The impact of parasite dispersal on antagonistic host–parasite coevolution

Authors


Michael A. Brockhurst, School of Biological Sciences, University of Liverpool, Crown Street, Liverpool, L69 7ZB, UK.
Tel.: +44 151 795 4557; fax: +44 151 795 4410; e-mail: michael.brockhurst@liverpool.ac.uk

Abstract

Coevolving populations of hosts and parasites are often subdivided into a set of patches connected by dispersal. Higher relative rates of parasite compared with host dispersal are expected to lead to parasite local adaptation. However, we know of no studies that have considered the implications of higher relative rates of parasite dispersal for other aspects of the coevolutionary process, such as the rate of coevolution and extent of evolutionary escalation of resistance and infectivity traits. We investigated the effect of phage dispersal on coevolution in experimental metapopulations of the bacterium Pseudomonas fluorescens SBW25 and its viral parasite, phage SBW25Φ2. Both the rate of coevolution and the breadth of evolved infectivity and resistance ranges peaked at intermediate rates of parasite dispersal. These results suggest that parasite dispersal can enhance the evolutionary potential of parasites through provision of novel genetic variation, but that high rates of parasite dispersal can impede the evolution of parasites by homogenizing genetic variation between patches, thereby constraining coevolution.

Introduction

Antagonistic host–parasite coevolution, the reciprocal evolution of enhanced host defence and parasite counter-defence, is pervasive in natural communities and is implicated in a wide range of ecological and evolutionary processes (Thompson, 2005; Woolhouse et al., 2002). Often populations of hosts and parasites are subdivided into a set of patches or demes connected by dispersal (a metapopulation). Under such conditions, the dynamics and outcomes of coevolution are likely to be influenced by the relative levels of dispersal between patches in each of the interacting species. All else being equal it is predicted that the species with the greater level of dispersal will have the upper hand in a given coevolutionary arms race (Gandon & Michalakis, 2002; Greischar & Koskella, 2007; Hoeksema & Forde, 2008). This arises because dispersal introduces novel genetic variation into the population, thereby enhancing its adaptive potential (Lenormand, 2002). However, theoretical and empirical studies suggest that very high levels of dispersal can have a detrimental effect on genetic variation and thereby adaptive potential (Garant et al., 2007). This arises through two mechanisms; first, high rates of dispersal can cause ‘genetic swamping’ by replacing locally adapted alleles with locally maladapted alleles common in the metapopulation as whole (Alleaume-Benharira et al., 2006); second, high rates of dispersal can homogenize genetic variation among patches thus reducing the supply of novel variation attainable through dispersal (Gandon & Michalakis, 2002). Combined, these processes lead to the prediction that the rate of adaptation is likely to peak at intermediate rates of dispersal.

A wide range of relative rates of gene flow, which is likely to correlate with dispersal rate, have been observed in natural antagonistic associations (encompassing host–parasite and predator–prey). Though some antagonistic associations show remarkably congruent patterns of gene flow (Jerome & Ford, 2002; Mulvey et al., 1991), in others, the rate of gene flow experienced by antagonists are decoupled, with either the host/prey (Delmotte et al., 1999; Martinez et al., 1999) or the parasite/predator displaying greater levels of gene flow (Davies et al., 1999; Dybdahl & Lively, 1996; Michalakis et al., 1994). Relatively greater rates of parasite compared with host dispersal or gene flow are likely to underlie patterns of parasite local adaptation (i.e. greater performance on sympatric compared with allopatric hosts) observed in natural populations through provision of genetic variation and thereby enhancement of the adaptive potential of parasites (Dybdahl & Lively, 1996; Gandon & Michalakis, 2002; Greischar & Koskella, 2007; Hoeksema & Forde, 2008; Lively & Dybdahl, 2000). Such local adaptation of organisms causing disease to humans or livestock and crops is of particular concern (Woolhouse et al., 2002), thus an understanding of the coevolutionary impact of greater relative rates of parasite compared with host gene flow is required.

In laboratory studies with bacteria and their viral parasites (phage), where rates of dispersal can be directly manipulated, dispersal between patches has emerged as a key determinant of the outcomes and dynamics of coevolution (Brockhurst et al., 2007a; Forde et al., 2004, 2007; Morgan et al., 2007, 2005). In the Pseudomonas fluorescens SBW25–SBW25Ф2 association, bacteria are locally adapted in the absence of dispersal (i.e. bacteria are more resistant to sympatric compared with allopatric phage populations) (Morgan et al., 2005). Moderate increases in the relative rate of bacterial dispersal (1–10%) have no effect on local adaptation; this is because bacteria already have the upper hand in the coevolutionary arms race. By contrast, moderate increases in the relative rate of phage dispersal (1–10%) reverse patterns of local adaptation such that phages are locally adapted. This arises because dispersal introduces novel genetic variation [genetic variation for both resistance and infectivity has been shown to readily evolve in coevolving populations of P. fluorescens and SBW25Ф2 (Poullain et al., 2008)] enhancing the adaptive potential of phage such that, on average, phages have the upper hand in the coevolutionary arms race (Morgan et al., 2005). Mechanisms, such as dispersal, that enhance the adaptive potential of the lagging partner in a coevolutionary association can have a significant impact upon the dynamics of coevolution because such reciprocal evolutionary change may only proceed as rapidly as the slowest adapting partner. Such ‘warming’ of coevolutionary cold-spots (Gomulkiewicz et al., 2000) through increased dispersal of the host (Brockhurst et al., 2007a) or host and parasite simultaneously (Forde et al., 2007; Morgan et al., 2007) has been observed in several studies. However, the effect of greater relative rates of parasite dispersal remains unconsidered.

Although empirical evidence suggests that greater relative rates of parasite compared with host dispersal lead to greater parasite local adaptation (Dybdahl & Lively, 1996; Lively & Dybdahl, 2000; Morgan et al., 2005), the impact on other aspects of the coevolutionary process such as the rate of coevolutionary change and the extent of coevolutionary escalation remain largely unexplored. The rate of coevolution has been shown to affect genetic diversity and population dynamics in coevolving populations (Buckling & Hodgson, 2007; Thompson, 2005), whereas the evolution of more broadly infective parasites has clear implications for disease (Woolhouse et al., 2002). Moreover, when compared with the wide range of relative rates of parasite dispersal observed in natural systems, only a very restricted range of relative rates has thus far been studied using experimental metapopulations (Morgan et al., 2005). To further investigate the effects of parasite dispersal on the coevolutionary process, we established replicate metapopulations of the common soil bacterium P. fluorescens SBW25 and its lytic viral bacteriophage SBW25Ф2, which were propagated by serial transfer. Within each metapopulation, phages were migrated from a migrant-pool at a range of different rates representative of those observed in natural systems, ranging from no dispersal in both host and parasite, to increasingly greater dispersal of parasites relative to hosts. Hosts were left unmigrated in all treatments. Previous studies with this host–parasite association have shown persistent cycles of coevolution imposing directional selection for increased infectivity and resistance ranges through time in phage and bacteria respectively (Buckling & Rainey, 2002), such ranges are a measure of the extent of coevolutionary escalation. We assayed levels of evolved bacterial resistance range and phage infectivity range (these are ‘global’ measures against both sympatric and allopatric antagonists), as well as the rate of coevolutionary change in one focal patch within each experimental metapopulation.

We hypothesized that phage dispersal would increase the adaptive potential of phages by introducing novel genetic variation, but that high levels of dispersal would impede adaptation by homogenizing genetic variation between patches and/or introducing locally maladaptive alleles. This leads to the prediction of a negative quadratic effect of phage gene-flow rate on the adaptive potential of phages. Because bacteria are ahead in the coevolutionary arms race in the absence of dispersal and phage adaptation is the rate-limiting-step of coevolution in this system, we further predicted: (1) a negative quadratic relationship between the rate of coevolution and phage dispersal rate; (2) a negative quadratic relationship of the extent of evolutionary escalation in resistance and infectivity ranges with phage dispersal rate.

Materials and methods

Culturing conditions

Cultures were grown in 30-mL glass universals with loose fitting plastic caps containing 6 mL of Kings B (KB) medium in a static incubator at 28 °C. Cultures were propagated by serial transfer, with 60 μL of culture being transferred to a fresh KB microcosm every 48 h. Samples of each culture were frozen in 20% glycerol and stored at −80 °C every two transfers throughout the course of the experiment.

Isolation of bacteria and phage

Phage samples were isolated during the experiment by centrifuging samples of culture (13 000 rpm/9500 g, 2 min) in 10% chloroform. This lysed and pelleted the bacterial cells, leaving a suspension of phage particles in the supernatant. Isolated phage samples were then stored at 5 °C. Bacteria were isolated by growing cultures overnight in a KB microcosm containing 0.37% Virkon® (a commercially available disinfectant). At this concentration Virkon® is toxic towards bacteriophage particles whereas being nontoxic towards P. fluorescens. 60 μL was then transferred to a fresh KB microcosm and incubated for a further 24 h. This treatment left bacteria viable and free from phage and Virkon®. Presence of phage following this procedure was routinely checked by assaying the infectivity of a sample of culture against ancestral bacteria, no phages were detected.

Initiating populations

18 KB microcosms were inoculated with approximately 107 isogenic cells of P. fluorescens isolate SBW25 and 105 isogenic particles of the lytic DNA phage, SBW25Ф2. Cultures were initially propagated for eight transfers to allow divergence between populations prior to migration.

Experimental treatments

Following divergence, populations were assigned to one of six replicate metapopulations, each consisting of three microcosms. Each replicate metapopulation was then used to found five further metapopulations, each of which was subjected to one of five different phage migration regimes (0%, 0.1%, 1%, 10% and 100% of phage population from migrant-pool) for 24 days (12 transfers) of culturing. Bacteria were left unmigrated in all treatments: at each transfer samples of bacteria were isolated from each population and 60 μL of this isolate was transferred to a fresh microcosm. By contrast transferred phage came from two sources: unmigrated phage isolated from the relevant population, and phage from a migrant-pool for each metapopulation, which consisted of equal proportions by volume of phages isolated from each constituent microcosm. The proportion of the total transferred volume (60 μL) added from each source was determined by the migration treatment, for example under the 1% migration regime, 0.6 μL of transferred phage came from the migrant pool and 59.4 μL came from the phage isolated from the relevant population.

Assays

Quantifying resistance and infectivity

Bacterial resistance was assayed as a binary trait, such that a given bacterial colony could be either susceptible or nonsusceptible to infection by phage. For each assayed population, ten individual bacterial colonies were isolated by plating on a KB agar plate. Colonies were then streaked across a 20 μL line of phage on a KB agar plate and incubated for 24 h at 28 °C. A colony was defined as susceptible if there was visible inhibition of growth upon crossing the line of phage. Resistance was recorded as the proportion of nonsusceptible bacteria per population, whereas infectivity was measured as the proportion of susceptible bacteria per population. Within each migration treatment, one focal population from each of the six replicate metapopulations was selected to undergo assays.

Rate of coevolution

To determine if directional antagonistic coevolution occurred in this experiment, we used stored population samples (see above) to measure how the infectivity of phage populations to a bacterial population changed through time. Specifically, at transfers 2, 4, 6, 8 and 10 we determined the resistance (proportion resistant colonies) of bacterial populations to past (two transfers previous), contemporary and future (two transfers subsequent) sympatric phage populations. If directional antagonistic coevolution was occurring then we would expect, for multiple time points, future phage to be better than contemporary phage, and for contemporary phage to be better than past phage at infecting contemporary bacteria, hence a positive slope of infectivity against time (past, contemporary and future). To determine the rate of coevolution, we calculated how much phage infectivity changed between past and future populations, given by the slope of infectivity against time, and averaged across time-points (Brockhurst et al., 2003). Because bacterial resistance to contemporary phage remains relatively constant across time-points, we can infer bacterial adaptation (Brockhurst et al., 2007b, 2003), hence when considered over multiple time-points this is a measure of coevolution, rather than simply phage infectivity evolution.

Resistance and infectivity ranges

The breadth of resistance and infectivity ranges was assayed every four transfers by determining the resistance/infectivity for each bacteria/phage population when assayed against all other focal populations from the other migration treatments that shared a founding metapopulation. This provides a ‘global’ measure of which treatment has produced the relatively most infectious and resistant populations, whereas controlling for the effect of founding metapopulation. Phage infectivity to their sympatric bacteria (i.e. the bacteria from the same microcosm and time point) was measured every two transfers throughout the course of the experiment.

Statistical analysis

Sympatric infectivity, rate of coevolution and breadth of infectivity and resistance ranges were averaged through time and analysed separately using General Linear Models carried out in Minitab. Founding metapopulation was fitted as a random factor and Log10 (migration rate + 0.01) was simultaneously fitted as both a linear and quadratic covariate. Whether the addition of a quadratic term significantly improved model fit over a simpler linear model was determined using partial F-tests. Resistance ranges through time were log10 transformed and infectivity ranges through time were square-root transformed to meet the necessary assumptions (normality, homogeneity of variance).

Results

As predicted, we observed a negative quadratic relationship between the rate of phage dispersal and the rate of coevolution which peaked at 1% (Fig. 1; founding metapopulation, F5,22 = 0.77, P = 0.579; linear effect, F1,22 = 0.37, P = 0.550; negative quadratic effect, F1,22 = 7.29, P = 0.013, partial F-test for inclusion of quadratic rate term, F1,22 = 7.29, P < 0.05). Because coevolution is predominantly directional in this system (Buckling & Rainey, 2002), more rapid coevolution is typically associated with the evolution of broader phage infectivity range. In line with this, a negative quadratic relationship between the rate of phage dispersal and phage infectivity range was observed which also peaked at 1% (Fig. 2; founding metapopulation, F5,22 = 3.26, P = 0.024; linear term, F1,22 = 26.18, P < 0.001; negative quadratic term, F1,22 = 38.51, P < 0.001; partial F-test for inclusion of quadratic rate term, F1,22 = 38.5, P < 0.01). These data are consistent with the hypothesis that phage dispersal between patches can increase genetic variation thereby enhancing phage evolutionary potential, but that high levels of dispersal (10–100%) may impede phage evolution, either through ‘genetic swamping’ or homogenization of genetic variation between patches, thereby limiting the rate and extent of coevolution attainable.

Figure 1.

 The effect of phage migration rate on the rate of coevolution. The rate of coevolution was given by the slope of the change in infectivity of a phage population through time. Bars show mean (+SEM) rate of coevolution averaged through time.

Figure 2.

 The effect of phage migration rate on phage infectivity range. The infectivity range was given by determining the infectivity of each phage population when assayed against bacteria from all other focal populations from the other migration treatments that shared a founding metapopulation, providing a measure of ‘global’ infectivity. Bars show mean (+SEM) infectivity range of phage populations averaged through time.

Because coevolution is a reciprocal process, bacterial resistance range was expected to evolve in response to changes in phage infectivity range. Bacterial resistance ranges also displayed a negative quadratic relationship with the rate of phage migration peaking at 1% (Fig. 3; founding metapopulation, F5,22 = 2.96, P = 0.034; linear term, F1,22 = 8.94, P = 0.007; negative quadratic term, F1,22 = 9.91, P = 0.005; partial F-test for inclusion of quadratic rate term, F1,22 = 9.94, P < 0.01) and were positively correlated with infectivity ranges (correlation of infectivity and resistance range means; Pearson’s = 0.935, P = 0.02). This suggests that bacterial resistance ranges were able to successfully evolve in response to the broadening of phage infectivity range through time despite a complete lack of dispersal. Taken together with previous studies (Brockhurst et al., 2007a; Morgan et al., 2007, 2005) this suggests that bacterial populations possess potential for coevolutionary escalation that remains unutilized in coevolving populations limited by the rate of phage adaptation.

Figure 3.

 The effect of phage migration rate on bacterial resistance range. The resistance range was given by determining the resistance of each bacteria population when assayed against phage from all other focal populations from the other migration treatments that shared a founding metapopulation, providing a measure of ‘global’ infectivity. Bars show mean (+SEM) resistance range of bacterial populations averaged through time.

The decline in the rate of coevolution and breadth of phage infectivity range at high rates of dispersal (10–100%) could have arisen through two mechanisms: ‘genetic swamping’ causing loss of locally beneficial alleles, or homogenization of genetic variation between patches. To assess whether locally beneficial infectivity alleles were lost at high phage gene-flow rates, infectivity of phages to their sympatric bacterial hosts was analysed; a decline in sympatric infectivity at high rates of dispersal would have been expected if locally beneficial infectivity alleles were being lost through ‘genetic swamping’. Phage dispersal led to higher levels of infectivity of phages on their contemporary sympatric bacterial hosts (Fig. 4; founding metapopulation, F5,22 = 3.31, P = 0.022; linear effect, F1,22 = 4.00, P = 0.058; negative quadratic effect, F1,22 = 7.97, P = 0.010; partial F-test for inclusion of quadratic rate term, F1,22 = 7.97, P < 0.01). Further analysis, excluding the 0% dispersal data, found no difference in sympatric infectivity between other rates of dispersal (Fig. 4; founding metapopulation, F1,16 = 3.29, P = 0.031; linear effect, F1,16 = 0.30, P = 0.593; negative quadratic effect, F1,16 = 0.00, P = 0.955). Because no decline in sympatric infectivity was observed with increasing dispersal rate, this suggests that, in this experimental system, high rates of dispersal do not significantly limit phage adaptation to local bacterial hosts through ‘genetic-swamping’. It seems likely therefore that the decline in the rate of coevolution and breadth of infectivity range observed at high rates of phage dispersal were due to homogenization of genetic variation between patches.

Figure 4.

 The effect of phage migration rate on sympatric infectivity. The sympatric infectivity was given by determining the infectivity of a phage population on bacteria from the same time-point and microcosm. Bars show mean (+SEM) infectivity of phage populations to contemporary sympatric bacterial populations averaged through time.

Discussion

Evidence from theory (Gandon & Michalakis, 2002), natural populations (Dybdahl & Lively, 1996; Lively & Dybdahl, 2000) and laboratory populations (Morgan et al., 2005) suggests that greater relative rates of dispersal in parasites compared with hosts should increase parasite local adaptation (Greischar & Koskella, 2007; Hoeksema & Forde, 2008). However, local adaptation provides only a contemporary ‘snap-shot’ of coevolutionary interactions, yielding little information about other aspects of the coevolutionary process. The results presented here extend local adaptation findings to consider the effect of a wide range of rates of parasite dispersal on the dynamics and outcomes of coevolution. We demonstrate that parasite dispersal can enhance the evolutionary potential of parasites increasing both the rate and extent of escalation attainable during antagonistic host–parasite coevolution. However, high rates of parasite dispersal can impede parasite adaptation, our results suggest that the most likely mechanism for this is through homogenizing genetic variation between patches, thereby constraining the coevolutionary process. In a previous study where bacteria and phage were migrated simultaneously (Morgan et al., 2007), evolved phage infectivity range did not decline at high rates of dispersal (10–50%) as observed here. This suggests that the decoupling of host and parasite dispersal can alter the outcome of coevolution by limiting the effects of dispersal on evolutionary potential to one or other antagonist.

Our results suggest that bacterial populations possess coevolutionary potential that remains unutilized in the absence of phage dispersal, posing the question: why if broader resistance ranges can be evolved do they not evolve in the absence of phage dispersal (as seen by the low evolved resistance ranges for 0% migration rate in Fig. 3)? The strong positive correlation between resistance range and infectivity range in this experiment suggests that selection favours the evolution of sufficient rather than maximal resistance ranges. This is likely to be due to the high cost of phage resistance mutations in this system (Brockhurst et al., 2004; Buckling et al., 2006), such that at any given time bacterial clones with broader than necessary resistance mutations are likely to be outcompeted by sufficiently resistant but fitter clones.

Acceleration of coevolution due to parasite dispersal is likely to be particularly apparent in coevolutionary systems where parasites are the lagging antagonist in the absence of dispersal. This is due to the rate of coevolutionary change being limited by the adaptive rate of the slowest partner. Under such conditions dispersal is likely to lead to the more rapid evolution of more infective parasites. The generality of the patterns of infectivity and resistance range evolution observed in this study may be somewhat limited to systems that undergo predominantly directional selection. Such systems include certain plant–pathogen interactions (see for example, Thrall & Burdon, 2003; Laine, 2006) and other host–parasite interactions that broadly comply with a multilocus gene-for-gene model of coevolutionary interaction, which allows for the evolution of generalist resistance and infectivity phenotypes in hosts and parasites respectively (Damgaard, 1999; Sasaki, 2000; Thompson & Burdon, 1992).

In this and previous studies with this host–parasite association, adaptation has consistently peaked at 1% dispersal despite differences in the precise ecological conditions used in each study (Brockhurst et al., 2007a; Morgan et al., 2007). However, it is unclear how, low, intermediate or high rates of dispersal should be defined for natural systems. Undoubtedly this is likely to be under the influence of a wide range of contributory factors that also affect genetic diversity (e.g. mutation rate, population size generation time, etc.). Given this proviso, these results could have implications for health and agriculture. Moderate increases in parasite dispersal associated with increased mobility of human populations and movement of livestock and crops could significantly alter coevolutionary dynamics leading to the more rapid emergence of more infective parasites. Both theoretical and empirical evidence suggests that through increasing transmission opportunities this is likely to be associated with an increase in the virulence of disease (Boots et al., 2004; Boots & Mealor, 2007; Boots & Sasaki, 1999; Herre, 1993). By contrast, very large increases in parasite dispersal rate are likely to erode the potential benefits to parasites of dispersal, leading to decline of parasite evolutionary potential, thereby limiting infectivity and virulence evolution.

Acknowledgements

We are grateful to Mike Begon, Steve Paterson, Angus Buckling and two anonymous reviewers for comments on earlier versions of this work. This work was funded by a NERC studentship to TV, a Royal Society research project grant to AF and a Wellcome Trust VIP award administered by the University of Liverpool’s Research Development Fund to MAB.

Ancillary