- Top of page
- Materials and Methods
- LITERATURE CITED
- Supporting Information
Theory indicates that spatial scale and habitat configuration are fundamental for coevolutionary dynamics and how diversity is maintained in host–pathogen interactions. Yet, we lack empirical data to translate the theory to natural host–parasite systems. In this study, we conduct a multiscale cross-inoculation study using the specialist wild plant pathogen Podosphaera plantaginis on its host plant Plantago lanceolata. We apply the same sampling scheme to a region with highly fragmented (Åland) and continuous (Saaremaa) host populations. Although theory predicts higher parasite virulence in continuous regions, we did not detect differences in traits conferring virulence among the regions. Patterns of adaptation were highly scale dependent. We detected parasite maladaptation among regions, and among populations separated by intermediate distances (6.0–40.0 km) within the fragmented region. In contrast, parasite performance did not vary significantly according to host origin in the continuous landscape. For both regions, differentiation among populations was much larger for genetic variation than for phenotypic variation, indicating balancing selection maintaining phenotypic variation within populations. Our findings illustrate the critical role of spatial scale and habitat configuration in driving host–parasite coevolution. The absence of more aggressive strains in the continuous landscape, in contrast to theoretical predictions, has major implications for long-term decision making in conservation, agriculture, and public health.
The strength and outcome of coevolutionary interactions is highly variable across space and time, ranging from hotspots with rapid reciprocal coevolution to coldspots where the two species do not coevolve (Laine 2009; Thompson 2013). Given such variable outcomes of coevolutionary interactions, the original question of whether natural selection plays a key role in host–parasite dynamics has recently shifted toward the question of when—and under what circumstances—we are most likely to witness evolutionary responses (Hereford 2009; Tack and Roslin 2010; Thompson 2013). Although the outcome of coevolution is generally expected to depend on the balance between selection, drift, and gene flow (Slatkin 1987; Lenormand 2002), few studies have explored how the relative strength of these factors—and hence the outcome of natural selection—depends on the spatial scale of the study and the configuration of the habitat. For example, although the meta-analytical approach has pinpointed several characteristics of the study system that may affect the strength of local adaptation (e.g., generalist vs. specialist or sessile vs. mobile parasites; Lajeunesse and Forbes 2002; Greischar and Koskella 2007; Hoeksema and Forde 2008), such an approach often ignores the fact that patterns of local adaptation may vary within a single species or community (Laine 2005; Tack and Roslin 2010). As a consequence, there is a clear need for evolutionary studies replicating experiments within a single pathosystem across spatial scales and across landscapes that differ in the configuration of the habitat.
Although few researchers have replicated local adaptation studies across multiple spatial scales within a single host–parasite system (Hanks and Denno 1994; Mopper et al. 1995; Thrall et al. 2002; Laine 2005), a few general patterns emerge from the studies to date. In a pioneering study, Mopper et al. (1995) demonstrated that local adaptation of a lepidopteran leafminer occurred at scales ranging from individual oak trees to oak populations separated by 65 km. In contrast, Laine (2005) demonstrated the presence of local adaptation of the powdery mildew Podosphaera plantaginis to its host plant Plantago lanceolata at the scale of tens of kilometers, whereas the pathogen showed no consistent pattern of adaptation at scales ranging from a few hundred meters to several kilometres. As the pathogen frequently dispersed up to a kilometer, the author suggested that local adaptation of the pathogen was swamped by gene flow at this small spatial scale. Corroborating this result, a cross-species comparison showed that plants are least resistant to local plant parasites and are most resistant to parasites collected several tens to hundreds kilometres away (Laine et al. 2011). However, the lack of gene flow among widely separated populations may also prevent adaptation at large spatial scales: whereas local hen flea populations were maladapted to local great tit populations as compared to nonlocal great tit populations on the same island (3.8–28.5 km between populations; Lemoine et al. 2012), another study did not find evidence for either local adaptation or maladaptation of flea populations separated by about 300 km (Dufva 1996). Overall, these studies may indicate that local adaptation is most likely to occur at “intermediate” spatial scales— where the definition of “intermediate” will depend on the balance between gene flow, relative dispersal ability of host and parasite, and the strength of natural selection (Gandon et al. 1996; Gandon 2002; Gandon and Michalakis 2002).
At any particular spatial scale, the evolutionary outcome of host–parasite interactions may strongly depend on habitat configuration (i.e., the spatial distribution of the habitat). Indeed, several theoretical studies have demonstrated the impact of habitat configuration on rapid and directional trait evolution (Rand et al. 1995; Boots and Sasaki 1999, 2000; Haraguchi and Sasaki 2000; Keeling 2000; O'Keefe and Antonovics 2002; van Baalen 2002; Boots et al. 2004; Kamo et al. 2007; Wild et al. 2009; Lion and Boots 2010; Best et al. 2011), trait diversity (Carlsson-Granér and Thrall 2002; Gandon and Michalakis 2002; Thrall and Burdon 2002; Kamo et al. 2007; Best et al. 2011), and local adaptation (Gandon et al. 1996; Gandon 2002; Gandon and Michalakis 2002) in host–parasite interactions. Although each of these models assesses trait evolution in a spatial perspective, the assumptions and ways of incorporating space vary widely (Lion and Boots 2010; Webb et al. 2013). For example, several theoretical studies investigate the impact of local and global dispersal or transmission on parasite evolution within a spatially substructured population, which generally leads to the prediction that virulence will decrease with more localized dispersal or transmission (Boots and Sasaki 1999; Haraguchi and Sasaki 2000; Best et al. 2011). However, as the majority of these studies do not consider host evolution (but see Best et al. 2011), they may not be suitable for deriving predictions when reciprocal evolution drives host–parasite dynamics. Coevolutionary models generally focus on the evolution of qualitative gene-for-gene interactions in a metapopulation characterized by infrequent dispersal among populations, and emphasize the general aspect that trait diversity can be maintained within metapopulations (Gandon et al. 1996; Nuismer et al. 2000; Thrall and Burdon 2002; Laine and Tellier 2008; Brown and Tellier 2011). Notably, specific outcomes may be affected by model assumptions including parasite life-history (e.g., O'Keefe and Antonovics 2002) and the postulation of trade-offs (and their shape) between parasite life-history traits (Anderson and May 1982; Kamo et al. 2007; Webb et al. 2013).
Although the theoretical prediction that parasite virulence, aggressiveness and diversity may evolve in response to changes in habitat configuration and increasing human movements are highly relevant for public health, agriculture, and conservation (Galvani 2003), theory has largely outpaced empirical studies in this field of research. A potential reason is the lack of any clear linkage between host–parasite systems as envisaged in silico and as observed in nature. In particular, the discrepancy between model assumptions and the complexity of real parasite life histories makes it challenging to summarize the diverse model outcomes and make a priori predictions for any specific natural host–parasite system. Nonetheless, two microevolutionary selection experiments have successfully validated model predictions on parasite trait evolution. Boots and Mealor (2007) showed that a high viscosity of the landscape (with resulting low movement rates and increased local interactions of the larvae of the moth Plodia interpunctella) selected for lower infectivity of a species-specific granulosis virus (PiGV). Kerr et al. (2006) found that localized dispersal in a phage-bacterial system increased dominance of competitively restrained “prudent” phage morphs, whereas “rapacious” phage evolved under unrestricted migration.
Even fewer studies have investigated the impact of the spatial configuration of the habitat on local adaptation. In one example, Tack and Roslin (2010) demonstrated that leaf miners and gallers were locally adapted to individual oak trees when immigration from neighboring trees was relatively low, whereas the insect community was nonadapted or maladapted when immigrants formed a large fraction of the local population. A bacteria-phage experiment further demonstrated that the shape of spatial dispersal networks may play a role in driving host–parasite coevolution and patterns of local adaptation (Vogwill et al. 2010). These studies then indicate that the configuration of the habitat, which provides the blueprint for gene flow across the landscape, may play a key role in host–parasite coevolution and local adaptation.
Finally, few studies have compared patterns of genetic and phenotypic differentiation among populations. In principle, such a comparison may reveal the spatial scale and type of natural selection (Merilä and Crnokrak 2001; Jorgensen et al. 2006; Tack et al. 2012). For example, if the main part of phenotypic diversity occurs within populations, whereas populations are genetically differentiated, this may indicate the maintenance of phenotypic trait variation by balancing selection within populations. In contrast, divergent selection would result in large phenotypic differentiation among populations as compared to genetic differentiation among populations.
In this article, we investigate the impact of both spatial scale and habitat configuration on parasite local adaptation of the powdery mildew P. plantaginis to its host plant P. lanceolata. Local adaptation, measured as higher parasite fitness on sympatric versus allopatric plants is taken as evidence for on-going coevolution (for other measurements of local adaptation, see Kawecki and Ebert 2004). Specifically, we investigate patterns of local adaptation and trait variation across three spatial scales: (i) among populations situated less than 1.6 km apart; (ii) among populations spaced 6–40 km apart; and (iii) among two regions (Åland and Saaremaa) set about 200 km apart and separated by a large body of water (Fig. 1). We employed an identical sampling scheme in both regions by collecting hosts and parasites at the same distances. As the regions differ in terms of the spatial configuration of the host populations (with Åland characterized by fragmented host populations, and Saaremaa by large continuous host populations), this design allows us to simultaneously test for impacts of habitat configuration on patterns of mean levels of phenotypic traits, trait diversity, and local adaptation.
Figure 1. Map of sampling locations. The large upper panel shows a map of northern Europe with the location of the two island systems (Åland and Saaremaa) indicated by arrows. The lower panels reflect Åland (left) and Saaremaa (right), with the clusters shown by circles and populations by dots within these circles. For each region, a single cluster is shown in detail, with the distribution of the host indicated by a black outline and the sampling locations by filled black squares and circles (to indicate focal and nonfocal populations, respectively; note that the black squares and circles partly overlap with the host distribution outline in Åland).
Download figure to PowerPoint
In an attempt to bridge the gap between the theoretical literature and empirical studies, we put forward a selection of a priori hypotheses derived from modeling work but hardly tested in natural systems:
- The scale of local adaptation. Local adaptation is expected to depend on the balance between selection and gene flow (Slatkin 1987; Lenormand 2002). Hence, we expect to find the strongest local adaptation at an intermediate spatial scale, where local adaptation is not swamped by gene flow, but interactions and movement are frequent enough for natural selection to play a role. Given the large dispersal range of aerially dispersed plant pathogens (Brown and Hovmøller 2002), we expect that “intermediate” distances may range from about 10 km to several hundreds of kilometres.
- Effects of habitat configuration on:
- Mean trait values. Lower pathogen virulence will prevent overexploitation in small populations (“self-shading” or “kin shading”) without displacement by the more aggressive pathogen strain (Rand et al. 1995; Boots and Sasaki 1999, 2000; Haraguchi and Sasaki 2000; Keeling 2000; O'Keefe and Antonovics 2002; Kamo et al. 2007; Wild et al. 2009; Best et al. 2011). From the plant perspective, clustering of the resistant host and its offspring will increase the benefit of higher host resistance (Best et al. 2011). Basically, these arguments go back to Hamilton's (1964) classic conjecture that spatial structure is beneficial to cooperation, because cooperators can gain additional benefits from being clustered (see also Lion and van Baalen 2008). Hence, we expect that pathogen aggressiveness, which is commonly correlated with virulence, will be lower in the fragmented populations of Åland than in the continuous populations of Saaremaa.
- Trait diversity. Nonspatial host–parasite models predict that coevolutionary dynamics and cycles may result in the loss of phenotypic variation in a single mixed population (e.g., Leonard 1977). Subsequent models have shown that spatial structure may increase trait diversity for qualitative traits like gene-for-gene interactions during infection (Thrall and Burdon 2002; Laine and Tellier 2008; Brown and Tellier 2011). Hence, we expect higher trait diversity in the fragmented region than in the continuous region.
- Local adaptation. Previous studies detected parasite local adaptation in our study system (Laine 2005, 2008). We then expect that the high extinction rates of parasite populations and population bottlenecks in the fragmented region are likely to wipe out or weaken local parasite adaptation (Bergstrom et al. 1999; Mopper et al. 2000). Hence, we expect stronger local adaptation in the continuous than in the fragmented region.
- (3) Genetic versus phenotypic differentiation among populations. Coevolutionary models and (more sparse) empirical data predict that negative frequency-dependent selection will maintain phenotypic diversity at a small spatial scale. This pattern is expected due to adaptation of the parasite to the most common local host genotypes and vice versa (Haldane 1949; Chaboudez and Burdon 1995; Dybdahl and Lively 1998; Lively and Dybdahl 2000; Brown and Tellier 2011). As neutral genetic variation is unaffected by such balancing selection, we may then expect more population differentiation among presumptively neutral genetic marker loci than among phenotypic traits.
- Top of page
- Materials and Methods
- LITERATURE CITED
- Supporting Information
Few previous studies have investigated the impact of spatial scale and habitat configuration on coevolutionary dynamics in wild host–parasite systems. The most serious lack relates to studies transferring clear-cut predictions derived in silico to real systems in nature. In this study, we specifically tested a series of explicit hypotheses derived from theory in a single, well-described host–parasite interaction as occurring across variable landscapes. In this context, we made several essential findings.
First, we detected few differences among the two regions in terms of mean parasite trait levels or parasite trait diversity; instead, both regions proved remarkably similar, with most variation in mean trait levels occurring among individual pathogen and plant genotypes within populations. Second, we detected parasite maladaptation among regions, and among populations separated by intermediate distances (6.0–40.0 km) within the fragmented region. Third, in both regions we detected strong genetic differentiation among populations, whereas the majority of phenotypic variation was found within populations. We discuss these findings in further detail in the sections later.
THE IMPACT OF HABITAT CONFIGURATION ON TRAIT EVOLUTION
Evolutionary epidemiology predicts that host ecology, like spatial structure, may strongly impact on the evolution of parasite traits. Indeed, both theory (Rand et al. 1995; Boots and Sasaki 1999, 2000; Haraguchi and Sasaki 2000; Keeling 2000; van Baalen 2002; Kamo et al. 2007; Lion and van Baalen 2008; Lion and Boots 2010; Best et al. 2011) and microevolutionary selection experiments (Kerr et al. 2006; Boots and Mealor 2007) indicate that virulence, transmission, and trait diversity may evolve in response to habitat configuration. From an applied perspective, such predictions are crucial to understand the long-term consequences of decision-making in natural, agricultural, and human systems (Galvani 2003). For instance, rapid changes in habitat configuration of wild-life habitat (in many cases decreasing connectivity among populations) may select for decreased disease virulence in natural systems (Galvani 2003). The construction of corridors among isolated habitat fragments with the aim to increase population viability of endangered species may have the negative side-effect of increasing selection for virulence in associated diseases. Similarly, the increasing mobility of the human population may increase disease virulence, with major implications for human health (Boots and Sasaki 1999; van Baalen 2002; Galvani 2003). In contrast with these in silico and in vitro predictions, our finding of no or few differences in parasite traits among the two regions suggests that habitat plays a minor role in driving trait diversity in naturam. Alternatively, there may be other factors that counteract or dilute the impact of habitat configuration on trait evolution. If parasite mean traits were largely driven by multiple infections we would not expect any variation in virulence among the two regions, as in our case coinfections were equally common in both regions (Alizon et al. 2013). A major challenge for future investigations may lie in identifying the relative importance of multiple factors in determining parasite trait evolution (e.g., Table 1 in Galvani 2003).
Another notable difference between our findings and those of previous studies may lie in the fact that the P. lanceolata–P. plantaginis system is characterized by reciprocal evolutionary dynamics (Laine 2005, 2006; Ovaskainen and Laine 2006; Laine 2008). In contrast, the majority of theoretical explorations and micro- and mesocosm experiments have involved systems where the host did not evolve (Kerr et al. 2006; Lion and Boots 2010). As such, the outcome of coevolutionary interactions may strongly deviate from that expected when only one of the parties is evolving (but see Best et al. 2011).
In summary, we do not find the expected variation in parasite life-history traits among the continuous and fragmented region. Instead, pathogen strains with highly variable infectivity, phenology, and aggressiveness coexist within populations in both regions.
THE IMPACT OF SPATIAL SCALE AND HABITAT CONFIGURATION ON LOCAL ADAPTATION
The spatial scale of the study has a strong impact on the patterns of local adaptation detected. In the global data set, we mainly detected parasite maladaptation at the scale of the region, indicating a coevolutionary disadvantage of the parasite at this large spatial scale. In the fragmented region, we also detected a weak but consistent sign of parasite maladaptation to plants from the local cluster, as compared to plants from more distant populations within the same region. In contrast, parasite performance did not vary significantly according to host origin in the continuous landscape. As previous studies in this system have shown a mosaic pattern of local adaptation with a tendency for the parasite population to gain the upper hand (Laine 2005, 2008), the current observation of pathogen maladaptation at both an intermediate (in Åland) and large (among the two regions) spatial scale seems surprising. In hindsight, one may argue that it is hard to predict who adapts to whom given the myriad numbers of factors affecting host–parasite coevolutionary dynamics (Greischar and Koskella 2007; Hoeksema and Forde 2008). Such prediction is further complicated by the fact that several of these factors are notoriously difficult to measure empirically (e.g., relative dispersal ability), and there is no straightforward manner to weigh different factors against each other. Importantly, the difference in the perception of who adapts to whom between this and previous studies suggests that the sign of local adaptation may vary in time. Such rapid temporal changes may not be surprising: a recent time-shift experiment by Thrall et al. (2012) demonstrates rapid evolution of flax resistance in response to the local flax rust population. Similarly, previous studies in our pathosystem have indicated that parasite selection pressures can induce rapid and localized increases in plant resistance (Laine 2006; Ovaskainen and Laine 2006). Crucially, such parasite selection pressure on the host plant may show strong temporal fluctuations due to yearly variation in drought stress, which strongly exacerbates parasite selection pressure in this system (Laine 2004). This is in line with a recent study, which shows that environmental conditions may mediate host–parasite coevolution and patterns of local adaptation (Laine 2008). Overall, both spatial and temporal variability in patterns of local adaptation may (partly) explain why two recent reviews have failed to confirm general patterns in terms of the existence or strength of parasite or host local adaptation, or to identify any consistent driving factors determining who adapts to whom in host–parasite interactions (Greischar and Koskella 2007; Hoeksema and Forde 2008).
Given the a posteriori knowledge that the plant here seems ahead in the coevolutionary race, we can reason why there is parasite maladaptation in the fragmented region, and not in the continuous region. Although the plant habitat is, like the pathogen habitat, highly fragmented, there is only minor turnover of plant populations (Nieminen et al. 2004). Hence, the genetic variation and evolutionary potential of plant populations may be high. In contrast, the pathogen faces rapid turnover due to high population extinction rates (Laine and Hanski 2006), which may reduce the evolutionary potential of the pathogen.
In summary, our study establishes the important notion that the existence, sign, and strength of local adaptation may vary with spatial scale, across regions that differ in habitat configuration, and through time. Although complex, such patterns may be essential in explaining the maintenance of phenotypic variation, and fit well with the predictions of the geographic mosaic of coevolution (Thompson 2005; Gandon and Nuismer 2009).
GENETIC AND PHENOTYPIC DIFFERENTIATION
Although our analysis revealed little genetic differentiation among the two regions, a large fraction of the variation (roughly half) occurred among focal populations within the region. In striking contrast, the majority of the phenotypic variation was found within populations. Such divergence between genetic and phenotypic variation may be explained by balancing selection maintaining phenotypic variation within local populations, whereas limited dispersal and genetic drift result in population differentiation in terms of neutral markers. Our data then support the long-standing theoretical prediction that negative frequency-dependent selection is a major evolutionary force maintaining phenotypic variation within populations (Haldane 1949; Brown and Tellier 2011).
- Top of page
- Materials and Methods
- LITERATURE CITED
- Supporting Information
Micro- and mesocosm experiments have a great tradition in revealing the ecology and evolution of species interactions (Gause 1934; Huffaker 1958; Bohannan and Lenski 2000; Jessup et al. 2004). Such approaches have thus far revealed many fascinating links between spatial structure and parasite evolution (e.g., Kerr et al. 2006; Boots and Mealor 2007). Nonetheless, although such experiments can test theory, reveal biological mechanisms and direct future research, the linkage between micro- and mesocosms and natural communities remains problematic, and this split has recently been reemphasized as one of the major challenges in ecology (Sutherland et al. 2013). Here we took the opposite approach of addressing big questions in the full complexity of a natural system. Naturally, such an approach comes with another set of limitations, the most severe of which relates to the number of replicates achievable. Indeed, although micro- and mesocosms can readily be replicated at the scale of an imaginary metapopulation, such replication is logistically more challenging (or even unfeasible) in natural host–parasite systems. In this study, although we used an optimized design to limit the number of inoculations necessary in the laboratory, we were still limited to comparing a single continuous region with a single fragmented region (cf. Burdon et al. 1999; Carlsson-Granér and Thrall 2002). Nevertheless, we argue that the present type of bold ventures into the natural complexity of real systems may offer the sole solution to ultimately linking theory, small-scale experiments and natural coevolutionary dynamics as playing out in the wild.
In summary, our study highlights the importance of spatial scale and habitat configuration in understanding host–parasite coevolution. Contrary to expectation, we detected a remarkable lack of trait differentiation and diversity among the two regions differing in host configuration, suggesting that factors other than habitat configuration may drive these patterns. Between the two regions we detected local adaptation, and we observed differentiation among the two regions in the strength of local adaptation. Together, these patterns suggest that both spatial scale and habitat configuration may play a key role in understanding coevolutionary outcomes, thereby giving rise to a geographic mosaic of coevolution (Thompson 2005).