1. The evolution of host range
The mechanisms that drive variation in specificity are elements of broader processes driving the evolution of specialization and the maintenance of biological diversity (Falconer, 1952; Levins, 1968; Price, 1980; Jaenike, 1990). In the absence of evolutionary constraints, specialization acts only to reduce the number of potential species with which both hosts and symbionts can successfully interact; a seeming disadvantage to the pathogen. However, specialization is a common and highly successful strategy (Fig. 1), suggesting that generalist strategies must be selected against under some ecological conditions. From the pathogen perspective, biotic heterogeneity, competition arising from multiple infection, disruptive selection and genetic trade-offs in adaptation to different hosts are all likely to be important in driving the emergence and maintenance of specialized lineages (Levins, 1968; Jaenike, 1990; Kawecki, 1998; Woolhouse et al., 2001) (Table 1).
Table 1. Ecological, evolutionary and mechanistic factors that have the potential to influence the evolution of host range
|Ecological||Host monocultures, high population/community relatedness||Host community complexity, low population/community relatedness|
|High host density, large populations, wide range||Ephemeral, rare or over-dispersed hosts|
|High transmission||Low transmission|
|Multiple infection, competition||Little direct competition for host resources|
|Genetic||Negative correlations between performances on different hosts||Lack of constraint, positive correlation between performances on different hosts|
|Coevolution||Strong effect of environment on performance|
Biotic heterogeneity (i.e. variation in community composition, diversity and spatial structure) is probably a key ecological driver of pathogen host range evolution (Thrall et al., 2007). As we have discussed above, epidemiological models and experimental work have demonstrated that host plant diversity plays an important role in the dynamics of infectious disease (e.g. the importance of pathogen spillover), but the consequences of such biotic complexity for the evolution of specialization within plant host–pathogen interactions remain largely unknown (Gandon, 2004; Thrall et al., 2007). Jaenike (1990) suggests that wide host ranges may be expected to evolve in pathogens when suitable hosts are rare, highly dispersed, or ephemeral. By contrast, specialists should be favoured when host diversity is restricted and hosts are abundant. General support for these predictions can be found in analysis of data from fish–metazoan and mammal–flea host–parasite interaction networks, which suggest that increased host abundance correlates directly with the proportion of specialist parasites found infecting that host (Vazquez et al., 2005). More recently, Thrall et al. (2007) developed a series of expectations for the evolution of specificity, based on the outcome (i.e. parasitism vs mutualism) of symbiotic interactions, evolutionary constraints and environmental quality. However, to our knowledge, no empirical studies have explicitly addressed the broader role of host community structure in driving host range evolution (or other traits such as transmission and virulence) in plant–pathogen interactions.
The evolution of pathogen specialization obviously depends in part on whether traits required to infect alternative hosts are positively or negatively correlated. One of the central hypotheses concerning the evolution of host range is that genetic trade-offs in adaptation to different hosts can drive the emergence of specialization (Falconer, 1952; Jaenike, 1990; Kawecki, 1998). Such trade-offs in components of fitness are typically thought to be driven by antagonistic pleiotropy – one or more genes that improve performance in one host or habitat but impair performance in another (Futuyma & Moreno, 1988; Jaenike, 1990). Theory predicts that, in the absence of genetic constraints, a generalist lineage should evolve to be as well adapted to each habitat as respective specialist lineages. Thus, in order for specialists to be favoured, the evolution of generalization necessarily entails a cost (Jaenike, 1990; Fry, 1996; Kawecki, 1998; Palaima, 2007).
While the generality of the trade-off hypothesis has been widely questioned (e.g. Kassen & Bell, 1998; Kawecki, 1998; Krasnov et al., 2004), empirical studies suggest that these mechanisms can be important for the evolution of specialization in plant–pathogen interactions. Agudelo-Romero et al. (2008) evolved replicate populations of the plant pathogen tobacco etch potyvirus, allowing viral populations to evolve for 15 generations on both a novel host and the native host species. In agreement with a pleiotropic cost of host specialization, lineages that evolved increases in virulence and growth in the novel host suffered negative effects for both these traits on the original host species. Similarly, Wallis et al. (2007) demonstrated that serial passaging of the plum pox potyvirus increased infectivity, growth rate and virulence in a novel host (Pisum sativum) with a concomitant reduction in transmission efficiency on the original peach (Prunus persica) host. Even so, there remains a lack of direct evidence that pleiotropic costs are responsible for these patterns.
Antagonistic pleiotropy is not the only genetic mechanism that can influence the evolution of host range. Disruptive selection, where different trait values are favoured on different hosts, combined with host–pathogen coevolution, can promote the formation of specialist lineages via assortative mating, even in the absence of trade-offs (Kawecki, 1998; Duffy et al., 2007). If hosts evolve new resistance following exposure to a pathogen, the impact of this host-induced selection will be weaker for a generalist than a specialist lineage, because only a subset of a generalist lineage is exposed to any given host, and gene flow will consistently counteract selection-driven changes in gene frequency (Fry, 1996; Whitlock, 1996; Kawecki, 1998). For similar reasons, specialization may also be promoted as a result of assortative mating followed by the accumulation of mutations and/or linkage between genes that reduce fitness in interactions with some hosts, but have little effect in others (Kawecki, 1994; Bergelson & Purrington, 1996; Fournier & Giraud, 2008), particularly for hosts that are encountered infrequently or irregularly.
From the host perspective, trade-offs in resistance to different pathogen species may also limit the generality of defence mechanisms. Loiseau et al. (2008) found for the bird species Passer domesticus (the common European house sparrow) that a single major histocompatibility complex (Mhc) class I allelic variant was associated with a 2.5-fold increase in susceptibility to a pathogenic Plasmodium strain, but with a 6.4-fold reduction in susceptibility to a Haemoproteus strain. Similar trade-offs in susceptibility and resistance may drive population-level maintenance of variants of the LOV1 gene in A. thaliana, which confers susceptibility to the fungal pathogen Cochliobolus victoriae (Sweat et al., 2008). Interestingly, loss-of-function mutations in this member of the nucleotide-binding-site leucine-rich-repeat (NBS-LRR) resistance gene family confer resistance to C. victoriae (Lorang et al., 2007), suggesting that the susceptible variant must perform some other important function, probably associated with resistance to an unidentified pathogen species (Sweat et al., 2008). Indeed, many R gene loci segregate for multiple allelic variants providing alternative specificities and susceptibilities (Ellis et al., 2000), limiting the capacity of a single host genotype to resist attack from multiple pathogen strains. This constraint to the evolution of plant defence could be removed if diversification for the same set of specificities had instead occurred at the level of the gene family, because different resistance specificities could be carried simultaneously, providing a greater range of resistance. However, fitness costs associated with carrying multiple genes or specificities simultaneously are one possible explanation for this apparent limitation (Bergelson & Purrington, 1996). Tian et al. (2003) demonstrated a 9% reduction in the fitness of A. thaliana plants carrying the R gene RPM1 relative to isolines lacking RPM1. Other context-dependent constraints may also limit the diversification of R gene families and the evolution of multiple resistance specificities. As one example, Bomblies et al. (2007) demonstrated a negative epistatic interaction in A. thaliana between an allele of an NBS-LRR disease resistance gene homologue and a specific allele at a second locus, which when combined prevent viable hybrids from forming.
2. Specialization and patterns of genetic variation within species
Hosts and pathogens interact within spatially variable environments. Even at relatively local scales, these can vary from single, largely homogeneous populations, to fragmented, spatially structured metapopulations. Within this framework, variation in different micro-evolutionary forces among demes may act to promote diversity and population divergence in different host and pathogen traits (e.g. Parker, 1985; Jarosz & Burdon, 1991; Laine, 2004; Thrall et al., 2005; Barrett et al., 2007). Local adaptation of pathogens to their hosts has been demonstrated as a strong driver of genetic structure in a number of host–pathogen interactions (see Greischar & Koskella, 2007 for review), and complex interactions between various host and pathogen traits in different populations may also arise. For example, in the interaction between the wild host plant Linum marginale and its rust pathogen, Melampsora lini, Thrall & Burdon (2003) demonstrated an intraspecific trade-off between host range and the mean number of infective spores produced by the pathogen, such that strains infecting a wider range of hosts are generally less fecund. Importantly, at the metapopulation scale, these patterns are implicated in maintaining diversity in both host range and virulence among local pathogen populations; selection favours pathogens that have a wide host range in resistant host populations, and more fecund, narrow host range pathogens in susceptible host populations. It is also likely that nonselective factors, such as random genetic drift and selection on linked traits, will influence host and pathogen evolution and drive genetic divergence among demes (Parker, 1991; Salathéet al., 2005; Barrett et al., 2008). Within this framework, rates of genetic recombination (Barrett et al., 2008), and the degree to which pathogen dispersal occurs at local scales relative to the metapopulation as a whole (Thrall & Burdon, 1999; Greischar & Koskella, 2007), will further influence how variation is maintained and distributed within the metapopulation.
The potential importance of coevolution and negative frequency-dependent selection in maintaining genetic diversity within host–pathogen interactions is well recognized (Hamilton, 1980; Clay & Kover, 1996). Theory underlying such dynamics is based upon assumptions of high levels of fidelity and genotype specificity within the interaction. However, multiple recognition specificities have been demonstrated in several gene-for-gene interactions. For example, dual recognition specificity has been shown for the Pseudomonas syringae pv. tomato (Pto) R gene in tomato (Martin et al., 1993), which independently recognizes two P. syringae effector proteins with little amino acid similarity, AvrPto (Ronald et al., 1992) and AvrPtoB (Kim et al., 2002). Similarly, convergent evolution underlies shared R gene specificities in A. thaliana (RPM1) and Glycine max (resistance to Pseudomonas syringae pv. glycinea (Rpg1-b)) to the type III effector protein AvrB from P. syringae (Ashfield et al., 2004). The role that more complex interactions such as these might play in the generation and maintenance of genetic variation within demes is currently unclear.
For more generalist pathogens, the generation and maintenance of adaptively significant genetic variation at local scales may be driven by adaptation among rather than within host species (e.g. Fournier & Giraud, 2008). Variation in pathogen specificity and virulence can further influence the development and maintenance of patterns of genetic diversity among pathogen demes. For example, Lajeunesse & Forbes (2002), using a meta-analysis approach, demonstrated that pathogens with narrow host ranges are more likely to be locally adapted to their hosts. Similarly, using a theoretical approach, Gandon (2002) demonstrated that both increasing specificity and virulence lead to greater local adaptation in pathogen species.
Mechanistic variation underlying resistance specificities may also influence intra-species patterns of genetic variation in both host and pathogen. As discussed in Section II, R genes sometimes interact directly with effectors, but in other cases interact with an intermediary host protein that is modified by the effector. Differences in the mode of recognition (direct or indirect) have been suggested to lead to qualitatively different outcomes in the diversity and specificity of host–pathogen interactions (Van der Hoorn et al., 2002). Specifically, in direct interactions, continual changes in pathogen Avr genes may be matched by changes in host R genes with a resultant increase in diversity and specificity of R and Avr allelic series (e.g. flax and flax rust: Dodds et al., 2006; Ellis et al., 2007). By contrast, indirect recognition appears to be often associated with simple, presence/absence polymorphisms for host resistance/susceptibility alleles and pathogen virulence/avirulence alleles that are maintained for long periods of time by balancing selection, as observed for the RPM1 and RPS5 loci in A. thaliana (Stahl et al., 1999; Tian et al., 2002).
3. The evolution of pathogen virulence
Complex interrelationships can develop between host and pathogen fitness, influenced by the effects that a pathogen has on its host and the specialization of the pathogen in question. Pathogen infection is often assumed to reduce host fitness as an inevitable consequence of pathogen growth within the host (Frank, 1996). Although growth is a critical element of pathogen fitness, increased host mortality resulting from pathogen growth will negatively impact the longevity and productivity of the pathogen's resource base, and thus influence the potential for persistence within populations and opportunities for among host transmission (Read, 1994). It is perhaps not surprising, then, that relationships between host and parasite fitness can change within interactions, depending on the identity of the partners (Salvaudon et al., 2005), and that general support for a positive relationship between virulence and pathogen growth in plant–pathogen systems is equivocal (Sacristan & Garcia-Arenal, 2008). Clearly, natural selection should strike a balance between the costs and benefits of harming hosts, and pathogens should be expected to vary in their impact on the host, depending on the context (Lenski & May, 1994).
Expectations for the evolution of pathogen virulence are dependent in part on the qualitative (e.g. killer, castrator or debilitator) and quantitative effects that a pathogen has on its host. Theory predicts that castrating, florally transmitted pathogens should be under strong selection to completely sterilize their hosts. By forcing hosts to invest all resources into vegetative growth rather than seed production, host resources are funnelled into clonal reproduction and flower production, thereby promoting pathogen transmission (Sloan et al., 2008). While the associated fitness effects of completely preventing seed production can be very high, this need not always be the case. The castrating fungal pathogen Epichloe glyceriae prompts its grass host Glyceria striata to invest more resources into clonal growth than uninfected plants. This promotes vertical transmission of the pathogen to new clones, but paradoxically increases host fitness under some ecological conditions (Pan & Clay, 2002; Pan & Clay, 2003).
The evolution of virulence can also depend on host range. Theory based on tightly coupled, obligate pathogens and their hosts suggests that natural selection will favour the maintenance of high levels of virulence whenever increased virulence is positively associated with increased growth and among-host transmission (May & Anderson, 1983; Frank, 1996). A generalist strategy affords the parasite with more opportunities for transmission and persistence, so that, in theory, decreasing specificity can be associated with increasing virulence (Kirchner & Roy, 2002). The same should be true for opportunistic pathogens and those with long-lived resting or dormant propagules (e.g. Bacillus anthracis), where survival for long periods in the absence of a host should enhance chances of transmission and permit high levels of virulence to evolve (Gandon, 1998; Caraco & Wang, 2008).
4. Environmentally mediated evolution of plant–pathogen interactions
The physical environments within which plants and pathogens interact are highly variable across space and time (Burdon, 1987). As we describe in Section III, environmental variation mediates the amount of pathogen growth, and the fitness effect of pathogen damage. Both phenomena can, in theory, influence the evolution of plant resistance and/or tolerance to microbial pathogens. First, environmental variation may alter the strength or direction of pathogen-mediated selection on plant resistance or tolerance (Burdon & Thrall, 1999). Secondly, environmental variation can reduce the proportion of phenotypic variation in resistance or tolerance that is genetic in nature (Levins, 1968; Reboud & Bell, 1997). One likely consequence of such spatial and temporal heterogeneity is that patterns and rates of evolutionary change will be highly variable, contributing to what Thompson (1994) has coined ‘the geographic mosaic of coevolution’. While important differences are typically envisioned to be most pronounced across large geographic regions (e.g. Barrett et al., 2008), micro-environmental differences can influence the dynamics and evolutionary outcomes of plant–pathogen interactions even at very small spatial scales (Laine, 2006).
Appreciation of such heterogeneity will be critical to developing a deeper understanding of important ecological and evolutionary phenomena including the stable coexistence of highly deleterious pathogens and their plant hosts, and the maintenance of genetic variation in plant resistance and related pathogen traits. Consider, for example, the killer class of pathogens, which are thought to be more strongly influenced by environmental variation than others (Jarosz & Davelos, 1995). While these pathogens can generate strong selection on plant resistance or tolerance, they also tend to have large host ranges and often require very specific environmental conditions to cause plant mortality (e.g. high host density, shade and humidity). Our understanding of how these factors trade off is extremely limited. However, it seems likely that, in a system such as this, local patterns of selection will depend on host plant diversity and the environment, and thus be highly heterogeneous.
The variable impact that infection has on host fitness means that it is important to pay attention not only to disease, but also to host fitness. Here we propose to extend the concept of the disease triangle to include a related plant fitness triangle (Figs 2, 3). From a fitness perspective, the disease triangle describes pathogen fitness, particularly when disease is quantified through some measure of pathogen growth or fecundity. The plant fitness triangle is not simply a mirror of the disease triangle (Fig. 2), and should be considered when the environment fundamentally alters the relationship between plant and pathogen fitness. In the plant fitness triangle, we consider the same three phenomena that influence the disease triangle: host genetic variation in resistance, pathogen genetic variation in growth, and environmental variation.
Figure 3. Corresponding, hypothetical plant disease (a) and fitness (b) triangles in plant–pathogen interactions. In (a), the dashed lines represent the potential parameter space where a pathogen could grow. Solid lines (inner triangle) delineate realized limits under which disease will develop given limited water availability (in this example, 50% of maximum water availability is the limiting parameter). Shading represents the relative intensity of disease resulting from interacting host, pathogen and environmental parameters. In (b), the dashed lines represent the potential parameter space within which a pathogen could reduce plant fitness. Solid lines again reflect realized limits given limited water availability. Shading represents the relative reduction in plant fitness resulting from interacting host, pathogen and environmental parameters. In this example, fitness effects and pathogen growth have become partially uncoupled, so that pathogen growth is reduced in environments with less water, but fitness effects of infection are more severe as a result of the direct effects of water limitations on the plant, interacting with water loss caused by infection.
Download figure to PowerPoint
Consider the following hypothetical disease plant fitness triangles. In an environment that is not conducive to growth, only the most well-adapted pathogen genotypes will grow, and only within the most susceptible plant genotypes. The disease triangle – the area of which describes the quantity of disease that develops – will be small relative to its potential (Fig. 3a). In an environment highly conducive to pathogen growth, pathogen genotypes will be able to infect all but the most resistant host genotypes, and the area of the triangle will be large relative to its maximum (Fig. 2). Now, consider a pathogen whose growth depends on an environment with abundant moisture, and whose negative fitness effects derive from rupturing the leaf surface and increasing water loss in the plant host. In a low-rainfall environment, pathogen growth is low, but even moderate growth reduces host fitness because water is scarce. When rain is plentiful, there may be more pathogen growth, but plant fitness is not impacted as strongly on a per-unit-damage basis. Thus, in this hypothetical example, the plant fitness triangle has become decoupled from the disease triangle, because in high-water environments, the interaction is inherently less pathogenic (Fig. 3).
In summary, we expect that environmentally induced variation in disease and the fitness effects of disease will create heterogeneities in the rate of evolution within and among natural populations. These heterogeneities will affect the evolution of resistance traits that reduce damage as well as tolerance traits that reduce the fitness effects of damage. However, it is not clear how environmentally induced variation is partitioned within and among populations, and whether it will tend to primarily slow plant–pathogen evolution within populations or make evolution more heterogeneous among populations. Given the widely recognized importance of the environment in plant disease, the potential influence of environmental variation on rates of evolutionary change in different plant–pathogen interactions is probably understudied.