Introduction

All organisms face attack by pathogens, parasites and predators. Understanding how adaptive changes in resistance and virulence influence the interactions between victim and natural enemy is a major challenge for population biologists. Critical to this task is an understanding of the proximate mechanisms through which resistance and virulence operate. While vertebrate immunity and resistance have been studied intensively for many years, the equivalent field of invertebrate immunity has lagged behind. Nevertheless, some important advances have been achieved in recent years, in part due to the ever-increasing importance of Drosophila melanogaster as a model system. The most important natural enemies of Drosophila in field populations are a suite of parasitoid wasps that attack the larvae or pupae. Here we review recent studies of the evolutionary biology of Drosophila–parasitoid interactions, and look forward to the integration of ecological, genetic and physiological approaches to the coevolution of resistance and virulence in this system.

Female D. melanogaster lay eggs on many fermenting substrates, and after hatching the larvae feed on yeasts and bacteria. It is to these substrates that their parasitoids are attracted. Most parasitoids are small wasps whose larvae develop on (ectoparasitoids) or in (endoparasitoids) other insects, and less frequently other arthropods (reviewed in Godfray, 1994). Successful parasitoid attack results in the death of the host. The parasitic Hymenoptera have two broad life-history strategies, defined by whether the host ceases development (idiobiont parasitoids) or continues development (koinobiont parasitoids) after attack (Godfray, 1994). Among the most common parasitoids of European D. melanogaster are koinobiont larval endoparasitoids of the genera Asobara (Braconidae: Alysiinae) and Leptopilina (Cynipoidea: Eucoilidae), but they are also attacked during the pupal stage by idiobiont ectoparasitoids such as the pteromalid, Pachycrepoideus vindemiae (Boulétreau, 1986; Carton et al., 1986). Levels of larval parasitoid attack suffered by D. melanogaster in natural populations vary both spatially and temporally. In southern England, attack rates average 6–8%, but can reach maxima of more than 50%, while in some Swiss populations up to 100% of larvae can be attacked on occasion (Carton et al., 1986). Not all of the larvae that are attacked will die, as some larvae can mount a successful immune response to the parasitoid egg.

The physiological battleground

Drosophila melanogaster larvae that successfully defend themselves against attack by koinobiont endoparasitoids employ a cellular immune response known as encapsulation (Salt, 1970). This immune reaction is thought to act independently of the humoral response mounted against invading pathogens (Coustau et al., 1996; Nicolas et al., 1996). Blood cells (haemocytes) circulating in the haemolymph are central to the immune response to parasitoid attack and are produced by stem cells in the insect haematopeic organ, the larval lymph glands (Shrestha & Gateff, 1982).

Cellular encapsulation in Drosophila has three main stages. First, the parasitoid egg is recognized as nonself. Just how this occurs is poorly characterized at present, but it is likely that receptors on the haemocyte surface are involved (Carton & Nappi, 1997; Asgari et al., 1998). There is some evidence that lectin (carbohydrate binding proteins or glycoproteins) recognition plays a role as D. melanogaster differing in ability to encapsulate the parasitoid Leptopilina heterotoma also differ in their ability to bind to lectins (Nappi & Silvers, 1984). This nonself recognition is followed by changes in the haemocyte cell surface membranes, with previously hidden molecules becoming exposed on the cell surface. Non-self recognition presumably triggers a signalling pathway and it has been suggested that the phenoloxidase cascade may be involved, although experimental evidence for this is weak (Nappi et al., 1991). In the second stage, there is a short-term increase in the number of circulating haemocytes produced by the lymph glands. Additionally, a class of specialized haemocytes termed lamellocytes differentiate from the plasmatocytes, the most abundant class of haemocytes in the larval haemolymph (Rizki & Rizki, 1992). This proliferation is thought to involve ligand binding by the Toll receptor on the haemocyte cell surface, triggering an internal signalling pathway involving cytoplasmic proteins, such as Cactus (Qui et al., 1998). The activated lamellocytes move to the parasitoid egg, flatten, and then adhere to the egg and each other, forming a multilayered capsule (Strand & Pech, 1995). This change in adhesiveness is thought to occur through the presentation of integrins (trans-membrane proteins) on the lamellocyte surface: these cell adhesion proteins bind to extracellular matrix molecules such as laminins, collagen IV and vitronectin, and can cause the cells to adhere together (Strand & Pech, 1995). Scavenger receptors, cell surface proteins involved in pattern-recognition pathways, may also play a role in cell–cell adhesion. During encapsulation, cell surface changes in the lamellocytes result in the exposure of phosphatidylserine (PS) on the outer cellular membrane (Asgari et al., 1997). The dSR-CI (Drosophila scavenger receptor class C) has several adhesive domains in its extracellular region, and since other scavenger receptors recognize the PS ligand, it has been suggested that dSR-CI may be involved in cell–cell binding (Pearson, 1996). The third stage involves a further class of haemocytes, the crystal cells. Lysis of these cells releases prophenoloxidase, an inactive zymogen of phenoloxidase, which is activated by protealytic cleavage at its amino terminus (Jiang et al., 1998). The phenoloxidase catalyses a cascade of reactions, resulting in the melanization of the capsule surface (Strand & Pech, 1995). The resulting rigid melanized capsule leads to the death of the parasitoid, either directly by asphyxiation (Salt, 1970) or through the release of cytotoxic compounds such as superoxide anions (Nappi et al., 1995) or hydroxy radicals (Nappi & Vass, 1998) by capsule components.

The number of haemocytes circulating in the larval haemocoel prior to parasitoid attack appears to be important in determining the encapsulation ability of Drosophila larvae. Comparative data show that Drosophila species that are better at encapsulating the braconid parasitoid Asobara tabida have a higher haemocyte count per unit volume of haemolymph than those that are less successful at resisting attack (Eslin & Prévost, 1996, 1998). In particular, (Eslin & Prévost 1998; (Table 1) showed that amongst six members of the melanogaster subgroup there was a strong and significant correlation between resistance and haemocyte counts. A potential criticism of their conclusion is that species are treated as independent data points and no account was taken of possible phylogenetic correlations. However, using the phylogeny of this group constructed by Jeffs et al. (1994) based on Adh gene sequences (Fig. 1), we re-analysed the data using independent contrasts (Harvey & Pagel, 1991), and confirmed Eslin & Prévost’s results (F1,4=10.6, r2=0.73, P=0.03; Fig. 2).

Table 1 Data used in the comparative analyses (from 1Eslin & Prévost, 1998; 2Berrigan & Hoang, 1999; 3Orr & Irving, 1997). The data on adult metabolic rate and body mass are mean values from both sexes
Full size table
Figure 1
figure 1

The phylogeny of Drosophila species used in the comparative analyses (after Jeffs et al., 1994).

Full size image
Figure 2
figure 2

The relationship between independent contrasts of total haemocyte count (mean of attacked larvae in Eslin & Prévost, 1998) and encapsulation ability. Data were transformed [log10(+1)] before contrasts were calculated and the regression is constrained to run through the origin.

Full size image

Across Europe, D. melanogaster shows substantial geographical variation in its ability to resist attack by two of its most important parasitoids, A. tabida and Leptopilina boulardi (Kraaijeveld & van Alphen, 1995; reviewed by Kraaijeveld & Godfray, 1999). Resistance to A. tabida is strongly correlated with latitude, with southerly populations encapsulating a greater proportion of parasitoids than larvae from populations further to the north. Drosophila melanogaster populations in central and southern Europe are most adept at defending themselves against L. boulardi, but the pattern of geographical variation here is less clear. Parasitoid virulence also varies geographically: A. tabida strains originating from the south are less likely to be encapsulated than those from northern European populations (Kraaijeveld & van Alphen, 1994). In all cases resistance and virulence were estimated by challenging flies or wasps from different localities with a single tester strain of parasitoid or host. Such a procedure would be misleading if there was coevolution between sympatric fly and wasp strains, but the available evidence suggests resistance and virulence are graded traits (like running speed or shell thickness) and that increased likelihood of success against one strain imparts an advantage against all (Kraaijeveld & Godfray, 1999).

Our understanding of the physiological basis of parasitoid virulence is still fragmentary although it is already evident that different parasitoid groups utilize a wide variety of stratagems to counter the host’s immune response. Amongst the main Drosophila parasitoids, Asobara tabida appears to adopt a passive means of avoiding encapsulation. The chorion of the egg is fibrous and causes it to adhere to host tissue such as the host’s fat body. As the host larva moves, the eggs become further embedded, hiding the egg from the host’s immune response (Kraaijeveld & van Alphen, 1994; Eslin et al., 1996). In contrast, L. boulardi and L. heterotoma have a more active means of defence. It has long been known that a substance injected by the female parasitoid at oviposition (lamellolysin) hinders the functioning of the lamellocytes. This has been shown to be a proteinaceous ‘virus-like particle’ (VLP) which is released from the egg chorion after oviposition, and interferes with the lamellocyte’s microtubule cytoskeleton, preventing capsule formation (Rizki & Rizki, 1990a; Dupas et al., 1996). Other non-Drosophila parasitoids inject viruses (rather than VLPs) containing DNA (polydnaviruses: poly disperse DNA viruses) into the host, the viral gene products attacking the host immune system in several different ways (Edson et al., 1981). The viral genome is integrated mainly within that of the parasitoid, although extra-chromosomal segments are also found (Shelby & Webb, 1999). Virions are assembled within the female calyx cells and infect host haemocytes and fat body cells (Shelby & Webb, 1999). The cysteine-rich proteins encoded by these viruses are first glycoslyated, and then bind to haemocytes, where one of their roles is to change the behaviour of the haemocyte surface lectin-binding sites, preventing haemocyte adhesion (Theopold & Schmidt, 1997). Reducing the dosage of polydnavirus reduces the titre of these proteins by a commensurate amount, and this in turn reduces their efficacy as a counter-defence against the host encapsulation response (Webb & Cui, 1998). This may explain why repeated oviposition by L. boulardi results in an increase in successful host immune response (Vass & Nappi, 1998a).

Genetics

Studies of the genetic basis of resistance in Drosophila (and, though less studied, of virulence in its parasitoids) have had two main goals. The first is to identify genes whose function is essential for successful resistance. Screening of sensitive mutants and the identification of the genes responsible offers a valuable tool for probing the physiological basis of resistance. The second aim is to understand the ecological genetics of adaptation, to determine the extent and nature of the additive genetic variation for resistance in natural populations of flies, and the genetic basis of differences between populations with varying abilities to encapsulate parasitoids. These two goals overlap to a certain degree.

A large number of loci that directly or indirectly influence encapsulation have now been catalogued (see Flybase at http://flybase.harvard.edu). Some of the more important include Black cells (tyrosinase gene associated with phenoloxidase structure; Rizki & Rizki, 1990b); lozenge (a transcription factor important in crystal cell formation; Rizki & Rizki, 1981; Rizki et al., 1985); serpent (required for haemocyte development; Rehorn et al., 1996); foraging (a protein kinase that determines larval foraging behaviour, which in turn influences parasitoid attack; Osborne et al., 1997); hopscotch (a Janus kinase required for haemocyte proliferation; Mathey-Prevot & Perrimon, 1998) and Ribosomal protein S6 (regulatory role in haematopoieisis; Watson et al., 1996). Mutations at many of these genes are associated with the occurrence of melanotic pseudotumours. These can be benign or malignant (Wright, 1987), and occur when capsule-like cell masses form around the tissues or organs of larval flies. Pseudotumours are presumably associated with a breakdown of self/nonself recognition mechanisms or the presence of aberrant cells. Some Black cells mutants (located at 55A1–55A4 on the D. melanogaster cytological map) are unable to melanize the haemolytic capsules formed around L. boulardi eggs, whereas other mutant alleles result in the formation of melanotic masses in the crystal cells (Rizki & Rizki, 1990b). The wizard mutation has a more direct influence on haematopoieisis, as the melanized phenotype is expressed in the lymph glands, crystal cells and larval head (Rodriguez et al., 1996). While mutations at many of these loci produce similar melanotic pseudotumours or capsules, the genes themselves have a multitude of roles. For example, the Toll/Cactus signal transduction pathway regulates larval haemocyte proliferation, and mutations at these sites result in an overabundance of melanized haemocytes (Qiu et al., 1998), whereas larvae bearing a mutant form of domino have melanized lymph glands, and as a result do not produce haemocytes at all (Braun et al., 1998). Such mutants are proving central to understanding the molecular and developmental genetics of resistance (Mathey-Prevot & Perrimon, 1998; the first goal identified above), but tell us rather little about the nature of genetic variation for resistance in natural populations (the genetics of adaptation; the second goal noted above).

An important technique for investigating the genes involved in encapsulation is the study of isofemale lines. Isofemale lines are formed by taking females that have been inseminated once, allowing them to lay eggs, and then inbreeding the descendants until most loci are homozygous. The genetic variation revealed by a comparison of isofemale lines derived from a single population is an amalgam of additive, dominance and epistatic components, and hence does not provide direct information about the ability of the population to respond to increased parasitoid selection pressures. Carton & Nappi (1997) reported that the difference between two isofemale lines differing in susceptibility to L. boulardi was due to a single locus, with the resistant allele completely dominant to the susceptible one. Chromosome substitution experiments allowed the locus to be located at the 55D–55F interval of the cytological map, on the long arm of the second chromosome. Further examination of these lines confirmed this result (Hita et al., 1999), and the authors suggest that this locus is involved specifically in L. boulardi recognition. Intriguingly, this is close to the location of the gene overgrown haematopeic organs 55DE. Mutants at this site produce diffuse lymph glands, which suggests that this gene may influence haemocyte production (Torok et al., 1993).

A similar experiment was performed by Benassi et al. (1998), who investigated the genetic basis of D. melanogaster resistance to A. tabida. They also found that differences in resistance amongst a set of isofemale lines appeared to be governed by a single locus, with the resistant allele showing complete dominance over the susceptible allele. This gene does not appear to be the same as that implicated in L. boulardi resistance (Carton & Nappi, 1997). The isofemale lines used to determine the genetic nature of resistance to L. boulardi were also assayed for their ability to encapsulate A. tabida. Both lines resistant and susceptible to L. boulardi were very successful at encapsulating A. tabida, which shows that the susceptible strains are not immune-deficient in general (Vass et al., 1993).

Orr & Irving (1997) investigated the genetic basis for geographical differences in D. melanogaster resistance to A. tabida. Using classical Drosophila genetic techniques, they showed that the majority of the variation amongst three populations of flies could be attributed to alleles segregating at a single locus, with the resistance allele completely dominant over the susceptible allele. The results of crossing experiments using larval phenotypic markers allowed them to locate the resistance gene near the centromere region of the second chromosome. Orr & Irving (1997) propose the dopa decarboxylase gene cluster as a strong candidate for the locus they have identified. The 18 loci in this gene cluster have important roles in the catalysis of the decarboxylation of dopa to dopamine, necessary for cuticle sclerotization, and mutations in 11 of these loci result in the production of melanotic tumours (Wright, 1996). Additionally, mutations at this site have been shown to compromise D. melanogaster encapsulation ability (Nappi et al., 1992).

Given that we are interested in understanding how changes in host resistance may be an adaptive response to parasitoid attack, we not only need to be aware of the genetics underpinning the trait, but also whether there is enough heritable variation in resistance within host populations for natural selection to act upon. There is now ample evidence for quite substantial additive genetic variation in resistance in D. melanogaster populations. Most of this evidence comes from experiments in which flies have been artificially selected for increased resistance (see Kraaijeveld et al., 1998 for a review of the older literature). Hughes & Sokolowski (1996) set up population cages of D. melanogaster with or without A. tabida (the fly populations were kept isolated within cages, but parasitoids were supplemented from other sources). They found that higher resistance evolved in cages where the wasp was present and where mean rates of parasitism averaged 7% over 19 generations. Larvae from cages containing parasitoids were almost twice (39% vs. 23%) as likely to encapsulate successfully a parasitoid egg as those from the control lines (see also below). Very strong artificial selection for increased resistance can be obtained by parasitizing a large number of larvae and breeding only from those individuals that have successfully encapsulated a parasitoid egg. Using this technique, Kraaijeveld & Godfray (1997) obtained an increase in survival against A. tabida from 5% to 50% in five generations, while Fellowes et al. (1998a) obtained an increase in survival against L. boulardi from 0.4% to 45% over the same period. In each case the response to selection was similar across four replicate lines. Narrow sense heritability for resistance was calculated using a model for threshold traits (which must be treated with some caution as it assumes a quantitative genetic background to the trait; Falconer & Mackay, 1996), giving an estimate of about 0.25 for resistance against either parasitoid.

As was discussed above, isofemale line studies have found genes that are specifically involved in resistance against A. tabida and L. boulardi. Studies of geographical variation in resistance, and of artificially selected lines, also have a bearing upon this question. First, no correlation was found between resistance to A. tabida and resistance to L. boulardi in a comparison of a large number of geographical separated populations of D. melanogaster (Kraaijeveld & van Alphen, 1995; Kraaijeveld & Godfray, 1999). Secondly, lines selected for resistance against A. tabida do not show increased resistance to L. boulardi, whereas lines selected for resistance against L. boulardi show a large increase in the likelihood of their successfully encapsulating A. tabida (Fellowes et al., 1999a). Both sets of lines show elevated resistance to a third parasitoid species, Leptopilina heterotoma. Asobara tabida and L. heterotoma attack a wide range of Drosophila species in fermenting substrates, whereas L. boulardi is restricted to D. melanogaster and its sibling species. Fellowes et al. (1999a) argued that their results could be explained by a two-component resistance model: a generalized increase in encapsulation ability, perhaps associated with elevated haemocyte densities, that increases protection against many parasitoid species, and a specific response against the countermeasures of the specialized parasitoid L. boulardi. Unfortunately, the genes responsible for the changes in resistance levels in the artificial selection lines have not yet been identified, and it is not possible to say whether they are the same as those identified in the isofemale line studies, or in Orr & Irving’s study of geographical variation.

Life-history trade-offs

The extensive additive genetic variation in resistance to parasitoids found in D. melanogaster is most likely to be maintained by a combination of fluctuating selection pressures, both temporal and spatial, and costs of resistance. The latter are central to most arguments about evolutionary interactions between hosts and parasites, or predators and prey, and insect parasitoids have proved a valuable model system for studying these issues.

There are two main issues in assessing the costs and benefits of investing in resistance. The first concerns what happens after a parasitoid attack: are there costs to the successful encapsulation of the parasitoid (or, equivalently, what are the benefits of resistance)? The second concerns the costs of maintaining the resistance machinery, costs that are borne irrespective of whether the host is attacked or not. By analogy, the two types of costs can be thought of as the penalties of fighting a war (actual defence) and of maintaining a standing army (resistance ability).

Actual defence

Since successful parasitoid attack will result in death, resistance will always be in the interest of the host, even if survival is associated with a reduction in fitness in comparison with an individual that had escaped attack. However, investment in resistance mechanisms will be less likely to occur if flies that survive have low reproductive success. Female flies that have survived parasitoid attack as larvae are smaller (both thorax and wing length) and less fecund (Carton & David, 1983; Fellowes et al., 1999b). Males also have smaller thoraxes, but there is no significant difference in wing length, and males experience only a small reduction in the number of offspring they sire, and then only under certain conditions (Fellowes et al., 1999b). Males clearly suffer less from resisting attack than females. This sexual dimorphism in the costs of parasitism has been shown to occur in many vertebrate taxa (Zuk, 1990; Møller et al., 1998), but we are aware of only one other study that has shown differences in an insect. Agnew et al. (1999) found that microsporidian infection of the mosquito Culex pipiens is more costly to females than males. Additionally, the pupae of larvae that have encapsulated a parasitoid have relatively thinner puparial walls (Fellowes et al., 1998b), and this may explain an unexpected indirect cost of encapsulating a parasitoid. The pupal parasitoid, P. vindemiae, will preferentially attack hosts that have previously survived attack by a larval parasitoid, possibly because of the reduction in puparium thickness, which will reduce the handling time required for successful attack (Fellowes et al., 1998b). Given that encapsulation is energetically costly and successful defence leaves fewer resources to be allocated to other physiologically demanding traits, it is unsurprising that resistance ability is reduced in D. melanogaster larvae exposed to environmental stress. Levels of successful defence falls when larvae are reared at high densities (Wajnberg et al., 1985), are exposed to insecticides (Delpuech et al., 1996), and are provided with a diet deficient in yeast (Vass & Nappi, 1998b).

Resistance ability

The costs of the ability to resist parasitoid attack are more subtle than the costs of actual defence. Lines selected for increased resistance to A. tabida (Kraaijeveld & Godfray, 1997) or L. boulardi (Fellowes et al., 1998a) show a decrease in larval competitive ability under conditions of resource stress. In both cases the loss of competitive ability is correlated with a reduction in the feeding rate of the more resistant larvae (Fellowes et al., 1999c).

When artificial selection was relaxed, the level of resistance in the lines selected against A. tabida quickly dropped from the maximum of 65% survival to approximately 30% (Kraaijeveld et al., 1998). Recall that in the last section we described how Hughes & Sokolowski (1996) found that D. melanogaster lines cultured with an A. tabida population that achieved a 7% attack rate had higher resistance levels than those cultured without parasitoids. In these experiments (designed to study the rover-sitter polymorphism), the nonparasitoid treatments were cultured at higher population densities. Also, the wild populations from which the base stock was isolated suffered higher parasitoid attack rates (16%) than the parasitoid treatments. These factors suggest an alternative explanation for their results, not that higher resistance evolved in the parasitoid treatments, but that costly resistance decayed faster in the nonparasitoid treatments.

Why should increased resistance be associated with a reduction in feeding rate, with a consequent reduction in competitive ability at high densities? Conceivably, the two responses could be pleiotropic effects of the same gene or genes, perhaps via increased investment in defensive functions, resulting in decreased investment in trophic functions such as feeding rate. Another possibility is that lower metabolic rates are selected for in the presence of high risks of parasitism. It has often been observed that Drosophila and other insects selected for tolerance to a variety of stresses evolve a lower metabolic rate (though the response of an individual to increased stress is normally an elevated metabolic rate; Hoffmann & Parsons, 1991). Possibly decreased metabolic rate and a consequent reduction in feeding rates is part of a suite of adaptations to prepare the fly for the stress of parasitoid attack.

We do not yet have data on the metabolic rates of flies from lines selected for increased resistance, but the idea can be tested by a cross-species comparison of encapsulation and metabolic rates. As discussed above, (Eslin & Prévost 1998; Table 1) reported the capacity of six members of the melanogaster subgroup to encapsulate A. tabida, and to this list we have added D. pseudoobscura which does not encapsulate this parasitoid species (Kraaijeveld & van Alphen, 1993; Orr & Irving, 1997). For metabolic rates, we used the data in Berrigan & Hoang (1999) which is based on respiration rates (Table 1). We tested the correlation using independent contrasts, again using the phylogeny produced by (Jeffs et al. 1994; Fig. 1). The data allowed six contrasts to be made, and these were sufficient to show that encapsulation is negatively associated with adult metabolic rate, after controlling for mass (partial correlation beta=− 0.79, t4=5.01, P < 0.01; Fig. 3). Clearly this is no more than an exploratory analysis (for example it assumes metabolic rates are correlated between the adult and larva) but it does suggest that the relationship between metabolic rate and resistance is worth further investigation.

Figure 3
figure 3

The relationship between independent contrasts of metabolic rate (here not corrected for size) and encapsulation ability. Data were transformed [log10 (+1)] before contrasts were calculated and the regression is constrained to run through the origin.

Full size image

Ecological consequences of resistance

Many hosts and parasitoids exist in tightly coupled ecological interactions, the parasitoid being the major regulatory factor affecting host densities (Hassell, 2000). Parasitoids thus act as a major selective force on the host, and at least theoretically there is the potential for significant feedback between population and genetic dynamics (Godfray & Hassell, 1991; Sasaki & Godfray, 1999; Godfray, 2000). For example, one joint dynamic–genetic model has two main outcomes: under certain conditions the host invests nothing in resistance mechanisms, in effect gambling the costs of resistance against the risk of being discovered by a parasitoid. Such an outcome might explain the inability of D. pseudoobscura to defend itself against any of its parasitoids. Under other conditions an arms race occurs with spiralling levels of costly resistance and virulence, until a point is reached when effective resistance is so costly that the host abandons defence, again gambling on not been found. At this point the parasitoid is selected to reduce its level of virulence, and the cycle begins again (Sasaki & Godfray, 1999). This result can explain persistent levels of additive genetic variation for resistance. Whether such models describe host–parasitoid interactions in nature is unclear, especially for D. melanogaster and its parasitoids where, perhaps surprisingly, the population dynamics of the host–parasitoid interaction have been little studied in the field.

Rapid evolutionary change in levels of resistance, on what perhaps would traditionally have been thought of as ecological timescales, may also influence the community structure of interacting species (Fellowes & Kraaijeveld, 1998; Thompson, 1998). For example, a D. melanogaster population subject to heavy parasitism by the specialist L. boulardi might evolve enhanced levels of resistance that would also protect it against A. tabida. Asobara tabida might then not be able to invade, or might as a consequence of the presence of L. boulardi be subject to selection for increased virulence. The asymmetric patterns of cross-resistance discussed above means that a D. melanogaster population that had evolved to protect itself against A. tabida would not be pre-adapted to withstand L. boulardi attack. Such indirect selection pressures may be common features of coevolutionary interactions (Miller & Travis, 1996), and in many ways can be considered analogous to apparent competition in ecological interactions (Holt, 1977). Effects such as this ‘apparent coevolution’, in addition to more direct evolutionary and coevolutionary changes, may be significant factors in determining the structure of insect-parasitoid and other resource-consumer communities (Fellowes & Kraaijeveld, 1998).

Conclusions

Parasitoids have provided valuable model systems for investigating a variety of problems in evolutionary biology, especially testing theories of adaptive sex ratios and other reproductive strategies (Godfray, 1994). Between 1940 and 1960, parasitoids were also used as a major genetic model system, especially the species Bracon hebetor and Nasonia (then called Mormoniella) vitripennis. Very few of the numerous mutant lines isolated during this period survive today. Drosophila and its parasitoids have been used extensively in behavioural and behavioural ecological work, and in studies of host–parasitoid physiology, but have been used rarely as ecological genetic models. This, however, is changing, largely spurred by the ever-increasing value of D. melanogaster as one of the animal models. Drosophila parasitoids have many advantages as model organisms; in particular, most species can be cultured as easily as their hosts, and their generation times are short, although about 50% longer than their host. A disadvantage is that the genetics of Drosophila parasitoids are as yet poorly understood, with very few visual or molecular markers available.

A challenge for biologists interested in the reciprocal evolutionary interactions between hosts and parasites and predators and prey is to understand the nature of the adaptive genetic changes, at all levels from the ecological to the molecular. We suggest that Drosophila and its parasitoids are likely to prove very valuable for disentangling these issues.