (see §3a) posed a technical limitation. Further experiments are needed to ask whether the fitness increase observed here is a general phenomenon.

An alternative hypothesis to explain the patterns observed here is that the uninfected siIII parental stock acquired a deleterious mutation prior to commencement of the population cage experiments. This alternative hypothesis seems unlikely as it would require a strongly deleterious mutation to fix in the short time between transinfecting the subline of siIII and commencing the population cages, a period of less than six months.

It is possible that the presence of Wolbachia simply marks the presence of other unknown symbiont or genetic factors that were the true cause of the interaction term observed here. Until Wolbachia can be cultured in isolation, any linkage to unknown symbionts cannot be adequately addressed. Regardless of true cause, the interaction term observed here reveals an additional layer of complexity when considering organismal fitness. Even within a single species, alternative host genetic backgrounds may manifest symbiont-associated fitness effects differently.