• Roshan K. Vijendravarma,

    1. NERC Centre for Population Biology, Imperial College London, Silwood Park Campus, Ascot, Berks SL5 7PY, United Kingdom
    2. E-mail: roshan.vijendravarma@unil.ch
    Search for more papers by this author
    • 2

      Current address: Department of Ecology and Evolution, University of Lausanne, Le Biophore, CH 1015 Lausanne, Switzerland

  • Alex R. Kraaijeveld,

    1. NERC Centre for Population Biology, Imperial College London, Silwood Park Campus, Ascot, Berks SL5 7PY, United Kingdom
    2. E-mail: a.r.kraaijeveld@soton.ac.uk
    Search for more papers by this author
    • 4

      Current address: School of Biological Sciences, University of Southampton, Boldrewood Campus, Southampton SO16 7PX, United Kingdom

  • H. C. J. Godfray

    1. NERC Centre for Population Biology, Imperial College London, Silwood Park Campus, Ascot, Berks SL5 7PY, United Kingdom
    2. E-mail: charles.godfray@zoo.ox.ac.uk
    Search for more papers by this author
    • 6

      Current address: Department of Zoology, University of Oxford, South Parks Road, Oxford OX1 3PS, United Kingdom


Most organisms experience strong selection to develop mechanisms to resist or tolerate their pathogens or parasites. Limits to adaptation are set by correlated responses to selection, for example reduced abilities to detect other parasites or trade-offs with other fitness components. For a few model systems it is now becoming possible to compare the evolutionary response to a broad range of natural enemies. In Drosophila, the evolutionary responses to ectoparasitic mites, parasitoids, and fungal and bacterial pathogens have previously been studied. Here replicate lines of D. melanogaster were exposed to the microsporidian parasite Tubulinosema kingi over a period of 61 weeks, with overlapping generations. Compared to controls, exposed lines had higher early-life fecundity and increased longevity when infected suggesting successful selection for resistance or tolerance. In the absence of the pathogen, exposed lines had lower fecundity when reared under harsh environmental conditions, and were poorer larval competitors than controls. They also had relatively higher densities of haemocytes, a component of the cellular immune system. Defense against this pathogen resembles more that against macroparasites than microsparasites, and this is interpreted in the light of what is known about the mechanisms of resistance to microsporidians.

Virtually all organisms are attacked by parasites and pathogens that often cause considerable mortality (Price 1980; Godfray 1994; Combes 2001). There are thus strong selection pressures to evolve methods of resistance or tolerance to reduce their deleterious effects. Despite this, genetic polymorphism for resistance and tolerance is pervasive and common. A frequent explanation for the presence of this genetic variation is that increased resistance leads to negative consequences for other components of fitness, in other words that there are trade-offs between resistance and the organism's other biological functions (Haldane 1949; Anderson and May 1982; Antonovics and Thrall 1994; Sheldon and Verhulst 1996; Gemmill and Read 1998; Fellowes and Godfray 2000; Agrawal and Lively 2002; Rigby et al. 2002; Little et al. 2008). Where investment in resistance requires the redirection of limiting resources it can be considered as one of many components of an organism's life-history strategy with the optimum allocation depending on the balance between the risks and costs of infection and the consequences of reallocation (Roff 1992; Stearns 1992). Alternatively, investment in defense against one parasite or pathogen may increase susceptibility to another because of qualitative switches in traits determining pathogen recognition (Rigby et al. 2002; Schmid-Hempel and Ebert 2003) or increase risk of auto-immune disease (Bischoff et al. 2006). The presence of trade-offs in resistance or tolerance to challenges that may vary in space or time, combined with the possibility of an antagonistic coevolutionary response by the pathogen, can explain substantial standing genetic variation (Anderson and May 1982; Antonovics and Thrall 1994; Sasaki and Godfray 1999).

When an organism is attacked by a parasite or pathogen, immune and other defense systems are typically upregulated and this may immediately lead to costs and other consequences, for examples reductions in fertility, fecundity or survival, or increased exposure to other pathogens (Fellowes et al. 1999c; Moret and Schmid-Hempel 2000; Rigby et al. 2002). Measuring this type of trade-off is relatively straightforward as the host can be exposed to experimental infection or to an artificial challenge that triggers a defensive response without the complication of the presence of a living parasite. A more difficult type of trade-off to study is the costs or other consequences of making constitutive changes to defense strategies. One approach is to look for genetic correlations between different traits using family or population-level studies (Reznick 1985) although this can be difficult for traits in which the chief response is binary (survive or die). An alternative powerful tool, as in the study of other aspects of life-history theory, is artificial selection experiments (Rose 1984; Partridge and Fowler 1992). An organism is selected to invest more in defense, or to make qualitative changes in the type of defense it deploys, and genetic trade-offs are identified as negative correlated responses to selection. This technique is not without problems as typically selection is strong and may disrupt the genetic architecture of the experimental population (Reznick 1985), but for most systems it is currently the best available. For organisms in which the precise mechanistic basis of defense is understood, and where gene manipulation technology is available, the consequences of different defense genotypes can be compared in a constant genetic background, a technique so far largely used in studies of plant pathogens (Bergelson et al. 1996; Goss and Bergelson 2006).

Studies of the consequences of enhanced resistance are beginning to accumulate. In the presence of elevated bacteriophage densities bacteria can evolve to become resistant, sometimes to numerous viral genotypes, and there is evidence that this is associated with costs, typically due to reduced replication rates because of poorer nutrient assimilation (Chao et al. 1977; Lenski and Levin 1985; Buckling and Rainey 2002; Morgan et al. 2005). Paramecium are attacked by bacteria (Holospora) and can evolve resistance that is most effective against local bacterial genotypes and that has costs in reduced replication rates (Lohse et al. 2006). In plants there is some evidence for costs associated with increased resistance, although many studies have failed to find any correlated reductions in fitness (Bergelson et al. 1996; Brown 2003; Goss and Bergelson 2006). Studies of animal systems are still relatively uncommon and have not always identified trade-offs (Ferrari et al. 2001). Outside of Drosophila, costs to increased resistance have been identified using artificial selection experiments in moths attacked by a virus (Boots and Begon 1993), mosquitoes infected by malaria (Yan et al. 1997; Hurd et al. 2005), and snails attacked by the schistosome parasite (Webster and Woolhouse 1999). Direct selection on phenoloxidase activity, an important part of the invertebrate immune system, resulted in reduced longevity under starvation in Scatophaga dung flies (Schwarzenbach and Ward 2006).

Drosophila species are attacked by a range of pathogens and parasites and sufficient studies have now been completed that it is beginning to be possible to compare their evolutionary responses to a range of natural enemies. Mites in the genus Macrocheles disperse between breeding sites by attaching themselves to Drosophila nigrospiracula adults that reduces the latter's fitness. Artificial selection for improved behavioral defenses leads to a reduction in fecundity, especially in stressed files (Luong and Polak 2007a,b) Drosophila larvae are attacked by specialist parasitoid wasps (Hymenoptera) that lay their eggs in early-instar larvae. The parasitoid does not progress beyond the first instar until its host has pupated and during this period of suspended development the egg and larva have to avoid encapsulation, the cellular immune response mounted by invertebrates to natural enemies too big to phagocytose (Meister and Lagueux 2003). Extensive within- and between-population variation in resistance to parasitoids has been observed in several Drosophila species (e.g., Kraaijeveld and van Alphen 1994; Dupas and Carton 1999; Kraaijeveld and Godfray 1999; Dupas et al. 2003; Fleury et al. 2004; Dubuffet et al. 2007). Replicate outbred lines of Drosophila melanogaster have been selected to become more efficient at encapsulating two species of parasitoid, and in both cases improved resistance was correlated with decreased fitness under conditions of severe larval competition for food (Kraaijeveld and Godfray 1997; Fellowes et al. 1998). Resistant lines had approximately twice the density of circulating haemocytes (Kraaijeveld et al. 2000), the cells responsible for encapsulation, but also had slower rates of food ingestion (Fellowes et al. 1999a). Genetic variation for susceptibility to the generalist fungal pathogen Beauveria bassiana has been found in both within- and between-population comparisons (Tinsley et al. 2006) although in artificial selection experiments little or no evidence was found for increased resistance, although possibly some increased tolerance (Kraaijeveld and Godfray 2008). Substantial genetic variation in the D. melanogaster humoral response to bacterial infection has been found (Corby-Harris and Promislow 2008), and the molecular basis of part of the response characterized (Lazzaro et al. 2004). Recently, negative genetic correlations between resistance to the bacterium Providencia rettgeri and several traits associated with fitness including fecundity have been demonstrated (McKean et al. 2008). Interestingly, these trade-offs were only manifest when the flies were in an environment in which food was limited. However, the genetic variation in responses against different species of bacteria appears to be largely uncorrelated (Lazzaro et al. 2006; Corby-Harris and Promislow 2008). The primarily vertically transmitted endosymbiont Wolbachia is common in Drosophila species in which it spreads by manipulating host reproduction, particularly causing cytoplasmic incompatibility (Werren 1997). It has complex effects on host fitness, which can be positive or negative, and are frequently age- or environment-dependent (Montenegro et al. 2006; Yamada et al. 2007). Theory indicates that hosts and symbionts in such systems should evolve to reduce costs (Turelli 1994), and there is now good evidence from Drosophila simulans for Wolbachia evolving to become less costly (Weeks et al. 2007). The best-studied virus attacking Drosophila is sigma virus (Brun and Plus 1980). Genetic variation in sigma virus resistance has been recorded, and is associated with a relatively small number of genes (Brun and Plus 1980; Contamine et al. 1989). We are not aware of studies of the costs of resistance to viruses.

Microsporidia are intracellular pathogens, now thought to be closely related to fungi (Hirt et al. 1999; James et al. 2006; Hibbett et al. 2007) that attack a broad range of eukaryote hosts including humans and Drosophila. They are immediately recognizable by the coiled polar filament in the spore stage responsible for horizontal infection. Some species are only transmitted horizontally whereas in others varying degrees of vertical inheritance also occur (Dunn and Smith 2001; Dunn et al. 2001). Species with almost exclusive vertical transmission spread by manipulating their hosts' reproductive system to produce more females than normal (Dunn and Smith 2001). Microsporidia can have a significant effect on host population dynamics and have been used in, or considered for, pest or vector control (Lewis 2002; Andreadis 2007; Rivero et al. 2007), or are significant pathogens of beneficial insects (Schmid-Hempel 1998). Hosts are known to vary in their resistance to Microsporidia. For example, field and laboratory experimental studies of several species attacking Daphnia have shown they can affect the clonal composition of populations (Capaul and Ebert 2003; Haag and Ebert 2004; Refardt and Ebert 2007).

Microsporidian pathogens have been recorded from several Drosophila species and are known to reduce host fitness, but whether the fly is able to evolve to become more resistant to the pathogen, and whether this is costly, has not been studied. A microsporidian infection appeared spontaneously in some of our cultured fly lines. The infection was recognized by its affect on Drosophila parasitoids in which it multiplies rapidly causing the abdomen to become swollen and distended. Molecular and ultrastructural studies (Franzen et al. 2006) showed it to belong to the genus Tubulinosema and although the original description (Kramer 1964) omits some details it is almost certainly conspecific with Tubulinosema kingi which was described from Drosophila. Aspects of the biology of T. kingi have been studied by Armstrong and colleagues (Armstrong 1976; Armstrong et al. 1986; Armstrong and Bass 1989a,b), and more recently by Futerman et al. (2006) and Vijendravarma et al. (2008) using the same fly population as that employed here. Adult flies infected by T. kingi suffered a 35–55% reduction in early-life fecundity, while development time and pupal mortality were also increased (Futerman et al. 2006). The larval stage is the only host life-stage susceptible to the pathogen (Vijendravarma et al. 2008) and transmission appeared chiefly through larval ingestion of spores derived from the cadavers of infected adults, with peri-ovarial vertical transmission also occurring but at low frequency (Futerman et al. 2006).

Here we select replicate lines of D. melanogaster to be more resistant to T. kingi. We ask whether there is heritable variation in resistance or tolerance to this pathogen, and whether this has correlated negative effects on fitness.



The D. melanogaster culture used for these experiments was derived from 250 wild flies captured near Leiden in the Netherlands in 1995. Since its establishment it has been maintained as a large laboratory population with steps taken to avoid inbreeding. This population has been used in our laboratory for artificial selection experiments designed to increase Drosophila resistance to other natural enemies (Kraaijeveld and Godfray 1997; Fellowes et al. 1998, 1999b; Sanders et al. 2005; Kraaijeveld and Godfray 2008).

Populations of flies guaranteed to be without infection were obtained by initiating a culture from eggs that had been surface sterilized with 0.6% NaOCl solution for 5 min followed by repeated washing. Giemsa-stained preparations of adult fly abdomens were made to check no spores were present, and the absence of infection was also confirmed by diagnostic PCR (see Futerman et al. (2006) for more details on Giemsa-staining and PCR). Microsporidia were maintained in a population of Drosophila subobscura (a good host for T. kingi) where the adults were allowed to die on and contaminate the larval medium to maximize transmission. Spores for the selection experiments were obtained by allowing the parasitoid Asobara tabida (Hymenoptera, Braconidae) to attack infected D. subobscura larvae. Individuals with physogastric abdomens were collected and microsporidial spores were harvested in 0.1% SDS (sodium dodecyl sulphate).


Five replicate pairs of selection and control lines were set up. Each pair of lines was treated in exactly the same way—for example cultured on the same day using the same batch of medium—except for exposure to the pathogen. The lines were initiated by placing 250 eggs laid by flies from the uninfected D. melanogaster population in culture bottles with standard Drosophila yeast/sugar medium and live yeast. The bottles were kept at 20°C in a 16:8 h light:dark regime until the adults emerged; the flies were then released into Perspex cages (20 × 20 × 20 cm) maintained under the same conditions in a controlled-temperature room at ambient humidity with constant access to honey and water. Twice a week the flies in each cage were allowed to oviposit for 24 h on standard medium and yeast in a sterile petri dish (9 cm). One hundred fifty eggs per plate were collected and transferred to rearing bottles and the resulting adult flies were released back into the appropriate line's Perspex cage. Dead flies in each cage were removed every second day. The water in the cages was changed every fortnight whereas the honey was replenished weekly.

Selection but not control populations were exposed to microsporidian infection. Approximately 107T. kingi spores in 0.1% SDS were added to the rearing bottles for the selection lines immediately after transfer of the eggs, while the control lines received the same volume of SDS without spores. Following an initial resistance assay (see below) the T. kingi spore dose used to infect the selection lines was increased to ∼108 from week 37 onward. The day before eggs were collected 10 flies were chosen randomly from the control cages and Giemsa-stained preparations of the abdomens made to check for infection. Ten flies from the selection cages were also removed to avoid any bias. No instances of cross-infection were found. The experiment was maintained for 73 weeks of overlapping generations although selection was suspended between weeks 19 and 31 due to unavoidable circumstances. During this 13-week period, the populations were maintained in standard culture bottles containing medium and live yeast with nonoverlapping generations.


The fecundity and survival schedules of flies from each line were assayed in the presence or absence of the microsporidian at week 34 and at the end of the experiment. Prior to the assays, flies of all lines were cured of any infection as described above; 15 flies per line were subsequently screened for T. kingi (see also above) but no infections were recorded. Both sets of populations were cultured in standard Drosophila bottles for one generation to remove any maternal effects arising from conditions in the population cages. We also measured differences between selection and control lines in two factors that might affect the consequences of microsporidian infection: the average spore density in infected flies and the density of circulating haemocytes (blood cells) in uninfected larvae.

Fecundity assay

Eight vials (80 × 22 mm) containing rearing medium were set up for each control or selection line. Fifty eggs were placed in each vial and in the first assay a dose of ∼2.5 × 106 spores in SDS buffer was added to four of the vials, the other four receiving the same volume of buffer. In the second assay the spore dose was increased 10-fold (as it had been in the selection protocol). The rearing vials were maintained under standard conditions until the adults emerged. Four females from each vial were randomly selected and placed individually with two males in further vials containing Drosophila medium. After 24 h the flies were transferred to a new vial and the number of eggs laid in the previous day was recorded. This procedure was repeated for 10 days. Any males that died were removed and replaced. All females were screened for T. kingi (by examining Giemsa-stained abdomen smears) to check that the infection treatment had been successful and that there had been no cross-contamination.

Survival assay

Four hundred eggs laid by flies from each of the control and selection lines were collected and distributed equally between two bottles containing standard medium and live yeast. In the first assay, ∼107T. kingi spores in 0.1% SDS were added to one bottle whereas the other received an equal volume of SDS. In the second assay the spore dose was increased 10-fold (as it had been in the selection protocol). The bottles were maintained at 20°C until the pupae had darkened. The pupae were then gently washed out of the bottles and 100 from each were placed individually in glass vials (50 × 12 mm) containing a small amount of honey. These were plugged with cotton wool that was kept damp and the vials were stored at 20°C under a 16:8 h light:dark regime. The longevity of each fly, to the nearest day, and its sex were recorded. Flies that failed to emerge from their pupae were not included in the analysis.

Spore density

To determine spore density in infected flies, 300 eggs were collected from flies in each line and distributed equally among six vials (80 × 22 mm) containing standard medium and live yeast. On the following day ∼2.5 × 107T. kingi spores in 0.1% SDS were added to each vial that was then incubated at 20°C until the adult flies emerged. Five flies were randomly selected from each vial and each individual homogenized in 100 μl 0.1% SDS. Spore densities were then estimated using a hemocytometer.

Hemocyte density

Flies were allowed to oviposit overnight in culture bottles containing Drosophila medium and live yeast at 25°C. The eggs were washed out of the bottles and for each control or selection line 50 eggs were added to each of four vials (80 × 22 mm) containing medium and live yeast. The vials were incubated at 20°C for 96 h. Fifteen third-instar larvae were taken from each vial and their hemolymph extracted and pooled. Three microliters of haemolymph was then pipetted onto a hemocytometer and the number of hemocytes (all classes) was counted using a light microscope at 40× magnification. Two estimates of hemocyte density were made for each replicate.


We explored whether selection for improved resistance or tolerance to Microsporidia led to correlated responses that negatively affected other aspects of fitness. Previous studies of life-history trade-offs (e.g., Partridge and Fowler 1992), and our own earlier work on selection for parasitoid resistance in Drosophila (Kraaijeveld and Godfray 1997; Fellowes et al. 1998), suggest that such correlated responses are most easy to detect when the organism faces challenging circumstances. We thus asked whether adult flies from selected lines had reduced fecundity compared to those from control lines when reared under conditions of strong larval resource competition. We also compared the survival of larvae from the two sets of lines when they competed for resources with a standard, genetically marked fly strain.

Adult fecundity in harsh environment

Flies from each line were allowed to oviposit in culture bottles containing medium and live yeast for 6 h at 25°C. The eggs were extracted by washing and for each line 50 eggs were placed in each of four vials (80 × 22 mm) that had been lined with agar and contained 0.25 mL of larval food (25 g live bakers yeast per 100 mL water). Previous studies had indicated that an average of 0.005 mL of food per individual was sufficient to stress the larvae and reduce survival and adult size. The vials were incubated at 20°C and when the adult flies emerged, four females from each vial were placed individually in new vials of the same size containing medium and live yeast together with two males from the same line. For the following 10 days flies were transferred daily to new vials and the number of eggs laid during the previous 24 h was recorded. Any dead males were removed and replaced.

Larval competition

Larvae from each of the lines were competed against a common strain carrying the sparkling poliert genetic marker whose adults have sparkling red eyes. Flies from the experimental and marker lines were allowed to oviposit overnight in bottles containing medium and live yeast at 25°C. The bottles were incubated at 20°C for 48 h after which the larvae were extracted by washing. Fifteen second-instar larvae from an experimental (selection or control) line were placed together with 15 second-instar larvae from the marker line in Petri dishes (5 cm diameter) lined with agar and either 0.2 mL or 0.1 mL of larval food (25 g live bakers yeast per 100 mL water). Our previous work (Kraaijeveld and Godfray 1997) had shown these resource levels represented weak or strong competition regimes, respectively. Fifteen replicates of each combination of experimental line and resource level were carried out. The Petri dishes were incubated at 20°C until the flies emerged and the number of experimental and marker flies that survived in each replicate was recorded. We analyzed the data by calculating a “competition index,” log(e/(m+ 1)), where e is the number of experimental and m the number of marker flies that emerged in each replicate (Santos et al. 1992).


Data were analyzed using standard analysis of variance (ANOVA) unless otherwise indicated.


Comparisons of survival and fecundity were first made after 34 weeks. These showed little differences in the performance of selection and control lines. Based on these results the spore dose used for artificial selection was increased as described in the Methods. We first describe briefly the results of the 34-week assays and then more fully the comparisons made at the end of the period of artificial selection.


Infected flies had significantly reduced fecundity in the first 10 days of life (534.1 ± 8.0 vs. 660.6 ± 5.9 for controls; F1,16= 143.9, P < 0.0001) confirming that the microsporidian is a pathogen. There was no difference in the average fitness of control and selection lines across the two infection treatments (F1,16= 0.09, P= 0.77), nor was there an interaction between selection history and infection treatment (F1,16= 0.60, P= 0.45).

We found no difference in longevity between males and females (F1,38= 0.51, P= 0.48) but flies exposed to the microsporidian had a significantly shorter life span (8.9 ± 0.3 vs. 11.4 ± 0.4 for controls; F1,16= 28.2, P < 0.001). There was also no difference in longevity between flies from the selection and control lines (F1,16= 2.4, P= 0.13), and no interaction between selection history and infection treatment (F1,16= 0.76, P= 0.39).

The week 34 bioassays thus showed no response to the selection imposed by exposing the flies to this level of microsporidian infection. This prompted us to increase the spore dose for the remainder of the experiment, as described in the Methods.


Fourteen females that failed to lay any eggs over the 10-day period of the fecundity assay were excluded from the analysis. There was no association between the occurrence of sterility and either selection history or infection treatment (Fisher's exact test, P= 0.11)

The average number of eggs laid per day per fly over the first 10 days of adult life is shown in Figure 1. There was a significant interaction between selection history and infection treatment (F1,16= 9.7, P= 0.007). Flies from the selection lines produced significantly more eggs when infected by the microsporidian compared to flies from the control lines (45.1 ± 2.4 vs. 39.1 ± 1.6). However, in the absence of infection, the pattern was reversed: flies from the control lines on average produced more eggs than those from the selection lines (70.7 ± 1.6 vs. 64.3 ± 2.1). As observed in the earlier assay, infection by the microsporidian had a major effect on fecundity with average number of eggs produced across all lines being 377.3 ± 19.1 and 673.1 ± 16.4 in the presence and absence of infection, respectively.

Figure 1.

The average daily fecundity over the first 10 days of adult life (mean ± SE) for infected and uninfected flies from the control and selection treatments.

The longevity of flies from the four selection history and treatment combinations is shown in Figure 2. No significant main or interaction effects involving fly sex were found and in the figure and in the analysis below data from males and females are combined. Overall, microsporidian infection significantly reduced longevity (F1,16= 32.3, P= < 0.001) and the interaction between selection history and infection treatment approached significance (F1,16= 3.0, P= 0.09). This was due to infected flies from control lines dying sooner than those from selection lines (3.00 ± 0.35 vs. 3.92 ± 0.07 days; F1,8= 6.58, P= 0.03) whereas there was no difference in the longevity of uninfected control and selection flies (5.22 ± 0.32 vs. 5.10 ± 0.35 days; F1,8= 0.06, P= 0.8).

Figure 2.

Longevity (mean ± SE) of infected and uninfected flies from the control and selection treatments.

At the end of the experiment infected flies from each line were homogenized and microsporidian spore densities were estimated using a hemocytometer. The mean spore densities from selection and control line flies were 7.9 ± 0.3 × 106 and 10.8 ± 0.4 × 106, respectively. Selection line flies thus had ∼27% fewer spores, a highly significant difference (F1,8= 31.7, P= 0.0004).

The densities of circulating haemocytes in uninfected larvae were estimated at the end of the experiment. Larvae from selected lines had on average 14% more blood cells than control lines (119 ± 2 vs. 104 ± 3 cells/count), a significant difference (F1,8= 15.1, P= 0.005).


Adult fecundity in harsh environment

Flies from selected lines that had been reared under conditions of harsh larval competition had significantly lower fecundity in the first 10 days of life (37.7 ± 0.9 eggs d−1) compared with flies from control lines (60.4 ± 1.7 eggs d−1; F1,8= 129.4, P < 0.001; Fig. 3).

Figure 3.

The average daily fecundity over the first 10 days of life (mean ± SE) of flies from selection and control flies (light bar) that had been reared under conditions of harsh larval competition.

Larval competition

Flies from the selection and control lines were competed against flies from a genetically marked line under low and high levels of resource stress. The relative competitive ability of the two classes of insect is shown in Figure 4. At low levels of competition there is no significant difference in the performance of fly larvae from control and selection lines (F1,9= 4.4, P= 0.07) but when competition is harsh the performance of selection line flies is severely impaired (F1,9= 80.2, P < 0.0001). Note, comparisons can only be made within the two competition levels as it is not known how the performance of the genetically marked line responds to increased competition.

Figure 4.

The relative competitive ability (mean ± SE of the competition index defined in the main text) of fly larvae from selection (dark bars) and control (light bars) lines under low and high levels of larval competition.


We used experimental evolution to determine whether there was heritable variation for resistance to microsporidian infection in D. melanogaster, and whether there were trade-offs between improved resistance and other components of fitness. We were able to select for increased resistance, although our initial efforts with a relatively low dose of microsporidian spores did not lead to detectable improvements in performance. After we increased spore dose we found that, when infected, selected flies had higher fecundity and longevity compared to controls. Infected adults from the selected lines had lower densities of microsporidian spores compared to controls, whereas their larvae had higher densities of circulating haemocytes. The possibility of costs to improved resistance was suggested by the observation that, when uninfected, selection line flies had lower fecundity than controls. In experiments explicitly designed to look for trade-offs we found that selection line flies reared in harsh environments had lower fecundity compared with controls, and that they were weaker larval competitors compared with controls when tested against the same genetically marked strain.

Wherever it has been looked for in D. melanogaster, heritable variation in resistance to natural enemies has been found (Contamine et al. 1989; Kraaijeveld and Godfray 1997; Fellowes et al. 1998; Lazzaro et al. 2006; Tinsley et al. 2006), and our study is no exception. There are, however, considerable differences in the magnitude of additive variation across different parasites and pathogens. Artificial selection for resistance to parasitoids leads to dramatic increases in host fitness (Kraaijeveld and Godfray 1997; Fellowes et al. 1998). For example, only ∼0.5% of the population of D. melanogaster used in these experiments could escape mortality from the figitid parasitoid Leptopilina boulardi yet after only five generations of selection survival had risen to ∼45%. Similar if not quite so dramatic increases in survival (from ∼5% to ∼55%) occurred after selection for resistance to a second unrelated parasitoid species. The response to selection is consistent with hereditabilities of resistance of the order of 20% In contrast, standing heritable variation for resistance to microparasites seems to be less strong. Our group found detectable but small responses to selection for resistance to a fungal pathogen (Kraaijeveld and Godfray 2008), whereas quantitative genetic studies of responses to fungi and bacteria have similarly shown significant but more modest variation (Lazzaro et al. 2004, 2006; Tinsley et al. 2006; McKean et al. 2008). However, comparisons of responses must be made cautiously as they involve different traits and different methods of measurement, and a systematic comparison has yet to be carried out. One difference between these two classes of natural enemies is that the parasitoids are Drosophila specialists (in the case of L. boulardi a specialist on melanogaster and related species) that are known to cause significant mortality in some but not all natural populations of D. melanogaster (Carton et al. 1987, 1991; Carton and Sokolowski 1992). In contrast, the bacteria and fungi that have been used are much more generalist, attacking a broad spectrum of insects. It is also not clear the degree to which they kill flies in the field (Tinsley et al. 2006; Corby-Harris and Promislow 2008), and whether what is being selected is a much more general microbial infection response. The additive genetic variation and the response to selection we observed to a microsporidian challenge was modest, more akin to the fungal and bacterial cases than the parasitoid. Unfortunately we know very little about the natural history of T. kingi that has chiefly been studied as an adventitious pathogen of laboratory Drosophila populations (Armstrong et al. 1986; Armstrong and Bass 1989b; Futerman et al. 2006; Vijendravarma et al. 2008). Similarly, we know very little about the extent to which D. melanogaster is attacked by T. kingi or other species of Microsporidia in the field– a small-scale field study our group conducted in Southern England (Futerman et al. 2006) failed to find Microsporidia in the local wild species of Drosophila. It is thus difficult to judge the degree to which D. melanogaster may have experienced recent selection in the field due to Microsporidia, and how this might affect additive genetic variation for resistance and tolerance.

Although the fitness of infected flies from the selection lines was only slightly higher than equivalent control flies, even this modest increase seemed to involve significant negative correlated reductions in other aspects of fitness. As has been found in other studies of the cost of resistance (and indeed other studies of life-history trade-offs), these costs are most manifest when the fly is stressed. When reared under harsh conditions, but in the absence of the microsporidian, the early-life fecundity of adult flies was reduced by ∼38% and they were much poorer competitors in the larval stage. Both studies demonstrating genetic correlations between resistance to bacteria and fitness traits (McKean et al. 2008) and correlated responses to selection for resistance to parasitoids (Kraaijeveld and Godfray 1997; Fellowes et al. 1998) have found similar significant genotype-by-environment effects.

Resistance to parasitoids is associated with an increase in the numbers of circulating haemocytes, the cells that form the capsules that lead to the death of parasitoid eggs. Densities of haemocytes may be twice as great in selected lines and it is known that this is associated with reduced ingestion rates (Fellowes et al. 1999a; Kraaijeveld et al. 2000). We have hypothesized that the link is most likely due to redirection of resources from trophic to defensive functions, or perhaps might be caused by change in the transport properties of the haemocoel fluid (the insect blood) due to increased cell density. In contrast to defense against parasitoids, that against fungi and bacteria is chiefly humoral involving now very well-characterized signaling cascades that lead to the production of defensive compounds, in particular anti-microbial peptides (Hultmark 2003; Lemaitre and Hoffmann 2007). Whereas improved defense against parasitoids seems to require investment in new cells, something that is likely to be costly in terms of other host function, the response to bacteria and fungi may mostly involve modification of recognition proteins (Leulier et al. 2003), changes that are less likely to be intrinsically costly (although may still lead to trade-offs if improved recognition of one type of pathogen reduces the probability of others being detected).

The mechanisms invertebrates use to combat microsporidian infections are still relatively poorly known (Biron et al. 2005). However, recent studies indicate that microsporidian infection can trigger both cellular and humoral immune responses in insects (Kurtz et al. 2000; Hoch et al. 2004). The role of the cellular arm of the immune system is supported by our previous demonstration that haemocyte densities in D. melanogaster larvae increase after infection with T. kingi (Vijendravarma et al. 2008) and our findings here that the selection lines evolved constitutively higher haemocyte densities. Similar cellular immune response has been reported from other insect-microsporidia systems (Hoch et al. 2004) although precisely how haemocytes attack Microsporidia is unclear. Although phagocytosis of microsporidian spores by insect haemocytes has been observed, by the time spores are produced the infection is already well established (Laigo and Paschke 1966; Nassonova et al. 2001; Hoch et al. 2004). Similarly, although melanization of infected tissue involving the phenol-oxidase cascade has been recorded its efficacy is not certain (Hoch et al. 2004). Roxstrom-Lindquist et al. (2004) used microarrays to analyze the transcriptional response of D. melanogaster infected by the microsporidian, Octosporea muscaedomesticae, a species that attacks a number of Diptera families. They observed 59 genes that were upregulated after microsporidian but not bacterial infection, although most of the genes were of unknown function and included no antimicrobial peptides nor components of the signaling cascades that regulate them. Note that this study involved feeding microsporidian spores to adult flies, whereas infection of larvae by eating medium contaminated by infected adult cadavers appears to be the more normal transmission route (Vijendravarma et al. 2008). A proteomic study of Aedes mosquitoes attacked by the Microsporidian Vavraia culicis did find antimicrobial peptides to be upregulated, although could not exclude secondary infection by other micro-organisms as a possible cause (Biron et al. 2005).

This study on microsporidian pathogens of Drosophila adds to our understanding of the response of a single model organism to a spectrum of natural enemies. Although there have been genetic and evolutionary studies of many D. melanogaster natural enemies, there are still major gaps in our knowledge, for example the role of viruses in D. melanogaster ecology and, at the other end of the size spectrum, response to nematode macroparasites. The latter have been studied in other Drosophila species (Jaenike and Dombeck 1998; Jaenike 2000; Perlman and Jaenike 2001) and although we have found them quite commonly in the field attacking obscura group species in Southern England (A. R. Kraaijeveld, unpublished results) we have not been able to get them into culture and so use them in experiments. In general, we still know relatively little about the importance of natural enemies (with the partial exception of parasitoids) in the ecology of natural D. melanogaster populations, both in its natural African range and in its modern anthropophilic habitats. This makes it hard to assess the significance of the different levels of additive genetic variation in resistance for different parasites and pathogens. Our lack of ecological background is particularly true for the natural enemy we studied here which is only known from laboratory strains (although microsporidian infections of wild D. melanogaster that may be this species have been observed), and only persists when cultures are maintained in such a way that adults die and contaminate larval food. Studies of T. kingi in its natural habitat would be particularly helpful, and it would be interesting to know whether the relatively minor effect the pathogen have on Drosophila compared to its highly virulent attack on its parasitoid is adaptive or just an accident of physiology.

We finish by noting that if we had not observed the effect of the microsporidian on the parasitoid, we might have overlooked its presence in our cultures, despite its effects on host fitness. Where flies are cultured in a manner that allows microsporidian persistence then screening for this pathogen is highly advisable before conducting other experiments on Drosophila biology.

Associate Editor: T. Chapman


We are grateful to M. Hopkins, G. Needham, and S. Narasimha for technical assistance, to D. Quicke and F. van Veen for advice and discussion, and to B. Wertheim (UCL) for providing the sparkling poliert strain of D. melanogaster.