In this study we investigated effect of simultaneous migration of coevolving bacteria and phages on the evolution of bacterial resistance and phage infectivity. Based on previous studies, that suggest that bacteria have a greater evolutionary potential than phages [17, 23], we hypothesised that phages should benefit more from migration than bacteria. Our data were consistent with this hypothesis. Analysis of sympatric resistance revealed a negative linear effect of migration rate (Fig. 1; F1,28 = 6.65, P = 0.015), no quadratic effect of migration (F1,28 = 0.34, P = 0.2) and no significant difference between founding populations (F5,28 = 1.52, P = 0.2). However the relationship between migration rate and sympatric resistance was likely to be solely due to the high rates of sympatric resistance at 0% migration; indeed, when the 0% migration treatment was excluded from the analysis, we found no significant effects of migration rate on sympatric infectivity (Fig. 1; P > 0.2 for linear and quadratic effects). Thus any migration benefited the more genetically limited and so less evolvable phages more than bacteria, regardless of rate; and changing the rate did not increase or decrease its advantage.
These data suggest that simultaneous migration is likely to benefit the least evolvable species most during host-parasite antagonistic arms races. In the context of bacteria-phage coevolution, bacteria tend to be ahead in the arms race [23], and so simultaneous migration is more likely to favour phages. It is unclear whether simultaneous migration will generally benefit hosts or parasites most. Patterns of local adaptation in natural populations suggest either hosts or parasites may have an evolutionary advantage, and not always parasites as common wisdom suggests (reviewed in Kaltz & Shykoff [24]).
We also addressed how migration rate affected global resistance of bacteria (i.e. the average resistance of bacteria from one treatment to phages from all other migration treatments) and global infectivity of phages (i.e. the average infectivity of phages from one treatment to bacteria from all other treatments), measuring infectivity and resistance across the range of migration regimes. We found a unimodal relationship between mean global bacterial resistance (to phages from all migration regimes) and the rate of migration, with resistance peaking at a migration rate of 1% per transfer (Fig. 2; quadratic effect of migration rate: F1,28 = 14.73, P = 0.001; there was no significant linear relationship between resistance and migration rate; linear effect of migration rate: F1,28 = 1.43, P = 0.241. Average levels of resistance were significantly different between replicates; effect of founding population: F5,28 = 2.82, P = 0.035). In this system, coevolution results in bacteria and phages evolving to be resistant and infective to an increasingly wide range of phage and bacterial genotypes, respectively [19, 25]; a situation broadly consistent with GFGM of coevolution [20, 26, 27]. As such, initial increases in migration rate will provide genetic variation that will allow a given resistance range to be reached more rapidly. However, further increases in migration caused the rate of resistance evolution to decline (Fig. 2). This is likely to be because migration allowed a globally fit bacterial clone at a given point in time to spread rapidly through all populations, at the expense of clones with resistance alleles that might have been beneficial in the future. Specifically, we suggest that clones with alleles conferring resistance to a wider range of phages than is currently useful, increase in frequency in isolated tubes. However, because of pleiotropic growth rate costs associated with wide resistance ranges, high rates of migration would increase the probability of these clones being competitively excluded by clones with resistance ranges that are narrower but sufficient to resist contemporary phage genotypes (clonal interference; [28]). Consistent with this hypothesis, previous work on this bacterium suggests the operation of trade-off between resistance and competitive ability in the absence of phages [29]. Similar costs of resistance have been reported in other bacteria-phage systems (reviewed in Bohannan & Lenski [30]).
By contrast, phage infectivity ranges (measured against bacteria from all migration regimes) showed a positive relationship with migration rate (Fig. 3; linear effect of migration rate: F1,28 = 19.40, P < 0.001), which plateaued at around 1% migration (quadratic effect of migration rate: F1,28 = 12.44, P = 0.001). There was no effect of starting population (F5,28 = 1.98, P = 0.113). As with the analysis of sympatric resistance above, the relationship between phage infectivity and migration rate disappeared when the no migration treatment was excluded from the analysis (P > 0.1 for both linear and quadratic terms). These patterns of global resistance and infectivity with respect to different rates of migration are broadly consistent with patterns of sympatric resistance. Sympatric resistance decreases with migration, hence migration benefits phages more than bacteria. Low levels of migration appear to increase both global bacterial resistance and phage infectivity, whereas further increases in migration do not affect global phage infectivity but decrease global bacterial resistance. The net effect is that migration also appears to benefit phages more than bacteria when resistance and infectivity traits are measured across meta-populations.
There are two plausible explanations as to why bacterial resistance declined, but phage infectivity plateaued at high migration rates (compare Figs 2 and 3). The first possibility is (as above) that phages are simply less evolvable than bacteria. Specifically, the likelihood of beneficial alleles being lost as a result of clonal interference may have been less in phage than bacteria populations, because phage populations contained fewer competing beneficial mutations at any given time. Thus, beneficial mutations became fixed in phage populations prior to new mutations arising. Second, increased phage infectivity range may not be associated with the same growth rate costs associated with bacterial resistance ranges. As such, an allele that conferred a broader infectivity range than was currently necessary at a given time point may not have been outcompeted by alleles conferring less broad infectivity ranges. However, we have observed that phages that have coevolved with bacteria tend to produce much smaller plaque sizes than ancestral phages, suggesting there is also a cost to increased infectivity ranges.
It is initially surprising that the 0% and 50% migration treatments generated such different results: such high migration may simply have the effect of producing a single 3-fold larger population. However, resistance and infectivity traits readily evolve within a single transfer (7 generations), and it is likely that this evolution will be divergent between populations [19]. This will increase genetic variation within the metapopulation beyond a simple population size effect.
Our measures of sympatric resistance with respect to migration are likely to hold true through coevolutionary time: if migration benefits phages more, then sympatric infectivity is likely to be higher. However, the global data analysed here clearly represents a very specific and non-equilibrium coevolutionary state. Bacteria and phage evolve increasing resistance and infectivity through time, but such increases are likely end at some point [27]. As such, comparisons of resistance and infectivity ranges as a function of migration rates may only be relevant to this particular period of coevolution. However, it is possible that migration regimes may affect average levels of resistance and infectivity ranges (a dynamic equilibrium state) measured over much longer time scales. Escalatory arms races, as observed here, may represent an ascending phase of a coevolutionary cycle, with selection for narrow resistance and infectivity ranges as costs of resistance and infectivity become too great. [27]. The continual supply of 'good' resistance and infectivity traits through migration may, for example, increase the magnitude of escalation before this happens. That aside, our measures of resistance and infectivity may be more generally applicable if they are equated with the rate of evolution of coevolving populations of hosts and parasites. This may be a reasonable assumption when it is considered that resistance and infectivity ranges increases through time [19], and where specifically investigated, we observe that populations that coevolve faster show broader resistance ranges [25]. Thus, assuming that selection fluctuates to some degree through time (which it always will, unless coevolution is a pure GFGM and resistance and infectivity evolution is cost-free) during antagonistic coevolution, we predict unimodal relationships between the rate of evolution and migration rate when clonal interference can occur. By contrast, in the absence of clonal interference (and populations are very mutation limited), we expect this relationship to be positive.