SEARCH

SEARCH BY CITATION

Keywords:

  • Aphis fabae;
  • genotype-by-genotype interactions;
  • host–parasitoid coevolution;
  • infectivity;
  • Lysiphlebus fabarum;
  • parthenogenesis;
  • resistance

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information

Antagonistic coevolution between hosts and parasites can result in negative frequency-dependent selection and may thus be an important mechanism maintaining genetic variation in populations. Negative frequency-dependence emerges readily if interactions between hosts and parasites are genotype-specific such that no host genotype is most resistant to all parasite genotypes, and no parasite genotype is most infective on all hosts. Although there is increasing evidence for genotype specificity in interactions between hosts and pathogens or microparasites, the picture is less clear for insect host–parasitoid interactions. Here, we addressed this question in the black bean aphid (Aphis fabae) and its most important parasitoid Lysiphlebus fabarum. Because both antagonists are capable of parthenogenetic reproduction, this system allows for powerful tests of genotype × genotype interactions. Our test consisted of exposing multiple host clones to different parthenogenetic lines of parasitoids in all combinations, and this experiment was repeated with animals from four different sites. All aphids were free of endosymbiotic bacteria known to increase resistance to parasitoids. We observed ample genetic variation for host resistance and parasitoid infectivity, but there was no significant host clone × parasitoid line interaction, and this result was consistent across the four sites. Thus, there is no evidence for genotype specificity in the interaction between A. fabae and L. fabarum, suggesting that the observed variation is based on rather general mechanisms of defence and attack.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information

Most organisms suffer from parasites, even many that are parasites themselves (Price, 1980). The ubiquity of parasitism and the fact that hosts and parasites are engaged in a coevolutionary arms race of adaptation and counter-adaptation led to the hypothesis that host–parasite interactions are an important mechanism maintaining genetic variation in populations (Judson, 1995), and may even select for sexual reproduction and recombination (Jaenike, 1978; Hamilton, 1980). This assumption hinges on a property that might be inherent in such interactions, namely that reciprocal selection between hosts and parasites is negative frequency-dependent, providing rare genotypes with a selective advantage that prevents their loss from the population. Negative frequency-dependence may arise either under high costs of resistance and infectivity, such that investment in defence is only favoured in a largely undefended population, for example, or if the interaction between hosts and parasites exhibits high genetic specificity (Hamilton et al., 1990; Frank, 1996; Sasaki, 2000; Agrawal & Lively, 2002). If this specificity is such that no individual host genotype is most resistant to all parasite genotypes and no parasite genotype is most infective on all host genotypes, negative frequency-dependence emerges very readily (Frank, 1994; Parker, 1994). The recognition of the evolutionary importance of genotype specificity has led to an increasing number of studies testing this assumption, many of which reported significant host genotype × parasite genotype (G × G) interactions (e.g. Webster & Woolhouse, 1998; Carius et al., 2001; Schulenburg & Ewbank, 2004; Lambrechts et al., 2005; Salvaudon et al., 2007). This suggests that the potential for negative frequency-dependent selection is realized in many host–parasite systems.

It would appear that reciprocal selection is particularly intense in insect host–parasitoid interactions, because their outcome is always fatal for one of the antagonists. Parasitoids are either killed by their hosts’ defences or they kill their hosts if these defences fail. The term virulence, typically defined as the reduction in host fitness resulting from infection by a parasite (Read, 1994), is therefore of limited use for parasitoids. If a parasitoid is able to overcome host defences, the host will invariably be killed and variation in virulence is restricted to more subtle differences such as time until killing or residual fecundity before death. Therefore, we will use the term infectivity for the ability of an insect parasitoid to parasitize its host. This is not completely satisfactory, as this term is normally used in the context of infectious diseases, but for lack of another term, we will also apply it to parasitoids. Although there is abundant evidence of genetic variation for resistance in hosts as well as infectivity in parasitoids (e.g. Henter, 1995; Henter & Via, 1995; Kraaijeveld & Godfray, 1999; Ferrari et al., 2001; Kraaijeveld et al., 2001; von Burg et al., 2008), surprisingly little is known about the degree of specificity in insect host–parasitoid interactions. Maybe best addressed is this issue in Drosophila melanogaster and its hymenopteran parasitoids Asobara tabida and Leptopilina boulardi. There is evidence for G × G interactions based on major gene effects for D. melanogaster and Le. boulardi, yet the interaction is such that universal infectivity is possible (Dupas et al., 2003). Selection in D. melanogaster for resistance to a specific strain of As. tabida resulted in higher resistance to other strains, too (Kraaijeveld & Godfray, 1999), and this increased resistance even extended to other species of parasitoids (Fellowes et al., 1999). Kraaijeveld & Godfray (1999) interpreted this as evidence for resistance and infectivity being quantitative traits that lack genetic specificity, although they acknowledged that the relevant experiments have yet to be done.

We have recently established the black bean aphid, Aphis fabae, and its most important parasitoid, Lysiphlebus fabarum, as a laboratory-based study system that is ideally suited to address the issue of genotype specificity in insect host–parasitoid interactions. Aphis fabae is a cyclical parthenogen and can be maintained clonally for any period of time under suitable conditions. Lysiphlebus fabarum, exceptionally among aphid parasitoids, also reproduces by parthenogenesis in most populations (Mackauer & Starý, 1967; Starý, 1988; Belshaw & Quicke, 2003). It is therefore possible to work with genetically uniform lines of host and parasitoid, which allows for powerful tests of genotype specificity in their interaction. Interestingly, aphids may harbour facultative endosymbiotic bacteria that are vertically transmitted and provide protection against parasitoids (Oliver et al., 2003). When Vorburger et al. (2009) exposed multiple clones of A. fabae with and without the defensive symbiont Hamiltonella defensa to two parthenogenetic lines of L. fabarum, they detected strongly increased resistance in clones harbouring H. defensa and a significant aphid clone × parasitoid line interaction on the proportion of aphids parasitized. However, this interaction was not observed when clones harbouring H. defensa were excluded from the analysis, suggesting that it may be because of specific interactions between symbiont and parasitoid genotypes, and that the direct interaction between aphids and parasitoids is characterized by a lack of genotype specificity (Vorburger et al., 2009). Yet with only two parasitoid lines tested, this result was far from conclusive. Here we present a more targeted and powerful test of G × G interactions between A. fabae and L. fabarum, yet we arrive at the same conclusion. If the aphids do not harbour any defensive endosymbionts, there is no evidence for genotype specificity in the interaction between A. fabae and L. fabarum.

Materials and methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information

Animals

All aphids and parasitoids were collected in June and July 2006 in the course of a Europe-wide sampling effort. The experiment reported here included animals from four different geographic origins, namely (i) the vicinity of Rennes in Britanny, France, (ii) the vicinity of Budweis in South Bohemia, Czech Republic, (iii) the Lower Rhine Valley in north-eastern Switzerland and (iv) the Lower Valais in south-western Switzerland.

In A. fabae, four different subspecies are described (Heie, 1986; Raymond et al., 2001). They use the same primary hosts (mainly Euonymus europaeus and Viburnum opulus), where the sexual females mate and lay overwintering eggs, but they differ in the range of secondary host plants used by the parthenogenetic, viviparous summer generations. Here we focus exclusively on the nominal subspecies A. f. fabae, which mainly uses broad bean (Vicia faba) and several Chenopodiaceae such as sugar beet (Beta vulgaris) or Chenopodium album as secondary hosts. Aphids were collected by clipping infested leaves or shoots of suitable secondary host plants, from which a single parthenogenetic female was used to establish a clonal line in the laboratory. We maintained clones on caged seedlings of broad bean (V. faba, Var. ‘Scirocco’) at 20 °C and a 16-h photoperiod. Under these conditions, A. fabae reproduces by continuous apomictic parthenogenesis. We genotyped all clones at eight microsatellite loci (Coeur d’Acier et al., 2004), and we screened them for the presence of facultative symbiotic bacteria as described by Vorburger et al. (2009). From the clones testing negative for facultative symbionts, we selected six from Rennes and five from each of the other three sites, all with different multilocus microsatellite genotypes. These genotypes and detailed collection information for all test clones are available in Table S1.

Parasitized colonies of aphids are easily recognized by the presence of ‘mummies’, i.e. dead aphids containing the parasitoid’s pupa. We sampled parasitoids by collecting colonies with mummies of known host species of L. fabarum into air-permeable containers, where we allowed the adult wasps to emerge. Single parthenogenetic females were then isolated from each sample and allowed to attack colonies of A. fabae on broad bean to found parthenogenetic isofemale lines. These lines are since maintained in the laboratory as mass cultures on a highly susceptible clone of A. fabae that was not included in the experiment. All females founding an isofemale line were genotyped at 11 microsatellite loci (Fauvergue et al., 2005; Sandrock et al., 2007). Unlike aphids, in which parthenogenetic reproduction only includes mitotic cell divisions (apomixis), parthenogenetic females of L. fabarum undergo a modified meiosis in which diploidy is restored by central fusion automixis (Belshaw & Quicke, 2003), which is why parthenogenetic isofemale lines should not be termed clones. Despite that these lines can be regarded as genetically uniform, because central fusion automixis rapidly leads to homozygosity distal to chiasmata and leaves nonrecombining areas of the genome unaffected. This was evidenced by the fact that when re-genotyped before the experiment in spring 2007, the microsatellite genotypes of parthenogenetic isofemale lines were still identical to those of the founding individuals in June/July 2006. We included three lines from Rennes and two lines from each of the other three sites in the experiment. Their microsatellite genotypes and collection details are provided in Table S2.

Experiment

Our general assay to estimate host susceptibility and parasitoid infectivity, respectively, was to expose known numbers of aphids to wasps for a fixed period of time, and determine the proportion of individuals successfully parasitized (Henter & Via, 1995). If this is done with several host clones and parasitoid lines in a fully crossed factorial design, a G × G interaction is detectable as a statistical interaction between host clone and parasitoid line on the proportion of aphids parasitized. The advantages of this approach are that we can determine the outcome of host–parasitoid encounters under realistic conditions in a reasonably complex environment, and that the simplicity of the assay allows for good replication. The disadvantage is that the approach is essentially blind to mechanism and cannot distinguish between pre- and post-ovipositional defences of hosts. For example, aphids may show behavioural avoidance of parasitoids (Foster et al., 2007), which is unlikely to be genotype-specific. However, Henter & Via (1995) have shown that a resistant and a susceptible clone of the pea aphid did not differ in the number of parasitoid ovipositions they suffered, and Vorburger et al. (in press) found that in A. fabae, up to three quarters of individuals on which parasitoid attacks have been observed may survive. This suggests that physiological defences of aphids can be quite effective. We therefore assume that the variation observed in our experiment will largely (but not exclusively) reflect the interaction between host and parasitoid after oviposition.

Because this experiment was concerned with the potential presence of G × G interactions as a prerequisite for negative frequency-dependent selection and not with local adaptation, we only exposed host and parasitoids from the same site to each other, i.e. we only worked with host/parasitoid combinations that could have occurred in the field. Thus we had a six host clones × three parasitoid lines infection matrix for Rennes and a five aphid clone × two parasitoid lines matrix for each of the other three sites, resulting in a total of 48 different combinations of host and parasitoid genotypes. Each combination was replicated ten times. We used more aphid clones than parasitoid lines in this cross-infection experiment because space constraints in the laboratory prevented us from keeping a larger number of parasitoid lines as mass cages.

At the start of the experiment, aphid stock cultures were split into the required number of colonies by placing two adult females on seedlings of broad bean grown in 0.07-L-plastic pots and covered with a small cage. These colonies were then distributed to random positions in ten plastic trays such that each tray contained one replicate of all host clone/parasitoid line combinations (randomized complete blocks). To avoid any inflation of among-clone differences by environmental maternal or grand-maternal effects carried over from the stock cultures, the replicated aphid colonies were maintained for two generations before exposure to parasitoids in the third generation. The test generation was started by placing seven second-generation adults from each aphid colony on new seedlings, where they reproduced for 24 h before being discarded. When their offsprings were 48–72-h old, we counted them (mean colony size = 51.9 ± 17.1 SD) and added two female parasitoids of the required lines to each cage. We removed the wasps again after 6 h and replaced the cage with a cellophane bag. Although the two wasps might interfere in the same cage and even superparasitize each other, preliminary trials showed that using two rather than a single wasp reduces the variation in mummification rates among replicates with the same combination of genotypes, possibly because this limits the influence of wasps that are unmotivated to sting. Nine days post-exposure to wasps, successfully parasitized aphids were recognizable as mummies and counted. To keep the daily work doable, we had to temporally stagger the experiment such that two complete blocks were handled per day over five consecutive days.

Statistical analyses

All analyses were carried out with the open source statistical software R 2.7.1 (R Development Core Team, 2008). Substantial overdispersion prevented us from analysing our proportion data as a success-failure vector using a generalized linear model with binomial errors. Instead, we arcsine square root transformed the proportions of aphids exposed to wasps that were mummified by parasitoids and analysed them with a linear mixed model, using the lmer procedure of lme4, a contributed library to R. We tested for the effects of site (fixed), block (random), host clone (random), parasitoid line (random) and the host clone × parasitoid line interaction (random). Site was treated as a fixed effect because with four levels only, the corresponding variance component would be estimated poorly. As the number of aphid nymphs exposed to parasitoids varied somewhat among replicates, we also included colony size as a covariate in the analysis. Tests of fixed effects were carried out with the pvals.fnc function of the languageR library. The function employs Markov Chain Monte Carlo sampling to obtain the highest posterior density (HPD) intervals and associated P-values for fixed effect parameters (Baayen, 2008). This offers a modern alternative to conventional significance tests of fixed effects in mixed models based on t or F statistics, which remain a contentious issue for ongoing disagreement about appropriate degrees of freedom (Baayen et al., 2008). To obtain an overall test of a fixed effect with more than two levels (i.e. site), the aovlmer.fnc function was used. The pvals.fnc function also provides 95% HPD intervals for random effects. We report these intervals with the estimates of variance components, but because they are constrained to never contain zero, the intervals cannot be used to infer statistical significance (Baayen, 2008; Baayen et al., 2008). Random effects were therefore tested by comparing models with and without the effects using likelihood ratio tests, i.e. by comparing the increase in scaled deviance resulting from removal of the term to a chi-square-distribution with d.f. = 1.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information

Colony size, i.e. the number of aphid nymphs exposed to parasitoids, did not significantly affect the proportion of individuals that were mummified, but there was a significant block effect (Table 1). Although the mean rates of successful parasitism did not differ significantly among the four sites included in the experiment, the variation in susceptibility among aphid clones within sites was large and highly significant (Table 1). This was most obvious in aphids from the Rennes area, where there was a more than five-fold difference in susceptibility between the least and the most resistant clone (Fig. 1a). We also found significant variation in infectivity among parasitoid lines (Table 1), but this variation appeared to be general rather than specific to the host clone they attacked. More infective lines tended to mummify a higher proportion of individuals in most aphid clones than less infective lines or, from the host’s perspective, more resistant aphid clones generally had a lower proportion of individuals mummified by all parasitoids (Fig. 1). Accordingly, there are only few crossing lines in the interaction plots (Fig. 1), and the host clone × parasitoid line interaction is not significant in the analysis (Table 1). We also ran separate analyses for each of the four sites and found this interaction to be nonsignificant in all cases (all > 0.1), suggesting that the lack of evidence for G × G interactions is consistent across four widely separated sites.

Table 1.   Results of linear mixed effects models for the proportion of aphids mummified by parasitoids. Proportions were arcsine square root transformed before analysis. P-values of random effects are based on likelihood ratio tests, P-values of fixed effects on the HPD intervals obtained from Markov Chain Monte Carlo (MCMC) sampling as implemented in the languageR library of R (Baayen, 2008).
SourceVariance components for random effects/coefficient for covariate (95% HPD)LR inline image P
  1. HPD, highest posterior density.

Colony size−0.0009 (−0.0024, 0.0006) 0.215
Block0.0025 (0.0004, 0.0095)8.8930.003
Site 0.237
Host clone (site)0.0195 (0.0065, 0.0253)104.070< 0.001
Parasitoid line (site)0.0071 (0.0015, 0.0272)23.408< 0.001
Host clone × parasitoid line (site)0.0013 (0.0000, 0.0034)0.5850.445
Residual0.0518 (0.0462, 0.0604)  
image

Figure 1.  Interaction plots depicting the susceptibility of clones of Aphis fabae from four geographic origins to syntopic lines of the parthenogenetic parasitoid Lysiphlebus fabarum. Aphid clones are ordered by increasing mean susceptibility. Each point represents the mean of ten replicate assays of the same combination.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information

In this study, we used the black bean aphid and parthenogenetic lines of its parasitoid L. fabarum to test if aphid host–parasitoid interactions exhibit genotype-specific interactions. We found no evidence for genotype specificity, but we cannot rule out that genotype specificity would be detected if additional host clones and parasitoid lines were tested. Yet considering that we tested 48 different combinations of host and parasitoid genotypes with ten replicates per combination, we feel confident in concluding that if present at all, such interactions have a weak influence on the outcome of host–parasitoid encounters compared to the substantial genetic variation for general resistance and infectivity we observed. However, this conclusion is limited to the ‘hard’ outcome of infection, i.e. whether a parasitoid is able to establish in the host and kill it. This is undoubtedly the trait with the most direct effect on fitness. But we cannot exclude that there may be genotype-by-genotype interactions on fitness-relevant traits of surviving parasitoids that were not studied here, such as the time until mummification, the rate of emergence from mummies or the body size of emerging parasitoids. In fact, there is some evidence for genotype-by-genotype interactions on parasitoid body size (Vorburger et al., in press). Similarly, we cannot rule out that there may be genotype-by-genotype interactions on traits of surviving aphids, e.g. fecundity. We only know that aphid fecundity is generally reduced upon resisting a parasitoid attack (Vorburger et al., 2008).

A lack of specificity in the defences of aphids that are not protected by endosymbionts was also suggested by the results of a previous study on the same system (Vorburger et al., 2009), as well as by a study on green peach aphids by von Burg et al. (2008), who exposed many aphid clones to two different species of parasitoids and detected no significant host clone × parasitoid species interaction. A caveat is in order here, however. Because these studies used similar assays as in the present experiment, it is also possible that part of the variation was because of behavioural mechanisms that are unlikely to be specific, such as differences in avoidance behaviour among host clones or variation in the motivation to sting among parasitoids.

Although based on somewhat different lines of evidence, Kraaijeveld & Godfray (1999) arrived at the similar conclusions for Drosophila: variation for resistance is general rather than specific to certain strains or species of parasitoids (but see Dupas et al., 2003). Thus it seems that in contrast to interactions between hosts and pathogens or microparasites, in which specificity is frequently observed (see Introduction), insect host–parasitoid interactions may be characterized by a low degree of genotype specificity. The similar conclusions from flies and aphids are also interesting because the mechanisms of defence against parasitoids are almost certainly different in the two systems. Drosophila resists parasitoids by encapsulation, a well-understood defence mechanism that is widespread in insects (Strand, 2008), but typically not observed in aphids (Henter & Via, 1995; Kraaijeveld et al., 2002). Very recent work by Oliver et al. (2009) has shown that aphid resistance conferred by the bacterial endosymbiont H. defensa is due to phage-encoded toxins, but the mechanistic basis of genetic variation in the aphids’ own resistance is still largely unknown.

The large amounts of genetic variation for resistance and infectivity we detected indicate ample scope for directional selection, yet it would be premature to conclude from the lack of G × G interactions that reciprocal selection between host and parasitoid cannot be frequency-dependent in the A. fabae/L. fabarum system. Models show that negative frequency-dependence may also emerge in host–parasite systems that lack strong specificity as long as increased resistance or infectivity come at a cost (Sasaki & Godfray, 1999; Sasaki, 2000; Agrawal & Lively, 2002). Nothing is known yet about costs of infectivity in aphid parasitoids, but there are studies looking for costs of resistance in aphids, and they provide only limited support for such costs. Gwynn et al. (2005) found more resistant clones of the pea aphid to be less fecund on average, but several other studies using larger numbers of aphid clones did not observe this correlation (Ferrari et al., 2001; von Burg et al., 2008; Vorburger et al., 2009). In the Drosophila/Asobara system, on the other hand, selection experiments provided evidence for evolutionary costs of resistance as well as costs of infectivity (Kraaijeveld & Godfray, 1997; Kraaijeveld et al., 2001).

Another factor to consider in aphids is defensive endosymbionts like H. defensa. Here we only used clones without H. defensa, but this bacterium infects a fraction of individuals in many species of aphids. The percentage of infected individuals may vary widely, with estimates ranging from 3% to almost 80% (Darby et al., 2001, 2003; Sandström et al., 2001; Haynes et al., 2003; Leonardo & Muiru, 2003; Russell et al., 2003; Ferrari et al., 2004; Vorburger et al., 2009), although populations uninfected with H. defensa have also been reported (Tsuchida et al., 2002; von Burg et al., 2008; Wille & Hartman, 2009). In the present study species, A. fabae, about one fourth of the individuals seem to be infected on average (Vorburger et al., 2009). Although this estimate is still based on rather limited sampling, it certainly indicates that L. fabarum is typically confronted with a mixture of hosts with and without H. defensa in the field. We have recently found preliminary evidence that H. defensa may increase not only the overall level but also the specificity of aphid resistance to parasitoids, possibly through symbiont × parasitoid genotype interactions (Vorburger et al., 2009; R. Rouchet & C. Vorburger, unpublished). As H. defensa is vertically transmitted, this would modify how reciprocal selection between hosts and parasitoids acts and therefore affect their coevolution. This interesting possibility deserves further research, for which it is important to know that the direct interaction between host and parasitoid genotypes in our study system is characterized by very limited specificity, as is shown here.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information

We thank B. Chaubet, J.-C. Simon and P. Starý for help in the field and J. Ferrari for endosymbiont screening. The manuscript benefitted from comments by J. Shykoff and two anonymous reviewers. Our work was supported by the Swiss National Science Foundation (grants 3100A0-109266 and PP00P3-123376 to C.V.).

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information

Table S1 Collection information and genotypes at eight microsatellite loci (Coeur d’Acier et  al., 2004) for the 21 clones of Aphis fabae used in this study.

Table S2 Collection information and genotypes at 11 microsatellite loci (Fauvergue et  al., 2005; Sandrock et  al., 2007) for the nine thelytokous lines of Lysiphlebus fabarum used in this study.

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer-reviewed and may be re-organized for online delivery, but are not copy-edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

FilenameFormatSizeDescription
JEB_1925_sm_TS1-TS2.pdf29KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.