Detecting local adaptation in a natural plant–pathogen metapopulation: a laboratory vs. field transplant approach


Anna-Liisa Laine, Metapopulation Research Group, Department of Biological and Environmental Sciences, PO Box 65 (Viikinkaari 1), FI-00014 University of Helsinki, Finland.
Tel.: +358 9 191 57743; fax: +358 9 191 57694;


Antagonistic coevolution between hosts and parasites in spatially structured populations can result in local adaptation of parasites. Traditionally parasite local adaptation has been investigated in field transplant experiments or in the laboratory under a constant environment. Despite the conceptual importance of local adaptation in studies of (co)evolution, to date no study has provided a comparative analysis of these two methods. Here, using information on pathogen population dynamics, I tested local adaptation of the specialist phytopathogen, Podosphaera plantaginis, to its host, Plantago lanceolata at three different spatial scales: sympatric host population, sympatric host metapopulation and allopatric host metapopulations. The experiment was carried out as a field transplant experiment with greenhouse-reared host plants from these three different origins introduced into four pathogen populations. In contrast to results of an earlier study performed with these same host and parasite populations under laboratory conditions, I did not find any evidence for parasite local adaptation. For interactions governed by strain-specific resistance, field studies may not be sensitive enough to detect mean parasite population virulence. Given that parasite transmission potential may be mediated by the abiotic environment and genotype-by-environment interactions, I suggest that relevant environmental variation should be incorporated into laboratory studies of parasite local adaptation.


Parasite local adaptation is taken as evidence of ongoing coevolution between hosts and parasites. Conceptually, local adaptation is indicated by a higher mean fitness of parasites on local vs. foreign hosts or by a higher mean fitness of local parasites than foreign parasites on local hosts (Gandon et al., 1998). Given the importance of parasites as a driving force of evolution by natural selection, and the threats that parasite evolution imposes on human health and agriculture, local adaptation has been repeatedly investigated for numerous host–parasite interactions. Traditionally, parasites are expected to be ahead of their hosts in the coevolutionary race due to their larger population sizes, shorter generation times and higher mutation rates (Hamilton et al., 1990; Kaltz & Shykoff, 1998; Gandon & Michalakis, 2002). Indeed, numerous studies have reported evidence for higher parasite fitness in sympatric combinations with its hosts (Parker, 1985; Lively, 1989; Ebert, 1994; Lively & Dybdahl, 2000). However, a number of studies have not been able to detect parasite local adaptation (Davelos et al., 1996; Dufva, 1996; Strauss, 1997; Mutikainen et al., 2000), and some studies have even found parasites to be maladapted to their sympatric hosts (Parker, 1989; Kaltz et al., 1999; Oppliger et al., 1999).

Two processes in particular may have an effect on whether parasites become locally adapted or not: first, the rate and scale of migration may have profound implications for parasite local adaptation (Gandon & Michalakis, 2002). Sufficient gene flow will provide necessary material on which natural selection can act, yet too high levels of gene flow may swamp the effects of local dynamics (Gandon et al., 1996; Morgan et al., 2005). Also, parasites that disperse more than their hosts are expected to become locally adapted (Gandon et al., 1996; Thrall et al., 2002; Morgan et al., 2005). Second, parasites are not likely to adapt only to their biotic environment (i.e. hosts) but also to their abiotic habitat (Kawecki & Ebert, 2004). Recent work on thermal biology of invertebrate–parasite interactions has highlighted the potential of the abiotic environment to influence the evolutionary trajectories of parasite life history and infectiousness (Thomas & Blanford, 2003; Mitchell et al., 2005; Fels & Kaltz, 2006). It has been shown that the effectiveness of certain insect parasitoids and endosymbionts can vary across the range of realistic temperatures experienced in the field and that different parasite genotypes may respond very differently to the temperature gradients, as evidenced by strong genotype-by-environment interactions (G × E; Blanford et al., 2003).

Traditionally, studies of local adaptation have been pursued in transplant experiments conducted in the field or in laboratory studies under a constant environment (Kawecki & Ebert, 2004). By excluding all environmental noise in the laboratory it may be possible to detect genotype-by-genotype interactions indicating local adaptation that may be missed under variable field conditions. However, the constant environment of the laboratory also has its drawbacks. Experiments are typically conducted with what is considered to be a representative sample of the natural local parasite population. Given our limited knowledge on the structure of natural parasite populations, rarely can we reliably estimate how representative our samples really are. Also, maintaining multiple parasite lines in the laboratory may be so laborious that practical constraints limit the number of genotypes we have representing each local parasite population. Finally, a profound limitation of studies conducted under constant laboratory conditions is that parasite fitness may be under- or overestimated if infection success is governed by G × E effects (e.g. Thomas & Blanford, 2003; Mitchell et al., 2005; Fels & Kaltz, 2006).

To date, we lack studies that provide comparisons of field transplant and laboratory experiments in detecting local adaptation. Given the central role of local adaptation in (co)evolutionary studies a comparison of these methods is called for. Here, I report findings of a field transplant experiment investigating local adaptation of Podosphaera plantaginis to its host Plantago lanceolata L. at three different spatial scales. The study design and the host and pathogen populations are the same as those used in a laboratory inoculation study that found Po. plantaginis populations to be locally adapted at the scale of their local host metapopulations (Laine, 2005). In the Laine (2005) study pathogen performance was assessed through inoculations performed in the laboratory. In this study, the host individuals were reciprocally transplanted into natural pathogen populations and their infection status was followed in the field. For the two experiments the host and pathogen populations were sampled in the same year to catch them at the same phase of the coevolutionary cycle. By comparing the results of these two experiments on the same host and pathogen populations conducted in field and under laboratory conditions, I wanted to test whether these two approaches differ in their ability to detect patterns of parasite local adaptation. Furthermore, I wanted to determine whether patterns of local adaptation detected in the field transplant experiment and in the laboratory cross-inoculation experiment vary between two key fitness components – virulence and aggressiveness. Throughout this paper, virulence refers to the ability of a given pathogen genotype to cause an infection, whereas aggressiveness refers to the severity of the symptoms.

Materials and methods

Study system

The host plant Pl. lanceolata (Plantaginaceae) is an herbaceous perennial plant whose inflorescences develop from the axils of the basal rosette leaves (Sagar & Harper, 1964). Plantago lanceolata is considered an obligate outcrosser, a trait that is maintained both by protogyny and by an S-RNase-driven self-incompatibility system (Ross, 1973). As the seeds of Pl. lanceolata ripen, they fall to the ground close to the mother plant (Bos, 1992; van Damme, 1992). Plantago lanceolata is capable of reproducing clonally via the production of side rosettes from the axillary meristems (Sagar & Harper, 1964).

Podosphaera plantaginis is an obligate powdery mildew fungus in the order Erysiphales within the Ascomycota (Yarwood, 1978) which in Finland appears to be restricted to Pl. lanceolata (A.-L. Laine, unpublished data). During the growing season, the pathogen is transmitted among hosts by clonally produced dispersal spores, conidia, that are carried by wind. In August, the sexually produced resting spores, cleistothecia, begin to appear. At the end of each growing season the pathogen goes through a major decline as most host individuals die back to root stock. The fungus completes its entire life cycle on the surface of the host plant where it is visible as localized (nonsystemic) white powdery lesions. The interaction between Po. plantaginis and Pl. lanceolata is characterized by strain-specific resistance as the same host individual is resistant to some strains of the pathogen and susceptible to others (Laine, 2004).

The study system in the Åland Islands comprises about 3000 local Pl. lanceolata populations. The host populations are typically small discrete entities that are naturally fragmented as they occur on dry meadows delimited by an unsuitable habitat (Hanski, 1999). The occurrence of Po. plantaginis in Pl. lanceolata populations in the Åland Islands has been systematically followed since 2001 (Laine & Hanski, 2006). Every year in early September, 30–40 students working in pairs survey the meadows for presence of the powdery mildew. At each infected site, students collect an infected leaf for subsequent microscopic identification. In early September, when the surveys are carried out the fungus is clearly visible on the host as white-greyish mycelial growth supporting black resting spores. Data from the regional surveys have shown Po. plantaginis to persist as a metapopulation in its host populations in Åland. Despite annual fluctuations, infection prevalence has remained low (∼5%) and there is considerable turnover of local pathogen populations (Laine & Hanski, 2006). Frequent colonization events and a spatially aggregated occurrence pattern of Po. plantaginis below the scale of 1 km suggest that there is substantial spore dispersal in this system among neighbouring host populations (Laine, 2005; Laine & Hanski, 2006).

Pathogen dynamics in four local metapopulations

Four metapopulations were chosen in 2002 to study the dynamics of Po. plantaginis in greater detail (Fig. 1). To objectively classify host populations into metapopulations, a hierarchic cluster analysis as implemented in the program SPOMSIM was used (Moilanen, 2004). Based on the degree of aggregation observed in pathogen occurrence patterns (see Laine, 2005), I chose four habitat patch networks with a radius of approximately 500 m, each of which consisted of five Pl. lanceolata populations. In 2002, the probability of finding an infected population at 500 m distance from a randomly chosen infected population is ∼4 times higher than the average infection prevalence in the Åland Islands in that year (Laine & Hanski, 2006). Hence, at this scale there is likely to be colonizations and spore dispersal among established populations. In each of the chosen metapopulations at least one of the host populations had to be infected in 2001. Six of the host population networks fulfilled these criteria, and four of them were chosen to represent different parts of the Åland Islands. Frequent colonizations and extinctions of local pathogen populations were observed in the four surveyed metapopulations, yet in at least one of these sites in each of the metapopulations one pathogen population persisted throughout the study period. In each of these four metapopulations, this population served as the site for field transplantations. Detailed results of the occurrence of Po. plantaginis in these host populations are reported in Laine (2005).

Figure 1.

 The distribution of Podosphaera plantaginis in Plantago lanceolata populations in the Åland Islands in 2001. Black symbols depict infected host populations and white symbols uninfected host populations. The circles show the location of the four local metapopulations in the Åland Islands from which host plants were sampled for the field experiments. In each metapopulation one population (IDs 114, 382, 2224 and 3270) served as the site of transplantation in summer of 2004.

Field transplant experiment

I tested local adaptation of four Po. plantaginis populations in a field transplant experiment at three different spatial scales: on hosts from the sympatric population, on hosts from the sympatric metapopulation and on hosts from three allopatric metapopulations. One host population that had been infected throughout 2001–2004 was chosen from each metapopulation as the site of transplantation (population IDs 114, 382, 2224 and 3270). In September 2003, seeds were collected from all 20 host populations in the four metapopulations. At each site, seeds from 14 haphazardly chosen individuals were collected into paper envelopes and stored at room temperature. On 19–20 March 2004, seeds were germinated by placing them in 0.8-L pots in a 30% vermiculate–70% potting soil mixture in greenhouse conditions of 16 h of light and at +22 °C. Of the maternal lines producing seedlings, 10 lines were chosen to represent each of the host populations in the experiment. Into each pathogen population, I transplanted 10 sympatric host plants, 10 plants from each of the four populations in the sympatric metapopulation and 10 plants from each of the three allopatric metapopulations. Hence, into each population, I transplanted altogether 80 host plants. The allopatric populations in the experiment were always those which also served as transplantation sites (i.e. population IDs 114, 382, 2224 and 3270). Altogether, the experiment consisted of 320 host individuals.

On 7–8 July 2004, the sites were visited to confirm that the mildew population had successfully over-wintered. On 9–10 July 2004, the plants were taken to the field. Transplanting individuals into soil at these sites would have been impossible because Pl. lanceolata populations in Åland occur on rocky outcrops with extremely shallow soils. Hence, the plants were kept in their pots for the duration of the experiment. In each population, 80 plastic cups (14 × 10.5 × 4.5 cm3) were placed in the field on the surface in the proximity of naturally infected Pl. lanceolata individuals. I avoided placing the cups on top of naturally growing host plants so as to interfere as little as possible with the proceeding of the local epidemic. In each population, the plants in their pots were then placed within these cups. The sites were visited every other day (first day populations 114 and 3270; second day populations 382 and 2224) and the plants were watered if necessary. To minimize the effect of spatial positioning (distance to infected individuals and position with respect to prevailing wind direction) on the infection probability and severity, the plants were randomized among the plastic cups at every visit. The plants were removed from the field on 13–14 August 2004.

Between 11 and 20 July, during visits to the sites at 2-day intervals, the following variables were recorded for each plant individual: an estimate of leaf area (the number of leaves, the length and width of the largest leaf), growth form (flat–medium–upright) and hairiness of the plant (yes–no). As it takes approximately 10 days for infection to become visible to the eye, mildew infection was assessed from 21 July onwards. At each visit, every plant was checked for mildew infection and the number of infected leaves and the number of all leaves were counted. The number of leaves a plant has may change throughout the season as some leaves wither and others may suffer herbivore damage.

Statistical analysis of the field transplant experiment

I tested infection success, measured as the number of infected individuals, as a function of sympatry–allopatry to determine whether, and at what scale, Po. plantaginis was adapted to its host plants in the field transplant experiment. The model was fitted by PROC GENMOD as implemented in SAS 9.1 (SAS Institute, 1999). Explanatory variables included pathogen population, host population and host origin type (1 = sympatric population, 2 = sympatric metapopulation and 3 = allopatric metapopulation) nested within the pathogen population. All these were categorical variables in the model. The response variable was whether or not the plant had become infected by the end of the experiment and hence the model assumed a binary distribution of errors and a logit link function.

I also tested local adaptation as aggressiveness of the pathogen populations with an anova as implemented by PROC GLM in SAS 9.1. The response variable was the proportion of infected leaves observed at the last visit in the field. These data were square root transformed to meet the model assumption of normally distributed errors. As categorical explanatory variables, I used pathogen population, host population and host origin type (1 = sympatric population, 2 = sympatric metapopulation, and 3 = allopatric metapopulation) nested within the pathogen population. The model of pathogen aggressiveness included only those plants that had become infected. The estimate of leaf area, plant growth form or hairiness of the plant did not significantly affect the probability of a plant becoming infected in the field experiment nor the proportion of infected leaves, and they were not included as explanatory variables in the final analyses.

Statistical comparison of field transplant and laboratory experiments

I compared whether the field transplant experiment and a laboratory cross-inoculation experiment differed in their ability to detect pathogen local adaptation. The study design and results of the laboratory study are reported in detail in Laine (2005). First, I looked at whether the estimates for virulence differed between the two experimental approaches. In the field, virulence was measured as the proportion of infected plants and in the laboratory as the proportion of host genotypes each pathogen strain was capable of infecting. It would also have been possible to estimate the proportion of infected plants for the laboratory experiment but nearly all of the host plants (97%) were susceptible to at least one of the pathogen strains, hence yielding a virulence estimate ‘1’. In the analysis, virulence was estimated across pathogen strains for each host population. The data were normally distributed and analysed with an anova as implemented by PROC GLM in SAS 9.1. As explanatory categorical variables in the model, I had pathogen population, host population and experiment type (1 = field transplant and 2 = laboratory inoculations) and host origin type (1 = sympatric population, 2 = sympatric metapopulation and 3 = allopatric metapopulation) nested within pathogen population.

I also compared these two different approaches in the estimates they provided for pathogen aggressiveness for the four pathogen populations. In the field transplant experiment, aggressiveness was estimated from the proportion of infected leaves on the individual (counted during the last visit in the field). Given that aggressiveness should reflect severity of infection for the host, this measure seems most appropriate. In the laboratory experiment, aggressiveness was estimated from pathogen developmental stage divided into four categories ranging from mycelial growth to heavy spore production [1 = sparse mycelium but no conidia, 1.5 = mycelium producing very few conidia and colonies visible only under a dissecting microscope, 2.5 = colonies visible with the naked eye but exhibiting sparse sporulation, 3 = profuse sporulation on colonies of moderate size (< 5 mm diameter) and 4 = profuse sporulation on large colonies (> 5 mm diameter): key adapted from Bevan et al., 1993a]. Although a proportion of leaf covered with mycelial growth may have been more comparable with the field-estimated aggressiveness, slight variation in spore density and distribution on the leaf during the brushing of spores would significantly affect a proportion of leaf covered by fungal growth, and hence would not appropriately reflect pathogen aggressiveness. In contrast, infection stages measured in the laboratory would have been impossible to estimate in the field because these estimates are highly dependent on the age of the infection, which would not have been possible to control for under field conditions. In the model, the response variable (proportion of infected leaves) averaged for each host population was normally distributed. Laboratory-estimated aggressiveness, averaged for each host population, was included as a covariate. The model was analysed by PROC GLM as implemented in SAS 9.1. In all models, nonsignificant interactions were excluded from the model in a backward stepwise manner.


Field-estimated local adaptation

Although the results indicate a trend for pathogen populations to infect a higher proportion of host plants originating from their sympatric site in the field than hosts from the sympatric metapopulation or allopatric metapopulations, the strength and even the direction of this trend varied among the pathogen populations and was not statistically significant (Fig. 2 black bars; Table 1). There were significant differences in plants from the different host populations in their probability of becoming infected in the four pathogen populations (Table 1), which suggests that there are population-level differences in host resistance/susceptibility. When the fitness of the pathogen population on hosts from different origins was measured as aggressiveness, no support for local adaptation emerged (Table 1). However, the four pathogen populations differed significantly in how efficiently they infected their hosts, regardless of the host origin type (Table 1). Also, severity of infection varied with host population identity (Table 1).

Figure 2.

 Virulence of four pathogen populations on hosts from their sympatric population, sympatric metapopulation and allopatric metapopulation. Black bars represent results of a field transplant experiment where virulence was measured as the proportion of individuals that were infected; grey bars represent the proportion of susceptible responses in a laboratory experiment conducted with these same populations (Laine, 2005). Error bars are based on standard errors of means.

Table 1.   Results of the generalized linear model and anova of the field-estimated virulence and aggressiveness of Podosphaera plantaginis.
χ 2 P F P
Pathogen population3,1674.220.2385.480.0013
Host population15,16729.190.0152.080.013

Field transplant experiment vs. laboratory inoculation experiment

The two methods differed significantly in their ability to detect pathogen local adaptation when pathogen fitness was measured as virulence (sympatry–allopatry × field vs. laboratory P = 0.006, Table 2; Fig. 2). For example, for pathogen population 114, the trend detected with these two methods is very similar but the effect size for the laboratory study is larger (Fig. 2). For pathogen population 382, the field experiment suggested pathogen maladaptation to its sympatric host population whereas the laboratory experiment suggests the pathogen population to be adapted at the scale of its sympatric population and metapopulation (Fig. 2). Host population identity had a direct effect on the estimated virulence but its effect also varied between the two methods as indicated by the significant ‘host population × field vs. laboratory’ interaction (P = 0.003, Table 2). The laboratory-estimated aggressiveness was not a good predictor of aggressiveness in the field. In fact, there was a slightly negative trend between field-estimated and laboratory-estimated pathogen aggressiveness (estimate = −0.054, F = 0.481,30, P = 0.494) so that the higher the aggressiveness in the laboratory, the lower its estimate in the field.

Table 2.   Results of an anova comparing the estimates of virulence in the field transplant experiment and laboratory experiment (for details on the laboratory experiment see Laine, 2005).
Source F P
Pathogen population3,541.910.14
Host population20,542.140.014
Field vs. laboratory1,540.500.484
Sympatry–allopatry × field vs. laboratory7,543.270.006
Host population × field vs. laboratory14,542.850.003


In contrast to the laboratory study (Laine, 2005), the field transplant experiment did not find evidence for local adaptation of Po. plantaginis to its sympatric host population or host metapopulation. This discrepancy for the same host and pathogen populations with these two different methods highlights the importance of the approach we take in the studies of parasite local adaptation.

Neither of the pathogen fitness measures showed Po. plantaginis to be locally adapted in the field experiment but the reasons for the discrepancy between the two traits may be quite different given their distinct genetic backgrounds. For plant–pathogen interactions characterized by strain-specific resistance, the specificity of the outcome is generally assumed to convey an underlying gene-for-gene (GFG) control (Thompson & Burdon, 1992). For a resistant reaction to occur (i.e. infection does not take place, as the host recognizes the presence of the pathogen), both the specific resistance gene and the avirulence gene in the pathogen must be present (Flor, 1955, 1956). In this context, virulence is defined as the ability of a pathogen to overcome a given host resistance gene. For compatible interactions, infection proceeds with pathogen aggressiveness, a measure of spore production and transmission potential. Aggressiveness is a trait governed both by the genotypic aggressiveness of the pathogen strain as well as polygenetically controlled strain nonspecific resistance of the host (Burdon, 1997). Furthermore, parasite transmission potential is often strongly mediated by G × E interactions (cf.Blanford et al., 2003; Thomas & Blanford, 2003; Fels & Kaltz, 2006). Hence, despite the limitations of field transplant experiments in detecting local adaptation for GFG systems discussed below, excluding all environmental variation in the laboratory may also hinder our ability to detect local adaptation.

In the Po. plantaginisPl. lanceolata interaction, virulence has been shown to be robust over realistic nutrient and temperature gradients (A.-L. Laine, unpublished data). However, aggressiveness is strongly influenced directly by variation in nutrient availability to the host and by variation in ambient incubation temperature, as well as indirectly through G × E interactions (A.-L. Laine, unpublished data). The role of the abiotic environment in governing pathogen aggressiveness will in part explain why the laboratory experiment, conducted under a constant environment, found local adaptation of Po. plantaginis to its sympatric host populations and metapopulations to be stronger when pathogen fitness was measured as virulence than it was for aggressiveness (Laine, 2005). There may be a variety of reasons for why aggressiveness measured in the field failed to produce evidence of local adaptation although the experiment was carried out under naturally varying environmental conditions. It is possible that aggressiveness is not a good measure of local adaptation for Po. plantagis, or that measuring it accurately in the field is very difficult.

The field and laboratory-estimated virulence of the pathogen populations differed but virulence was not consistently higher or lower in the field than in the laboratory. Under laboratory conditions, we can distinguish what proportion of host genotypes each pathogen genotype is capable of infecting. However, in the field all infections may be caused by a single highly virulent pathogen strain. As there is enough evidence to assume that local pathogen populations are polymorphic in their virulence (de Nooij & van Damme, 1988; Bevan et al., 1993b; Burdon & Roberts, 1995), this will be a poor estimate of mean population virulence. Local adaptation is assumed to be a trait at the population level (Gandon et al., 1998), hence the performance of single highly virulent pathogen strains is not sufficient evidence for or against local adaptation. Furthermore, for any individual not infected in the field, we cannot tell whether this is the result of a resistant response or because a spore never happened to land on that host during conditions that favoured germination.

Although much of the discrepancy in the results between field and laboratory approaches may be attributed to methodological issues discussed above (in the field genotype-by-genotype interactions are hard to tease apart and microscopic parasite life-history stages are difficult to follow), we may also be missing some fundamental components of local adaptation in laboratory studies. For example, both hosts and parasites may have phenological and physiological adaptations linked with their native habitat, that will also eventually affect the probability of infection (cf.Kawecki & Ebert, 2004). Furthermore, a considerable limitation of inoculation studies conducted in the laboratory is that they only measure the final infection outcome, i.e. whether or not the host has been successfully infected, while ignoring mechanisms of the transmission process, which in itself can be a key component of local adaptation (Ridenhour & Nuismer, 2007). This could be especially important for parasites with active searching behaviour and the need to cope with their hosts’ behavioural defence strategies. For many passively dispersed parasites, such as the conidial spores of Po. plantaginis, we may expect to find less adaptation to the transmission process, as successful transmission may be best ensured by the production of sufficient levels of transmission propagules. Indeed, the suitability of laboratory and field methods may very much depend on the characteristics of the host–parasite interaction, as some studies have reported a tight correspondence between infection rates for specific host genotypes in natural populations and in laboratory infection experiments (Dybdahl & Lively, 1998; Little & Ebert, 2000, 2001). However, for interactions governed by strain-specific resistance, I suggest that field transplant experiments are not sufficiently sensitive to detect the mean virulence or aggressiveness of local pathogen populations.


Based on the results of this study, it appears that for many host–parasite interactions field transplant experiments may fail to detect parasite local adaptation due to difficulties in detecting G × G interactions and accurately measuring parasite life-history traits. I recommend three steps we should take when designing laboratory/common garden approaches to study parasite local adaptation: (1) a rough estimate of the diversity of local host and parasite populations is required to know what constitutes a representative sample of the local populations; (2) we need to identify key environmental variables for the interactions and test for possible (G ×) G × E interactions; (3) the relevant environmental variation should be incorporated in the study design.

The process of parasite adaptation to sympatric hosts, and the resulting differences on parasite fitness on local vs. foreign hosts, will be important in both basic and applied biology. It is becoming clear that the spatial population structure and environmental heterogeneity can have profound implications for trajectories of coevolution (Thompson, 1999, 2005). Incorporating enough spatial reality into future studies of coevolution is a key future challenge if we want to provide reliable predictions of parasite virulence and aggressiveness evolution. Whereas theoretical studies and empirical work with laboratory model systems will be valuable in reaching this goal, natural host–parasite systems are irreplaceable for testing the generated predictions.


I want to thank Jason Hoeksema, Juha Merilä, Saskya van Nouhuys and two anonymous reviewers for valuable comments on the manuscript. Otso Ovaskainen, Saila Kuokkanen, Riitta Ovaska and Mikko Putkonen assisted in the field work. This study was funded by LUOVA Graduate School (Ministry of Education) and Academy of Finland (Grant no. 20286 and 213457 to I. Hanski, Finnish Centre of Excellence Programme 2003–2005 and 2006–2008).