Antibiotic resistance in bacterial communities

Abstract

Bacteria are single-celled organisms, but the survival of microbial communities relies on complex dynamics at the molecular, cellular, and ecosystem scales. Antibiotic resistance, in particular, is not just a property of individual bacteria or even single-strain populations, but depends heavily on the community context. Collective community dynamics can lead to counterintuitive eco-evolutionary effects like survival of less resistant bacterial populations, slowing of resistance evolution, or population collapse, yet these surprising behaviors are often captured by simple mathematical models. In this review, we highlight recent progress — in many cases, advances driven by elegant combinations of quantitative experiments and theoretical models — in understanding how interactions between bacteria and with the environment affect antibiotic resistance, from single-species populations to multispecies communities embedded in an ecosystem.

Introduction

Antibiotic resistance is a growing threat to public health, as the rise of resistance outpaces the development of new antimicrobial drugs. Decades of progress in molecular biology, microbiology, and biophysics has driven us deep within the bacterial cell, where we increasingly understand the molecular events — the enzymatic reactions that hydrolyze drugs, or the molecular machines shuttling toxic molecules out of the cell — that separate bacterial survival from death. Yet bacteria do not live as isolated cells, but as ever-evolving members of complex communities with potentially billions of neighbors. Within these communities, physical and chemical interactions between cells impart complex spatial and temporal dynamics to the miniature bacterial ‘civilizations’ that comprise human infections [1–4]. These collective effects may have a profound impact on strategies to eradicate or control bacterial growth. While antibiotics directly target intracellular functions, such as protein synthesis, that are required for survival of a single cell, these drugs frequently originated as tools of molecular warfare pioneered by microorganisms [5], and are thus themselves a byproduct of community interactions. A deeper understanding of the interactions that shape the microbial communities may therefore hold clues for slowing, and even reversing, antibiotic resistance [6].

Predicting how bacterial communities and antibiotic treatments will interplay in a specific environment is a significant challenge, in part because the complexity of multispecies communities grows exponentially as the number of microbial and chemical players increases [7]. Yet a growing body of research implicates collective behaviors of bacterial populations (single species) and communities (multiple species) in the emergence of antibiotic resistance. We are far from a comprehensive, mechanistic model of these complex communities, but recent advances in both experiment and theory are highlighting ecological and evolutionary principles that shape community composition (Figure 1a) and may eventually be leveraged to prolong the effectiveness of currently available antibiotics. These breakthroughs reveal interconnected dynamics across length scales, tying physical interactions between single cells to the fate of communities on evolutionary horizons. In this review, we highlight a subset of recent advances — with particular emphasis on those inspired by mathematical and physical models — with the potential to inspire new (bacterial) community-based strategies for slowing resistance.

Figure 1.

Antibiotic resistance integrates properties across different scales of bacterial life. (a) Molecular mechanisms at the single-organism level (e.g. enzymes or efflux pumps) are an important component of antibiotic resistance (top left). Antibiotic resistance is, however, also influenced by interactions in both homogeneous and physiologically or genetically heterogeneous single-species populations, such as competition and cooperation (top right). Multispecies communities of bacteria can contain manifold interspecies interactions, such as cross-feeding, antibiotic protection or active mutual inhibition, whose balance significantly determines community composition and response to antibiotic (bottom right). In addition, how communities aggregate and organize in space, as well as how they interact with other components of their ecosystem, are important contributing factors to observed antibiotic resistance (bottom left). (b–d) Examples of community effects include antagonistic resistance in heterogeneous populations [18] (b), weakest-link dynamics in cross-feeding communities [30] (c), and antibiotic-induced aggregation and expulsion of bacterial aggregates from the host gut [46] (d).

Collective effects in single-species populations

Antibiotic resistance occurs across different scales of bacterial life. At the single-cell level, a suite of molecular defense mechanisms — for example, drug degrading enzymes or efflux pumps — protect bacteria from antibiotic [8]. But as microbiologists have long recognized, environmental context can change the bacterial response to antibiotic. Perhaps the simplest example is the well-known inoculum effect: The minimum inhibitory concentration of an antibiotic depends on the initial (inoculating) density of bacteria in media [9–11]. Although typically cell density increases antibiotic resistance, it decreases resistance to some antibiotics [10]. The mechanisms for this phenotype are diverse — including enzymatic degradation of drug [12], growth bistability driven by the heat-shock response [11], or modulation of environmental pH [10], an effect potentially exacerbated in multispecies communities [13,14].

The density dependence of antibiotic activity may also impact ecological and evolutionary dynamics among subpopulations with varying levels of resistance. Antibiotic-resistant cells can either reduce [15–17] or accentuate [18] the direct effects of antibiotics on nearby susceptible cells. Interestingly, these effects do not always require the secretion of diffusible molecules into the extracellular media; for example, β-lactamase producing Streptococcus pneumoniae [13] and Enterococcus faecalis [19] function as local reaction sinks for diffusing antibiotic, even when the enzyme remains associated to the producing cells. Similarly, drug efflux from individual Escherichia coli can increase the local drug concentration for neighboring cells [18] (Figure 1b).

‘Cooperative’ resistant cells can also raise the antibiotic threshold for the spread of resistance-gene carrying plasmids to susceptible cells [20], and subtle feedback loops between cooperating subpopulations can trigger counterintuitive dynamics such as population collapse [21]. Conversely, resource-consuming susceptible cells are able to indirectly slow the growth of resistant populations, and time-dependent drug dosing schemes can, in principle, exploit this competition. For example, recent experiments in computer-controlled bioreactors showed that doxycycline-resistant E. coli populations can be held in check by doxycycline when dosing is designed to maintain a competing subpopulation of sensitive cells [22]. Mathematical models suggest this competitive suppression may apply beyond the laboratory, and it serves as the basis for adaptive therapies targeting infectious diseases and cancer [23,24].

Multispecies community interactions

Interactions between susceptible and resistant bacteria can also modulate the dynamics of more complex communities, ranging from small, well-defined communities with a few species to highly complex microbiome assemblies. Cooperative resistance can pose a particular problem for antibiotic treatment when endogenous bacterial species protect susceptible pathogens [25]. For example, Stenotrophomonas maltophilia have been shown to offer enzyme-mediated protection from the antibiotic imipenem to pathogenic Pseudomonas aeruginosa in a multispecies model of cystic fibrosis [26]. This effect is dependent on a drug-specific trade-off between killing efficacy and the rate of enzymatic degradation of antibiotic. Exchanging imipenem for a closely related carbapenem alters this trade-off, leading to a strikingly different outcome where the susceptible species is no longer protected [26]. In contrast, mathematical models indicate that antibiotic resistance in commensals can be beneficial for control of a more susceptible pathogen during coinfection when the interaction between commensals and pathogens is competitive or exploitative [28].

In some cases, interactions between cohabiting species can lead to counterintuitive outcomes such as the more susceptible species taking over the community. For example, short-term exposure to a range of antibiotics (chloramphenicol, gentamycin, or ampicillin) is able to drive a bistable community of mutually inhibiting species (Corynebacterium ammoniagenes and Lactobacillus plantarum) to recover to a state dominated by the more drug susceptible, but faster growing species (C. ammoniagenes) [27]. A similar effect can also be driven by antibiotic tolerance (characterized by growth after lag time) rather than genetic resistance. In mixed cultures of Bacillus subtilis and E. coli, only the latter proliferates in the presence of β-lactams; by contrast, in mono-cultures B. subtilis is tolerant while E. coli is susceptible [29]. These ‘antithetic’ responses can be predicted from a kinetic model that captures binding of β-lactam to its intracellular targets (penicillin binding proteins) in the two species, which exhibit differential rates of antibiotic deactivation. In mixed cultures, the common pool of drug molecules harms the species that inactivates antibiotic faster (B. subtilis) and benefits the slow inactivator (E. coli). As a whole, these results indicate that the susceptibility profiles of individual species alone do not determine community-level dynamics, and illustrate the power of simple ecological and biophysical models for predicting these emergent community behaviors.

In addition to ecological dynamics, community-level interactions can directly affect how antibiotic resistance evolves. Microbial communities with mutualistic cross-feeding interactions can exhibit ‘weakest link’ dynamics, in which the most antibiotic susceptible community member defines the antibiotic susceptibility of the whole community [30]. In a cross-feeding community of E. coli and Salmonella enterica, this ecological property of mutualism changed the evolutionary emergence of resistance. Under two different antibiotics, ampicillin and rifampin, cocultures displayed a slower increase in resistance than monocultures of each species. Under ampicillin, both the speed and molecular mechanisms of adaptation differed in co- versus mono-culture. A mathematical model suggests that the slowing of resistance evolution by mutualism may be a general mechanism, because rare large-effect mutations are less advantageous in coculture, when susceptibility is still dependent on a more susceptible partner species [31] (Figure 1c).

Translating results from simple communities with a handful of species — where the dynamics, and increasingly, even the underlying molecular or mathematical mechanisms can be teased apart — to complex natural communities is an enormous challenge. However, a number of recent breakthroughs are pointing a way forward by uncovering surprisingly simple principles — based, for example, on pairwise approximations, geometric constraints, or stochasticity — that drive population-level outcomes, including composition [32–36], multicellularity [37], and complex metabolic inter-dependencies [38,39]. Whether similar simplifying principles describe the response of complex communities to antibiotics is an open question.

Innovative top-down approaches have started to decipher how pathogen interactions and antibiotic resistance behave within natural microbiome communities, which are notable not only for their immense complexity, but also for the fact that many species cannot even be cultured individually in the lab. For example, multiple studies have found that microbiomes can reduce antibiotic resistance evolution or selection to different antibiotics in E. coli [40,41], and community influences can uncouple selection for resistance from antibiotic concentrations above a certain threshold [42]. Suggested mechanisms included a higher cost of resistance, but also protection of susceptible bacteria, which reduces selective pressure. Understanding these complex communities will require continued efforts that draw on the success of both bottom-up and top-down approaches to ecological and evolutionary dynamics.

Spatial, mechanical, and environmental effects on resistance

Bacterial populations exist within larger, complex ecosystems, and thus the populations’ compositional complexity is shaped by the physical environment in which they are embedded. Classical ecological interactions — such as cross-feeding or enzymatic protection — are governed, in part, by diffusion of intercellular signaling molecules, which can be heavily influenced by spatial and temporal heterogeneity.

Further, most bacterial life exists in biofilms, dense surface-attached populations. The biofilm lifestyle has long been known to influence how cells respond to antibiotics [43], in part because of physical limitations imposed on diffusive signaling; antibiotic exposure can also affect biofilm structure [44,45]. Recent work in zebrafish has shown that antibiotics can modulate aggregation of gut bacteria, which in turn increases their expulsion from the gut [46] (Figure 1d). Mathematical models suggest that statistical properties of these gut aggregates arise from a set of simple, and common, biological ‘ingredients’: the growth of bacteria within clusters, the escape of cells from the clusters, and a multicluster aggregation process that drives the system to a single large cluster, making these populations fascinating examples of living gels [47]. Aggregation has also been observed to occur as a community interaction between Staphylococcus aureus and P. aeruginosa in co-culture model of cystic fibrosis, which alters the community’s antibiotic resistance [48].

Spatial segregation between species of a community is considered an important evolutionary source of cooperation — allowing producers of public goods, such as antibiotic-degrading enzymes, to preferentially benefit [49,50]. The spatial arrangement of resistant cells may also provide strong synergistic ecological benefit [51], allowing mixed communities to exhibit properties of purely resistant colonies when only a small fraction of the colony surface includes resistant cells [19]. A wide range of physical interactions, including mechanical crowding [52], cell motility [53], and electrical signaling [54], are increasingly recognized as drivers of ecological and evolutionary dynamics. It remains to be seen whether, and how, these interactions shape the response of cells to antibiotics.

Conclusions and future directions

In this review, we aimed to highlight recent advancements in our understanding of the eco-evolutionary dynamics of bacterial communities. We focused specifically on the interactions of these communities with antibiotics, and on studies that used experiments and modeling to extract quantitative and predictive insights. The work, as a whole, underscores the importance of interrogating antimicrobial resistance at a systems level, where interactions across scales give rise to complex, and often counterintuitive, dynamics. While simple mathematical models have proven remarkably successful at describing and predicting microbial dynamics, particularly in simple populations, these coarse-grained models are largely agnostic to the species-specific properties that are undoubtedly important as we look on finer scales. Connecting the molecular and biophysical events that occur within the cell to emergent community level properties will continue to be an important challenge, and one that, in the long-run, may inspire new therapeutic strategies — for example, personalized antibiotic regimens based on machine learning [55] or control theoretic principles [56] — aimed at eradicating or controlling microbial communities.

Data Availability

No data were used for the research described in the article.