Bacteria-Phage Antagonistic Coevolution in Soil

See allHide authors and affiliations

Science  01 Apr 2011:
Vol. 332, Issue 6025, pp. 106-109
DOI: 10.1126/science.1198767


Bacteria and their viruses (phages) undergo rapid coevolution in test tubes, but the relevance to natural environments is unclear. By using a “mark-recapture” approach, we showed rapid coevolution of bacteria and phages in a soil community. Unlike coevolution in vitro, which is characterized by increases in infectivity and resistance through time (arms race dynamics), coevolution in soil resulted in hosts more resistant to their contemporary than past and future parasites (fluctuating selection dynamics). Fluctuating selection dynamics, which can potentially continue indefinitely, can be explained by fitness costs constraining the evolution of high levels of resistance in soil. These results suggest that rapid coevolution between bacteria and phage is likely to play a key role in structuring natural microbial communities.

Host-parasite antagonistic coevolution—the reciprocal evolution of host defense and parasite counter-defense—is theoretically crucial to a range of ecological and evolutionary processes, including population dynamics and extinction risk, the evolution of diversity and speciation, the evolution of sex and mutation rates, and the evolutionary ecology of pathogen virulence (14). A number of excellent studies have inferred the operation of host-parasite coevolution in natural populations from patterns of local adaptation of parasites to their hosts in space and time (510). However, genetic variation between the host populations in space and time may be driven by parasite-imposed selection but could equally be driven by neutral process or additional selection pressures. A direct demonstration of coevolution requires evidence of host adaptation to parasites as well as parasite adaptation to hosts.

Antagonistic coevolution between bacteria and their ubiquitous parasites, bacteriophage (phage), is likely to be of particularly broad importance because of their extremely rapid rates of evolution (3, 11)—the key role played by bacteria in ecosystem functioning—and the therapeutic use of phages as “evolving” antibiotics in agricultural and clinical contexts (11). Both the dynamics and consequences of coevolution between bacteria and viruses have been extensively studied in the laboratory (12, 13), but little is known about the extent or role of rapid bacteria-phage coevolution in natural populations (14, 15). Given that phages are typically highly specific to bacteria species and even genotypes (11) and the massive amount of diversity present in microbial communities (16), a given interacting bacteria and phage population is likely to make up a tiny fraction of the microbial community. It is therefore unclear whether phages, which only encounter hosts passively, impose sufficient selection on bacteria for rapid coevolution to occur.

We used a “mark-recapture” approach (17) to follow the ecological and evolutionary dynamics of a soil bacteria clone, Pseudomonas fluorescens SBW25, and a naturally associated lytic bacteriophage clone SBW25ϕ2 (18) in soil microcosms. Despite these organisms having been used extensively for in vitro evolution studies (13), they were frozen shortly after their original isolation and hence would have undergone little laboratory adaptation before our experiment. The phage, which is infectious to SBW25, can only be transmitted by lysis of the bacterial cell, resulting in the reciprocal evolution of resistant bacteria and infectious phages in nutrient media (18). Having established that no culturable bacteria in our soil were resistant to the antibiotic gentamicin (or could be infected by the experimental phage), we used an engineered gentamicin-resistant strain of P. fluorescens SBW25 (3) to inoculate the soil microcosms. All recovered bacteria that could grow on gentamicin therefore descended from the inoculated ancestral population. Similarly, we established that there were no phages in the soil that were able to infect P. fluorescens SBW25; hence, all recovered phages that could infect P. fluorescens SBW25 descended from the inoculated phage population. In an attempt to disentangle effects of the physical and biotic components of the soil environment, we cultured our bacteria and phage in sterilized soil, in which the resident microbial community had been reinoculated in some replicate communities but not in others. Samples from the replicate soil communities (eight per treatment combination) were collected every 4 days for 24 days, with a further sample taken after 48 days. At each time point, we determined densities and resistance and infectivity of the focal bacteria and phage, as well as densities of the culturable component of the total bacterial community, by plating onto antibiotic-free agar.

Phages reduced the mean density of the P. fluorescens population in the presence of the natural community, whereas in the absence of the natural community phages actually increased mean P. fluorescens density (generalized linear mixed model, interaction between presence/absence of natural community and phages: F1,28 = 60, P < 0.0001) (Fig. 1). In cases in which both bacteria and phages were co-inoculated, the presence of the natural community significantly reduced densities of bacteria (F1,14 = 23.5, P < 0.001) and phages (F1,14 = 132.4, P < 0.001) (Fig. 1B). The density-increasing effect of phages in the absence of the natural community is surprising, and its potential implications for the therapeutic use of phages warrants extensive future investigation. However, we speculated that phages selectively killed growing and not stationary cells (19), and the presence of high frequencies of growing cells [growth advantage in stationary phase (GASP) mutants] under nutrient limitation reduced population sizes (20). In contrast, selective elimination of growing P. fluorescens cells by phages did not increase P. fluorescens density in the presence of the natural community because this community also inevitably contained growing cells of different species.

Fig. 1

Population dynamics of the bacterial and phage populations. Connected symbols show densities through time [mean log10(colony forming units/g soil) for bacteria or log10(plaque forming units/g soil) for phage, ± SEM] of phage SBW25ϕ2 (●); P. fluorescens SBW25 evolved in the presence (■) or absence (□) of phage; and the culturable fraction of the natural community (▲). Populations were evolved in the (A) absence and (B) presence of the natural community.

Coevolution between this bacteria and phage in laboratory media typically results in the evolution of increased resistance and infectivity ranges through time (arms race dynamics) (13). Specifically, bacteria are more resistant to past than contemporary phages and more resistant to contemporary than future phages (13, 21); the same pattern holds for phage infectivity. To determine whether this coevolutionary dynamic occurred in soil, we measured the infectivity to bacteria isolated from day 24 of ancestral (day 0) phages (past) and phages isolated from days 24 (contemporary) and 48 (future) within each replicate. Similarly, we measured the resistance of ancestral (day 0) bacteria (past) and bacteria from days 24 (contemporary) and 48 (future) to phages from day 24. In contrast to results in laboratory media, bacteria were more resistant to their contemporary than past and future phages (Friedman test of effect of time, with replicate fitted as a blocking factor: P < 0.01; n = 16 populations) (Fig. 2, A and B), and phages were least infective to contemporary than past and future bacteria (P < 0.01; n = 16 populations) (Fig. 2, C and D), with no effect of the presence of the natural community on resistance/infectivity patterns (Mann-Whitney tests: P > 0.2 for resistance/infectivity at each time point, and changes in time). This significantly greater resistance of bacteria to contemporary than to noncontemporary phages (and corresponding lower infectivity of phages) held when comparisons were made only between days 24 and 48 (Wilcoxon paired sample test: P = 0.02; n = 16 populations). We extended our measurements to all sampled time points and found the frequency of bacteria resistant to the ancestral (time zero) and contemporary phages to be approximately 2 and 10%, respectively (paired t tests of resistance to contemporary versus ancestral phages, averaged through time: t = 3.5, P < 0.01; t = 10.2, P < 0.001, n = 16 populations, for absence and presence of natural community, respectively) (Fig. 3), with the difference in resistance to contemporary and ancestral phages greater for populations evolved in the presence of the natural community (independent sample t test: t = 6.0, P < 0.01, n = 16 populations). Taken together, these results demonstrate that bacteria and phage rapidly evolve, but that they are not undergoing arms race coevolution.

Fig. 2

Coevolutionary dynamics of bacteria and phages. Proportion of P. fluorescens isolated from day 24 resistant to ancestral (day 0) phage, contemporary (day 24), and future (day 48) phage populations evolved in the (A) absence and (B) presence of the natural community. Proportion of ancestral (day 0), day 24, and day 48 P. fluorescens susceptible to day 24 phages evolved in the (C) absence and (D) presence of the natural community. Individual thin lines indicate the eight separate replicates in each treatment; bold lines indicate means for all populations.

Fig. 3

Resistance of bacteria to ancestral and contemporary phages. Proportion of P. fluorescens resistant to the common ancestral (day 0) (○) and contemporary (●) phage populations evolved in the (A) absence and (B) presence of the natural community for all populations through time. No resistance to phages was detected in P. fluorescens populations evolved in the absence of phages.

The data are, however, consistent with coevolution in which parasite genotypes specialize on host genotypes, and fitness of a given genotype fluctuates through time (fluctuating selection dynamics) (2, 5, 6). If parasites adapt more rapidly than do hosts, fluctuating selection dynamics are on average expected to result in parasites better adapted to their contemporary compared with past and future hosts, and vice versa if hosts adapt more rapidly than do parasites (6). Our data show the latter pattern, unequivocally demonstrating adaptation of bacteria populations to the evolving phage populations. Phages require susceptible hosts to propagate; hence, it is almost certain that evolutionary changes in phage infectivity also result from adaptation to evolving hosts. The operation of fluctuating selection dynamics is important because this type of coevolution can potentially continue indefinitely and maintain genetic diversity within populations (2), whereas arms race dynamics purge diversity and require a continual supply of novel mutations, so may be short-lived.

In an attempt to obtain more direct evidence of phage adaptation to evolving bacteria, we measured interactions between bacteria and phages in space as well as time by comparing the infectivity of phages to their local bacteria with infectivity to bacteria from replicate microcosms. We found that phages were significantly locally adapted, performing consistently better on local as compared with foreign bacteria (paired t test of infectivity to local versus foreign phages, averaged through time: t = 3.4, P < 0.01, n = 16 populations) (Fig. 4), with no difference between communities evolved in the presence or absence of natural communities (t test: P > 0.2). These data unequivocally demonstrate that phages adapt to changes in bacterial resistance.

Fig. 4

Phage local adaptation. Mean local adaptation (proportion of local to foreign clones of bacteria susceptible to local phages) ± SEM measured at days 24 and 48 in the (A) absence and (B) presence of the natural community.

Consistent local adaptation in space is commonly used to infer greater rates of adaptation of one evolving partner over the other in putative coevolutionary interactions (22, 23). Our spatial data would therefore suggest phages adapt more rapidly than do bacteria. However, this interpretation is in apparent contrast to the direct measurements of adaptation in time, which suggest that bacteria adapt more rapidly than do phages. These results suggest that host-parasite adaptation in space and time need not be positively correlated (6) and that local adaptation in space may tell us little about relative rates of adaptation.

Despite direct evidence for bacteria-phage coevolution in soil, the fluctuating coevolutionary dynamics are markedly different from the arms race dynamics observed in high-nutrient broth, in which bacteria and phages with increasingly broad resistance and infectivity ranges, respectively, are favored through time. What limits the evolution of bacterial resistance and phage infectivity in soil, resulting in the change from arms race dynamics in vitro to fluctuating selection dynamics in soil? Theory suggests that growth-rate costs associated with resistance and infectivity might be important (24, 25), and there is evidence for bacteria bearing costs of resistance to phage (26, 27), and that these costs are greater in less productive environments (28). We therefore hypothesized that resistance was more costly in soil, which typically supports lower microbial biomass than does high-nutrient media. To test this hypothesis, we examined the competitive ability (relative to ancestral bacteria) of soil-evolved phage-resistant and -sensitive clones of bacteria in both broth and soil environments in the absence of phages. After competing for 2 days in high-nutrient broth in the absence of phages, no competitive costs associated with resistance were detected [(generalized linear model (GLM), phage-resistant versus -sensitive clones: P > 0.2) (fig. S1). In contrast, there was an approximately 36% reduction in relative fitness associated with resistance in soil (GLM, phage-resistant versus sensitive clones: F1,21 = 7.6, P = 0.01) (fig. S1), which did not differ between resistant clones that had evolved in the presence and absence of the natural community (presence versus absence of natural community: P > 0.2) (fig. S1). The costs associated with phage resistance in soil may explain why the evolution of resistance in soil is limited to co-occurring phages rather than to a wide range of previously encountered phage populations observed during coevolution in nutrient broth (18).

Despite considerable differences between the population dynamics of the focal bacteria and phages in the presence versus the absence of the natural microbial community, coevolutionary dynamics were very similar. There was no difference in the magnitude of parasite local adaptation in space (Fig. 4) and only small differences in patterns of adaptation in time, with an approximately 4% greater resistance to contemporary phages in the presence of the natural community (Fig. 3). If anything, we expected weaker selection for resistance, and less coevolution, in the presence compared with absence of the natural community, in part because of the reduced densities of the focal bacteria and phages (Fig. 1) (28). The limited impact of the microbial community on coevolution suggests that intraspecific competition and parasitism play a more substantial role than interspecific competition in driving evolution within microbial soil communities.

Our results show that bacteria and phage rapidly coevolve in soil, with very similar dynamics in the presence and absence of the natural microbial community. Coevolution altered ecological population dynamics and resulted in bacteria adapted to phages in time and phages adapted to bacteria in space. Unlike coevolution in high-nutrient broth, coevolutionary dynamics in soil appear to be driven by fluctuating selection—a form of coevolution that can potentially continue indefinitely. These results suggest that rapid bacteria-phage coevolution, and not just purely ecological interactions, are likely to be crucial in explaining the structure, population dynamics, and ultimately the function of natural microbial communities.

Supporting Online Material

Materials and Methods

Fig. S1


References and Notes

  1. Materials and methods are available as supporting material on Science Online.
  2. We thank T. Bell and A. Hall for comments on the manuscript. The work was supported by the European Research Council. P.G. was supported from Ministerio de Ciencia e Innovación (MICINN, Spain) by National Mobility Program of Human Resources “Jose Castillejo.” Partial 16S ribosomal RNA sequences have been assigned to the European Nucleotide Archive, European Molecular Biology Laboratory­–European Bioinformatics Institute, under accession numbers FR746065 to FR746094.
View Abstract

Navigate This Article