Relative number of generations of hosts and parasites does not influence parasite local adaptation in coevolving populations of bacteria and phages

Authors


Andrew D. Morgan, Max Planck Institute for Developmental Biology, Spemannstrasse 35, Tuebingen 72076, Germany.
Tel.: +49 7071 601840; fax: +49 7071 601305; e-mail: andrew.morgan@tuebingen.mpg.de

Abstract

A potential consequence of host–parasite coevolution in spatially structured populations is parasite local adaptation: local parasites perform better than foreign parasites on their local host populations. It has been suggested that the generally shorter generation times of parasites compared with their hosts contributes to parasites, rather than hosts, being locally adapted. We tested the hypothesis that relative generation times of hosts and parasites affect local adaptation of hosts and parasites, using the bacterium Pseudomonas fluorescens and a lytic phage as host and parasite, respectively. Generation times were not directly manipulated, but instead one of the coevolving partners was regularly removed and replaced with a population from an earlier time point. Thus, one partner underwent more generations than the other. Manipulations were carried out at both early and later periods of coevolutionary interactions. At early stages of coevolution, host and parasites that underwent relatively more generations displayed higher levels of resistance and infectivity, respectively. However, the relative number of generations that bacteria and phages underwent did not change the level of local adaptation relative to control populations. This is likely because generalist hosts and parasites are favoured during early stages of coevolution, preventing local adaptation. By contrast, at later stages manipulations had no effect on either average levels of resistance or infectivity, or alter the level of local adaptation relative to the controls, possibly because traits other than resistance and infectivity were under strong selection. Taken together, these data suggest that the relative generation times of hosts and parasites may not be an important determinant of local adaptation in this system.

Introduction

Antagonistic coevolution between hosts and parasites, the reciprocal evolution of host defence and parasite counter-defence, is ubiquitous (Thompson, 1994), and can result in parasites becoming adapted to their local host populations (Gandon et al., 1996; Lively, 1999; Gandon & Michalakis, 2002; Gandon, 2002). Parasite local adaptation has important implications. For example, parasites will be better able to infect host populations that they are most likely to encounter, and local adaptation will contribute to the maintenance of genetic diversity in both host and parasite populations. Parasite local adaptation is clearly undesirable in the context of human and agricultural pathogens, but may be of benefit when pathogens are used as biocontrol agents, or even in natural populations where maintaining genetic diversity is of primary importance.

Assuming host–parasite specificity allows local adaptation to occur (Lively, 1999), parasite populations will on average be locally adapted if they evolve faster than their hosts (Gandon et al., 1996; Gandon & Michalakis, 2002; Gandon, 2002). Likewise, hosts will on average be locally adapted if they evolve faster than parasites. Theory suggests two factors are important for determining whether hosts or parasites are locally adapted. First, the relative amount of within population genetic variation of host and parasite population: the greater the genetic variation, the more rapidly a population can adapt. Within-population genetic variation will be determined by population size, mutation rates (Gandon et al., 1996; Gandon & Michalakis, 2002) and migration rates (Gandon et al., 1996; Lively, 1999; Gandon, 2002; Gandon & Michalakis, 2002; Morgan et al., 2005). Secondly, the relative generation time of host and parasite (Lively, 1999; Gandon & Michalakis, 2002). Organisms with short generation times can adapt rapidly, but this may itself purge genetic variation from the population, providing it is not replaced by migration or mutation, retarding subsequent adaptation (Gandon, 2002). Here, we address the role of relative generation times of hosts and parasites in determining local adaptation.

Experiments in natural populations have demonstrated local adaptation of parasites (Parker, 1985; Lively, 1989; Ebert, 1994; Lively & Dybdahl, 2000), host local adaptation (Kaltz et al., 1999) and no local adaptation at all (Parker, 1989). In all cases it is difficult to demonstrate the causes: for example both mutation and generation times of parasites may be high and the relative migration rates of hosts and parasites may be difficult to determine.

Microbes can be useful for the study of evolutionary and ecological processes. Large population sizes and short generation times allow rapid observable evolution (Lenski et al., 1991). Microbes can be stored in a freezer in ‘suspended animation’ and coevolutionary change can be measured through time. In such systems variables can be kept constant, and independently manipulated; and experiments can be easily replicated. Our study system is the bacterium Pseudomonas fluorescens SBW25 (Rainey & Bailey, 1996) and the parasitic bacteriophage SBW25Φ2 (Buckling & Rainey, 2002). We have previously demonstrated their reciprocal evolution of bacterial resistance and phage infectivity (Buckling & Rainey, 2002; Brockhurst et al., 2003), and increasing phage migration rates increases phage local adaptation (Morgan et al., 2005). Directly manipulating relative generation time is unfortunately not possible in this system: phage replication rates depend on the replication rate of the bacteria (although we do know to what degree in this system) due to its utilization of the bacterial replication machinery (Goldberg et al., 1994). Instead, the evolution of one partner was prevented for a period, before allowing a period of coevolution. This resulted in one of the coevolving partners going through more generations over the duration of the experiment: a situation analogous to direct manipulation of generation time

Materials and methods

Generation of divergent populations

Six replicate populations were initiated from approximately 108 cells of isogenic P. fluorescens SBW25 and approximately 105 particles of SBW25Φ2 and grown for 48 h at 28 °C in 6 mL of King's Media B in static 30 mL glass universals with loose lids. After the 48 h had elapsed, 60 μL (1%) of culture was transferred to a fresh microcosm. In Experiments 1a and 1b, populations were transferred for a total of 12 transfers (c. 80 generations); and in Experiments 2a and 2b populations were transferred for a total of 73 transfers (c. 500 generations). Note that the transfer regime is unlikely to create extreme bottlenecks, with an average of approximately 104 and 107 phages and bacteria, respectively, transferred.

Separation of bacteria and phages

To isolate phages from bacteria, 100 μL of chloroform was added to 900 μL of culture and vortexed to lyse the bacterial cells, then spun at 11 356 g for 2 min to pellet the bacterial debris, leaving a suspension of phages in the supernatant. To isolate bacteria from phages a 5% solution of Virkon (a commercially available disinfectant) in water was made. This was added to 6 mL of Kings Media B in 30 mL glass universals to make a final concentration of 0.37% Virkon. Sixty microlitres of culture was added to these microcosms and grown up static at 28 °C for 24 h with loosened caps. Sixty microlitres of culture was transferred to 6 mL of King's Media B in static 30 mL glass universals and grown for a further 24 h at 28 °C (Morgan et al., 2005). This left a phage and Virkon free stock of bacteria. The presence of phages in these cultures was periodically checked for by plating out on to semi-solid King's Agar B seeded with P. fluorescens SBW25 and incubated for 24 h at 28 °C. The presence of plaques would indicate the presence of phages, but in all cases the test proved negative.

Manipulating generation time

We carried out four separate experiments. In the first two (1a and 1b) the starting populations were diverged for approximately 80 generations, and in the second two for approximately 500 generations. We allowed the starting populations to diverge for differing amounts of time, as this may affect the conditions for local adaptation.

Experiments 1a and 2a

Experiments 1a and 2a consisted of three treatments: (i) control, (ii) phages have fewer generations (analogous to phages having a longer generation time, or the bacteria having a shorter generation time) and (iii) bacteria have fewer generations (analogous to the bacteria having a longer generation time, or the phages having a shorter generation time). Sixty microlitres of each diverged culture was used to seed three new microcosms per diverged culture, one for each treatment (a total of 18 microcosms, six replicates per treatment, three treatments in both Experiments 1a and 2a). The start of the experiment was defined as T0. After 48 h, bacteria and phages were isolated from each other (as described above), and transferred to a fresh microcosm (T1). In the control treatments 60 μL of isolated bacteria and 60 μL of isolated phages were added to the fresh microcosm. In the ‘phages have fewer generations treatment’, 60 μL of bacteria were added to a fresh microcosm, and the isolated phages were discarded. Instead, 60 μL of T0 phages were added to the microcosm (this occurred at every transfer). Similarly in the ‘bacteria have fewer generations’ treatment, 60 μL of isolated phages were used to inoculate a fresh microcosm; the isolated bacteria were discarded, and T0 bacteria were added to the microcosm (Table 1). This was continued for a total of 12 experimental transfers. Cultures were frozen every four transfers at −85 °C in 20% glycerol. Phages were isolated every four transfers and stored at 4 °C for the local adaptation assays.

Table 1.   Table of transfer regimes showing the effective transfer number of each organism added at each transfer.
 T0T1T2T3T4T5T6T7T8T9T10T11T12
  1. The numbers in subscript after T indicates the actual transfer number. The number is subscript after B and P is their effective transfer number.

  2. T, transfer; B, bacteria; P, phage; fewer gens, fewer generations.

ControlP0, B0P1, B1P2, B2P3, B3P4, B4P5, B5P6, B6P7, B7P8, B8P9, B9P10, B10P11, B11Stop
1a & 2a (P fewer gens)P0, B0P0, B1P0, B2P0, B3P0, B4P0, B5P0, B6P0, B7P0, B8P0, B9P0, B10P0, B11Stop
1a & 2a (B fewer gens)P0, B0P1, B0P2, B0P3, B0P4, B0P5, B0P6, B0P7, B0P8, B0P9, B0P10, B0P11, B0Stop
1b & 2b (B fewer gens)P0, B0P0, B1P0, B2P1, B3P1, B4P1, B5P2, B6P2, B7P2, B8P3, B9P3, B10P3, B11Stop
1b & 2b (P fewer gens)P0, B0P1, B0P2, B0P3, B1P4, B1P5, B1P6, B2P7, B2P8, B2P9, B3P10, B3P11, B3Stop

Experiments 1b and 2b

These experiments were similar to the above, but we allowed intermittent periods of coevolution between periods where only one species evolved. The experimental regime was identical to the same treatments in Experiments 1a and 2a, except the in the ‘bacteria have fewer generations’, and ‘phages have fewer generations’ treatments, every third transfer the bacteria and phages were allowed to coevolve together and were not replaced by either the ancestral bacteria or phages. For the two transfers after the round of coevolution: either the bacteria (in the ‘bacteria have fewer generations’ treatment), or the phages (in the ‘phages have fewer generations’ treatment) are replaced by the bacteria, or phages, respectively from the transfer before the previous round of coevolution. For example if the bacteria have fewer generations at T1 and T2 the bacteria are discarded and replaced with T0 bacteria, at T3 the bacteria is allowed to coevolve with the phages, at transfers T4 and T5 the bacteria is discarded and replaced with the bacteria from T3 (the transfer before they coevolved together). See Table 1. This was continued for a total of 12 experimental transfers. Cultures were frozen every four transfers at −85 °C in 20% glycerol. Phages were isolated every four transfers and stored at 4 °C for the local adaptation assays.

Local adaptation assay

Local adaptation was measured at various points in time: adaptation to a new host genotype is not instantaneous and there will inevitably be lags in local adaptation (Morand et al., 1996; Kaltz & Shykoff, 1998; Gandon & Michalakis, 2002). Local adaptation was assessed in terms of phage infectivity and parasite resistance. This was achieved by isolating 20 independent bacterial colonies from each replicate and streaking them across a 20 μL line of phages that had been previously dried onto the plate (Buckling & Rainey, 2002). If there was any inhibition of the bacteria it was classed as sensitive, if not, it was classed as resistant. Within a treatment, bacteria from each replicate was streaked across the phages from all replicates so there was a total of 36 (6 × 6) bacteria-phage pair wise interactions within a treatment (Morgan et al., 2005). Local adaptation was measured in all experiments at transfers 4, 8 and 12 of the experimental manipulation. Note that unless otherwise stated, bacteria–phage interactions were measured between contemporary populations of bacteria and phages, i.e. the population of bacteria or phages they had just previously coevolved with, before they were frozen.

Statistical analyses

Local adaptation for each replicate was calculated by subtracting the performance of foreign parasites from the performance of local parasites (parasite performance is the proportion of bacterial colonies sensitive to phages [phage infectivity] (Kawecki & Ebert, 2004)). A positive value (high infectivity of local phages, low infectivity of foreign phages) indicates parasite local adaptation. A negative value (low infecivity of local phages, high infectivity of foreign phages) indicates parasite local maladaptation (Morgan et al., 2005). Note that parasite maladaptation does not necessarily imply host local adaptation when considering a single population, but when averaged across all populations they are equal to each other (Morgan et al., 2005). These values were averaged across the three time points (T4, T8 and T12) and analysed as General Linear Models using Minitab, fitting treatment as a three level factor (control, phages evolve slower, bacteria evolve slower) and line as a six-level factor. Line was treated as a random factor, although the lack of either nesting or replication for the line-by-treatment interaction in the study, means that error MS is used as the denominator for calculating F-ratios in all cases (Sokal & Rohlf 1995). In other words, line was treated in the analyses as a fixed effect. Note that Experiment 1a was terminated after four transfers due to extreme results (see below), hence only a single time point was used in this analysis.

The treatments may affect phage infectivity to local and foreign hosts as well as explicitly affecting local adaptation. We independently analysed sympatric and allopatric infectivity averaged across time in General Linear Models, fitting treatment as a three level factor and starting sympatric or allopatric resistance as a covariate (Morgan et al., 2005).

Results

Results are summarized in Table 2. In Experiment 1a, populations diverged for 12 experimental transfers (approximately 80 generations) prior to experimental treatments. This experiment was continued for only four subsequent transfers, due to extreme results. Average phage infectivity was almost 100% in the ‘bacteria have fewer generations’, 10% in the ‘phages have fewer generations’, and at an intermediate level (c. 50%) in control treatments, in terms of both sympatric (F2,10 = 14.51, P < 0.001, Fig. 1a) and allopatric infectivity (F2,10 = 35.08, P < 0.001). Local adaptation was however not affected by the treatments (F2,10 = 0.05, P = 0.95, Fig. 1b). Line was not significant (P > 0.2) in all cases.

Table 2.   Summary of results.
ExperimentLocal adaptationSympatric infectivityAllopatric infectivity
  1. A summary of the statistical tests. Sig indicates a significant result P < 0.05, N/S indicates a non-significant result P > 0.05. A significant result is a significant effect of treatment, rather than a significant level of local adaptation, sympatric infectivity, or allopatric infectivity. Experiment 1a, starting populations diverged for c. 80 generations, the organism with fewer generations replaced with T0 at each transfer. Experiment 1b, starting populations diverged for c. 80 generations, the organism with the fewer generations allowed to coevolve for one in three transfers. Experiment 2a, starting populations diverged for c. 500 generations, organism with fewer generations replaced with T0 at each transfer. Experiment 2a (tested on T0), as Experiment 2a, but faster evolving bacteria and phage streaked against T0 phage and bacteria respectively. Experiment 2b starting populations diverged for c. 500 generations, slower generating organism allowed to coevolve for one in three transfers.

1aN/SSigSig
1bN/SSigSig
2aN/SN/SN/S
2a (tested on T0)N/SN/SN/S
2bN/SN/SN/S
Figure 1.

 (a) Mean phage infectivity, Experiment 1a. Initial populations diverged for 12 transfers (c. 80 generations), organism with fewer generations replaced with T0 at each transfer. Shown are the sympatric infectivity (black bars, proportion of sensitive bacteria to sympatric phages) and allopatric infectivity (grey bars, average proportion of bacteria sensitive to the five allopatric phage populations) of phages from three different treatments, measured at T4 only. Results are mean ± SEM across replicates. Controls = equal generation rates of bacteria and phages, P Fewer Gens = phages have fewer generations than bacteria, B Fewer Gens = bacteria have fewer generations than the phages. (b) Mean phage local adaptation, Experiment 1a. Averaged across six replicates, measured at T4 only. Results are mean ± SEM across replicates.

The treatments in Experiment 1a were too extreme to favour the occurrence of local adaptation, hence the experiment was repeated, but allowing the organism with fewer generations one round of coevolution every third transfer (Table 1; Experiment 1b). The effect of treatment on sympatric infectivity showed a similar pattern as the previous experiment (phage infectivity was highest in ‘bacteria have fewer generations’ treatment, lowest in ‘phages have fewer generations treatment’, and control populations at an intermediate level (F2,10 = 7.06, P < 0.05, Fig. 2a). Similar results were also found for the effect of treatment on allopatric infectivity (F2,10 = 4.92, P < 0.05, Fig. 2a). There was no significant effect of treatment (F2,10 = 0.13, P = 0.88, Fig. 2b) on local adaptation. Line was not significant (P > 0.2) in all cases.

Figure 2.

 (a) Mean phage infectivity, Experiment 1b. Initial populations diverged for 12 transfers (c. 80 generations), slower evolving organism allowed to coevolve for one in three transfers. Figure legends as Fig. 1a, except sympatric and allopatric infectivities are averaged over six replicates and three time points (T4, T8 and T12). Results are mean ± SEM across replicates. (b) Mean phage local adaptation, Experiment 1b. Figure legends as Fig. 1b except local adaptation averaged across six replicates, and three time points (T4, T8 and T12). Results are mean ± SEM across replicates.

We hypothesized that the absence of local adaptation in the above experiments was because populations had not diverged sufficiently. The early stages of coevolution in the system are consistent with directional gene-for-gene dynamics, which favours the evolution of generalists (Parker, 1994) and hence there is no specialization that is essential for local adaptation (Gandon et al., 1996; Lively, 1999). In Experiments 2a and 2b the starting populations were allowed to diverge for approximately 500 generations. Local adaptation has been observed in this system after 350 bacterial generations (Buckling & Rainey, 2002; Morgan et al., 2005), suggesting generalist evolution is constrained. Experiment 2a had an identical experimental regime to Experiment 1a: at each transfer the slower evolving organism was replaced by the ancestral (T0) species. There was no effect of treatment on either sympatric infectivity (F2,10 = 0.59, P = 0.57; Fig. 3a), or allopatric infectivity (F2,10 = 0.11, P = 0.9; Fig. 3a). There was no significant effect of treatment on local adaptation (F2,10 = 0.37, P = 0.7, Fig. 3b). Line was significant for allopatric infectivity and local adaptation (P < 0.01), but not for sympatric infectivity (P = 0.2).

Figure 3.

 (a) Mean phage infectivity, Experiment 2a. Initial populations diverged for 73 transfers (c. 500 generations), the organism with fewer generations replaced with T0 at each transfer. Figure legends as Fig. 2a. (b) Mean phage local adaptation, Experiment 2a. Figure legends as Fig. 2b.

In Experiment 2a, we also determined the infectivity of the T0 phages to the bacteria from the ‘phages have fewer generations’ treatment at transfers 4, 8 and 12, and the infectivity of the phages from the ‘bacteria have fewer generations’ treatment to the T0 bacteria at transfers 4, 8 and 12. As the bacteria from the ‘phages evolve slower’ treatments had constantly evolved on these T0 phages; and the phages from the ‘bacteria have fewer generations’ were constantly evolved on the T0 bacteria, we hypothesized that such interactions may provide the clearest indication of local adaptation. Surprisingly, the nonevolving organism did not appear to suffer an evolutionary disadvantage: there was no significant effect of treatment upon sympatric (F2,10 = 2.4, P = 0.14, Fig. 4a) or allopatric infectivity (F2,10 = 0.93, P = 0.42, Fig. 4a); line did not have any significant effects (P = 0.02 allo & 0.16, respectively). Given that treatments had no significant impact on phage infectivity, it is not unexpected that treatment had no significant effect upon local adaptation (F2,10 = 0.42, P = 0.67, Fig. 4b). Line was significant for allopatric infectivity and local adaptation (P < 0.05 &<0.01, respectively), but not for sympatric infectivity (P = 0.15).

Figure 4.

 (a) Mean phage infectivity, Experiment 2a, on T0 bacteria and phages. Initial populations diverged for 73 transfers (c. 500 generations), the organism with fewer generations replaced with T0 at each transfer. Figure legends as Fig. 2a. (b) Mean phage local adaptation, Experiment 2a, on T0 bacteria and phages. Figure legends as Fig. 2b.

In Experiment 2b, we allowed the organism with fewer generations to coevolve for one in every three transfers to compare with Experiment 1b. Treatment did not have any significant effects upon any of the hypothesized outcomes in this experiment: (i) sympatric infectivity (F2,10 = 0.81, P = 0.47, Fig. 5a), (ii) allopatric infectivity (F2,10 = 0.75, P = 0.5, Fig. 5a) and (iii) local adaptation (F2,10 = 0.32, P = 0.74, Fig. 5b). Line was significant for sympatric infectivity, but not allopatric infectivity or local adaptation (P < 0.05, = 0.36 & = 0.06 respectively).

Figure 5.

 (a) Mean phage infectivity, Experiment 2b. Initial populations diverged for 73 transfers (c. 500 generations), the organism with fewer generations allowed to coevolve for one in three transfers. Figure legends as Fig. 2a. (b) Mean phage local adaptation, Experiment 2b. Figure legends as Fig. 2b.

To test whether the dynamics of the system did indeed change to becoming more favourable for local adaptation at later stages of coevolution, levels of parasite maladaptation were tested at both early and late stages of coevolution. In this system parasites have a tendency to be locally maladapted at later stages of coevolution (Buckling & Rainey, 2002; Morgan et al., 2005). Average levels of local maladaptation through time (at T0, T4, T8 and T12) were compared with zero in control populations of Experiments 1b (starting populations diverged for c. 80 generations) and 2b (starting populations diverged for c. 500 generations). Levels of parasite local maladaptation in control populations of 1b were not significantly different to zero (one-sample t-test, t = 0.1, n = 6, P > 0.5). Parasites were locally maladapted at later stages of coevolution in control populations of Experiment 2b (one-sample t-test, t = 2.53, n = 6, P = 0.05).

Discussion

Here we investigated the impact of the relative number of generations experienced by bacteria and phages during coevolutionary interactions on their levels of local adaptation. Manipulating the relative number of generations experienced by bacteria and phages had no impact on local adaptation in any of the four experiments. However, decreasing the relative numbers of generations experienced by bacteria and phages increased average phage infectivity and bacterial resistance, respectively, when populations had been allowed to diverge for approximately 80 generations prior to manipulations. No effects on infectivity or resistance were detected when treatments were carried out after populations had diverged for approximately 500 generations. These results suggest that the relative generation time of hosts and parasites are unlikely to be important determinants of local adaptation in this system.

Theoretical studies suggest that the potentially faster rate of adaptation resulting from shorter generation times may be counteracted by the resultant loss of genetic variation (Gandon & Michalakis, 2002). As such, shorter relative generation times may not provide a net evolutionary advantage in a coevolutionary arms race, and hence not influence the probability of local adaptation. This mechanism is however unlikely to be responsible for a lack of effect of relative generation time on local adaptation in any of our experiments. First, the experiments carried out after populations had been coevolving for a short time (80 generations) demonstrate that increasing the relative numbers of generation experienced by one of the species can provide an evolutionary advantage, but still did not affect local adaptation. Secondly, in one of the experiments carried out at later stages of coevolution, either the bacteria or phages were not allowed to evolve at all (Experiment 2b). Any loss of genetic diversity in the evolving bacteria or phage populations should then not have a negative consequence on resistance or infectivity, respectively, because the other species is being held constant. The lack of purging of genetic diversity is probably the result of large bacteria and phage population sizes (an average of approximately 109 and 106 respectively).

A likely explanation for the lack of an effect of the generation time manipulations on local adaptation in the early stages of coevolution, is simply that conditions were not favourable for local adaptation. The dynamics of coevolution during the early stages are consistent with Gene for Gene Model of coevolution (GFGM) (Thompson & Burdon, 1992; Parker, 1994; Sasaki, 2000), where there are variable levels of specificity among the different host and parasite genotypes, allowing for the evolution of generalists. Under a pure GFGM the spread of universally resistant hosts and universally infective parasites prevents the emergence of local adaptation. The degree of host–parasite specificity can however be viewed as continuum (Agrawal & Lively, 2002; Morgan et al., 2005), with GFGM at one extreme and a model of equal host–parasite specificity, which allows only specialist genotypes (Matching Alleles Model; MAM) (Hamilton, 1980; Nee, 1989; Frank, 1994), at the other. As coevolution proceeds, the coevolutionary dynamics increasingly resemble a MAM, as is apparent from evidence of local adaptation in previous studies (Buckling & Rainey, 2002; Morgan et al., 2005). This shift towards MAM is likely to result from elevated costs of bacterial resistance and phage infectivity (Buckling & Rainey, 2002). Note that in the present study, and consistent with previous studies with this system (Buckling & Rainey, 2002; Morgan et al., 2005), parasites were locally maladapted at later stages of coevolution, and hence dynamics were more MAM-like.

Despite local adaptation being possible in the later stages of coevolution, and purging of genetic diversity being an unlikely explanation, we did not observe an effect of changing the relative number of generations of bacteria and phages on local adaptation or average levels of resistance and infectivity. There are two, not mutually exclusive, explanations for these results. First, resistance and infectivity evolution proceeds much more slowly after long periods of coevolution (Buckling & Rainey, 2002), possibly because of elevated fitness costs associated with bacterial resistance and phage infectivity. Secondly, selection may act more strongly on other traits important for host–parasite interactions in addition to gross measures of resistance and infectivity. For example, rather then evolving to infect additional bacterial genotypes, phages may instead evolve to be more efficient at exploiting the hosts they can infect. Similarly, bacteria may evolve an increased probability of avoiding infection, while not being entirely resistant. Our simple population level assays were unable to detect changes in these life history traits.

It is prudent to consider the biological relevance of our experimental manipulations, and whether unconsidered effects of these manipulations could explain our results. We always allowed a period of coevolution (either 80 or 500 generations), and then allowed one species to evolve whereas the other was held constant, and then allowed a short period of coevolution (except in the second assays in Experiment 2a). These manipulations clearly differ from explicit manipulations of generation time, because species are undergoing periods were only one evolves, followed by bursts of coevolution. The closest (although by no means perfect) natural parallel to these manipulations is when parasites undergo many generations within a host population, before the host population undergoes any generations. Furthermore, the transfer regime creates fluctuations in population size unlikely to be experienced in natural populations. Despite the somewhat artificial nature of our experimental system, we believe that the results are likely to have more general applicability. Ultimately, differences in generation time are predicted to affect local adaptation simply because shorter generation times can provide an evolutionary advantage (Gandon & Michalakis, 2002). Consistent with these predictions, our manipulations resulted on average in a detectable evolutionary advantage to the species that underwent the most generations, and therefore provided a good analogy for directly manipulating generation time. An evolutionary advantage of one coevolving species is predicted to result in local adaptation of that species, if the specificity of interaction allows it (Gandon et al., 1996; Lively, 1999; Gandon, 2002; Gandon & Michalakis, 2002; Morgan et al., 2005). Of course we cannot entirely rule out the possibility that our manipulations somehow affected this specificity, although we cannot think how.

This is the first study to attempt to manipulate the generation time, although indirectly, of coevolving host and parasite populations. At the earlier stages of coevolution, increasing the relative number of generations experienced by host or parasite did provide that species with the expected evolutionary advantage. This suggests that relative generation times may affect the likelihood of local adaptation in host–parasite systems that are more amenable in detecting local adaptation. However, theory suggests that relative generation times are less important determinants of local adaptation than variables that increase within-population genetic variation, such as mutation and migration rates (Gandon & Michalakis, 2002). Data from this system are consistent with this hypothesis. Generation time manipulations had no impact on local adaptation after approximately 500 bacterial generations of coevolution, whereas increasing the relative rate of migration of phages increased phage local adaptation after a similar period of time (Morgan et al., 2005).

Acknowledgments

We thank Ben Sheldon, Arjan de Visser and two anonymous referees for comments on the manuscript. The work was funded by the Royal Society.

Ancillary