1. Introduction
Whether found in the soil [
1,
2], the leaf [
3,
4], the ocean [
5,
6], or the human body [
7,
8], microbial communities are proving to be more dynamic and diverse than could have been predicted. This incredible diversity is seen both within environments and among environments [
7], but how it is generated and maintained is unclear. Early ideas about microbial diversity posited that “Everything is everywhere, but the environment selects [
9].” However, this view of the microbial world remains hotly debated [
10]. Much of the work testing this tenet has focused on the abiotic environment, such as salinity [
11] or soil types [
12], but the biotic environment is likely to be just as, if not more, important in shaping selection on microbial populations. Bacteria in any given environment face strong selection pressures from other microbes, predators, viruses, and in the case of bacteria living within another organism, the host immune response. In contrast to the abiotic environment, biotic “environments” have the potential to evolve in response to any changes in the microbial community, making them highly dynamic and capable of driving divergence among populations [
13,
14]. Bacteriophages (phages) represent perhaps the most ubiquitous of these biotic drivers [
15,
16,
17]. To understand the role of phage-mediated selection in generating diversity, however, we need good insight into how specific phages are to their bacterial hosts.
For phages to alter the composition of a microbial community, there must exist a degree of specificity such that some hosts are more resistant to local phages than others or are better able to respond to phage-mediated selection. There is clear evidence that not all bacteria are infected by all phages, and indeed that most phages can only infect a subset of bacterial species (
Table 1; [
18]), but our understanding of phage host range is far from complete. Can phages easily adapt to infect new bacterial types as they become common? Can the same phage lineage shift from one bacterial species to another? These questions are far from new [
19], but the development of recent techniques and the power of comparative genomics are moving us towards more satisfying answers. Experimental evolution (
Box 1) provides one powerful approach to address these knowledge gaps, as the bacteria-phage interaction can be observed in the absence of other abiotic or biotic selection pressures; as such it has offered key advances in our understanding of the evolution of bacterial resistance to phages and reciprocal adaptations of phages to overcome such resistance. However, there are many reasons that the outcome of coevolution in a test tube might not be predictive of coevolution in nature, given the added biotic and abiotic complexity of most microbial ecosystems. For example, although experimental evolution studies have almost exclusively focused on phage adaptation within a population of one, or at most a few, bacterial species, most bacterial communities are highly rich, and therefore most bacterial species to which phage are adapting are rare. A comparison of culture-independent sequencing studies of microbial communities from the leaf surface, soil, atmosphere and the human body shows that the most dominant species in each given community represents a mere 2–5% of sequences [
20]. Given the heterogeneity and diversity of these microbial communities, it is unclear how a phage with narrow host range could evolve and be maintained.
For virulent bacteriophages,
i.e., those that reproduce within and then lyse their host cells, success depends on the chance event of encountering a susceptible host cell in the environment, and is most certainly reduced as a function of community diversity, dispersal, and exposure to the harsh conditions outside of a cell. It makes intuitive sense, therefore, that those phages with a larger host range should be at an inherent advantage. The data gathered so far, however, do not clearly support this intuition. First, many phages seem to be specific to a single bacterial species, and are often specific to only a few strains within that species [
21,
22,
23] (
Table 1). Second, there is building evidence that phages are “locally adapted” to their bacterial hosts [
3,
24], indicating a degree of specialization to common bacterial strains or species in a given population. Third, although phages do tend to increase their host range during the initial stages of coevolution, there is evidence that this expansion is short-lived [
25]. The underlying mechanisms of phage infectivity and bacterial resistance are of course key to the evolution of phage host range [
22,
26], and have been the focus of extensive review elsewhere [
21,
27]. The data make it clear that phage infectivity is a complex function of adsorption [
28], structural change of both host and phage [
29], transport of nucleic acid into the bacterial cell, and avoidance of degradation once inside the cell [
30], and is thus a result of both phage and bacterial phenotype. In addition, host susceptibility/resistance to phage can be determined by plasmids hosted by the bacterial cell [
31,
32]. This form of phage-plasmid interaction could lead to broad phage host range due to horizontal transfer of the plasmid among bacterial species within a community. A more thorough understanding of these interacting mechanisms will allow us to better predict the potential for host range expansion/contraction and therefore the effect of phages on microbial communities under both natural and therapeutic settings.
The more general question of why parasites specialize is of course not specific to bacteria-phage interactions, and we can apply much of the current coevolutionary theory to understanding the evolution of phage specificity. Many phages act as obligate parasites, as they are both unable to reproduce outside of their host cells and require cell lysis to transmit, thus killing their hosts. However, we acknowledge that, although virulent phages are obligate killers, other phages integrate into the host genome and their fitness relies on host reproduction. In these cases, the acquisition of a prophage can confer beneficial phenotypic change to the bacterial hosts, and therefore this latter relationship acts more synergistically than antagonistically. In either case, the question of host range for phages that are in the lytic cycle, and being transmitted among cells can be broken into two parts: first, the specificity of host resistance against infecting parasites; and second, the specificity of parasites on different hosts. It is often difficult to tease these two processes apart, but a recent review suggests that the failure to infect “nonhost” species (
i.e., those not considered to be hosts for the pathogen in question) may be the result of pathogen evolution leading to specialization on its own source host species and not the result of host evolution for resistance [
33]. By reviewing studies across many host-parasite systems, the authors find a general trend towards decreasing parasite infection success on hosts of increasing genetic distance from the focal host. It remains to be determined whether this pattern is ubiquitous for bacteria-phage interactions, especially given the broad host ranges of some phages (
Table 1).
The most supported evolutionary argument for why parasites specialize on given hosts, despite the clear advantage of a broad host range, is that there exists a trade-off between fitness and the breadth of parasite infectivity or host resistance [
34]: In other words, the idea that “a jack of all trades is a master of none.” This can be explained either by antagonistic pleiotropy, a situation where an adaptation that is advantageous in one host is deleterious in another, or else by selection for a less efficient but more general mechanism of infection. Support for this trade-off has been found for phage ϕ2, where individuals with broader host range within populations of its bacterial host,
Pseudomonas fluorescens, were shown to pay a cost for this increased breadth relative to phages with narrow host ranges [
35]. Specific evidence of antagonistic pleiotropy has also been found; during experimental host range expansion of phage ϕ6, spontaneous mutants able to infect novel hosts were found to be less infective to their native hosts in seven out of nine cases [
36]. These trade-offs are also likely to be common in host populations. Indeed, recent results from experimental evolution (see
Box 1) of
Prochlorococcus hosts and their associated phages demonstrate that resistance to one phage genotype often came with the added cost of increased susceptibility to another phage genotype [
37]. Similarly, experimental evolution of
P. syringae in either single phage or multiple phage environments shows that bacteria evolved with multiple phages paid a higher cost of resistance than those evolving with single phages [
38].
Box 1. Experimental evolution of phage specificity.
Experimental evolution of bacteriophage specificity has offered some key insights into the underlying process, evolutionary consequences, and fitness costs of host range expansion. The power of this method is that it allows replicate lines, started with genetically identical phages, to be passaged on homogeneous or heterogeneous hosts populations under a range of conditions (such as density and resources) for many thousands of generations (
Figure 1). During this time, both the host bacterium and the phage can be frozen in time and resurrected at the end of the experiment, at which point [
39] the fitness of evolved phages can be compared directly to both the ancestral types and phages experiencing a different selection regimen.
This approach has been used to demonstrate a number of key features of phage specificity, and has gone some way in explaining both when and how phage host range is likely to expand. First, in terms of range expansion within a host population (
i.e., the evolution of “generalist” phages capable of infecting more genotypes of a given bacterial species), there is evidence that phage ϕ2 is more likely to increase its host range during experimental coevolution with its bacterial host,
P. fluorescens, than when the bacterial population is held constant [
35]. Similar results were found in coevolving populations of phage SBW25ϕ2 and
P. fluorescens [
40]. Furthermore, the emergence of evolved “generalist” phenotypes of both bacteria and phages during experimental coevolution has been demonstrated both in a marine cyanobacteria and cyanophage system [
5] and in a
P. fluorescens and phage SBW25ϕ2 system [
41]. Second, in terms of host range expansion to novel hosts, experimental evolution has provided evidence that phage ϕ6 populations are more likely to evolve expanded host range when there is strong competition for hosts (
i.e., when the focal host is rare in a population) [
42]. It has also been shown that during the early stages of such a host shift, the likelihood of successful adaptation to a novel host is increased when contact with the native host is maintained, as this prevents extinction of the phage [
43].
An experimental coevolution approach can also be taken to identify the mutations underlying gains or losses of host types. For example, host range expansion of phage ϕ6 to a novel host was found to be associated with a single nucleotide change [
44]. Furthermore, phages experimentally coevolved with
P. fluorescens hosts evolved increased host range over time and the phage genotypes with the broadest host ranges were found to have the most nonsynonymous amino acid changes, especially in the phage tail fiber gene [
45]. Finally, experimental evolution of phage λ on populations of
Escherichia coli that had lost the receptor used for phage attachment were found to evolve the ability to infect the bacterial host via a novel receptor following the spread of key precursor mutations, suggesting that phage host shifting can occur via entirely new innovations [
46]. Clearly, the power of experimental evolution in understanding phage host range has not been fully exploited and moving forward this approach will offer further insight to the evolution of phage specificity in complex bacterial communities, fitness trade-offs between broad and narrow host ranges, and the potential limits of host shifting among phages.
Figure 1.
Illustrative example of experimental evolution of phage host range, where: (A) independent lines of genetically identical phage populations are propagated under different treatment regimens (e.g., different bacterial host species); and then (B) tested for infectivity on focal and alternate hosts. (C) Outcomes of these experiments might be a directional change towards increased host range over time (a), an initially increasing but then stable host range, perhaps indicative of coevolutionary response by the host population (b), or a decrease in host range associated with antagonistic pleiotropy during specialization on the focal host (c).
Figure 1.
Illustrative example of experimental evolution of phage host range, where: (A) independent lines of genetically identical phage populations are propagated under different treatment regimens (e.g., different bacterial host species); and then (B) tested for infectivity on focal and alternate hosts. (C) Outcomes of these experiments might be a directional change towards increased host range over time (a), an initially increasing but then stable host range, perhaps indicative of coevolutionary response by the host population (b), or a decrease in host range associated with antagonistic pleiotropy during specialization on the focal host (c).
1.1. The Structure of Bacteria-Phage Interaction Networks
Studies of phages in natural populations have confirmed that, like many other parasites, phages are well adapted to their local host populations [
3,
24,
47]. However, whether individual phages are specialized on certain genotypes/species, the frequencies of which differ across space, or whether coevolution is driving among-population divergence remains to be determined. Importantly, phage specificity can only be meaningfully evaluated within a culturable reference set of host and/or parasite genotypes; and a different reference panel of host or pathogen genotypes may reveal different levels of specificity. Choosing an appropriate reference set is often challenging, as the interaction networks and species ranges of most phages is not known, but it is of central importance to choose a panel that is biologically meaningful and/or informative to the predictions being tested.
Those studies that have looked at host range of individual phages from the environment demonstrate large variation in specificity, both within and across bacterial species. In fact, some phages that appear to be “generalist,” in the sense that they can infect bacteria spanning genera, fail to infect a subset of strains or species within genera (
Table 1). This apparent inconsistency, which is likely the result of both specific phage adaptations and the subsequent evolution of bacterial resistance in some lineages, including via transfer of plasmids, makes it difficult to decipher specific rules regarding phage host range. For example, populations of
Candidatus isolated from two separate sludge bioreactors were found to differ primarily in genomic regions encoding phage defense mechanisms, despite global dispersal of the strains among the two sites [
48]. It is only when many of these studies are compared and datasets are compiled that clear patterns emerge. A recent meta-analysis of the statistical structure of bacteria-phage interactions shows that the infection networks are non-random and are typically nested [
18]. This means that the highly resistant bacterial strains/species are only infected by broad host range phages, whereas the highly susceptible bacteria are infected by phages with narrow to broad host ranges. The observed lack of modularity may be suggestive of a true continuum of phage host range. Alternatively it might reflect the fact that most studies included in the analysis examine interactions either within a single bacterial species (
i.e., across multiple strains/genotypes) or between phages from one environment on bacteria from entirely different communities, habitats, and even continents. To determine the statistical structure of bacteria and phage communities in nature requires analysis of phage host range on representative bacterial hosts from the same local habitat.
1.2. The Evolutionary Implications of Phage Host Range
Given that phages are relatively specific, are capable of rapid adaptation, and are often obligate killers of their host cells, they can impose strong selection on bacterial populations and shape microbial communities. The “Kill the Winner” hypothesis posits that phages adapt to and preferentially infect the lineages of bacteria with the highest frequencies in the population, for example, those with higher metabolic fitness. Evidence for this hypothesis has been collected in a number of ways, including experimental coevolution of phages and bacteria in a test tube [
49], monitoring of population change over time [
50,
51], and using metagenomic approaches [
16]. Killing the winner is a form of negative frequency-dependent selection, as bacterial fitness is an indirect function of its frequency in the community. Specifically, bacterial species that are rare and free from phage attack will increase in frequency until the point at which an infective phage is introduced, either via mutation or migration, and spreads through the population. At this point, the common bacterial species will be at a relative disadvantage and may decrease in frequency.
Clearly, this type of dynamic is only possible if specificity underlying infection exists, but it also requires a time lag during which previously rare bacterial species can increase in frequency and remain free from phage attack. The length of this lag,
i.e., the time it takes for a phage mutant to arise by mutation or immigration, will dictate how common bacterial species can become before being targeted by coevolving phages. There is also likely to be a lag in the time it takes for a rare phage genotype to increase in frequency as its respective host becomes common, especially in populations where bacterial densities, and therefore rate of encounter, are low. Extending this theory to bacteria-phage interactions, we might predict that oscillatory dynamics should be more pervasive in relatively closed microbial communities, such as the human gut, than in highly connected communities, such as the ocean. This can be extended to predict that microbial communities with higher gene flow of bacteria and/or phages should show greater species evenness than closed communities, where frequencies are fluctuating over time. The data on phage infection of common bacterial species remains scarce, but the evidence we do have is in line with the above prediction. Estimates from marine communities suggest that cyanobacterial cell lysis by cyanophages ranges from a mere 0.005% to 3.2% per day, depending on the season [
52]. Similarly, for bacterial isolates collected from the surface of tree leaves, a habitat that is open to constant immigration, only 3% of bacteria were found to be susceptible to local phages. This is in stark contrast to the interior of those same leaves, a more protected and closed microenvironment, where 45% of bacterial isolates were found to be susceptible to local phages [
3]. It is important to note, however, that a number of differences exist among these habitats beyond the potential for immigration [
53]. A comparable result was found in natural soil samples, where 33 to 40% of bacteria could be lysed by phages from the same sample [
24]. Given the paucity of studies that have explored bacteria-phage dynamics in nature, the ubiquity of phage-mediated negative frequency-dependent selection remains unclear. Future studies exploring the natural prevalence of phage infection, coevolutionary dynamics of bacteria and phages over time, and the evolution of phage host range in either natural or experimental communities are still needed to predict how phages influence bacterial communities.
1.3. The Applied Implications of Phage Host Range
In addition to their potential role in shaping bacterial community composition, phages are key players in shaping the evolution of bacterial genomes [
54]. As lytic bacteriophages reproduce within the host and reassemble, bacterial chromosomal DNA can be inadvertently packaged into the viral capsid along with the viral DNA. This mistake will lead to
generalized transduction (in contrast to
specialized transduction by prophages) and can move chromosomal DNA from one bacterial host to another. During this movement among hosts, phages can transfer genes encoding toxins or virulence factors, and thus critically alter the bacterial phenotype. For example, phage-mediated transfer of pathogenicity islands between
Listeria monocytogenes and
Staphylococcus aureus has been demonstrated in raw milk [
55] and phage-mediated transfer of antibiotic resistance has been demonstrated among species of
Enterococcus [
56]. It is increasingly clear that transduction can occur across distantly related bacterial species, and even the seemingly highly conserved 16S rRNA gene has been found within the genome of a broad host range transducing phage [
57]. Thus understanding phage specificity among bacterial strains and species is key to predicting potential movement of genes across bacterial species and habitats, and thus the potential emergence of novel pathogens.
An understanding of phage specificity is also central to predicting the success and consequences of phage therapy,
i.e., the use of phage or cocktails of phages to control the growth and/or virulence of pathogenic bacteria. The utility of phage therapy is often called into question because of the apparent specificity of phages [
58]. However, this specificity is also a clear advantage of phage therapy over more general treatments, such as antibiotics, since the non-target bacterial populations should remain relatively undisturbed. The first steps in testing the benefits of a potentially therapeutic phage are to test a) whether the phage is too specific to be effective against the standing strain variation of a pathogen in a host population and b) the likelihood that the phage will affect other non-pathogenic bacteria, either immediately due to a large host range or over short evolutionary timescales as the phage evolves. For example, recent work from silage of dairy farms found a great deal of strain-to-strain variation in susceptibility of the food-borne pathogen,
L. monocytogenes, to phages collected from silage. They tested the host range of 114 listeriaphages and found that 12% of these phages had narrow host ranges and could infect fewer than half of the strains tested, representing the nine major serotypes of
L. monocytogenes. However, another 29% of the phages were capable of infecting nearly all of the strains tested, suggesting that these phages would be good candidates for therapeutic control of the pathogen [
59]. Furthermore, given the ease of full genome sequencing, it is now possible to scan the phage genome for virulence factors and known toxin-encoding genes to ensure the phage will not act to increase the harm caused by a given pathogen.