Bacteria-Phage Interactions across Time and Space: Merging Local Adaptation and Time-Shift Experiments to Understand Phage Evolution
*
Abstract
The study of parasite local adaptation to host populations offers important insight into the spatial scale of host-parasite interactions. For parasites adapting to local hosts, the process is continually driven by change in the host population, in response to either the parasite or alternative selection pressures. In the case of reciprocal coevolutionary change, this adaptation should generate a pattern whereby parasites are most fit against hosts from the recent past (which have not yet responded to parasite-mediated selection) and least fit against future host populations (with increased resistance). I argue that combining data on local adaptation across space with data on evolutionary responses over time can offer novel insight into the process of adaptation. Using bacteriophages from horse chestnut trees, I compare infectivity on bacterial hosts isolated from either the same tree or different trees over multiple months of the growing season and find that phage adaptation to local hosts is most pronounced on bacterial hosts from the recent past. These results confirm that phages are well adapted to bacterial populations living within eukaryotes and more broadly suggest that local adaptation studies may underestimate the magnitude of parasite evolution, as host and parasite adaptation are confounded within contemporary time points.
Introduction
The ubiquity of disease across natural systems indicates that hosts are unable to evolve complete resistance against all parasites and/or that parasites are often capable of overcoming any evolved resistance (Antonovics et al. 2012). Much of our understanding of parasite evolution, as well as of host evolution in response to parasite-mediated selection, comes from studies that test parasite infectivity/host resistance over either space or time (reviewed in Greischar and Koskella 2007; Hoeksema and Forde 2008; Gaba and Ebert 2009; Brockhurst and Koskella 2013). Specifically, local adaptation studies set out to test whether a parasite is better at infecting hosts from the same geographic location (or population) relative to those from a different location. The idea behind this approach is that selection should benefit any parasite capable of infecting common host genotypes within its local habitat and, therefore, that a well-adapted parasite population is one that has responded to such selection and is capable of infecting a relatively high subset of host genotypes in the local population. Assuming some spatial structure among host and parasite populations (i.e., through limited migration among sites) and genetic specificity underlying parasite infectivity and host resistance (such that there are no universally infective parasites or universally resistant hosts), the process of parasite adaptation should result in parasites that perform, on average, better on local host genotypes and worse on genotypes from other host populations. Indeed, two systematic reviews of local adaptation studies lend support to the idea that parasites are often well adapted to their local host populations, especially when compared to the paucity of evidence for host adaptation to local parasites (which equates to parasite maladaptation when infection and resistance are the metrics of fitness; Greischar and Koskella 2007; Hoeksema and Forde 2008). However, this trend is unlikely to represent a lack of evolution in host populations but rather reflects greater, more rapid adaptation of parasite populations relative to host populations due to, for example, increased additive genetic variation as a result of higher rates of migration, shorter generation times, or increased mutation rate (Lively 1999; Gandon 2002).
Local adaptation studies alone do not provide direct evidence for host-parasite coevolution. Instead, they test the potential for parasite-mediated selection on host populations and vice versa, especially for parasites that kill or sterilize their hosts on infection (Lively 1999; Gandon and Nuismer 2009). Whether or not the host/parasite population is capable of responding to such selection will depend on a number of factors, including the genetic variation within the population; the strength of other selection pressures, including other parasites/hosts; and the costs associated with resistance/infectivity (Thompson 1994, 2005). To directly demonstrate coevolution, one must measure reciprocal changes in infectivity and resistance, for example, by monitoring changes in genotype frequencies and relative infection prevalence among genotypes in host populations over time (Jokela et al. 2009).
One powerful experimental tool for measuring such coevolutionary change is the use of time-shift experiments (Gaba and Ebert 2009), where parasite infectivity/host resistance is tested on host/parasite populations from previous, contemporary, and future points in time. For example, this approach has been used to demonstrate rapid coevolution between the water flea, Daphnia magna, and its bacterial parasite, Pasteuria ramosa (Decaestecker et al. 2007). By crossing hosts and parasites from different layers of a pond sediment core, the researchers found that bacterial populations were most infective to contemporary host populations and that the water flea populations were able to evolve resistance to the bacteria over time. However, one limitation of time-shift experiments is that cross-inoculations are typically performed within interacting host-parasite populations, and therefore, there is no way to determine whether any observed coevolutionary change is the result of specific or general adaptations. For example, parasite-mediated selection could lead to a more general resistance mechanism in host populations that would be just as effective against allopatric parasite populations as sympatric parasite populations and perhaps even as effective against other types of parasites. Therefore, just as classic local adaptation studies are unable to provide direct support for coevolutionary change, time-shift experiments performed within populations do not provide information regarding the scale and/or specificity of the species interaction.
Due to the inherent limitations of local adaptation or time-shift studies alone, there has been recent interest in combining the two approaches to get a more complete understanding of the coevolutionary process (Forde et al. 2004; Gandon et al. 2008). For example, theoretical work exploring patterns of adaptation across space and time has emphasized that the two measures should be similar, as both describe the fit of a population to its local environment, but that diverging selection pressures among populations can lead to relatively stronger spatial versus temporal adaptation (Blanquart and Gandon 2013). This is nicely illustrated with data from experimental exposure of the human immunodeficiency virus to antibodies from either the same patient or other patients performed at multiple points in time, as the host mounts an immune response (Blanquart and Gandon 2013). The authors first examined temporal adaptation within patients and found that, although the mean fitness of the virus against contemporary antibodies did not change over time (which they ascribe to confounded effects of viral evolution and increased patient immune response, described as “environmental degradation”), the virus was generally most fit in the presence of antibodies from the recent past (evidence for an adaptive immune response to the virus). The authors then examined spatial adaptation by comparing viral fitness in the presence of contemporary antibodies from either the same or different patients. In this case, they found evidence of significant viral maladaptation that increases in magnitude over the course of the study, suggesting that the specific immune response of the host was well adapted to its particular viral population. This combined approach can therefore be used to gain nice insight into the changing patterns of spatial adaptation across time (Forde et al. 2004; Blanquart and Gandon 2013).
Here, I extend this argument to suggest that studies in which parasite adaptation is measured on hosts from sympatric versus allopatric populations from across multiple points in time can lend novel insight into the speed and strength of parasite evolution by decoupling the signature of parasite adaptation from that of its host. This approach moves beyond both classical local adaptation studies performed across contemporary combinations of sympatric versus allopatric populations, which necessarily confound host and parasite local adaptations, and time-shift experiments performed within populations, which are unable to distinguish the specificity and/or scale of adaptation. By measuring the infectivity of natural bacteriophage populations from a single time point on bacterial hosts from either sympatric or allopatric populations across multiple months, I test the prediction that the signature of phage local adaptation should be stronger on host populations from the recent past than on contemporary host populations that may themselves be adapted to the local parasite. Such a pattern would suggest that the true strength of parasite local adaptation is typically masked by reciprocal adaptations in the local host populations. As such, the combined analysis of adaptation over time and space would act as a great complement to the existing toolbox of approaches for measuring coevolution.
Methods
Study System
Lytic bacteriophage viruses (phages) are obligate killers of their bacterial host cells, as the next generation of phage particles is released into the environment only on host cell lysis. Infection of bacterial cells occurs when a phage recognizes receptors on the cell surface and attaches to the receptor in a “lock and key” fashion (Labrie et al. 2010). Once inside the host cell, lytic phages hijack the metabolic machinery of their host to replicate and mature, eventually bursting the cell open. The fitness of a given phage genotype thus necessarily depends on the density of susceptible host cells within the local environment, and there should be strong selection on phage populations to adapt to local bacterial hosts. Indeed, phage local adaptation has been documented in a number of field and laboratory environments (Forde et al. 2004; Morgan et al. 2005; Vos et al. 2009; Koskella et al. 2011). Although phages often have negligible direct effects on bacterial densities due to the rapid emergence of resistant bacterial strains (Lenski and Levin 1985; Waterbury and Valois 1993), they can have large effects on bacterial population dynamics (Pantastico-Caldas et al. 1992; Kuno et al. 2012), diversity (Rodriguez-Valera et al. 2009), and bacterial fitness (Bohannan et al. 2002).
Given the relatively short generation times of both bacteria and phages and the ability to easily manipulate the interaction in the laboratory, much of what we understand about bacteria-phage interactions comes from experimental coevolution studies. These studies allow initially identical populations of bacteria and phages to coevolve over time, and the evolutionary and coevolutionary responses of both antagonists can be measured throughout. This approach has been used to demonstrate the importance of relative migration rates among populations in shaping phage local adaptation (Morgan et al. 2005). Specifically, while experimental coevolution studies with no phage dispersal among replicate microcosms have typically found host local adaptation, suggesting that bacterial hosts have the evolutionary advantage (Buckling and Rainey 2002), experiments in which relative phage dispersal is increased tend to find patterns of phage local adaptation (Forde et al. 2004; Morgan et al. 2005). The experimental coevolution approach has also been used to demonstrate that patterns of phage local adaptation can be more pronounced across populations with differing resource levels relative to those with similar levels (Lopez Pascua et al. 2012). This phage adaptation should lead to the evolution of host resistance, and bacteria-phage systems have led the way in applying time-shift experiments to demonstrate this coevolutionary change (reviewed in Gaba and Ebert 2009). Because both bacteria and phages can be frozen throughout experimental coevolution and resurrected at a later time, cross-inoculations among bacteria and phages from past, contemporary, and future time points are relatively straightforward. Bacteria-phage coevolution has been demonstrated using this approach by showing that the fitness of each antagonist is highest against past populations of the other and lowest against future populations (Buckling and Rainey 2002). Furthermore, the rate of this coevolutionary change can be altered by changing the resource availability (Lopez-Pascua and Buckling 2008) and dispersal rates (Brockhurst et al. 2003).
The plant phyllosphere (i.e., the aboveground plant tissue) presents a particularly useful environment in which to study interactions between bacteria and phages in nature, as each plant can be expected to harbor a separate community of interacting microbial species. Given both the limited phage dispersal among versus within plants and the heterogeneity of resources and defenses among plants, we might expect phages to be well adapted to bacterial populations (and/or communities) from the same plant relative to those from other plants. In line with this prediction, phages isolated from leaves of the horse chestnut tree, Aesculus hippocastanum, were found to be consistently more infective to bacterial hosts from the same tree relative to hosts from nearby trees (separated by between 25 and 450 m; Koskella et al. 2011). Interestingly, no signature of phage local adaptation was observed among leaves from the same tree, suggesting a role for the tree environment in shaping the scale of the bacteria-phage interaction. Given that phages in the phyllosphere are generally well adapted to their local bacterial hosts, we might further predict that they should impose strong selection for bacterial resistance. In line with this prediction, a time-shift experiment run over the course of a single growing season (May–September) found that bacterial hosts were typically most resistant to phage populations from the prior month and least resistant to phage populations from 1 month in the future (Koskella 2013).
Both of these previous studies (Koskella et al. 2011; Koskella 2013) examined the susceptibility of a given bacterial isolate from the phyllosphere (i.e., a single clone) to a population/community of phages (i.e., all phages from a given leaf bulked into a single inoculum). One important step forward in understanding phage adaptation in this system would therefore be to examine the infectivity of single phage clones. Here, I take this approach by examining the infectivity of six independent phage clones, isolated from six separate trees, on the community of culturable bacteria from either sympatric or allopatric trees over the course of the season. By performing a reciprocal cross-inoculation over both space (i.e., bacteria from six different trees) and time (i.e., bacteria from four different months), I am able to tease apart the signature of phage adaptation from any reciprocal adaptation of contemporary host populations to their local phages.
Collection of Bacteria and Phage Isolates
Bacterial isolates were sampled from individual leaflets (the second-largest leaflet in the leaf) from the same branch of each of six trees, separated by between 25 and 450 m, within a park in Oxfordshire, United Kingdom (51.751125, −1.244974) during the growing season of 2011. Two leaflets per tree branch were sampled approximately every 4 weeks from April until September and brought back to the laboratory, where they were surface sterilized using a 10% bleach, 0.01% Tween detergent solution, rinsed with sterile water, and placed in a 15-mL Falcon tube containing 0.1 M potassium phosphate (pH 7.2) and 20% glycerol buffer. Tubes were immediately frozen at −20°C and stored until the end of the season. At the end of the season, leaves were rapidly thawed at 38°C and homogenized with a FastPrep-24 Instrument (MP Biomedicals, Cambridge, UK) using three ceramic beads at 4.0 rotations per second for 60 s. The leaf homogenate was then dilution plated on 1.2% King’s broth (KB) agar plates and incubated for 24 h at 28°C. This allowed growth of individual bacterial colonies, each initiated by a single bacterial cell from the leaf. Bacteria were isolated from only June, July, August, and September due to low densities in May (reported in Koskella 2013). For the remaining four time points, 48 bacterial colonies were isolated from each of two leaflets by randomly assigning a point on each plate and selecting the colonies that were closest to the point, regardless of size or color. Two leaflets per tree were used in order to generate the required number of independent bacterial isolates for the experiment. Previous work has demonstrated that phages from a given leaf are just as infective to bacteria from that leaf as they are to bacteria from other leaves within the same tree (Koskella et al. 2011); therefore, bacterial samples from across leaves were pooled for all analyses. Each of the chosen colonies (96 per tree for each of four time points) was used to initiate overnight cultures in 1 well of a deep, 96-well plate containing 800 μL of KB. After 24 h of incubation, glycerol was added to each well to a final concentration of 20%, and the plate was frozen at −80°C.
Phages were isolated from inocula generated as part of a previous study (Koskella 2013) by centrifuging 9 mL of each leaf homogenate at 550 g for 10 min and then filtering the supernatant through a 0.45-μ syringe filter (Millipore Millex, Watford, UK). This method is commonly used to separate bacteria and phage particles, and filtered leaf homogenates were stored in the dark at 4°C after separation. Homogenates were then tested against each of 24 bacterial isolates (a subset of the 96 isolates described above) from September from the same tree in order to isolate individual phage clones. To do this, overnight cultures of bacterial isolates were grown up and mixed into soft agar, poured into a square petri dish, spotted with 10 μL of each leaf homogenate, and incubated overnight at room temperature (this soft agar overlay technique is a common method for visualizing phage infection; Buckling and Rainey 2002; Morgan et al. 2005). Bacterial lawns were then scored for plaque formation (i.e., localized absence of bacterial growth due to phage infection and spread), and individual phage plaques were isolated to seed six independent phage stocks (one phage clone from each of six trees). In cases where phage density was too high to visualize individual, fully separated plaques, a dilution series was run and the soft agar overlay was repeated until individual plaques could be sampled.
To amplify the phage clones and generate inocula, phage samples were coinoculated into a tube containing KB and a low density of host cells, corresponding to the host from which each was originally isolated. Tubes were then incubated overnight, centrifuged, and filtered to generate one bulk inoculum for each phage clone. These inocula were then tested against their known bacterial hosts both to confirm that the isolation was successful and to measure and standardize the density of each inocula (i.e., the number of plaque-forming units per milliliter). In this way, six independent phage clones were chosen, each isolated from leaves collected in September from one of six different horse chestnut trees. Note that tree numbering is not continuous (i.e., phages were isolated from trees 1, 2, 4, 5, 7, and 8). This numbering corresponds to the eight trees used in a previous study (Koskella 2013) and reflects the six trees from which phage clones were successfully isolated from September.
Cross-Inoculation Experiment
In order to allow high-throughput testing of bacterial sensitivity to each of the six phages, a “streaking assay” was used where 40 μL of each phage inoculum was dripped down the length of a square petri dish containing solidified 1.2% KB agar and allowed to dry. A 12-well pin replicator was then dipped into a row of bacterial overnight cultures and streaked across the line of phage (fig. 1). Bacterial growth was compared in the zone prior to the phage drip with growth within and beyond the phage zone. In most cases, bacteria were scored as sensitive to a given phage when there was a complete absence of growth within the phage zone. However, some hosts were able to grow at very low densities over the phage by producing a clear mucoid layer over the phage zone (e.g., host/phage 1/1, 4/4, and 7/7 in fig. 1). As the difference in growth was clearly visible, these isolates were also considered to be sensitive. A total of 9,990 crosses were successfully scored (out of the 13,824 that were attempted, i.e., 96 bacteria from each of six trees from each of four time points, crossed against six phages) to measure sensitivity of each of the bacterial isolates to each of the six phages (one from the same tree, referred to as “sympatric,” and five from different trees, referred to as “allopatric”). Bacterial hosts were scored as either resistant, if they were able to successfully grow over the phage zone, or susceptible, if their growth was inhibited.
Demonstration of streaking assay to measure bacterial sensitivity to each of the six phages. Red bars indicate zone of phage, where inocula were dripped vertically down the plate. Bacterial cultures were then streaked from left to right, so inhibition of growth within and/or beyond (to the right of) the phage zone is indicative of susceptibility to a given phage clone. Each of the six phages is tested here against the bacterium from which it was initially isolated as well as five allopatric bacterial hosts. Note that phages 2 and 5 are infective to both hosts 2 and 5, suggesting that they have a similar host range despite being isolated from separate trees.
Statistical Analyses
To compare local adaptation across months, I first examined variation in bacterial susceptibility using a generalized linear mixed model (GLMM) with binomial errors and the logit-link function to test whether phage local adaptation was affected by the month in which bacteria were collected. In this analysis, tree was included as a random effect and phage clone, bacterial month, and sympatry (same tree vs. different tree) were included as fixed factors. Pairwise contrasts were then run to compare the success of each phage on sympatric versus allopatric hosts from each of the 4 months, and tests were adjusted for multiple comparisons using the sequential Bonferroni adjustment. A similar analysis was run using the proportion of susceptible bacteria per tree per time point in a mixed between/within subjects repeated measures ANOVA, where month was the within-subject variable and phage and sympatry were included as between-subject variables. In this case, local adaptation within each month was calculated with a series of Mann-Whitney tests. In addition, I examined the pattern of phage adaptation using a reciprocal pairwise method that simultaneously compares the difference between sympatric and allopatric combinations (Vos et al. 2009). For this analysis, local adaptation was calculated for each tree pair within a given time point as the mean difference in the proportion of sympatric (S) and allopatric (A) bacterial isolates (S − A) infected by the two phage clones. For this analysis, a measure of zero indicates no phage local adaptation, a negative value indicates phage local maladaptation (equivalent to host local adaptation), and a positive value indicates phage local adaptation. The mean measure of local adaptation across trees was then compared across months using a one-way ANOVA with least significant difference (LSD) post hoc tests. Finally, I examined the similarities of host range for each of the six phages using Pearson’s ϕ coefficient of association, which allows for comparison of two binary variables. All statistical analyses were run in IBM SPSS, version 21.
Results
Local adaptation of phages collected from September was first examined by comparing the infectivity of a given phage clone on bacterial hosts either from the sympatric tree or from allopatric trees across all time points. I examined the effect of sympatry (sympatric/allopatric) and bacterial time point (June, July, August, or September) on the probability that a bacterial isolate would be susceptible to each phage. The results of a GLMM reveal a significant interaction between phage clone, sympatry, and month, suggesting that the magnitude of local adaptation differs among both phages and time points (fig. 2; table 1). This three-way interaction is driven by the strong differences among phages regarding when the peak signature of local adaptation is observed for each phage (fig. 3). A series of pairwise contrasts suggest that while significant phage local adaptation is observed for only one tree in June (the most distant time point) and two trees in September (the contemporary time point), all but phages from tree 5 are found to be locally adapted in either July or August (table 1). A similar pattern is observed when proportion infected is compared across time using a repeated measures ANOVA. In this case, there is again a significant three-way interaction between phage clone, sympatry, and month (
, 
, 
), as well as a significant main effect of month (
, 
, 
) and significant interactions between phage and month (Wilks’s λ = 0.355, 
, 
) and sympatry and month (
, 
, 
). Furthermore, this analysis uncovers significant main effects of both between-subject variables, sympatry (
, 
, 
), and phage clone (
, 
, 
), as well as a significant interaction between the two (
, 
, 
). Mann-Whitney tests further show that, overall, phages are significantly locally adapted to bacterial hosts from July (
, 
) and August (
, 
) but not June (
, 
) or September (
, 
). As all phage clones were isolated from leaves collected in September, this result indicates that phage local adaptation is time lagged such that phages perform relatively better on sympatric bacteria from the recent past than on contemporary bacterial hosts (fig. 2). Finally, using the methods of Vos et al. (2009) to compare the mean difference in local adaptation across time points, phages were found to be locally adapted to bacterial hosts from their same tree at each of the four time points (table 2), but the magnitude of local adaptation differed across months (one-way ANOVA comparing mean 
over time, 
, 
). The LSD post hoc analyses suggest that the signature of phage local adaptation is higher in August than in June (
) or September (
) but only marginally higher than in July (
).
Local adaptation of phages from September on bacteria from sympatric versus allopatric trees across 4 months (June–September) of the growing season. Black lines represent the mean proportion of bacterial isolates from the same tree that were susceptible, while dashed gray lines represent the mean proportion of bacterial isolates from the other five trees that were susceptible to infection. Error bars indicate ±1 SEM.
Generalized linear mixed model of variation in bacterial susceptibility to September phages from either sympatric or allopatric trees across four time points
| F | df1 | df2 | P | |
|---|---|---|---|---|
| Fixed effects: | ||||
| Phage clone | 16.559 | 5 | 5,054 | <.001 |
| Sympatry | .001 | 1 | 9,942 | .980 |
| Month | .010 | 3 | 9,942 | .999 |
| Phage clone × sympatry | 3.595 | 5 | 1,082 | .003 |
| Phage clone × month | 1.763 | 15 | 9,942 | .034 |
| Sympatry × month | .016 | 3 | 9,942 | .997 |
| Phage clone × sympatry × month | 2.489 | 18 | 9,942 | <.001 |
| Estimate | t value | df | Adjusted P | |
| Pairwise contrast (sympatric − allopatric): | ||||
| June: | ||||
| Tree 4 | 5.103 | 2.259 | 2,080 | .024 |
| Tree 7 | 5.844 | −2.417 | 1,348 | .016 |
| July: | ||||
| Tree 1 | 6.269 | 2.504 | 1,132 | .012 |
| Tree 2 | 7.536 | 2.745 | 639 | .006 |
| Tree 7 | 11.956 | 3.458 | 386 | .001 |
| August: | ||||
| Tree 1 | 15.546 | 3.943 | 361 | <.001 |
| Tree 4 | 44.105 | 6.641 | 117 | <.001 |
| Tree 7 | 7.426 | 2.725 | 795 | .007 |
| Tree 8 | 5.843 | 2.417 | 1,110 | .016 |
| September: | ||||
| Tree 1 | 6.733 | −2.595 | 983 | .01 |
| Tree 7 | 16.116 | 4.015 | 246 | <.001 |
| Tree 8 | 4.937 | 2.222 | 1,557 | .026 |
| Estimate | SE | P | 95% CI | |
| Random effect: | ||||
| Tree | .047 | .041 | .253 | .008, .261 |
Infection patterns of each of the six phage isolates (each collected from a different tree in September) on either sympatric (black lines) or allopatric (dashed gray lines) bacterial hosts from 4 months (June–September) of the growing season. Exact 95% confidence intervals based on the binomial distribution are shown.
Mean measure of local adaptation (sympatric − allopatric [S−A] for each pairwise phage-bacterial population interaction) across the four time points and results of one-sample t-test
| Bacterial time point | Mean S − A | 95% CI | t14 | P |
|---|---|---|---|---|
| June | .045 | .031 | 2.797 | .014 |
| July | .088 | .067 | 5.048 | <.001 |
| August | .136 | .048 | 5.595 | <.001 |
| September | .040 | .030 | 2.642 | .019 |
Although the phages were each isolated from a separate tree, there was some clear overlap in host range among them (fig. 4). Of the 517 bacteria that were found to be susceptible to phage, 303 were susceptible to only one of the six phages, 205 were susceptible to two phages, 7 were susceptible to three phages, and 2 were susceptible to four of the phages. By far, the majority of the bacterial hosts that were susceptible to multiple phages were susceptible to phages 2 and 5 (92%). Interestingly, phage 5 performed particularly poorly on its sympatric bacterial hosts (fig. 3), suggesting it may have been a recent migrant from another population. Furthermore, these two phages were the most likely to infect bacterial isolates from other tree communities, with phage 2 infecting an average 20.3% of allopatric isolates across time and phage 5 infecting 14.8% of allopatric isolates. This suggests the possibility that these phages either have particularly large host ranges or are infecting a universally common bacterial species. However, I can rule out the possibility that the two represent exactly the same phage genotype, as there are cases of each infecting an independent subset of bacterial hosts. Overall, there was a significant positive association for phages 2 and 5 and significantly negative associations between phages 1 and 2, phages 1 and 5, phages 2 and 4, phages 2 and 7, phages 2 and 8, phages 4 and 5, and phages 5 and 7 with regard to host infectivity range (fig. 4; table 3).
Infection matrix of bacterial communities from each of six trees across 4 months (June–September) of the growing season. Each row represents a given phage sensitivity profile (i.e., all bacterial isolates with susceptibility to the same subset of phage collapsed into a single row), while the color of the box represents how common this profile was in the community. Gray boxes represent phage from allopatric trees, and red boxes represent susceptibility to phage from the sympatric tree.
Associations in host range among the six phages tested across the 517 susceptible bacterial isolates
| Phage 1 | Phage 2 | Phage 4 | Phage 5 | Phage 7 | Phage 8 | |
|---|---|---|---|---|---|---|
| Phage 1 | … | … | … | … | … | … |
| Phage 2 | −.417** | … | … | … | … | … |
| Phage 4 | −.016 | −.527** | … | … | … | … |
| Phage 5 | −.190** | .364** | −.241** | … | … | … |
| Phage 7 | .018 | −.421** | −.088* | −.230** | … | … |
| Phage 8 | −.048 | −.154** | .050 | −.071 | .036 | … |
Discussion
Previous work on bacteria-phage interactions in the horse chestnut phyllosphere has demonstrated that the scale of phage local adaptation is the host tree; phages were found to be well adapted to bacterial hosts both from their own leaf and from other leaves within the same tree but performed poorly on bacterial hosts from neighboring trees (Koskella et al. 2011). In addition, when adaptation of phage populations was measured within sympatric bacteria-phage populations from each of eight trees over the course of a growing season, phages were found to be most infective on bacterial hosts from the recent past and least infective to hosts from the future (Koskella 2013). The former study demonstrates significant spatial phage adaptation within a single point in time, while the latter demonstrates significant temporal phage adaptation within sympatric combinations of bacteria and phage populations. Based on these previous results, I predicted that the signature of phage local adaptation across space should be most pronounced for phages tested on sympatric versus allopatric bacterial hosts from the recent past. In line with this prediction, phages were found to perform, on average, 4% better on their sympatric hosts than on bacteria from other trees within the contemporary time point (September) but 9% and 14% better on sympatric bacteria than allopatric bacteria from earlier months (July and August, respectively; fig. 4; table 2). Phage local adaptation to bacterial hosts from 3 months earlier (June) was again lower, with phages infecting on average 4% more of their sympatric hosts than allopatric hosts. These results suggest that phage local adaptation at contemporary time points may be masked by bacterial local adaptation to phages and can therefore be uncovered by comparing local adaptation to hosts from the recent past.
Theoretical work exploring host-parasite coevolution predicts a time lag in parasite adaptation to host populations, as parasite populations respond to either selection imposed by newly arisen host resistance (in the case of arms race dynamics) or changes in the composition of host populations (in the case of fluctuating selection dynamics; Jaenike 1978; Lively 1999; Sasaki 2000). Indeed, evidence for such time-lagged adaptation of parasites to hosts has come from both experimental (Buckling and Rainey 2002; Koskella and Lively 2007) and natural (Dybdahl and Lively 1998; Koskella 2013) populations. Both models of host-parasite coevolution predict that parasites should perform best on host populations from the recent past. However, the contrast between the models comes in the success of parasites on hosts from the more distant past, where parasites should still perform better under arms race coevolution (as these hosts will have much fewer resistance adaptations than contemporary hosts) but have lowered performance under fluctuating selection dynamics (as these past host populations are less likely to be composed of host genotypes to which the contemporary parasite is well adapted; Gandon et al. 2008; Gaba and Ebert 2009). The observed decrease in local adaptation at the furthest time point in this study is more in line with theoretical predictions of fluctuating selection dynamics (Hamilton 1980; Gandon et al. 2008), where fitness of a given host or parasite fluctuates through time in response to negative frequency-dependent selection. This pattern could be explained by the limited host range of the phages, such that they lose the ability to infect previous hosts as they adapt to new hosts, or by changes in the bacterial population and/or community over time via migration. Furthermore, it has been suggested that the timing of the peak in temporal adaptation can be used to infer the speed of adaptation, as influenced by both the additive genetic variation in the population and the strength of selection imposed (Blanquart and Gandon 2013). Under this logic, the differences among the six phages in peak local adaptation (fig. 3) might suggest differences among populations in the speed at which coevolution is progressing.
That phages are most infective to bacterial hosts from the recent past is in line with data from experimental coevolution studies where replicate microcosms are subsampled over time and then time shifted to test for coevolution (reviewed in Brockhurst and Koskella 2013). However, when moving from controlled laboratory settings into natural environments, it becomes important to control for changes external to the bacteria-phage interaction. In particular, caution must be employed when interpreting any decreased success of parasites on host populations from the distant past, as loss of infectivity might also represent effects of a “degrading environment” (Blanquart and Gandon 2013). By integrating time-shift experiments with local adaptation experiments, it is possible to estimate the effect of the changing environment on bacterial and phage populations (as this should affect sympatric and allopatric populations equally) and to parse out the signatures of adaptation to the antagonist and adaptation to the environment (Gandon et al. 2008; Blanquart and Gandon 2013). For bacteria and phages in the phyllosphere, it is known that both population density (Koskella 2013) and community composition (Jackson and Denney 2011) can vary dramatically over the course of a season. However, allopatric hosts from previous time points can be used as control populations to account for any variation in environmental changes through time. Using this approach, the results of the current study suggest that, although seasonality might account for small changes in host susceptibility (fig. 2), the pattern of time-lagged phage adaptation was significantly more pronounced for sympatric phage-bacteria combinations than allopatric combinations.
The observation that phages are able to infect only a small proportion of their local contemporary bacteria, especially relative to infectivity on hosts from the recent past, suggests that the bacterial hosts had evolved resistance over the course of the previous month or months (depending on the peak in infectivity). This confirms the results of a previous study (Koskella 2013) and could be explained by evolutionary change within particular bacterial populations, via either the evolution of resistance within lineages or the introduction of resistant genotypes into the population via migration. Alternatively, this response could represent a change in the species composition of the bacterial community over time, again due to either a selective advantage for rare, resistant species or immigration of new resistant species. Previous work suggests that phages are capable of adapting to and therefore selecting against both common bacterial genotypes within a population (as predicted by the Red Queen hypothesis for coevolution; Jaenike 1978; Hutson and Law 1981) and common bacterial species within a community (as predicted by the “kill the winner” hypothesis; Thingstad and Lignell 1997; Weinbauer and Rassoulzadegan 2004). For example, experimental coevolution of Pseudomonas fluorescens and phage SBW25Φ2 was initially dominated by selective sweeps of highly resistant/infective types, but populations tended to move toward the fluctuating dynamics that are indicative of negative-frequency dependent selection over experimental time (Hall et al. 2011). Similarly, bacteria and phages from aquaculture environments were found to be highly dynamic at the strain level, despite being relatively static at the community level (Rodriguez-Brito et al. 2010). On the other hand, microbial community composition of bioreactor environments was found to change frequently, with correlated change in the phages associated with each dominant taxa (Shapiro et al. 2010), and the presence of phages in mixed bacterial communities cultured in seawater was found to reduce the dominance of common bacterial taxa (Fuhrman and Schwalbach 2003). Whether the observed change in host susceptibility is the result of population- or community-level change is further complicated by the potential for phages with broad host ranges. Previous characterization of phages from the horse chestnut phyllosphere suggests that phages can have very broad host ranges, even infecting bacteria across genera, and that many hosts are susceptible to multiple phages simultaneously (Koskella and Meaden 2013). This, therefore, emphasizes that coevolution is unlikely to proceed in a simple pairwise fashion within these complex communities.
The pattern of phage local adaptation uncovered both here and in a previous study of the same system (Koskella et al. 2011) could also be the result of either population- or community-level processes. For example, it could be that each tree is harboring a significantly different microbial community from the next and that each phage clone is specifically adapted to the common bacterial species within its own tree. Alternatively, it could be that the microbial community composition is relatively similar among trees but that phages are well adapted to particular genotypes within each bacterial population. It is likely that both mechanisms are at work simultaneously, given the previous evidence that (1) the communities of culturable bacteria (i.e., those that could be grown and isolated on media in the laboratory) are remarkably different across trees and (2) the signature of local adaptation among trees remains strong when only bacterial isolates of the species Pseudomonas syringae are examined (Koskella et al. 2011). Regardless of the level at which phage-mediated selection is acting, if changes in the host population/community during the course of the season were such that bacterial communities from two trees were more similar to one another at a single point in the season than communities from the same tree from two different points in the season, then we would expect the signature of phage local adaptation to be lower for noncontemporary combinations. The current results suggest instead that bacterial populations/communities from the same tree over the course of the season are more similar to one another in terms of susceptibility to local phages than communities from multiple trees within the same month. However, the fact that four of the six phages were found to be capable of infecting a substantial number of bacterial hosts from other trees, often peaking at earlier time points, suggests that phages may well be moving among the trees over the course of the season (fig. 3).
Importantly, the method of phage isolation may have biased the results toward phages infective to sympatric, contemporary bacteria, as phages were isolated from a previous experiment in which crosses were run between only bacteria and phages from the same tree (Koskella 2013). However, the primary question being addressed here was not whether phages were locally adapted, as this has been previously demonstrated in the same system (Koskella et al. 2011), but rather whether the signature of local adaptation differs across time-shifted bacterial hosts. Therefore, the method of isolation should, if anything, have biased the results to strong contemporary local adaptation and is, as such, a conservative test of the hypothesis. The relatively small effect of phage local adaptation observed for contemporary bacteria-phage combinations indicates that individual phage clones may not be particularly well adapted to their bacterial hosts. This is in stark contrast to the patterns observed for whole phage populations, where phage inocula generated from leaf material were found to be consistently well adapted to local bacteria (Koskella et al. 2011). Models of pairwise host-parasite coevolution suggest that parasites should be locally adapted on average but that an absence of local adaptation or even parasite maladaptation should also be commonly observed due to the time-lagged nature of parasite adaptation (Kaltz and Shykoff 1998). It is therefore likely that, while individual phage clones show variation in adaptation to their local host populations (fig. 3), the pattern of local adaptation observed for the phage population as a whole reflects the average signature of local adaptation across phages and is therefore more likely to be large (Koskella et al. 2011). If this interpretation is correct, then it suggests another key advance of the current method: it allows for a direct test of time-lagged and specific adaptation of individual parasite lineages to their local hosts even when these parasites have lost their advantage due to an evolutionary response of the host population.
Finally, it is perhaps surprising that these six phages, all isolated in September, were sampled from their respective leaves, given the relatively small proportion of their contemporary host community each was capable of infecting, particularly for phages 1, 4, and 5 (fig. 4). One intriguing explanation could be that the isolated phages were in fact temperate phages that happened to be in the lytic cycle at the time of sampling. Temperate phages are viruses that integrate into the genome of their hosts and that can be vertically transmitted during bacterial reproduction. These phages retain the ability to enter the lytic cycle, lysing their host cells and transmitting horizontally, and often do so under stressful conditions (Echols 1972). Temperate phages can spread within a host population if they provide some kind of selective benefit to the bacterium (Rankin et al. 2010), and one such advantage could be protection against superinfection by the lytic form of the phage (Berngruber et al. 2013). Under this scenario, temperate phages that once again enter the lytic cycle to infect new hosts should be most infective to bacteria from earlier in coevolutionary time, as these hosts are less likely to be carrying the phage in their genome.
Conclusions
Although parasite local adaptation is often observed, it is far from a ubiquitous pattern; a systematic review of published studies found consistent evidence for parasite local adaptation in only 18 out of 54 studies (Greischar and Koskella 2007). The results presented here suggest that these studies may underestimate the magnitude of parasite—and, indeed, host—local adaptation as they confound the two effects. If parasites are adapting to their host populations and, reciprocally, hosts are responding to these changes by becoming more resistant to their local parasites over time, then the overall pattern of local adaptation observed may reflect which partner is ahead in the coevolutionary battle but does not necessarily indicate lack of adaptation of the other. I therefore put forward the idea that testing for local adaptation within a time-shift framework presents a powerful way to disentangle the effects of parasite and host adaptation and suggest that doing so can uncover the true effect of adaptation across space.
I thank M. Boots, A. Buckling, S. Gandon, S. Meaden, P. Thrall, M. Vos, and the disease group at University of Exeter’s Cornwall campus for helpful discussion in developing this idea and project; J. Jokela and D. Vergara for advice on statistical analyses; and F. Débarre, T. Taylor, and two reviewers for comments on previous versions of the manuscript. I especially thank C. M. Lively for inviting my contribution to this symposium, for his great insight and dedication to the field of disease ecology and evolution, and for years of support and enthusiasm. This study was funded by a Natural Environment Research Council research fellowship (NE/K00879X/1).
Notes
*This issue originated as the 2013 Vice Presidential Symposium presented at the annual meetings of the American Society of Naturalists.
Literature Cited
Antonovics, J., M. Boots, D. Ebert, B. Koskella, M. Poss, and B. M. Sadd. 2012. The origin of specificity by means of natural selection: evolved and non-host resistance in host-pathogen interactions. Evolution 67:1–9. Berngruber, T. W., R. Froissart, M. Choisy, and S. Gandon. 2013. Evolution of virulence in emerging epidemics. PLoS Pathogens 9:e1003209. Blanquart, F., and S. Gandon. 2013. Time-shift experiments and patterns of adaptation across time and space. Ecology Letters 16:31–38. Bohannan, B., B. Kerr, C. Jessup, J. Hughes, and G. Sandvik. 2002. Trade-offs and coexistence in microbial microcosms. Antonie Van Leeuwenhoek 81:107–115. Brockhurst, M. A., and B. Koskella. 2013. Experimental coevolution of species interactions. Trends in Ecology and Evolution 28:367–375. Brockhurst, M. A., A. D. Morgan, P. B. Rainey, and A. Buckling. 2003. Population mixing accelerates coevolution. Ecology Letters 6:975–979. Buckling, A., and P. B. Rainey. 2002. Antagonistic coevolution between a bacterium and a bacteriophage. Proceedings of the Royal Society B: Biological Sciences 269:931–936. Decaestecker, E., S. Gaba, J. A. M. Raeymaekers, R. Stoks, L. Van Kerckhoven, D. Ebert, and L. De Meester. 2007. Host-parasite “Red Queen” dynamics archived in pond sediment. Nature 450:870–873. Dybdahl, M. F., and C. M. Lively. 1998. Host-parasite coevolution: evidence for rare advantage and time-lagged selection in a natural population. Evolution 52:1057–1066. Echols, H. 1972. Developmental pathways for the temperate phage: lysis vs lysogeny. Annual Review of Genetics 6:157–190. Forde, S. E., J. N. Thompson, and B. J. M. Bohannan. 2004. Adaptation varies through space and time in a coevolving host-parasitoid interaction. Nature 431:841–844. Fuhrman, J. A., and M. Schwalbach. 2003. Viral influence on aquatic bacterial communities. Biological Bulletin 204:192–195. Gaba, S., and D. Ebert. 2009. Time-shift experiments as a tool to study antagonistic coevolution. Trends in Ecology and Evolution 24:226–232. Gandon, S. 2002. Local adaptation and the geometry of host-parasite coevolution. Ecology Letters 5:246–256. Gandon, S., A. Buckling, E. Decaestecker, and T. Day. 2008. Host-parasite coevolution and patterns of adaptation across time and space. Journal of Evolutionary Biology 21:1861–1866. Gandon, S., and S. L. Nuismer. 2009. Interactions between genetic drift, gene flow, and selection mosaics drive parasite local adaptation. American Naturalist 173:212–224. Greischar, M. A., and B. Koskella. 2007. A synthesis of experimental work on parasite local adaptation. Ecology Letters 10:418–434. Hall, A. R., P. D. Scanlan, A. D. Morgan, and A. Buckling. 2011. Host–parasite coevolutionary arms races give way to fluctuating selection. Ecology Letters 14:635–642. Hamilton, W. D. 1980. Sex versus non-sex versus parasite. Oikos 35:282–290. Hoeksema, J. D., and S. E. Forde. 2008. A meta-analysis of factors affecting local adaptation between interacting species. American Naturalist 171:275–290. Hutson, V., and R. Law. 1981. Evolution of recombination in populations experiencing frequency-dependent selection with time delay. Proceedings of the Royal Society B: Biological Sciences 213:345–359. Jackson, C. R., and W. C. Denney. 2011. Annual and seasonal variation in the phyllosphere bacterial community associated with leaves of the southern magnolia ( Magnolia grandiflora). Microbial Ecology 61:113–122. Jaenike, J. 1978. An hypothesis to account for the maintenance of sex within populations. Evolutionary Theory 3:191–194. Jokela, J., M. F. Dybdahl, and C. M. Lively. 2009. The maintenance of sex, clonal dynamics, and host-parasite coevolution in a mixed population of sexual and asexual snails. American Naturalist 174(suppl.):S43–S53. Kaltz, O., and J. A. Shykoff. 1998. Local adaptation in host-parasite systems. Heredity 81:361–370. Koskella, B. 2013. Phage-mediated selection on microbiota of a long-lived host. Current Biology 23:1256–1260. Koskella, B., and C. M. Lively. 2007. Advice of the rose: experimental coevolution of a trematode parasite and its snail host. Evolution 61:152–159. Koskella, B., and S. Meaden. 2013. Understanding bacteriophage specificity in natural microbial communities. Viruses 5:806–823. Koskella, B., J. N. Thompson, G. M. Preston, and A. Buckling. 2011. Local biotic environment shapes the spatial scale of bacteriophage adaptation to bacteria. American Naturalist 177:440–451. Kuno, S., T. Yoshida, T. Kaneko, and Y. Sako. 2012. Intricate interactions between the bloom-forming cyanobacterium Microcystis aeruginosa and foreign genetic elements, revealed by diversified clustered regularly interspaced short palindromic repeat (CRISPR) signatures. Applied and Environmental Microbiology 78:5353–5360. Labrie, S. J., J. E. Samson, and S. Moineau. 2010. Bacteriophage resistance mechanisms. Nature Reviews Microbiology 8:317–327. Lenski, R. E., and B. R. Levin. 1985. Constraints on the coevolution of bacteria and virulent phage: a model, some experiments, and predictions for natural communities. American Naturalist 125:585–602. Lively, C. M. 1999. Migration, virulence, and the geographic mosaic of adaptation by parasites. American Naturalist 153(suppl.):S34–S47. Lopez Pascua, L., S. Gandon, and A. Buckling. 2012. Abiotic heterogeneity drives parasite local adaptation in coevolving bacteria and phages. Journal of Evolutionary Biology 25:187–195. Lopez-Pascua, L. D. C., and A. Buckling. 2008. Increasing productivity accelerates host-parasite coevolution. Journal of Evolutionary Biology 21:853–860. Morgan, A. D., S. Gandon, and A. Buckling. 2005. The effect of migration on local adaptation in a coevolving host-parasite system. Nature 437:253–256. Pantastico-Caldas, M., K. E. Duncan, C. A. Istock, and J. A. Bell. 1992. Population dynamics of bacteriophage and Bacillus subtilis in soil. Ecology 73:1888–1902. Rankin, D., E. Rocha, and S. Brown. 2010. What traits are carried on mobile genetic elements, and why? Heredity 106:1–10. Rodriguez-Brito, B., L. Li, L. Wegley, M. Furlan, F. Angly, M. Breitbart, J. Buchanan, et al. 2010. Viral and microbial community dynamics in four aquatic environments. ISME Journal 4:739–751. Rodriguez-Valera, F., A.-B. Martin-Cuadrado, B. Rodriguez-Brito, L. Pasic, T. F. Thingstad, F. Rohwer, and A. Mira. 2009. Explaining microbial population genomics through phage predation. Nature Reviews Microbiology 7:828–836. Sasaki, A. 2000. Host-parasite coevolution in a multilocus gene-for-gene system. Proceedings of the Royal Society B: Biological Sciences 267:2183–2188. Shapiro, O. H., A. Kushmaro, and A. Brenner. 2010. Bacteriophage predation regulates microbial abundance and diversity in a full-scale bioreactor treating industrial wastewater. ISME Journal 4:327–336. Thingstad, T., and R. Lignell. 1997. Theoretical models for the control of bacterial growth rate, abundance, diversity and carbon demand. Aquatic Microbial Ecology 13:19–27. Thompson, J. N. 1994. The coevolutionary process. University of Chicago Press, Chicago. ———. 2005. The geographic mosaic of coevolution: interspecific interactions. University of Chicago Press, Chicago. Vos, M., P. J. Birkett, E. Birch, R. I. Griffiths, and A. Buckling. 2009. Local adaptation of bacteriophages to their bacterial hosts in soil. Science 325:833. Waterbury, J. B., and F. W. Valois. 1993. Resistance to co-occurring phages enables marine Synechococcus communities to coexist with cyanophages abundant in seawater. Applied and Environmental Microbiology 59:3393–3399. Weinbauer, M. G., and F. Rassoulzadegan. 2004. Are viruses driving microbial diversification and diversity? Environmental Microbiology 6:1–11.
Print of horse chestnut leaves (Aesculus hippocastanum) on nutrient agar (after 48 h of incubation) showing bacterial and fungal inhabitants. Photo credit: Britt Koskella.
Symposium Editor: Curtis M. Lively
