Dispersal provides the opportunity to escape harm and colonize new patches, enabling populations to expand and persist. However, the benefits of dispersal associated with escaping harm will be dependent on the structure of the environment and the likelihood of escape. Here, we empirically investigate how the spatial distribution of a parasite influences the evolution of host dispersal. Bacteriophages are a strong and common threat for bacteria in natural environments and offer a good system with which to explore parasite-mediated selection on host dispersal. We used two transposon mutants of the opportunistic bacteria, Pseudomonas aeruginosa, which varied in their motility (a disperser and a nondisperser), and the lytic bacteriophage ФKZ. The phage was distributed either in the central point of colony inoculation only, thus offering an escape route for the dispersing bacteria; or, present throughout the agar, where benefits of dispersal might be lost. Surprisingly, we found dispersal to be equally advantageous under both phage conditions relative to when phages were absent. A general explanation is that dispersal decreased the spatial structuring of host population, reducing opportunities for parasite transmission, but other more idiosyncratic mechanisms may also have contributed. This study highlights the crucial role the parasites can play on the evolution of dispersal and, more specifically, that bacteriophages, which are ubiquitous, are likely to select for bacterial motility.
Dispersal provides a number of potential benefits including reduced kin competition (Hamilton & May, 1977; Comins et al., 1980; Taylor & Frank, 1996; Gandon & Michalakis,1999; Rousset & Gandon, 2002; Taylor & Buckling, 2010), reduced inbreeding (Bengtsson, 1978) and increased probability of survival in temporally heterogeneous environments (Van Valen, 1971; McPeek & Holt, 1992). Dispersal will be selected against if there are associated high costs, such as reduced reproduction or increased mortality (Rousset & Gandon, 2002). One potential cost comes from the probability of encountering harm imposed by natural enemies (Lion et al., 2006). However, dispersal does not only have the potential to increase encounters, but also offers the opportunity for escape in heterogeneous environments (Augspurger, 1983; Brown et al., 1995). Whether or not dispersal is beneficial in the presence of enemies is therefore likely to be influenced by the spatial distribution of the enemies in the environment.
Where parasites are highly localized, such as during intermittent epidemics, dispersal is likely to be individually advantageous because it allows escape from the disease outbreak, however, where parasites are endemic and distributed throughout the environment, the costs and benefits of dispersal are less clear. Dispersal reduces host population structuring, which can potentially increase parasite transmission across the whole environment as there are greater opportunities for parasites to move between patches of hosts (Rhodes & Anderson, 1996; Watve & Jog, 1997; Boots & Sasaki, 1999; Wilson et al., 2003). However, reducing host structuring also reduces the size of high-density host patches, and parasite transmission is likely to be high within high-density patches (Watve & Jog, 1997). As a result, parasite transmission is predicted to be minimized at an intermediate level of host dispersal (Watve & Jog, 1997).
Here, we empirically investigate how spatial distribution of parasites in the environment influences host dispersal behaviour using the opportunistically pathogenic bacteria Pseudomonas aeruginosa and an associated virus (bacteriophage; phage). Understanding the selective pressures that shape bacterial motility is particularly important for pathogenic species, as motility traits have been identified as important virulence factors (Drake & Montie, 1988; Josenhans & Suerbaum, 2002). Phages are ubiquitous and will act as strong natural selective agents for bacteria. However, they are not actively motile, and thus, bacteria have the opportunity to escape phages in a structured environment (i.e. where dispersal cannot be passive). Lytic bacteriophage infection begins with the attachment of the phage to a specific surface receptor on the bacterial host. The phage injects its DNA into the host cell, which then hijacks the host's replication machinery to propagate new phage particles. The bacterial cell then lyses, releasing the viral progeny into the environment (Lenski, 1988).
Phages can affect selection for bacterial dispersal in numerous ways. Phages are known to exploit motility surface structures (flagella and pili), and experiments have shown that bacteria can gain resistance to pilus-specific and flagellatropic phages through alteration of these motility structures, resulting in reduced bacterial motility (Icho & Iino, 1978; Mattick, 2002; Bradley & Altizer, 2005; Brockhurst et al., 2005). To investigate the ecological generality of this association, in a previous study, Koskella et al. (2011) used experimental evolution combined with an observational survey to look at the relationship between bacterial motility and phage resistance in natural populations of Pseudomonas spp. and a community of phages isolated from horse chestnut trees. The survey data suggested that bacterial isolates that were highly resistant to phages had higher motility than their more susceptible counterparts; therefore, natural phages may have been imposing selection for motility in these natural populations. Subsequent evolution experiments found no net effect of these natural phage communities on bacterial motility, potentially because of the conflicting selection pressures: motility may be generally beneficial, but phages often target motility organelles. Here, we explicitly test how a phage that does not target motility organelles imposes selection for dispersal in bacteria under conditions where there is an ‘escape route’ and where phages are present everywhere.
Materials and methods
Strain details and phage amplification
Two transposon mutants defective in type IV pili and generated from a wild-type strain of P. aeruginosa (PAO1) were used: a PilU (nondisperser) mutant which is able to express but unable to retract pili (hyperpiliated) and a PilA (disperser) mutant which is absent of pili (D'Argenio et al., 2001; Taylor & Buckling, 2010). In semi-solid agar, the PilA mutants acted as the ‘dispersers’ as they were able to freely move through the substrate using their flagella, whereas the PilU mutants were the ‘nondispersers’, as the permanently extruded pili caused the bacteria to become stuck, restricting movement. Cultures were grown overnight in a homogeneous (shaken) environment (0.9 g) at 37 °C, in 30 mL universal vials containing 6 mL fresh medium (King's B (KB): 20 g L−1 protease peptone, 10 g L−1 glycerol, 1.5 g L−1 potassium phosphate and 1.5 g L−1 magnesium sulphate heptahydrate).
The phage ФKZ was chosen because it targets the LPS component of the bacterial outer cell membrane (Ceyssens & Lavigne, 2010) and not motility structures (such as pili), which is important in the context of this experiment because we wanted the motility phenotype and not the presence or absence of motility structures to determine the fitness of the genotypes. An amplified phage stock solution of ФKZ was created via passaging through an ancestral host, followed by chloroform treatment and centrifugation, to burst and separate bacterial cells, respectively (Buckling & Rainey, 2002). The phage stock solution containing approximately 2.2 × 1010 phage particles/mL was stored at 4 °C. Phage was spotted onto a bacterial lawn to test the infectivity to each of the motility variants, and both types were found to be sensitive.
To test the optimal dispersal strategy of P. aeruginosa in the presence of ФKZ, we measured the fitness of the two motility variants (disperser and nondisperser), under two phage treatment conditions: local, whereby the phages were local only to the inoculation site; and global, whereby the phages were present throughout the environment. 25 mL of semi-solid KB agar (0.6% wt/vol agar) was poured into a 9 cm diameter petri dish and allowed to dry in the laminar flow hood for 20 min. Agar plates being used for the global treatment group were supplemented with 1 mL phage stock solution (agar was cooled prior to phage addition and briefly vortexed to ensure complete mixing). Inoculum for starting bacterial populations under local treatment was created via a 1 : 1 mix (mixed immediately prior to inoculation onto agar plate) of ФKZ phage stock and one of the dispersal variants. For the global treatment group, pure disperser and nondisperser cultures were used as the inoculum. 2.5 μL of inoculum was pipetted into the centre of the agar plate, just below the agar line. Plates were left for 48 h. There were four replicates in each treatment group.
To control for differences in susceptibility to phages and growth rates not directly caused by motility differences, we grew the dispersers and nondispersers with phage ФKZ in both shaken KB liquid broth (where both motility variants should have entirely unrestricted movement) and hard KB agar (1.2 % wt/vol) (where both motility variants should have entirely restricted movement). In addition, we grew the motility variants in semi-solid KB agar in the absence of any phages. To account for any differences in dispersal distance due to growth rate, we measured the maximum exponential growth rate, Vmax (the maximum rate of change of optical density during log growth), for the disperser and nondisperser and found no significant difference between the strains (Taylor & Buckling, 2010).
Checking for genetic changes
To account for any evolutionary changes that may have occurred during the course of the experiment, we tested for difference in motility and phage resistance phenotype after the bacteria had been exposed to each treatment group. Firstly, we looked for evolved motility differences: we sampled from the edges of colonies which had been exposed to each of the treatment groups (local and global) and inoculated fresh KB soft agar plates without any phages. We measured the area of dispersal for each treatment group, and noted the spreading pattern. Next, we looked at the proportion of resistant clones present in each population after phage treatment. We washed-off bacteria from plates within each treatment group (three replicates within each) after 48 h exposure to phage (or control) treatment, diluted to an appropriate dilution to allow colony differentiation and spread 50 μL evenly over a KB agar plate. We picked 10 colonies from each plate and streaked each across a line of ФKZ phage which had been dried across the middle of the plate. We left the plates to grow overnight and looked for susceptibility to phage (which was scored as binary).
Data collection and statistical analysis
Methods were as in Taylor & Buckling (2010). The area covered by the bacteria was calculated (using ImageJ, Rasband, http://imagej.nih.gov/ij/). Samples were taken using a 1-mL pipette (Finnpipette), at regular 5-mm intervals along the radius of the colony. The samples were then washed in M9 buffer (12.8 g L−1 Na2HPO4, 3 g L−1 KHPO4, 0.5 g L−1 NaCl, 1 g L−1 NH4Cl), diluted to an appropriate dilution to allow colony differentiation and plated to count colony-forming units (CFUs). Total cell numbers on the plate were estimated by scaling up the colony-forming unit counts from the area of the pipette tip.
Analyses and figures were produced on PASW Statistics 18 (SPSS). Significance of treatment on cell density was analysed parametrically using generalized linear models (GLMs). Terms used in the model are defined as the following: ‘Area dispersed’ [response variable], the square-rooted average area covered by bacterial growth after 48 h; ‘Cell Density’ [response variable], the total mean CFUs (log10-transformed); ‘Treatment’ [explanatory variable, factor with three levels], the phage treatment (global, local and phage absent); and, ‘Strain’ [explanatory variable, factor with two levels], the motility variant (disperser or nondisperser). Replicate is treated as a random factor. T-tests were performed between treatment groups within strain to test for differences in cell density.
We first established how motility behaviour was affected by phages. The average area dispersed (± 1 SD) of the motility variants in soft agar (KB 0.6 % w/v agar) in the absence of the phages given 48 h for growth, and dispersal was 8.82 cm2 (± 1.66) for the disperser, and 1.73 cm2 (± 0.12) for the nondisperser (Fig. 1c). Under both global and local phage treatments, the disperser covered a larger area than the nondisperser, with apparent directional movement (chemotaxis) out towards the edge of the plate under the global treatment (Fig. 1a, b; Fig. 2a; GLM: Genotype as main-effect; F1,3 = 249.129, P =0.001). The disperser covered a larger area under the global treatment compared with the local treatment, whereas the nondisperser covered an equivalent area across both (GLM: Genotype*Treatment interaction; F1,3 = 62.06, P =0.004). These results suggest that phages are not qualitatively affecting motility behaviour, that is, dispersers always disperse more. Note that when both strains were grown in hard agar, however, (and thus essentially immobilized), in the presence of phages, there was no difference in area dispersed, suggesting that the motility type expressed by the dispersers on soft agar is likely to be flagella dependent (Fig. 1d; Fig. 2b; t4 = 2.035, P =0.112; Henrichsen, 1972; Harshey, 2003).
We then wanted to determine whether motility conferred a selective advantage in the presence of phages, under conditions where bacteria could escape from local phage populations, and when phages were equally distributed throughout the environment. We compared densities after 48 h growth and dispersal of high (dispersers) and low dispersing (nondispersers) bacteria in the absence of phages, when phages were present in the inoculation site (local) and when phages were present everywhere (global). Crucially, we find that the effect of phage on bacterial density differed depending on dispersal behaviour of the bacteria (Genotype*Treatment interaction F2,4 = 92.783, P <0.001). Pairwise comparisons between treatments show that the disperser does not suffer a reduction in cell density under either the local treatment compared with the global or phage-absent treatment (T-test: Local vs. Global, t6 = 1.912, P =0.116; Local vs. Phage Absent, t5 = 0.611, P =0.568), although there is a slight reduction in cell density in the global compared with the phage-absent treatment (t5 = 3.172, P =0.032). However, nondispersers suffer a large reduction in density in the presence of phages (by more than two orders of magnitude) under both treatment conditions (Global vs. Phage Absent, t5 = 15.887, P <0.001; Local vs. Phage Absent, t5 = 25.713, P <0.001) and do equally poorly in the presence of both phage treatments (Local vs. Global, t6 = 0.795, P =0.497). In other words, the disperser has a higher fitness relative to the nondisperser in the presence (both local and global distribution) compared with in the absence of phages (One-way anova: F1,2 = 221.851, P =0.004; Fig. 3). Note that the disperser also has a higher fitness in the absence of phages (Fig. 3; t4 = 9.574, P =0.001), a result consistent with our previous work under these conditions (Taylor & Buckling, 2010).
We next considered whether reduced density in the presence of phages of the nondisperser is simply the result of increased susceptibility. To test this, we measured density of bacterial cells, exposed to phages, when swimming motility was unlikely to be important, in hard agar (where the cells are immobilized) and shaken liquid (which creates a homogeneous environment which the cells do not need motility to move though). After correcting for multiple comparisons (using the Bonferroni correction), in both cases, there were no differences between total cell density of motility variants (T-test: in liquid broth, t10 = 2.371, P =0.077; in hard agar t4 = 2.035, P =0.112; data not shown). This suggests that the fitness differences noted above are only true when there is a disparity in motility, but not when they are equally motile.
Finally, we checked whether any differences between treatments were due to evolutionary (genetic) changes that occurred during the short experiments. We found no differences in the area of dispersal between bacteria which had and had not been exposed to phages, when grown in the absence of phage between all treatment groups (GLM: F2,4 = 1.204, P =0.390). Moreover, the morphological phenotype (i.e. the spreading pattern) was the same as the ancestral strain. Next, we tested whether the any observed patterns could be due to evolved resistance. All bacteria grown in the absence of phage (control treatment) were susceptible to phage (proportion of resistance clones: disperser = 0.0, nondisperser = 0.0), and there was only a small increase in proportion of resistant clones across local and global treatment groups (Local: disperser = 0.067, nondisperser = 0.167; Global: disperser = 0.067, nondisperser = 0.133), although these differences were not significant between treatment (GLM: F1,2 = 0.250; P = 0.667), or strain (GLM: F1,2 = 1.563; P =0.338).Therefore, any differences in response to phage observed during the experiment appear to be due to phenotypic and not genetic differences.
Our results suggest that within a semi-solid environment, motile bacteria have a fitness advantage over nonmotile bacteria in the presence (vs. the absence) of a parasite (bacteriophage), even when phages are present everywhere. Although the precise mechanism behind this effect is unknown, we can rule out intrinsic differences in susceptibility to phages, or different potential to evolve resistance, because when motility was prevented, there were no differences in fitness between motile and nonmotile mutants. A general (but vague) explanation for our results is that motility creates a host population structure that reduces parasite transmission (Watve & Jog, 1997). However, there are more idiosyncratic features of bacterial–phage interactions that may also be important. First, it may be harder for phages to attach to motile bacteria. Indeed, we have found that phage attachment is reduced when bacterial cultures are shaken very vigorously (Wang et al., 2001). Second, flagella-dependent ‘swarming’ motility, the particular type of motility investigated in the present study, results in bacteria moving together, increasing bacteria density in colonized patches, and such behaviour has been found to confer elevated resistance to antibiotics (Butler et al., 2010). The precise mechanism of elevated resistance has not been determined, but it is thought that the bacteria at the edges of the swarm and at the bottom of the swarm (i.e. those in direct contact with the agar and thus the antagonist) may provide a physical barrier that protects those within the cluster. This selfish group defence strategy has parallels to ‘safety in numbers’, whereby an individual is more likely to survive an attack in a large group because the likelihood is, the job of defence will fall to someone else (Waterman, 1997, 2002; Clutton-Brock et al., 1999a,b; Cant et al., 2001). Moreover, bacteria at high densities are more likely to show reduced replication rates by entering stationary phase, and this physiological state can limit phage proliferation (Dennehy et al., 2007).
Although we focus on the advantages of host motility for hosts in this study, there is a possibility that in a natural setting, this behaviour may ultimately also be advantageous to the parasite. Although motility conferred an increased benefit in the presence vs. the absence of parasites in this study, dispersal also resulted in higher host densities under all treatments, which will ultimately benefit parasite transmission (Anderson & May, 1979). Moreover, motility of infected hosts will allow the parasite to spread to previously uninfected patches of hosts (Rhodes & Anderson, 1996).
Host motility is likely to have important consequences for host–parasite coevolutionary interactions. For example, increased mixing of host and parasite populations, a likely consequence of elevated host motility can accelerate antagonistic host–parasite coevolutionary dynamics (Brockhurst et al., 2003) because of increased selection for resistance and subsequent infectivity. Moreover, motility will increase gene flow between populations, and this can provide significant evolutionary advantages to the species experiencing a greater supply of genetic variation, resulting in local adaptation (Gandon et al., 1996; Morgan et al., 2005). Indeed, whether or not hosts or parasites are locally adapted is likely to in turn influence subsequent selection for motility. Host–parasite mixing can also have important influences on parasite life-history strategies (Frank, 1996), for example whether selection primarily favours horizontal or vertical transmission, which in turn can shape the extent to which coevolutionary interactions are antagonistic or mutualistic (Messenger et al., 1999).
This study also has potential implications for specifics regarding P. aeruginosa's pathogenic life history. ФKZ is known to infect P. aeruginosa strains pathogenic to humans and is therefore attractive for phage therapy research. Phage therapy uses phages as an antibacterial agent for the treatment of pathogenic bacterial infections, and although it is not currently used in western medical practice, its potential applications are an area of intensive current research (Skurnik & Strauch, 2006). There is a large amount of literature that identifies motility as an important virulence factor (Drake & Montie, 1988; Josenhans & Suerbaum, 2002), and it is therefore important to determine whether behavioural responses to phages, such as increased motility, will have correlated consequences for virulence factors, and importantly, to understand what these responses are likely to be in an environment similar to within host conditions. This study suggests that the optimal dispersal strategy for P. aeruginosa in the presence of LPS phages is to always disperse, even if phages are present throughout the environment. This result is only apparent when the bacteria are observed in a structured environment. The study may help to explain diversity of bacterial motility in natural populations that coexist with phages that commonly target motility organelles. More generally, this work highlights the crucial role the parasites can play in the evolution of dispersal and the implications of dispersal for coevolution with respect to genetic structure of host and parasite populations.
We thank the European Research Council and the Leverhulme Trust for funding (AB).