Antibiotic resilience: a necessary concept to complement antibiotic resistance?

Abstract

Resilience is the capacity of systems to recover their initial state or functions after a disturbance. The concepts of resilience and resistance are complementary in ecology and both represent different aspects of the stability of ecosystems. However, antibiotic resilience is not used in clinical bacteriology whereas antibiotic resistance is a recognized major problem. To join the fields of ecology and clinical bacteriology, we first review the resilience concept from ecology, socio-ecological systems and microbiology where it is widely developed. We then review resilience-related concepts in microbiology, including bacterial tolerance and persistence, phenotypic heterogeneity and collective tolerance and resistance. We discuss how antibiotic resilience could be defined and argue that the use of this concept largely relies on its experimental measure and its clinical relevance. We review indicators in microbiology which could be used to reflect antibiotic resilience and used as valuable indicators to anticipate the capacity of bacteria to recover from antibiotic treatments.

1. Introduction

Bacterial systems are complex and adaptive. When faced with disturbances such as antibiotic treatments, they survive, recover and evolve. The capacity of systems to recover their initial state, some of their functions or composition after a disturbance is called resilience. In ecology, the concepts of resilience and resistance are complementary: while resistance refers to the insensitivity to disturbance, resilience focuses on the recovery of the disturbed system [1,2]. Both resistance and resilience are two different aspects of systems' stability. However, in clinical bacteriology, antibiotic resilience is not used whereas antibiotic resistance is overwhelmingly studied because multidrug-resistance is currently a major societal challenge [3]. Instead of using resilience, in clinical bacteriology, several other terms have been developed to describe bacterial responses to antibiotic treatments such as tolerance, persistence, phenotypic resistance or recalcitrance. The interest of the scientific community in these mechanisms is gaining importance because they may precede the emergence of antibiotic resistance [4–6].

One reason for the lack of interest in the resilience concept in clinical bacteriology is the fact that this concept is often ambiguous. The resilience-based framework, initially developed by Holling [7], suggested numerous concepts and ideas to structure the reactions of systems to disturbances. However, the concepts used, mainly resistance and resilience, are vague and potentially conflicting if they are not clearly defined [8,9]. Ecology and socio-ecology have assigned various definitions to resistance and resilience. For instance, resistance can be ‘staying essentially unchanged despite the presence of disturbances’ and resilience ‘returning to the reference state (or dynamic) after a temporary disturbance’ [10, p. 325]. Elsewhere, Walker et al. [11, p. 1] define resilience as ‘the capacity of a system to absorb disturbance and reorganize while undergoing change so as to retain essentially the same function, structure, identity and feedback’. Another paradigm of resilience is the viability-based framework of resilience which focuses on the recovery and maintenance of properties of interest instead of pre-disturbance states [12–14]. In microbial ecology, the focus is usually on the capacity of microbial communities to maintain or recover their pre-disturbance functions and species composition [1]. In clinical bacteriology, Meredith et al. [15] recently developed a framework of antibiotic resistance and resilience and applied it to characterize the response of bacterial populations expressing β-lactamases and exposed to β-lactam treatments. While the framework of resistance and resilience they propose fits well with their case study, its general application remains questionable.

To develop the concept of antibiotic resilience for clinical bacteriology, we review the concept of resilience used in the fields of socio-ecological systems and microbiology. We then review resilience-related concepts in microbiology, including bacterial tolerance and persistence, phenotypic heterogeneity and collective tolerance and resistance. We argue that the applicability of the concept of antibiotic resilience largely relies on representative experimental indicators and we discuss how antibiotic resilience could be measured and used to classify the response of bacterial strains to antibiotic treatments. Such an indicator, in addition to the commonly used minimal inhibitory concentration (MIC) to assess antibiotic resistance, could be a valuable tool to anticipate the success or failure of antibiotic treatments.

2. What is resilience?

The resilience concept is widely developed in the field of socio-ecology and has various meanings. In microbiology, the use of this concept is more recent and mostly restrained to the field of microbial ecology. To have a global view of resilience, we review the main concepts from both fields.

(a). Borrowing concepts from socio-ecological systems

Resilience can take fundamentally different meanings. Holling [16] differentiates two kinds of resilience: engineering resilience and ecological resilience. Engineering resilience is the speed with which a set of given variables recovers after a disturbance. This definition assumes a single equilibrium and does not take into account possible shifts in system behaviours. By contrast, ecological resilience is the amount of disturbance a system can endure before it shifts to an alternative regime with different properties. It assumes that ecological systems can have multiple equilibria and shift from one to another under the effect of disturbances [11,17]. In this case, resilience is characterized by the size of the system's basin of attraction, i.e. the set of states from which the system tends to a unique equilibrium, also called a stable state or attractor [11,17].

Carpenter et al. [18] add that it is necessary to specify what is resilient to what disturbances: ‘resilience of what to what?’ To be operational, it is necessary to state clearly which properties of a system we take into account, and what are the disturbances we consider. Hence, resilience is not the resilience of a whole system but the resilience of a restricted number of properties. This has also been called specific resilience [19]. For Folke [19, p. 260], resilience is more useful as an abstract approach. He differentiates between ecological resilience, focused on the persistence of different equilibria, and socio-ecological resilience, focused on adaptation, transformation, learning and innovation. He states that ‘the resilience approach is concerned with how to persist through continuous development in the face of change, and how to innovate and transform into new more desirable configurations'. In this view, disturbances are necessary to create opportunities and recombination, themselves necessary for systems to evolve.

The viability-based framework of resilience, which is not specific to socio-ecological systems, is oriented towards a property of interest instead of attractors [20]. Unlike specific resilience, which is centred on attractors, viability-based resilience is defined as the capacity to recover and maintain a property of interest after a disturbance [14], implying that this property has been lost at some point. Thus, systems can adopt various distinct behaviours, stable or not, and still be resilient as regards the chosen property of interest. Considering this definition, the capacity of a system to change its behaviour, i.e. its composition and interactions, is a key factor of resilience.

Different meanings of resilience are represented in figure 1. Two main streams emerge from the socio-ecological literature on resilience. One focuses on the speed of recovery (engineering resilience) or maintenance (ecological resilience) of a reference state, and the other on the resilience of properties of interest that can be lost and recovered and allows changes in the system's state.

Figure 1.

Different meanings of resilience. Red lightning: disturbance. (a) Engineering resilience: time necessary to recover the initial state after disturbance. (b) Ecological resilience: amount of disturbance necessary to shift the system to a different state. (c) Viability-based resilience: capacity to recover and maintain a property of interest (here the green-shaded area) after having lost it. The state of the system may change. Here, the system moves from state A to state B, both desirable as they maintain the property of interest. (Online version in colour.)

(b). Resilience in microbiology

In microbiology, the resistance/resilience framework is mainly used in the field of microbial ecology. This field usually focuses on the resistance and resilience of the composition and the functions of microbial communities faced with particular disturbances [1,21,22]. Resistance is defined as the insensitivity to a disturbance and resilience as the time required for a community to recover its former composition and functions after a disturbance. Both resistance and resilience are parts of the concept of robustness, i.e. the capacity of a biological system to maintain particular functions despite disturbances [23]. However, the compositions and functions chosen vary greatly between studies owing to the great diversity of microbial communities. For instance, resilience is much used for microbiome studies to refer to the capacity of the gut microbiota to recover its former composition and functions after exposure to antibiotics [24].

Song et al. [25] recently suggested grouping together ecological and engineering resilience by defining resilience as the recovery of a function of interest of microbial communities. This point of view is close to the viability-based view of resilience which defines resilience as the capacity to recover and maintain a property of interest [20]. If we elect to study resilience relative to a limited function or a property of interest, changes in system state may favour resilience, i.e. the recovery of this particular function or property. For example, if the function of interest is a substrate conversion rate, the microbial community can vary its composition to recover this function and be considered resilient.

In a clinical context, resilience sometimes refers to the capacity of a patient to recover from an infection [26], resilience is not taken from the perspective of bacteria like antibiotic resistance is. In a recent paper, Meredith and colleagues applied the ecological concepts of resistance and resilience to study the response of bacterial pathogens to β-lactam treatments [15]. They propose antibiotic resilience as the ability of a bacterial population to recover after an antibiotic stress, even though the treatment may kill most of the cells. They measure resilience as the time needed for an untreated bacterial population to reach 50% of its maximal biomass compared to an antibiotic treated population, which in their application greatly depends on the population size. In addition, their framework overlaps other concepts in microbiology including the classical definition of antibiotic resistance (defined for single cells), bacterial persistence and collective resistance. In microbiology, several concepts related to antibiotic resilience already exist and we review the main ones below.

3. Resilience-related concepts in microbiology

Instead of using resilience, a great diversity of concepts have been developed in microbiology such as bacterial tolerance, persistence, recalcitrance and so on (referenced in table 1 and detailed below). Like resilience, these concepts are sometimes blurry and depend on the authors and on the context. These concepts tend to reflect the great diversity of mechanisms used by bacteria to withstand antibiotic treatments and other disturbances.

Table 1.

Resilience-related terms used in microbiology.

term	definition
bacterial tolerance	the capacity of a bacterial cell or population to survive long exposure to high concentrations of bactericidal antibiotics while being unable to grow [27,28]
bacterial persistence	the formation of a subpopulation of slow-growing persister cells tolerant to antibiotics. It results in a heterogeneous response to treatments and a characteristic biphasic killing curve under antibiotic exposure [28,29]
phenotypic heterogeneity	the heterogeneity of an isogenic bacterial population, possibly leading to its heterogeneous response to antibiotic exposure [30,31]
phenotypic resistance (heteroresistance)	non-genetic phenotypic heterogeneity that temporarily enables subpopulations of bacteria to grow in the presence of a bactericidal antibiotic [32,33]. Also called heteroresistance [34–36]
adaptive resistance	a transient phenotypic state induced by environmental conditions which increases the survivability of bacteria [37,38]
collective tolerance	the capacity of a sufficiently large bacterial population to tolerate the presence of antibiotic although single cells are susceptible [39]
collective resistance	the capacity of a sufficiently large bacterial population to grow in the presence of antibiotic although single cells are susceptible [40,41]
recalcitrance	the global capacity of a bacterial population to withstand an antibiotic treatment [27,42]

(a). Bacterial tolerance and persistence

Bacterial antibiotic tolerance is the ability of bacteria to survive long exposure to high antibiotic concentrations [43]. Unlike antibiotic resistance, tolerant bacteria cannot grow in the presence of high antibiotic concentrations. However, tolerant bacteria have a small death rate compared to susceptible bacteria, allowing a tolerant bacterial population to survive hours or days longer than a susceptible population.

Bacterial persistence is the presence of a subpopulation of tolerant cells in a susceptible population of bacteria [44,45]. These tolerant cells were first described by Bigger [46] and have been named ‘persisters’. The presence of a minority of persisters leads to a characteristic biphasic kill curve under antibiotic exposure. That is, susceptible cells die quickly during the first treatment phase then persisters survive for a long period. Persisters were first described to be non-growing dormant cells in an isogenic population. Because most antibiotics are effective against actively growing cells, persisters survive antibiotic treatments while the majority of the bacterial population dies quickly. Persisters can resume growth spontaneously on standard media. After antibiotic removal, they allow the recovery of the bacterial population, which is again composed of a majority of susceptible cells and a subpopulation of persisters. Persisters could be formed by stochastic switches in the equilibrium of toxin/antitoxin modules although this hypothesis is still debated [47–49], or induced by stressful conditions such as entry into stationary phase [50]. Stress responses have been reported to be involved in the formation of persisters: stationary phase, starvation or low antibiotic concentrations often increase the level of persisters in bacterial populations [44]. Thus, the persister phenotype is actively maintained by bacteria and not merely a non-growing state [51].

(b). Phenotypic heterogeneity

Although the term ‘persister’ was first used to refer to dormant cells, its signification evolved with our knowledge of persistence mechanisms. Recent reviews emphasize that persisters do not comprise one single phenotype but, rather, are heterogeneous [44]. That is, persisters surviving one type of stress may express a different phenotype than persisters surviving another type of stress, although being genetically identical. It is sometimes unclear if bacterial persistence refers to the formation of a subpopulation of dormant cells or more generally to the phenotypic heterogeneity of a bacterial populations leading to their partial survival during treatments. The phenotypic heterogeneity of an isogenic bacterial population can emerge from ‘persisters’ but also from various mechanisms. Heterogeneity in enzyme production or in efflux pump activity have been reported to induce heterogeneous stress tolerance in bacterial populations [34,37]. In Escherichia coli, cells could escape the action of antibiotics by switching their phenotype from walled state to L-form [52]. Cells could also escape the action of β-lactam antibiotics by being protected by the debris of lysed cells [53]. Fluctuations in transcription [54], gene copy number, location of genes, promoter regions and mutation all affect gene expression and can randomly vary from cell to cell. Sometimes, a minority of cells in isogenic populations is able to grow during antibiotic treatments. Because these cells are normally susceptible (unlike resistant cells) and are actively growing (unlike persisters), this phenomenon has been called ‘phenotypic resistance’, ‘adaptive resistance’ or ‘heteroresistance’ [35–37,55]. Phenotypic heterogeneity allows bacterial populations to adopt a ‘bet-hedging’ strategy in unpredictable environments, increasing the chances that a part of the population will survive disturbances such as antibiotic treatments [34,37,40]. While this heterogeneity can be actively maintained by bacteria such as the formation of dormant cells, it can also emerge from the inherent randomness at the molecular level [56].

(c). Collective tolerance and resistance

Antibiotic resistance is usually defined for single cells. However, the effectiveness of antibiotics can be reduced in large bacterial populations. This phenomenon is know as the ‘inoculum effect’, ‘collective resistance’ or ‘collective tolerance’ [39,40]. This phenomenon can occur when bacteria deactivate the antibiotic, e.g. by secreting an enzyme such as β-lactamase, which becomes effective when bacteria are sufficiently numerous [15,41]. In a microbial community and under specific conditions, susceptible cells can also be protected by their resistant neighbours [41,57].

A common form of bacterial collective behaviour leading to treatment failure is the formation of biofilms. A biofilm is composed of bacteria agglomerated in a matrix of extra-polymeric substances. This lifestyle allows bacteria to be grouped in a particular place and behave collectively in a more complex way than their planktonic counterpart. Being organized in biofilms allows bacteria to overcome harsh environments that single cells could not [58]. Biofilms enhance the survivability of bacterial populations by a variety of mechanisms including limited diffusion of solutes and slow growth rates [40,59,60]. In addition, the spatial heterogeneity of mature biofilms leads to various phenotypes depending on the position of a bacterium in the biofilm. Even when the environment is favourable at the surface of the biofilm, some bacteria are stressed by substrate limitation, metabolite accumulation or pH inside the biofilm and express a tolerant phenotype. Consequently, the response of a biofilm to antibiotic treatments is fundamentally heterogeneous.

4. Which definition of antibiotic resilience for clinical bacteriology?

The diversity of survival mechanisms found in bacteria led to a diversity of terms referring to different survival strategies (table 1). This profusion of concepts is difficult to handle and would benefit to be united under a single definition of resilience. On the other hand, the variety of resilience definitions in the literature highlights that resilience depends on what is considered resilient, the disturbances and the context considered (figure 2). Meredith et al. [15, p. 1] propose resilience as ‘the ability to recover after being perturbed by an antibiotic’. Although this general definition is easily apprehended, in practice, they apply the resistance-resilience framework to the particular case of a bacterial population producing β-lactamases and treated with β-lactams, and their framework overlaps with the existing concepts of bacterial persistence and collective tolerance and resistance.

Figure 2.

The context of antibiotic resilience. While the general definition of antibiotic resilience is easily apprehensible—the ability to recover from antibiotic exposure (regardless of the mechanism)—in practice, antibiotic resilience widely depends on the bacterial system under consideration, on the disturbance applied and on the metric used to quantify resilience.

To make operational the resilience concepts, Martin et al. [20] from socio-ecological science and Song et al. [25] from microbial ecology both suggest to define resilience as the capacity to recover a property/function of interest after it has been completely or partially lost owing to a disturbance. In a complementary way, resistance is the capacity to maintain this property/function despite disturbances. Because antibiotic resistance refers to the capacity of a bacteria to grow in the presence of high concentrations of bactericidal antibiotic, it is tempting to define antibiotic resilience as the capacity of bacteria to resume growth after an antibiotic treatment. However, with this definition, antibiotic resilience would still depend heavily on experimental conditions such as cell density, substrates available and the pre-treatment state of the bacterial population such as its lifestyle (e.g. planktonic or biofilm) and growth phase. While the general definition of antibiotic resilience seems straightforward—‘the ability to recover after being perturbed by an antibiotic’ [15]—its application readily faces the same problematics as found in socio-ecology, that is a very context-dependent definition.

Antibiotic resistance is mainly based on the measure of the MIC which is then compared to the resistance threshold MIC provided by health institutions. This simple measure makes antibiotic resistance easily understandable, independently of the resistance mechanism at stake. Perhaps antibiotic resilience should also be based on a simple measure or set of measures to be readily applicable by microbiologists, regardless of the various resilience mechanisms and dynamics involved.

5. Which indicator for antibiotic resilience?

Owing to the dynamical nature of resilience, measuring it remains a challenge in ecology [61]. However, having simple, reliable and general indicators of antibiotic resilience is necessary to make this concept applicable. As an indicator of antibiotic resistance, the MIC is the most commonly used in laboratories, that is the capacity of a strain to grow in the presence of a given concentration of antibiotic. However, when the inoculum effect is strong, Artemova et al. suggest to use the single-cell MIC instead of the classical MIC [62]. Concerning antibiotic resilience, because temporal evolution, phenotypic and spatial heterogeneity are important factors, its measure is not straightforward. The MIC does not give any information on the capacity of a strain to partially survive and recover from antibiotic treatments. To complement the MIC, other measures have been developed. The minimal bactericidal concentration is the lowest antibiotic concentration that kills greater than 99.9% of the cells in a planktonic culture overnight [63]. For biofilms, the minimum biofilms eradication concentration is the lowest antibiotic concentration necessary to eradicate a biofilm [27], which is often orders of magnitude higher than the MIC [64]. With this method, biofilms are grown on pegs and then dislodged by sonication before the addition of the antibiotic.

Brauner et al. [28] proposed an experimental workflow to differentiate between susceptible, resistant, tolerant and persistent strains. This method is based on the measure of the MIC and the minimal duration for killing (MDK). The MDK₉₉ is the exposure time to an antibiotic required to eliminate greater than 99% or more of the bacteria. The MDK is relevant to identify strains which are tolerant or form subpopulations of dormant cells able to survive long period of antibiotic exposure. In their application of the resilience concept to antibiotic treatments, Meredith et al. [15] measure resilience as the time needed for a population to reach 50% of its maximal biomass in an unstressed environment compared to an environment disturbed by a given antibiotic. However, this measure implies that the recovery occurs while the antibiotic is still present and that the trajectory of the recovery over time is known.

These measures are interesting efforts to quantify the capacity of bacterial populations to overcome antibiotic treatments without resistance mechanisms. However, to account for the dynamical nature of antibiotic resilience and for the bacterial heterogeneity involved is a difficult task. For instance, biofilms can grow on many different surfaces and have different stages of maturity. In addition, the proportion of persister cells varies greatly depending on the growth phase of bacterial cultures [65]. Regarding the diversity of biological mechanisms bacterial populations can use to survive antibiotic treatments, the MIC used to determine the resistance or susceptibility of a bacterial strain is clearly limited. The MDK may be the most relevant indicator of antibiotic resilience because the ability to recover from antibiotic treatment depends on the ability of bacteria to withstand high concentrations for an extended period. The MDK of bacteria in exponential and stationary phase should provide a first valuable insight into the antibiotic resilience of a bacterial strain, in a similar way that the MIC provides a first valuable insight into its antibiotic resistance, regardless of the specific mechanisms at stake. However, the MDK threshold value that should be considered worrying in a clinical context remains to be evaluated.

6. Concluding remarks

Resilience and resistance are complementary concepts in ecology. Similarly, the concept of antibiotic resilience could be complementary to antibiotic resistance. While antibiotic resistance focuses on the ability of a strain to grow under high concentrations of antibiotic, antibiotic resilience focuses on the ability of bacteria to recover from antibiotic exposure. This capacity largely relies on the phenotypic heterogeneity of bacterial populations. It joins other ecological concepts such as response diversity [66,67], i.e. the diversity of responses to disturbances, and bet-hedging strategies [31], i.e. random variations increasing the chance to respond adequately to environmental change; strategies which may play a major role in the evolutionary success of bacteria [68,69].

The involvement of biological mechanisms related to antibiotic resilience in the emergence of antibiotic resistance is gaining concern in the scientific community [6,38,70], and the development of strategies to oppose these mechanisms with specific molecules is in progress [43,58,71]. Antibiotic resilience could complement antibiotic resistance in our understanding of bacterial response to treatments. However, the use of the concept of antibiotic resilience in clinical bacteriology is critically related to its measurement. We urge microbiologists to use new indicators complementary to the MIC such as the MDK and to evaluate their relevance in clinical settings [28,72]. The measure of antibiotic resilience should be a useful indicator to identify and classify strains able to overcome antibiotic treatments despite lacking antibiotic resistance mechanisms.

Data accessibility

This article does not contain any additional data.

Authors' contributions

C.F. and J.-D.M. obtained the funding. All authors conceived the study. G.C. wrote the manuscript. All authors gave fruitful comments during the research process and revised the manuscript.

Competing interests

We declare we have no competing interests.

Funding

This work was supported by a Contrat Plan Etot Région from the Région Auvergne, France.