Trading-off and trading-up in the world of bacteria–phage evolution

Abstract

In 1979, Richard Law introduced the conceptual idea of the ‘Darwinian Demon’: an organism that simultaneously maximizes all fitness traits [1]. Such an organism would dominate an ecosystem, displacing any competitors and collapsing biodiversity to only a singular species. Surveying the tremendous species diversity of bacteria in the microbial world reveals that Darwinian Demons do not exist on Earth, and the popular notion is that fitness trade-offs generally constrain such possible evolution. However, the trade-offs faced by evolving bacterial populations presumably hinder their adaptation in ways that are not fully understood. In some cases, bacteria show evolved trade-ups, whereby selection causes multiple fitness components to improve simultaneously. Understanding these trade-offs and trade-ups, as well as their prevalence and roles in shaping microbial fitness, is key to elucidating how the incredible diversity of the Bacteria domain came to be, what maintains that diversity, and whether such diversity can be leveraged for technologies that improve human health and protect environments.

Law applied his Darwinian-Demon thought exercise mainly to animal prey populations facing selection imposed by their predators [1]. He reasoned that predators would exert selection pressure on prey populations, forcing life-history evolution along a trade-off between reproduction and survival. For example, prey may evolve various traits that improve defense or avoidance of predator attack, thus necessarily devoting resources (metabolic and physiological effort) to survival that otherwise could be used to enhance reproduction. Analogous selection occurs in bacterial populations facing attack by ‘predatory’ lytic bacteriophages: viruses that specifically infect and lyse (kill) their host cells. Indeed, bacteria–phage evolution experiments are frequently used as empirical models to study the generalities of such life-history trade-offs [2–5]. In contrast, these microbial systems are less often harnessed to explore the potential role of phage-imposed selection for evolved trade-ups in host bacteria.

Here, we discuss how lytic phages can select for evolution of phage resistance in bacteria, and how this evolution entails changes to other phenotypic traits — an evolutionary effect known as pleiotropy. Pleiotropy can occur when evolution of a single gene impacts multiple traits; for example, when evolved phage resistance in bacteria decreases uptake of resources, thereby creating a trade-off (‘antagonistic pleiotropy’). We discuss impacts of lytic-phage selection on bacteria, especially how these interactions can influence evolution of greater or lesser bacterial resistance to antibiotics. Finally, we consider the ‘big questions’ that future research should address to better understand how bacterial resistance to phages and antibiotics can evolutionarily interact to influence human health and natural ecosystems.

Pleiotropy in the evolution of bacteria–phage interactions

Trade-offs between phage resistance and other key traits in bacteria have been studied for nearly a century and occur when bacterial evolution results in phage resistance at the cost of another fitness-generating trait. Fitness components most commonly altered by evolution of phage resistance include slowed rates of cell division, reduced nutrient uptake, lowered virulence, altered antibiotic resistance, diminished biofilm formation and lessened cell motility. (See [6] for a review of other affected traits, especially those relevant in biomedicine.) These types of trade-offs are manifest because phage gain entry to their host cells by utilizing specific structures that generally function directly in other bacterial traits [6,7].

Trade-offs between bacterial resistance to phages and to antibiotics are particularly relevant for both therapeutic applications and as models for general evolutionary processes. As lethal selective agents, both lytic phage and bactericidal antibiotics offer the potential to kill bacteria, thereby exerting strong selection for bacteria to evolve resistance. Thus, phage–antibiotic–bacteria interactions create powerful systems for the generalized study of pleiotropy in well-controlled laboratory microcosms [8].

Of particular interest to basic and applied trade-off biology are phages that rely directly and specifically on cellular receptors that function in antibiotic resistance, such as multi-drug efflux pumps (Figure 1). These protein complexes span the cell membranes and are capable of actively removing various antibiotics, thus contributing to the widespread failure of antibiotics to control bacterial infections. When phages use efflux pump components (such as the outer membrane proteins TolC and OprM) as receptors, bacteria face selection for lost or altered pump alleles that eliminate or slow phage adsorption. When the evolution of resistance takes this route, the antibiotic efflux function may also be impaired, potentially sensitizing bacteria to antibiotics.

Figure 1. Examples of pleiotropy involving phage resistance and antibiotic resistance.

Scenarios A, B, and C represent examples of evolutionary trade-offs, whereas scenario D is an example of a trade-up. (A) Import–export pleiotropy occurs when a single structure (for example, an antibiotic efflux pump encoded by ‘allele A’) mediates phage genome entry into the cell and antibiotic export from the cell. Loss or modification of the structure (‘allele B’) results in phage resistance and antibiotic sensitivity. (B) Import–import pleiotropy occurs when a single structure (in this case, capsular polysaccharide encoded by ‘allele A’) mediates phage genome entry and serves as a barrier against an antibiotic. (C) Import–barrier pleiotropy occurs when a single structure (for example, a specific form of the barrier molecule LPS) mediates phage genome entry and has intrinsic resistance to antibiotic disruption; another form of the molecule (conferred by ‘allele B’) may block phage genome entry but can be disrupted by the antibiotic directly. (D) A second example of import–import pleiotropy in which the same structure (again LPS, ‘allele A’) mediates phage entry, but in this case also mediates antibiotic entry, either through cellular permeability or regulation of antibiotic import proteins; the alternative LPS encoded by ‘allele’ B blocks both phage entry and antibiotic entry.

Generally, the relationship between phage selection and the evolution of antibiotic-resistant bacterial phenotypes has only recently been studied, and a few well-characterized phages show the ability to drive these evolutionary trade-offs. The first-such characterized phage, TLS, adsorbs to TolC component of Escherichia coli efflux pumps, exerting selection for evolution of phage-resistant bacterial mutants with altered TolC proteins [9]. Similarly, phage U136B also directly relies on the TolC protein of E. coli to initiate infection, selecting for bacteria that have evolved resistance via loss or modification of tolC, which then confers altered resistance to the antibiotics tetracycline and colistin [8] (Figure 1A). Phages LL5 (in E. coli) [10] and ST27, ST29, and ST37 (in Salmonella enterica serovar Typhimurium) [11] also rely on TolC for infection, and phage OMKO1 associates with the homologous OprM efflux-pump protein in Pseudomonas aeruginosa [12,13]. In these cases, it’s predicted that tolC and oprM mutants may be selected for, either because phages use them directly as receptors or because these genes positively regulate the phage receptor genes.

Aside from efflux pumps, a variety of other cell-surface structures in bacteria relate less-directly to antibiotic resistance traits, but show the potential to mediate trade-offs between evolved phage resistance and antibiotic resistance. For example, phages øFG02 and øCO01 infect Acinetobacter baumannii (an opportunistic pathogen that infects the lungs, bloodstream, wounds, and urinary tract) through its capsular polysaccharide, which confers protection against beta-lactam antibiotics [14]. Selection by øFG02 and øCO01 resulted in the evolution of phage-resistant A. baumannii mutants with increased sensitivity to beta-lactam antibiotics, as well as consequential changes to other clinically relevant traits [14] (Figure 1B). Other phages rely on binding to another cell-surface structure — the bacterial lipopolysaccharide (LPS) [15], a structure that can indirectly determine antibiotic resistance. LPS can mediate cell-membrane permeability to antibiotics and also interacts with other membrane structures that transport antibiotics into the cell. Beyond possible interactions with antibiotics, LPS crucially determines which host species a bacterial pathogen can infect, as well as clinically relevant phenotypic traits such as antigenicity and virulence. Understanding the evolutionary implications of phage–LPS interactions will be important for predicting if and when phage selection on LPS changes bacterial virulence and drug resistance phenotypes and how these trade-offs are mediated in nature. For example, phage U136B relies on both TolC and LPS of E. coli, and phage resistance can readily evolve through changes to LPS without any effects on TolC efflux function [8]. These mutants also have greater sensitivity to the antibiotic colistin (Figure 1C). This trade-off between LPS-based antibiotic resistance and phage sensitivity can be bidirectional: in the opportunistic bacterial pathogen Klebsiella pneumoniae, selection by colistin results in evolution of increased drug resistance that coincides with greater sensitivity to phage øNJS1 [16].

Despite clear links between antibiotic resistance structures and phage infectivity, trade-offs do not always evolve predictably. In contrast to phage-resistance mutations that incur an antibiotic-resistance trade-off, other mutations that confer phage resistance can include those that allow escape from phage attack while maintaining wild-type antibiotic resistance levels. These outcomes suggest that bacterial populations might readily evade phage–antibiotic resistance trade-offs through specific mutations [8,9]. Additionally, phages may rely on more than one cell structure for cell entry, such that evolved phage resistance may take varied (and unexpected) mutational paths, arising through selection on multiple different targets.

In contrast to genetic trade-offs, trade-ups have been less well studied and characterized. One explanation is that much of adaptive evolution is fundamentally constrained by antagonistic pleiotropy, such that trade-ups — and the Darwinian Demons they could potentiate — are unlikely to evolve. Or, trade-ups may be simply under-studied because they tend to be less consequential for development and use of bacteria in industrial applications. Moreover, trade-ups may be less readily observed in the laboratory. For example, many trade-offs may be obvious to experimenters, through observations of small colony variants, reduced growth rates, and abnormal growth of overnight cultures. Furthermore, trade-offs may present lethal evolutionary options to bacterial populations: for lytic phages and bactericidal antibiotics, resistance can be ‘all-or-none,’ an easy-to-screen phenotype. In contrast, the fitness gains obtained with trade-ups may be much smaller and therefore harder to detect, perhaps only revealed through quantitative measures of fitness components.

Nevertheless, some examples of trade-ups between evolved phage resistance and antibiotic resistance in bacteria are known, and there are certainly many more waiting to be discovered. For example, selection using the phage 14/1 can result in the evolution of phage resistance that coincides with increased antibiotic resistance in the opportunistic lung and wound pathogen P. aeruginosa [17]. Trade-ups may also evolve even in cases where phage directly exploit antibiotic resistance genes, such that both trade-offs and trade-ups can happen in the same system. This is true for phage U136B of E. coli, which relies on both TolC and LPS for cell binding; unlike with tolC mutations (which confer antibiotic susceptibility, as described above), in cases where phage resistance arises via LPS changes, bacterial mutants show increased resistance to tetracycline [8] (Figure 1D). Such trade-ups pose a clear challenge for the use of phages in human therapy, where the evolution of phage resistance is inevitable, as discussed further below.

How common are trade-offs and trade-ups, and how might we find them?

To better understand how interactions between antibiotic resistance and phage resistance in bacteria may play out in ecosystems, we need to examine possible ecological drivers that dictate prevalence and co-occurrence of these phenotypes. There have been relatively few studies documenting phages that rely directly or indirectly on efflux pumps and other structures that impact antibiotic resistance and sensitivity in bacteria. However, these viruses can be readily found when screening modest-sized collections of natural samples [8]. Therefore, we suspect that their poor representation in the literature reflects an under-characterization of phage biodiversity, rather than low abundance in natural microbial communities. For example, phage U136B is an isolate from a Connecticut swine farm, an agricultural setting where antibiotics are conventionally applied and sometimes over-used. Would such an environment constitute a favorable one to bioprospect for naturally occurring phages with useful biotechnology properties — that is, a phage capable of synergistically interacting with antibiotics to kill target bacteria while selecting for antibiotic re-sensitivity? This is currently unknown, making it uncertain whether presence of the phage reflects an underlying pattern in phage evolution in certain environments, versus a serendipitous discovery that comes from efforts to catalog phage biodiversity. The answer could come from studies that systematically compare large numbers of phages drawn from geographies with perennially low versus high use of antibiotics. More generally, such hypothesis-driven screens would be useful for future efforts to discover naturally occurring phages with potential utility in therapy and other applied goals.

In addition to looking for phages that use specific factors such as efflux pump proteins that function in antibiotic resistance, it is crucial to examine the potential pleiotropy of phage resistance in relation to loci throughout the genomes of bacteria. In E. coli, for instance, tolC is just one of 283 genes that contribute to antibiotic resistance [8]. Each of those other genes may have the potential to interact with phage infectivity, for example, by coding for cell-surface structures used by phage when attaching to the host, or those that affect phage reproduction through central metabolism of the cell. High-throughput and large-scale screens of bacterial libraries challenged to grow in the presence of phage collections and medically approved antibiotics are required to begin mapping both trade-offs and trade-ups across the bacterial genome. Knockout libraries (for example, the E. coli Keio collection [18,19], the Salmonella enterica sv Typhimurium knockout collection [20], and Tn-seq libraries [19,21]) will be especially useful tools in this pursuit. We predict that these types of screens might facilitate identification of phages with the potential to pleiotropically select against drug-resistance genes in bacteria. At the same time, such screens would help address more fundamental and basic questions in phage–bacteria interactions, especially whether trade-offs between evolved phage resistance and antibiotic resistance commonly occur in the microbial world versus in only a handful of circumstances that have been fortuitously observed in recent studies.

Can phage-driven trade-offs be applied at a large scale?

If sufficiently strong pleiotropic interactions between phage resistance and antibiotic resistance can be identified, they offer the potential for applications in human health, particularly to reduce densities of infectious bacteria while also driving evolution of restored antibiotic susceptibility [6,8,12,14]. However, given the complexities of possible evolutionary responses and antibiotic specificities, further basic research is needed to understand how environmental conditions might influence these interactions. In particular, biofilm structures, nutrient availabilities, host immune systems, and microbial communities (microbiomes) may influence the outcomes of phage selection. Those outcomes may include alternative phage-resistance mutations in bacteria that may have unique costs to normal cellular functions. It will also be highly useful to compare classic mutation screens with outcomes of experimental evolution studies. Evolution experiments reveal that different pleiotropic outcomes can emerge when multiple bacterial mutants must compete to fix in the population [8]. Such results reveal that the mutational options governing phage resistance in a genetic screen are narrowed when bacteria must vie for resources [8]. Understanding this more complicated ecological context will be key for figuring out how we might nudge evolution in a preferred direction to achieve the desired trade-off in target bacteria, and how to steer evolution away from possible trade-off evasion or the occurrence of evolved trade-ups.

Once basic principles are established, further laboratory experiments can be used to refine predictions of how phage applications and their outcomes will scale-up to clinical settings. Scaling up to in vivo trials offers a major challenge for any new drug therapy. The particular challenge to developing predictable evolution-based phage therapies is the possibility for genetic responses to exceed those observed in laboratory settings, due in part to the larger population sizes of target bacteria and administered phages. Importantly, any phage-based therapy will also encounter the complexity of the host immune system and interact with hundreds to thousands of other bacterial and phage species present in the human body, perhaps interacting with other concurrent therapies as well. All of these factors have the potential to shift the direction of bacterial evolution in ways that have not been explored in the laboratory. Having a better understanding of this genetic uncertainty is crucial not only to the outcome of a specific therapeutic intervention, but also has public health considerations, as bacteria that evolve under a specific treatment can later spread to other patients and the community. To address these challenges, it would be ideal to monitor the phenotypic and genomic evolution of target bacteria within clinical-trial patients before, during and after phage therapy. Mapping the potential adaptive landscapes of bacteria in this way will help build confidence that the scaling-up of novel therapeutic approaches can be done safely and effectively. In particular, monitoring patients’ microbiotas — both the commensal and pathogen populations — over time promises to reveal the variation not observable in constrained laboratory settings. These molecular and phenotypic data should offer insights into the potential for phages to drive trade-offs and trade-ups.

How does pleiotropy throughout the natural world affect how diversity is evolved, maintained, and lost?

In addition to the potential applications to improve human health, phages and the trade-offs they generate can also serve as model systems for studying how biological diversity is generated and maintained, not only in microbial ecosystems, but also in traditional macroscopic ecosystems. To understand how life evolved from the last universal common ancestor of cellular life, we must study the evolutionary processes that led to the origin and maintenance of earth’s biodiversity. Bacteria and phages have long been used as model systems to study core evolutionary and ecological processes, including the generation of phenotypic and genetic diversity and the origin of novelty [22–25]. The use of phages that drive trade-offs and trade-ups with antibiotic-resistance now offers an opportunity to study the role of pleiotropy in the evolution of diversity and its maintenance. Through mutation screens and evolution experiments conducted under varied conditions, we can now begin to understand how bacteria and phage influence one another’s natural histories. Such experiments can also be integrated with microbial ecology by introducing more complex bacterial community interactions in the lab. These experiments will help us understand how focal antagonistic interactions impact ecosystem diversity, function, and robustness. From there, we can continue to incorporate valuable but missing nuances to our models of the history of life on Earth, and we can make predictions about how it will continue to evolve.