Laura del Carmen Lopez Pascua, Department of Zoology, University of Oxford, Oxford OX1 3PS, UK. Tel.: +44 1865 271100; fax: +44 1865 310447; e-mail: firstname.lastname@example.org
Host–parasite coevolution is believed to influence a range of evolutionary and ecological processes, including population dynamics, evolution of diversity, sexual reproduction and parasite virulence. The impact of coevolution on these processes will depend on its rate, which is likely to be affected by the energy flowing through an ecosystem, or productivity. We addressed how productivity affected rates of coevolution during a coevolutionary arms race between experimental populations of bacteria and their parasitic viruses (phages). As hypothesized, the rate of coevolution between bacterial resistance and phage infectivity increased with increased productivity. This relationship can in part be explained by reduced competitiveness of resistant bacteria in low compared with high productivity environments, leading to weaker selection for resistance in the former. The data further suggest that variation in productivity can generate variation in selection for resistance across landscapes, a result that is crucial to the geographic mosaic theory of coevolution.
Host–parasite antagonistic coevolution, the reciprocal evolution of host defence and parasite counter-defence (Janzen, 1980; Thompson, 1994), is believed to affect a range of ecological and evolutionary processes, including host and parasite population dynamics (Thompson, 1998) and genetic diversity (Buckling & Hodgson, 2007) and the evolution of host sexual reproduction (Hamilton, 1980) and parasite virulence (Bull, 1994). The effect of coevolution on these processes will, however, be dependent on its rate, which is likely to be crucially dependent on the energy flow through an ecosystem (or productivity) (Rosenzweig, 1995). Productivity varies across landscapes (Rosenzweig, 1995) and is already believed to play a major role in determining coevolutionary dynamics (Hochberg & van Baalen, 1998, 2000; Thompson, 2005). However, the relationship between the rate of coevolution and productivity has yet to be explicitly addressed, either theoretically or empirically. Here, we report an experimental evolution study using bacteria and parasitic viruses that investigates this relationship.
Increasing productivity may affect the rate of coevolution in numerous ways. The most obvious effect will be an increase in host and parasite population sizes, which is likely to increase the rate of coevolution for two reasons. First, there will be an increase in the supply of genetic variation on which selection for defence and counter-defence can act. Second, encounter rates between hosts and parasites will be greater, imposing stronger selection for host defence, and hence parasite counter-defence (Hochberg & Holt, 1995; Hochberg & van Baalen, 1998).
The rate of coevolution may be further affected by productivity in cases where coevolutionary dynamics are characterized by ‘arms races’, whereby hosts and parasites evolve to be progressively more resistant and infective respectively. Resistance to parasites is often associated with reduced competitiveness for resources (Zhong et al., 2005; Hasu et al., 2006; Luong & Polak, 2007), and increasing productivity may increase the strength of selection for resistance as a result of reduced competition for resources (Janmaat & Myers, 2005; Kraaijeveld & Godfray, 1997). Increased positive selection acting on resistance mutations will increase their substitution rate, and in turn increase the strength of selection on parasites to overcome host resistance. Assuming a coevolutionary arms race characterized by increased costs with increased resistance (or infectivity), the rate of fixation of subsequent resistance and infectivity mutations will be affected similarly, resulting in accelerated coevolution in productive environments.
The effect of productivity on coevolution has not been explicitly addressed in natural populations because of the timescales required to measure coevolutionary change, and lack of experimental control. These problems can be overcome to some extent by employing experimental microbial systems (Lenski & Levin, 1985; Jessup et al., 2004; Brockhurst et al., 2007) which, thanks to their large population sizes and their short generation times, allow evolution to be observed in real time. Populations can also be stored in suspended animation in the freezer, allowing for direct comparison of ancestors and descendants (Lenski & Levin, 1985; Jessup et al., 2004). The interaction between lytic bacteriophage and bacteria has proved particularly useful for the study of antagonistic coevolution (Lenski & Levin, 1985). Lytic bacteriophages infect their bacterial hosts, multiply and then induce cell lysis to release phage progeny. This imposes very strong selection for bacterial resistance to phages, which consequently imparts strong selection for phage infectivity.
There have been a number of excellent studies addressing the effect of productivity on evolutionary interactions between the bacterium Escherichia coli and associated bacteriophages (T2, T4 and T7) (Bohannan & Lenski, 1997, 1999, 2000a,b; Forde et al., 2004, 2007). However, it is not clear from these studies what impact variation in productivity has on the rate of coevolution. In two of the systems (T2 and T4), coevolution does not occur, as phages seemed unable to evolve to overcome bacterial resistance (Bohannan & Lenski, 2000b). In the third system (T7), coevolution does occur (Chao et al., 1977), but studies to date have been too short term to address the relationship between productivity and rates of coevolution. Resistant mutants (that are the result of spontaneous mutations) increased in frequency most readily at high productivities (Forde et al., 2004, 2007), but the subsequent evolution of phages that can infect resistant bacteria appeared to be maximized at intermediate productivities (Forde et al., 2007).
We addressed how productivity affected coevolution between the common plant-colonizing gram-negative bacterium Pseudomonas fluorescens and its associated lytic DNA phage SBW25Φ2 (Buckling & Rainey, 2002). These organisms undergo persistent coevolution, with multiple cycles of resistance and infectivity evolution in nutrient-rich media (Buckling & Rainey, 2002). Evolved bacteria and phages tend to retain the ability to resist and infect, respectively, many of the previously encountered genotypes, resulting in bacteria and phages evolving increasingly wide resistance and infectivity ranges over time in nutrient-rich media (Buckling & Rainey, 2002). These coevolutionary dynamics are broadly consistent with a multi-locus gene-for-gene model of coevolutionary interactions (Frank, 1996; Sasaki, 2000). Furthermore, as is sometimes observed in other bacteria–phage systems (Bohannan & Lenski, 2000b), there is a competitive cost to the bacteria associated with phage resistance (Buckling et al., 2006), and an increased cost (in terms of reduced growth rates on susceptible bacteria) associated with increasing phage infectivity ranges (Poullain et al., 2008). We coevolved replicate populations of bacteria and phage under a range of productivities for approximately 120 bacterial generations, and determined rates of evolution of bacterial resistance and phage infectivity over time.
Materials and methods
Bacteria and phage were cultured in 6 mL of M9 salt solution supplemented with differing amounts of glycerol and proteose peptone (0.01, 0.1, 0.5 and 1 times the standard concentration for King’s Media B (KB) of 10 g L−1 glycerol and 20 g L−1 of proteose peptone no. 3), in 25 mL of glass universals with loose plastic lids (microcosms) (Kassen et al., 2000). Bacteria were derived from a single P. fluorescens SBW25 (Rainey & Bailey, 1996) clone grown overnight in KB broth, at 28 °C, in an orbital shaker at 200 rpm (0.9 g). We determined that productivity had the expected ecological consequences on host population densities: there was a positive relationship between productivity and the density of bacteria after 48 h in the absence of phages (F1,10 = 34, P < 0.001; n = 3 per media type), with 1× media resulting in an approximately 10-fold greater density than 0.01× media. Bacterial densities were estimated by plating culture onto KB agar and determining the number of colony-forming units (CFUs) after 48 h of growth at 28 °C. Starting bacterial densities were as described below.
Six replicate microcosms for each treatment were inoculated with 108 bacterial cells and 105 clonal phage particles, obtained from a single plaque of a clone of phage SBW25φ2 (Buckling & Rainey, 2002). The bacteria and phage were then allowed to coevolve in a static incubator at 28 °C for a 48-h time period before 1% (60 μL) of the total population was transferred into a fresh microcosm, following vortex mixing to homogenize the culture. This process was repeated for a total of 16 transfers (approximately 120 generations). Every second transfer, samples (600 μL) of the total populations were frozen at −86 °C in 20% v:v glycerol:KB solution, and a sample of phage isolated from bacteria by adding 100 μL of chloroform to 900 μL of the culture, and centrifuging at 13 000 rpm for 3 min (13800 g), to lyse and pellet bacteria.
Coevolution can affect a range of traits, but for simplicity we focus only on bacterial resistance and phage infectivity at the population level.
Evolution can be directly measured by determining the fitness of populations in the selective environment before and after selection. This is more problematic when measuring evolution in a coevolutionary context, because the selective environment is necessarily changing. However, it is possible to estimate the extent of phage infectivity evolution between two time points by measuring the resistance of a bacterial population (see below) from an intermediate time point to the phage populations from before and after. Specifically, we measured the resistance of a population of bacteria to phages from two transfers in the past and two transfers in future. For example, the resistance of bacteria from transfer 4 was determined against phages from transfer 2 and phages from transfer 6. The difference in resistance provides a measure of phage infectivity evolution over this four transfer periods (Brockhurst et al., 2003). The equivalent method was used to measure the evolution of bacterial resistance: the resistance of past and future bacterial populations was determined against the phage population from the intermediate time point. Rates of evolution of infectivity and resistance were estimated every second transfer. Note that we also measured resistance of bacteria to their contemporary and the ancestral phage populations.
Resistance of a bacterial population against a given phage population was determined by streaking 20 independent bacterial colonies across a perpendicular line of phage (20 μL) that had previously been streaked and dried onto a KB agar plate. A colony was defined as resistant if there was no inhibition of growth; otherwise it was defined as sensitive. Population resistances were measured as proportions.
Measuring the effects of productivity on the strength of selection for resistance
We estimated the relative strength of selection for bacterial resistance under different productivities in two ways. First, we simply determined the proportion of bacterial cells that were resistant to the ancestral phages during the course of the experiment. Assuming that resistance evolution is not constrained by a lack of genetic variation, a greater proportion of resistant cells was expected as selection for resistance increased. This measure does not, however, distinguish between increased selection for resistance caused by: (a) increased encounter rates between bacteria and phages; or (b) reduced competition for resources.
Second, to determine if competitive ability of resistant bacteria differed between environments, we carried out competition experiments between evolved and ancestral bacteria in the absence of phages. We isolated the bacteria from treatments 1 and 0.1 from their phage. A 5%VirkonTM (Antec International, Sudbury, England; a commercially available disinfectant)/water solution was made, and added to KB to a concentration of 0.375% VirkonTM. Sixty microlitres of culture was added to 6 mL of the VirkonTM/KB solution in glass universals, and left static for 24 h at 28 °C (Morgan et al., 2005). This procedure left the bacteria viable (although there is between 5% and 40% mortality, Pal et al., 2007; and completely phage free, Morgan et al., 2005). Sixty microlitres of the VirkonTM cultures was added to a fresh static glass universal containing 6 mL of KB, and grown for 1 day at 28 °C, to give a phage-free and VirkonTM-free stock. The potential presence of phage in VirkonTM-treated cultures was tested for by plating out the cultures on to semi-solid agar seeded with P. fluorescens strain SBW25 and incubated at 28 °C for 24 h; plaques would indicate the presence of phage. One of the populations evolved in 0.1× media could not be isolated from the phage (it is unclear why); so, they were not included in these experiments.
We established cultures of the 11 phage-free populations, ancestral SBW25 and an isogenic strain of ancestral SBW25 with a lacZ insert (Bailey et al., 1995), and grew these overnight at 28 °C, shaken at 200 rpm (0.9 g). All cultures reached approximately the same densities; so, it was assumed they were in comparable physiological states (Lenski et al., 1991). Equal numbers of evolved bacteria and SBW25::lacZ, and the SBW25 and SBW25::lacZ (approximately 5 × 107; starting densities were directly measured as determined below) were inoculated into fresh media and grown in a static incubator for 48 h; the duration of a transfer. Bacteria were then plated on Luria–Bertani (LB) agar supplemented with X-gal (1 μL of 20 mg mL−1 solution per mL of LB) & IPTG (1 μL of 100 mM IPTG per mL of LB) to determine the outcome of the competition. The SBW25::lacZ strain formed dark blue colonies, whereas the wild-type strain produced yellow colonies. Relative fitness (W) was calculated from the ratio of the estimated malthusian parameters (m) of the evolved and SBW25::lacZ strains, m = ln(Nf/N0), where N0 is the initial and Nf is the final density (Lenski et al., 1991). All competitions were replicated three times, and analysed values were the means of these replicates. We needed to compare the relative fitness of our evolved lines to the ancestor between environments, but unfortunately we found that the relative fitness of SBW25::lacZ varied between environments. We therefore obtained an estimate of the selection coefficient (S) of the evolved lines (relative to the ancestor) by subtracting W of SBW25 in a given environment from the W values of each of the evolved line fitness in the same environment, i.e.
Note that selection coefficients are typically calculated between evolved and ancestral genotypes in direct competition (i.e. within the same tube) from (Lenski et al., 1991)
The impact of productivity on: (1) the rate of change of phage infectivity; (2) the rate of change of bacterial resistance; (3) resistance of bacteria to the ancestral phage; and (4) resistance of bacteria to their sympatric phage populations was analysed using mixed linear model (REML) repeated measures analysis in GenStat (v.8.1; VSN International, Hemel Hempstead). Replicate was fitted as a random factor, productivity fitted as a fixed factor and time as a fixed effect covariate. Uniform correlations within replicates across time were assumed. The proportions of bacteria that were resistant to ancestral and sympatric phages were angular transformed to meet the assumptions of normality and homoscedasticity. Significance of fixed effects was determined from Wald statistics.
Competition experiments (values of S) were analysed as a GLM (in GenStat 8.1) with test and selective environment fitted as factors, and genotype (random factor) nested in selection environment. To determine if there was an absolute cost of resistance relative to the ancestor, we also carried out 1-sample t-tests of the relative fitness measures against the control relative fitness within each environment.
The aim of this study was to address how productivity affected host–parasite antagonistic coevolution in experimental populations of bacteria and phage. We can infer coevolution between bacterial resistance and phage infectivity if there are recurrent short-term increases in phage infectivity and bacterial resistance. This can be measured by determining the change in infectivity of phage populations (to contemporary bacteria) between two transfers in the past and two transfers in the future; and changes in bacterial resistance (to contemporary phage populations) over the same timescales. The coevolutionary dynamics of the most rapidly coevolving population (from the 1× treatment) is shown in Fig. 1.
We determined if coevolution occurred in the different treatments by analysing changes in resistance and infectivity over time in separate analyses. Short-term increases in both resistance and infectivity occurred in the 0.1×, 0.5× and 1× treatments (Fig. 2; means for each treatment were greater than 0; P < 0.01 in all cases from t-tests), strongly suggesting that coevolution between resistance and infectivity occurred. Despite a rapid increase in bacterial resistance to the ancestral phage by the second transfer in the 0.01× treatment, mean (over time) changes in infectivity and resistance were not significantly greater than 0. This suggests little coevolutionary change occurred in the lowest productivity media, at least in terms of reciprocal increases in resistance and infectivity. Note that some populations showed occasional decreases in resistance and infectivity (as indicated by negative values in Fig. 2), but it is difficult to determine if these occasional results are typically the result of sampling error or selection.
We next determined how productivity affected the rate of coevolution. As expected, we found that increased productivity increased the rate of phage infectivity (Fig. 2a; = 11.23; P < 0.001) and bacterial resistance (Fig. 2b, = 7.12; P < 0.01) evolution. There was also a tendency for the rate of evolution of both infectivity (Fig. 2a; = 20.31, P < 0.001) and resistance (Fig. 2b; = 16.56, P < 0.001) to decrease over time. Furthermore, we found that the mean rates over time of increases in resistance and infectivity showed a highly significant positive correlation (Fig. 3; Pearson’s correlation: r = 0.81, n = 25, P < 0.001): rapid changes in phage infectivity were associated with rapid changes in bacterial resistance. This provides further evidence for the strong reciprocal selection experienced by bacteria and phages.
We also addressed whether different productivities created any asymmetries in the arms race between bacteria and phages. Specifically, we determined how resistance of bacteria to their sympatric phages varied across treatments. We observed no difference in levels of sympatric resistance across productivity treatments (Fig. 4a; = 1.21, P > 0.5). There was a tendency for sympatric resistance to increase over time (Fig. 4a; = 44.26, P < 0.001), although this pattern is not typical of other studies on this system (Buckling & Rainey, 2002; Brockhurst et al., 2003; Buckling et al., 2006).
We speculated that variation between environments in strength of selection for bacterial resistance may have contributed to the positive relationship between rates of coevolution and productivity. To address this possibility, we determined the frequency of bacteria that were resistant to the ancestral bacteria during the course of the experiment. Resistance to the ancestral phage had increased from not detectably greater than 0% to between 70% and 95% in all populations by the second transfer (Fig. 4b), suggesting that resistance evolution was not constrained by a lack of genetic variation in any of the media. However, the frequency of cells resistant to the ancestral phage throughout the experiment showed a significant positive relationship with productivity (Fig. 4b; = 50.06; P < 0.001). These data suggest that selection for resistance to the ancestral phage increased with increasing productivity.
The above result suggests that selection for bacterial resistance is greater in productive environments, but it is not clear whether this is solely due to increased encounter rates between bacteria and phages, or if selection against resistance is reduced in productive environments because of reduced resource competition. We addressed this latter possibility by isolating bacteria from the 0.1× and 1× productivity media (evolution media) from transfer 8, and removing any associated phage populations. These phage-free populations were then competed against a marked competitor in 0.1× and 1× media (test media). To first address whether the evolved populations showed a cost of resistance, we carried out one-sample t-tests of the fitness of evolved lines vs. the ancestor in each environments: there was a significant cost in 1× media (approximately 5% reduction in competitive ability; t = 1.95, P < 0.05), and a highly significant cost in 0.1× media (approximately 25% reduction in competitive ability; t = 5.11, P < 0.001). We carried out a generalized linear model to further explore these costs. We found no evidence of media-specific adaptation (Fig. 5; interaction between test and evolution media: F1,9 = 0.6, P > 0.2), and furthermore, costs of resistance (relative to the ancestor) did not differ between bacteria evolved in the different media (Fig. 5; main effect of evolution media: F1,10 = 0.24, P > 0.2, with interaction removed from model). However, there was significantly greater selection against resistance when bacteria were competed in the 0.1× media compared with 1× media (Fig. 5; main effect of test media: F1,10 = 16.42, P = 0.002).
In this study, we investigated the effect of environmental productivity on the rate of coevolution between experimental populations of bacteria and phages. We found that the rate of change of both bacterial resistance and phage infectivity increased with increasing productivity, strongly suggesting that the rate of coevolution increased with increasing productivity. Further support for this conjecture is the positive correlation between rates of change of resistance and infectivity.
Three mechanisms are likely to have contributed to the increased rates of coevolution with increased productivity. Increased productivity may have increased population sizes of bacteria and phages, which would first, increase the supply of genetic variation on which selection acted; and, second, increase the strength of selection for resistance by increasing the encounter rate of bacteria and phages (increased selection for resistance would have increased the substitution rate of resistance mutations, and hence coevolution). Third, selection for resistance may be further increased in productive environments because of reduced competition for resources with conspecifics. This latter mechanism relies on resistant bacteria showing reduced competitiveness, which is apparent in both this and other (Buckling et al., 2006) studies.
Unfortunately, we were not able to reliably estimate phage densities in this study. We attempted to estimate phage densities by counting the number of plaque-forming units on a bacterial lawn; each round plaque is formed when a single phage establishes a successful infection and subsequently replicates. These plaques were sometimes extremely small in some evolved phage populations, making counting them too unreliable. This reduction in plaque size formed by evolved phages is consistent with a recent study that demonstrates a reduction in growth rate on the ancestral bacteria of phages that have evolved to infect a wide range of bacterial genotypes (Poullain et al., 2008). Note, however, that it is likely that population sizes of bacteria or phage or both increased with productivity: in the absence of phages, increased productivity increased bacterial population density (see Materials and methods).
We did, however, find evidence that the strength of selection for resistance increased with increased productivity. First, the proportion of bacteria resistant to the ancestor increased with productivity. This result demonstrates increased selection for resistance (to the ancestral phage, at least), but does not distinguish between two possible ecological mechanisms: increased encounter rates and reduced competition for resources. Second, competition experiments in the absence of phages explicitly demonstrated reduced competitiveness in less productive environments. We do not know the mechanistic basis for this reduction in competitiveness, but a plausible explanation is that resistance to phages involves alteration of nutrient uptake, potentially reducing the efficiency of nutrient uptake, particularly in nutrient-poor environments. Note that the demonstration of reduced competitiveness in less productive environments does not rule out the importance of changes in encounter rates. Indeed, in a previous study on this experimental system that reported an increase in the rate of coevolution as a result of increased population mixing by occasional shaking of test tubes, increased encounter rates were the most plausible mechanism (Brockhurst et al., 2003).
We also addressed whether different productivities created any asymmetries in the coevolutionary arms race between bacteria and phages. Specifically, we determined if levels of bacterial resistance to sympatric phages varied between treatments; increasing resistance would indicate increasing relative evolutionary success of the bacteria. We, however, found no difference between productivity treatments in the proportion of bacteria resistant to their sympatric phage populations. This result is broadly consistent with a theoretical study that explicitly considered coevolution between host resistance and parasite infectivity: largely parallel increases in costly resistance and infectivity were predicted with increasing productivity (Hochberg & van Baalen, 1998), although model details can alter this to some extent. Furthermore, a study of natural populations of plants and their fungal parasites suggest a correlation between increased host resistance and the evolution of more virulent parasites (Thrall & Burdon, 2003). These results can be intuitively explained by a change in selection for resistance being accompanied by a reciprocal change of selection on parasite infectivity. By apparent contrast, other theoretical (e.g. Leibold, 1996) and empirical (e.g. Bohannan & Lenski, 2000b) work predicts an increase in the frequency of a costly resistance mutation, and hence higher sympatric resistance, with increasing productivity. This is because increased productivity is likely to increase the benefit of host resistance through increased encounter rates. However, these models and experiments do not involve an evolutionary response of the parasite to the resistant host; in other words, there is no coevolution.
Our results also provide some support for the geographic mosaic theory of coevolution, which suggests that the commonly observed mismatch of phenotypes between locally interacting hosts and parasites (or any coevolving species) is the result of gene flow across a ‘selection mosaic’ (Thompson, 2005). Selection mosaics specifically refer to variation in selection acting on different genotypes across landscapes (Thompson, 2005; Gomulkiewicz et al., 2007); i.e. a genotype might be selected for in one part of the landscape, but selected against in another. Productivity commonly varies across landscapes (Rosenzweig, 1995), and our experiments demonstrate that such variation may create selection mosaics: specifically, selection for bacterial resistance is stronger in more productive environments. Migration across a productivity gradient may therefore result in bacteria and phage in environments with locally maladapted levels of resistance and infectivity.
In summary, we have shown that increasing environmental productivity accelerates the rate of coevolution between experimental populations of bacteria and phages. This relationship can in part be explained by reduced competitiveness of resistant bacteria in low compared with high productivity environments, leading to a weaker selection for resistance in the former. These results have two important implications. First, the impact of coevolution on other ecological and evolutionary processes is likely to be greater in more productive environments. Second, productivity gradients are likely to create selection mosaics in natural populations.
This work was funded by the Leverhulme Trust & the Royal Society. We thank Mike Brockhurst, Ben Ridenhour and Sylvain Gandon for useful discussions and comments on the manuscript.