HOW DOES SPATIAL DISPERSAL NETWORK AFFECT THE EVOLUTION OF PARASITE LOCAL ADAPTATION?

Authors


Abstract

Studying patterns of parasite local adaptation can provide insights into the spatiotemporal dynamics of host–parasite coevolution. Many factors, both biotic and abiotic, have been identified that influence parasite local adaptation. In particular, dispersal and population structuring are considered important determinants of local adaptation. We investigated how the shape of the spatial dispersal network within experimental landscapes affected local adaptation of a bacteriophage parasite to its bacterial host. Regardless of landscape topology, dispersal always led to the evolution of phages with broader infectivity range. However, when the spatial dispersal network resulted in spatial variation in the breadth of phage infectivity range, significant levels of parasite local adaptation and local maladaptation were detected within the same landscape using the local versus foreign definition of local adaptation. By contrast, local adaptation was not detected using the home versus away or local versus global definitions of local adaptation. This suggests that spatial dispersal networks may play an important role in driving parasite local adaptation, particularly when the shape of the dispersal network generates nonuniform levels of host resistance or parasite infectivity throughout a species’ range.

Studying patterns of parasite local adaptation can reveal the spatiotemporal dynamics of host–parasite coevolution (Thrall et al. 2002). Historically, parasites were often predicted to have an evolutionary advantage over their hosts, due to their generally shorter generation times and larger population sizes, and as such were expected to be locally adapted (i.e., able to overcome immune responses of local hosts) (Ebert 1994; Kawecki and Ebert 2004; Morgan et al. 2005). However, empirical studies have revealed that this is not always the case and examples of parasite local maladaptation or no spatial variation in adaptation are common in the literature (Greischar and Koskella 2007; Hoeksema and Forde 2008).

Fundamentally, local adaptation in host–parasite systems is driven by spatially asynchronous coevolutionary dynamics, and is therefore influenced by the degree of dispersal between populations (Gandon et al. 1998). In the absence of dispersal, this asynchrony can be maintained by the coevolutionary process itself, producing spatially variable selection (Gandon et al. 1998; Gandon 2002). In the presence of dispersal, asynchronous allele frequencies are still predicted to be maintained under certain conditions, despite dispersal acting as a synchronizing force (Gandon and Nuismer 2009). This can be via either stochasticity, with small population sizes coupled with genetic drift preventing synchrony (Burdon 1992; Thompson and Burdon 1992; Gandon 2002), or by deterministic processes but under somewhat stricter conditions—low migration, many populations, and strong selection (Gavrilets and Michalakis 2008; Gandon and Nuismer 2009). Alternatively, selection mosaics, where the strength of the reciprocal selection between host and parasite is spatially variable (Thompson 2005), are predicted to be able to prevent homogenization by selecting for differing alleles in different parts of a species’ range (Nuismer 2006; Gandon and Nuismer 2009).

Although dispersal is known to play a central role in the evolution of local adaptation (Greischar and Koskella 2007), the role that landscape topology, or the spatial dispersal network, plays in driving the evolution of local adaptation during host–parasite coevolution has been largely overlooked. The arrangement of subpopulations or patches within a landscape affects its connectivity (Bull et al. 2006)—the number or the identity of the populations that are connected by dispersal. As such, it can influence the diversity of migrants a patch receives, which is considered a major determinant of the evolutionary potential of a population (Gandon and Michalakis 2002; Garant et al. 2007; Morgan et al. 2007; Vogwill et al. 2008). Furthermore, certain arrangements of patches may result in some patches receiving greater diversity of migrants than other patches within the same landscape, potentially affecting evolutionary potential between patches within landscapes. Here we manipulate the topology of six-patch experimental landscapes consisting of the common soil bacterium Pseudomonas fluorescens SBW25 and its lytic bacteriophage parasite SBW25Φ2 (Buckling and Rainey 2002; Brockhurst et al. 2007), to investigate how different landscape topologies affect host–parasite coevolution and, in particular, the presence of parasite local adaptation or local maladaptation.

Materials and Methods

INITIATING POPULATIONS

Twenty-four microcosms were inoculated with approximately 107 isogenic cells of P. fluorescens isolate SBW25 and 105 isogenic clonal particles the lytic DNA phage, SBW25Φ2. Microcosms consisted of a 30 mL glass universal with loose fitting plastic caps containing 6 mL of Kings B (KB) medium grown in a static incubator at 28°C. Cultures were propagated by serial transfer, whereby 60 μL of culture (1% of the population) was transferred to a fresh KB microcosm every 48 h. Cultures were initially propagated for eight experimental transfers to allow divergence between populations. Cultures were then grouped into four experimental landscapes of six microcosms each. Each of these initial landscapes was used to found four replicate landscapes, each of which was exposed to a different landscape shape and propagated for a further 12 experimental transfers.

EXPERIMENTAL DESIGN

Experimental transfers after the initial divergence also involved the transfer of 60 μL of culture to a fresh microcosm, but migration was simulated by 1% of this inoculum coming from other populations within the same experimental landscape. Which population provided these migrants depended on the topology of the landscape. Four different landscape topologies were used: isolated, linear, circular, and global. In the isolated treatment, populations were maintained without any migration between populations within a landscape. Migration in the linear dispersal treatment consisted of unidirectional stepping-stone migration along a linear string of populations, whereby populations received migrants from the population immediately upstream of them at each transfer. Similarly, the circular dispersal treatment consisted of unidirectional stepping-stone migration along a linear string, but the string was wrapped so that the end of the string connected to the beginning. In the global dispersal treatment, all populations contributed to a pool of migrants that were redistributed to all patches at every transfer. At the end of the experiment, phages were found to have persisted in all populations in all treatments.

SAMPLING POPULATIONS

Cultures were frozen every two transfers throughout the course of the experiment in 20% glycerol and stored at −80°C. Phage populations were isolated by centrifuging in 10% chloroform and then stored at 5°C. Bacterial populations were isolated by plating on agar.

INFECTIVITY ASSAYS

Infectivity of a particular phage population against a particular bacterial population is here defined as the proportion of bacterial colonies that the bacteriophages are capable of infecting (the proportion of susceptible bacteria). Similarly, bacterial resistance is measured as the proportion of colonies which are resistant. This is assayed by first streaking a line of the phage population onto an agar plate and then streaking 10 bacterial colonies perpendicularly across it. Bacteria are deemed susceptible if they suffer any reduction in growth upon encountering the line.

Streaking assays were used to measure levels of phage infectivity within landscapes. Populations of phage and bacteria were crossed against each other from positions 2, 4, and 6 within each metapopulation after 2, 4, 6, 8, 10, and 12 experimental transfers. Populations 2, 4, and 6 represent the second, fourth, and sixth population along the string for linear landscapes, and were selected on this basis so that population 1, which receives no gene flow, would not be assayed. Comparisons between populations receiving no dispersal and some dispersal are common in the literature, and this was not the goal of this article. Numbering of populations is arbitrary in other landscapes, but is based on having a shared founding population with the corresponding population from the linear landscape.

Overall phage infectivity range is measured as the average level of infectivity against all host assay populations from within its own landscape. Likewise, the average resistance range of a bacterial population to all phage assay populations within its own landscape it used as a measure of overall bacterial resistance.

Patterns of local adaptation were also calculated using streaking assays. Both commonly used definitions of local adaptation were used (Kaltz and Shykoff 1998; Kawecki and Ebert 2004; Hoeksema and Forde 2008). Specifically, these are home against away and local versus foreign. In the former this is comparing local phage performance on “home” hosts against local phage performance on “away” hosts, whereas in the latter it is comparing local phage performance on home hosts against “foreign” phage performance on home hosts. Parasite local adaptation was also analyzed by comparing average local fitness (the fitness of a parasite population against its local host population) and average global fitness (the fitness of a parasite population against all host populations) at the landscape or metapopulation scale (Nuismer and Gandon 2008). Local adaptation, or local maladaptation, is inferred when local fitness and global fitness significantly differ, implying an underlying association between the fitness of a parasite genotype and interactions with particular host genotypes.

STATISTICAL ANALYSIS

To detect within-metapopulation variation in resistance and infectivity, levels of bacterial resistance and phage infectivity were averaged over all time points and analyzed using a general linear model performed in Minitab, with position within landscape fitted as a fixed factor and experimental landscape identity (replicate) fitted as a random factor. The three different methodologies for detecting local adaptation were each analyzed separately using repeated measures linear mixed models performed in SPSS (SPSS Inc., Chicago, IL). Levels of bacterial resistance and phage infectivity were again averaged across all time points and then analyzed to detect any local adaptation. Although this precludes studying any temporal variation in local adaptation, it increases the statistical power of the analysis. For the local versus foreign definition of local adaptation, host population was used as the subject and phage infectivity was used as the repeated measure; phage origin and position-within-landscape were treated as fixed factors and experimental landscape identity as a random factor. For the home versus away definition, phage population was used as the subject and phage infectivity was used as the repeated measure; host origin and position-within-landscape were treated as fixed factors and experimental landscape identity as a random factor. For the comparison of average local fitness and average global fitness at the landscape scale, landscape identity was used as the subject and infectivity context (local or global) as the repeated measure, with infectivity context also fitted as a fixed factor. Degrees of freedom were estimated using the Satterthwaite approximation.

Results

PATTERNS OF INFECTIVITY AND RESISTANCE RANGE

With no dispersal, phage infectivity ranges were relatively low compared to bacterial resistance ranges (Fig. 1A), but with no consistent variation by position within each landscape for either infectivity ranges (Table 1; position: F2,6= 0.02, P= 0.978) or resistance ranges (Table 1; position: F2,6= 0.91, P= 0.451). Conversely, infectivity ranges were generally higher than bacterial resistance ranges for populations exposed to global dispersal or circular dispersal (Fig. 1B,C), but again there was no variation with position in either infectivity range (Table 1; position in global treatment: F2,6= 0.32, P= 0.738; Position in circular treatment: F2,6= 0.31, P= 0.743) or resistance range (Table 1; position in global treatment: F2,6= 0.03, P= 0.968; Position in circular treatment: F2,6= 0.43, P= 0.670). However, infectivity ranges were found to significantly increase with linear dispersal in the same direction as dispersal (Fig. 1D, Table 1; Position: F2,6= 15.30, P < 0.01). In contrast, no increase in bacterial resistance ranges was observed with increasing distance from the beginning of the landscape (Fig. 1D, Table 1; Position: F2,6= 1.12, P= 0.386).

Figure 1.

Levels of bacterial resistance (gray bars, proportion resistant colonies ± standard error) and phage infectivity (white bars, proportion susceptible colonies ± standard error) from positions 2, 4, and 6 within landscapes. Panel (A) no dispersal; (B) global dispersal; (C) circular dispersal; (D) linear dispersal.

Table 1.  Variation in host resistance and phage infectivity by position within landscape. Results of general linear model with position within landscape as a fixed factor, and experimental landscape identity (replicate) as a random factor.
Dispersal network dfFP
 Resistance
  NonePosition2,60.910.451
 Replicate3,61.360.342
  GlobalPosition2,60.030.968
Replicate3,65.95<0.05 
  CircularPosition2,60.430.67 
 Replicate3,63.810.077
  LinearPosition2,61.120.386
Replicate3,69.41<0.05 
 Infectivity
  NonePosition2,60.020.978
Replicate3,60.090.963
  GlobalPosition2,60.320.738
 Replicate3,621.34<0.005
  CircularPosition2,60.310.743
Replicate3,69.64<0.05 
  LinearPosition2,615.3 <0.01 
 Replicate3,66.73<0.05 

LOCAL ADAPTATION 1: LOCAL VERSUS FOREIGN

No consistent variation in the ability of local and foreign phages to infect local hosts was detected for no-dispersal landscapes (phage origin: F1,17= 0.28, P= 0.607; position: F2,17= 0.26, P= 0.776; interaction: F2,17= 0.24, P= 0.792; Fig. 2A). Similarly, no variation in the performance of local and foreign phages on local hosts was detected for either circular or global dispersal (Fig. 2B,C), with both landscapes producing generally highly infectious phages (global dispersal: phage origin: F1,17= 1.20, P= 0.291; position: F2,15= 0.06, P= 0.940; interaction: F2,15= 0.27, P= 0.766. circular dispersal: phage origin: F1,15= 0.65, P= 0.434; position: F2,15= 0.84, P= 0.451; interaction: F2,15= 0.07, P= 0.936). However, significant differences were detected between local and foreign phage infectivity against local hosts from landscapes subjected to linear dispersal (Fig. 2D), but this depended on the population's position within the string (phage origin: F1,13= 2.50, P= 0.136; position: F2,13= 0.76, P= 0.489; interaction: F2,13= 7.73, P < 0.01). Specifically, phage populations in position 2 tended to be locally maladapted, whereas populations 4 and 6 both contained locally adapted phage populations.

Figure 2.

Infectivity (proportion susceptible colonies ± standard error) of local phages (gray bars) and foreign phages (white bars) against host population from positions 2, 4, and 6 within landscapes. Panel (A) no dispersal; (B) global dispersal; (C) circular dispersal; (D) linear dispersal.

LOCAL ADAPTATION 2: HOME VERSUS AWAY

No significant variation in the ability of phages to infect home or away hosts from no-dispersal landscapes was detected (Fig. 3A; host population: F1,17= 0.17, P= 0.685; position: F2,17= 0.06, P= 0.941; interaction: F2,17= 0.10, P= 0.907). Nor was there any variation in infectivity against home or away hosts for phage populations from either global or circular dispersal treatments (Fig. 3B,C; global dispersal: host population: F1,13= 1.57, P= 0.234; position: F2,13= 0.22, P= 0.806; interaction: F2,13= 0.07, P= 0.936. circular dispersal: host population: F1,14= 1.04, P= 0.326; position: F2,14= 0.39, P= 0.685; interaction: F2,14= 0.87, P= 0.440), with high levels of infectivity against both. Neither was there significant home versus away local adaptation for phages with linear dispersal (Fig. 3D), but phages did still show a significant increase in infectivity in the direction of dispersal (host population: F1,11= 1.55, P= 0.239; position: F2,11= 8.97, P < 0.01; interaction: F2,11= 0.06, P= 0.940).

Figure 3.

Infectivity (proportion susceptible colonies ± standard error) of phages from positions 2, 4, and 6 against home hosts (gray bars) and away hosts (white bars). Panel (A) no dispersal; (B) global dispersal; (C) circular dispersal; (D) linear dispersal.

LOCAL ADAPTATION 3: LOCAL VERSUS GLOBAL

No significant difference between average local infectivity and average global infectivity was detected for any dispersal treatment (Fig. 4; global dispersal, F1, 6= 0.106, P= 0.756; circular dispersal, F1, 6= 0.005, P= 0.946, linear dispersal, F1, 6= 0.274, P= 0.620; no dispersal, F1, 5= 0.475, P= 0.519). This implies that phage local fitness and phage global fitness was closely related in all treatments, and no parasite local adaptation is present in the form of a spatial covariance between host and parasite genotypes.

Figure 4.

Average local infectivity (proportion susceptible colonies ± standard error, grey bars) and average global infectivity (proportion susceptible colonies ± standard error, white bars) for each dispersal network.

Discussion

Patterns of phage infectivity and bacterial resistance ranges were markedly different for isolated landscapes when compared to the dispersal treatments: dispersal consistently led to the evolution of broader phage infectivity ranges relative to bacterial resistance ranges (Fig. 1). This supports previous work using this system which suggested that phages benefit more from dispersal than their bacterial hosts, due to their lower within-population evolutionary potential in the absence of migration (Morgan et al. 2005, 2007). Moreover, within linear landscapes infectivity was found to increase with the direction of dispersal, suggesting that dispersal further increased the evolutionary potential of phage populations with each “step” along the landscape. Phage local adaptation and phage local maladaptation were both detected within the same experimental landscape, but only for linear dispersal networks and only using the local versus foreign definition. Previous work has shown that SBW25Φ2 is unlikely to show local adaptation during the early, directional stages of coevolution (Morgan et al. 2005; Morgan and Buckling 2006), but no previous work had examined landscapes that incorporated spatial variation in infectivity.

Significant parasite local adaptation or local maladaptation was only detected using the local versus foreign definition, and is likely to have arisen from the within-landscape spatial variation in phage fitness coupled with the lack of variation in host fitness, rather than reflect “true” local adaptation (sensu local versus global local adaptation [Nuismer and Gandon 2008]). As the local versus foreign definition of local adaptation compares the performance of local and foreign phages against a common host population, the lack of quantitative variance in host performance causes all variation in local adaptation to be a result of the main effect of phage infectivity. For example, parasites with high local infectivity will also possess high global infectivity, and appear locally adapted when compared to parasites with lower infectivity from foreign locations. In contrast, if host resistance was spatially variable but phage infectivity was not, it is likely that the home versus away definition of local adaptation would produce both significant local adaptation and local maladaptation. For example, parasites whose local hosts possess higher global resistance will appear locally maladapted, due to improved parasite performance on less resistant hosts from away locations. As such, both conventionally used definitions of local adaption will only give congruent results in which neither infectivity nor resistance shows significant spatial variation (Thrall et al. 2002).

Although neither of the commonly used definitions of parasite local adaptation will always detect true local adaptation, they will still be a measure of the coevolutionary dynamics at a landscape or metapopulation level, particularly if coevolved traits vary geographically. Geographic variation in the strength of coevolutionary interaction has been reported for a wide range of systems (Kraaijeveld and Godfray 1999; Brodie et al. 2002; Thrall et al. 2002; Benkman et al. 2003; Thompson 2005; Laine 2006; Toju and Sota 2006; Hanifin et al. 2008). These situations will likely result in a mix of locally adapted and local maladapted populations: populations whose geographic location results in greater selection for coevolutionary traits will tend to show local adaptation, whereas those populations experiencing ecological conditions not favoring escalated infectivity and defence traits will be locally maladapted.

As with similar work using bacteria and phage, the experiment reported here migrated bacteria and phage at the same rate, and the results may be more relevant to host–parasite systems in which host and parasite display similar degrees of population structuring (Forde et al. 2004, 2007; Morgan et al. 2007). Furthermore, dispersal only occurred at one rate in this experiment (1% of founding population at each transfer), which previous work on this system suggests is an intermediate level of dispersal (Morgan et al. 2005, 2007; Vogwill et al. 2008). Local adaptation is predicted to be strongest at intermediate levels of dispersal (Gandon 2002; Gandon and Michalakis 2002); low rates of dispersal impair adaptation by constraining genetic diversity, whereas high rates prevent local adaptation by swamping local conditions.

Both theoretical work and empirical studies suggest that the degree of dispersal has a major affect on both host–parasite coevolution and parasite local adaptation. However, the findings presented here also demonstrated that spatial dispersal networks can be major drivers of coevolutionary dynamics. Specifically, we have shown that geographic variation in infectivity, as generated by unidirectional dispersal can drive the evolution of parasite local adaptation and local maladaptation, in the absence of other ecological differences between patches (Morgan et al. 2005; Morgan and Buckling 2006). This demonstrates that spatial dispersal networks can have a major effect on host–parasite coevolution, and provides further support to the notion that host–parasite interactions can only be understood in a spatially explicit context.


Associate Editor: S. Nuismer

ACKNOWLEDGMENTS

This work was funded by a NERC studentship to TV, a Royal Society Project Grant to AF, and a Wellcome Trust VIP award to MAB. We are grateful to Scott Nuismer and two anonymous reviewers for helpful comments on earlier versions of this work.

Ancillary