OUP user menu

Emergence and spread of antibiotic resistance following exposure to antibiotics

Rafael Cantón, María-Isabel Morosini
DOI: http://dx.doi.org/10.1111/j.1574-6976.2011.00295.x 977-991 First published online: 1 September 2011

Abstract

Within a susceptible wild-type population, a small fraction of cells, even <10−9, is not affected when challenged by an antimicrobial agent. This subpopulation has mutations that impede antimicrobial action, allowing their selection during clinical treatment. Emergence of resistance occurs in the frame of a selective compartment termed a mutant selection window (MSW). The lower margin corresponds to the minimum inhibitory concentration of the susceptible cells, whereas the upper boundary, named the mutant prevention concentration (MPC), restricts the growth of the entire population, including that of the resistant mutants. By combining pharmacokinetic/pharmacodynamic concepts and an MPC strategy, the selection of resistant mutants can be limited. Early treatment avoiding an increase of the inoculum size as well as a regimen restricting the time within the MSW can reduce the probability of emergence of the resistant mutants. Physiological and, possibly, genetic adaptation in biofilms and a high proportion of mutator clones that may arise during chronic infections influence the emergence of resistant mutants. Moreover, a resistant population can emerge in a specific selective compartment after acquiring a resistance trait by horizontal gene transfer, but this may also be avoided to some extent when the MPC is reached. Known linkage between antimicrobial use and resistance should encourage actions for the design of antimicrobial treatment regimens that minimize the emergence of resistance.

Keywords
  • resistant mutants
  • mutant prevention concentration
  • mutant selection window
  • horizontal gene transfer
  • quinolone resistance
  • β-lactam resistance

Introduction

Since the introduction of antimicrobial agents in medical practice, many facets of the resulting antimicrobial resistance problems have been published. Most reports essentially describe the emergence of microorganisms that are resistant to different antimicrobial agents, the frequencies of antibiotic resistances, and include surveillance studies collecting minimum inhibitory concentration (MIC) values of different drugs. With the popularization of molecular techniques, including typing methods, resistance mechanisms and the corresponding resistance genes have been widely documented.

By comparison with this huge quantity of information, descriptions of the in vivo emergence of resistant bacteria under antibiotic exposure are less detailed. However, in vitro studies and to a lesser extent animal models have led to a better knowledge about the emergence of resistant bacteria (Craig, 1998; Sykes, 2010). In recent years, special attention has been paid to different concepts including the mutant selection window (MSW), which comprises a range of concentrations where resistant bacteria can be selectable under antimicrobial selective forces (Drlica, 2001) (see Glossary for definitions). This concept assembles the pharmacodynamic knowledge and the criteria for the potential selection of resistant mutants. In addition, both large- and small-scale studies monitoring antibiotic use have demonstrated linkage with resistance development. Interestingly, these studies have also confirmed important difficulties in curtailing resistance once this event has occurred (Enne et al., 2001; Rahal et al., 2002). In many cases, this is due to the coresistance phenomenon: another antimicrobial class is also unable to kill or suppress bacterial growth because the targeted microorganism is insensitive to the new compound as well. Furthermore, once resistance genes have become fixed in a bacterium, it is difficult to eliminate them (Cantón, 2003a; Sundqvist et al., 2010; Andersson & Hughes, 2011).

Defining antibiotic resistance

From a clinical point of view, resistance is defined as a state in which a patient, when infected with a specific pathogen, is treated with an adequate antimicrobial dosage and administration schedule, but clinical criteria of cure (at a clinical and/or a microbiological level) are not reached. In the microbiology laboratory, clinical resistance is defined using the clinical breakpoints (Turnidge & Paterson, 2007). A clinical breakpoint is an MIC value that correlates with the clinical outcome and that separates those isolates that are considered as clinically susceptible or associated with a high likelihood of therapeutic success from those that are considered as clinically resistant or associated with a high likelihood of therapeutic failure (ISO, 2006).

Clinical breakpoints are calculated taking into account different criteria and are mainly influenced by pharmacokinetic/pharmacodynamic (PK/PD) parameters. For antibacterial agents, PK/PD studies have defined the relationship between the concentration at the site of infection (i.e. a PK variable) and microbial inhibition or killing in vivo (i.e. a PD property of the drug). The PK/PD breakpoints (also expressed as concentration cut-off values) state the probability of a target attainment associated with a high probability of clinical outcome. They are mainly generated in animal models and then extrapolated to humans using mathematical or statistical techniques (Mouton et al., 2002; Turnidge & Paterson, 2007). Once the PK/PD breakpoints have been established, they are adjusted by correlating MIC values with clinical outcomes; practical examples have been illustrated by Turnidge & Paterson (2007). For new antimicrobial agents, these data are mainly obtained from clinical trials. For older antimicrobials, PK/PD knowledge has been scarcely applied and a process of revision of old breakpoints is currently ongoing (Kahlmeter et al., 2003; Macgowan et al., 2008).

From a microbiological point of view, resistance is defined as a state in which an isolate has a resistance mechanism rendering it less susceptible than other members of the same species lacking any resistance mechanism. This definition is valid irrespective of the level of resistance (i.e. low or high level of resistance) and does not necessarily correlate with clinical resistance. Isolates that are microbiologically resistant can be phenotypically recognized using the so-called epidemiological cut-off (ECOFF) value, an MIC value that separates the wild-type population from those isolates that have developed resistance, either because of mutations or as a consequence of horizontal gene transfer, and independent of whether the level of resistance has clinical relevance (Brown & Cantón, 2010). Figure 1 illustrates a wild-type population separated by the corresponding ECOFF from a bacterial population with resistance mechanisms. This figure also includes clinical breakpoints (susceptible and resistant), according to the criteria established by the two main committees defining these values: the Clinical and Laboratory Standards Institute (CLSI, 2011) and the European Committee for Antimicrobial Susceptibility Testing (EUCAST, 2011). In the example of this figure, the clinically susceptible population (below the clinical susceptible breakpoint) includes part of the microbiologically resistant population and the wild-type population. It is important to remark that clinical breakpoints (as defined by CLSI and EUCAST) do not try to detect the resistant organisms and are basically defined for treating patients. Bacteria may have a resistance mechanism and may be considered microbiologically resistant, but are susceptible from a clinical point of view.

Figure 1

Ciprofloxacin MIC distribution of Escherichia coli isolates (http://www.eucast.org). Epidemiological cut-off (ECOFF) values and clinical susceptible (S) and resistant (R) breakpoints from CLSI and EUCAST committees are indicated. The clinically susceptible population (below the clinical susceptible breakpoint) includes part of the microbiologically resistant population (low-level resistant bacteria, presumably expressing qnr-like genes or other PMQR mechanisms or first step gyrA mutations) and the wild-type population (below the ECOFF value and presumably without resistance mechanisms). The clinically resistant population (beyond the clinical resistant breakpoint) includes isolates with high-level resistance mechanisms (most probably double-step gyrA mutants or a combination of gyrA with parC mutations).

Emergence of antibiotic resistance

In this review, antibiotic resistance is understood as acquired antibiotic resistance, i.e. resistance that emerges from susceptible bacterial isolates either by mutation or by acquisition of resistance genes (Sykes, 2010). Exposure to antibiotics has been considered as the most important factor influencing the emergence and spread of antibiotic resistance. This view emphasizes the influence of natural (Darwinian) selection in the evolution of resistance, such that antibiotic-resistant organisms survive and have a progeny, while their susceptible counterparts become extinct (Baquero & Cantón, 2009). However, reality is more complex. The emergence of antibiotic resistance should be understood as the emergence of an abnormal resistance trait in a particular microorganism, generally a bacterial pathogen. However, emergence is frequently observed as a phenomenon only when there is a sufficiently high frequency of the new resistance trait in an organism and, therefore, the first occurrence of resistance might often remain cryptic.

Antibiotic selection essentially enriches the number of resistance genes in a particular setting, but these genes already exist before selection operates. In this context, it is important to recall the concept of the resistome, i.e. the ensemble of potential resistance traits that can be identified in an organism (termed intrinsic resistome) or in a complex microbial system (termed metagenomic resistome) (D'Costa et al., 2006, 2007; Alvarez-Ortega et al., 2011). A number of bacterial genes, many of which are involved in housekeeping functions, may be considered as preresistance genes. Evolution of housekeeping genes to antibiotic resistance can be favoured by gene duplication, an unexpectedly frequent event whereby one of the genes retains the old function, whereas a modified copy evolves to acquire a novel antibiotic resistance function (Andersson & Hughes, 2009). When a resistance trait has become widespread in different organisms, the use of different antibiotics belonging to the same class strongly contributes to genetic diversification and to concomitant substrate specificity (Novais et al., 2010). Such an evolution may be favoured by antibiotics that enhance mutation or recombination rates.

However, most clinically used antibiotics are not considered as mutagenic for bacteria (Sierra et al., 2005; Hayasaki et al., 2006). While these antimicrobials can produce a number of effects on bacterial populations, even at low concentrations, a small fraction of cells (normally ranging from 10−8 to 10−9) is not specifically affected by the antimicrobial challenge. This insensitive subpopulation might have mutations causing resistance to the antimicrobial drug and can be selected during antimicrobial treatment. A similar selection process exerted by antimicrobials also operates when a susceptible population acquires resistance determinants by horizontal gene transfer (Sykes, 2010).

Basic concepts of the selection process

Antibiotic resistance within the patient in the course of therapy is maximized by delays in establishing therapy, low dosages and long periods between dosages (inadequate pharmacokinetics). In infections caused by some pathogens, single-drug administration may also favour the emergence of resistance (Bauernfeind et al., 1995). Therefore, one of the main objectives is to reduce and prevent, as early as possible, the size of the viable bacterial target population, as a large population size increases the likelihood of resistant clones arising after mutation or acquisition of antibiotic resistance by horizontal gene transfer. Importantly, exposure to very low antibiotic concentrations can select for low-level resistant mutants, which eventually serve as stepping stones paving the way for high-level resistance (Baquero, 2001). This concept will be discussed later in more detail. Failure to limit the emergence of resistance in a patient also constitutes a risk for other patients. Indeed, the spread of antibiotic resistance depends on the transmission between patients and on the persistence of resistant microorganisms. Moreover, when resistance occurs in commensal bacteria, transmission between healthy individuals and patients plays a role (Martínez & Baquero, 2002; Valverde et al., 2004, 2008). The importance of so-called high-risk clones in the spread of the antibiotic resistance genes has also been demonstrated (Peirano & Pitout, 2010).

Emergence of resistant bacteria under antibiotic exposure in vivo is not understood in detail. By contrast, in vitro studies and, to a lesser extent, animal models have provided current knowledge of how resistant bacteria arise (Craig, 1998; Sykes, 2010). Recently, special attention has been paid to different concepts that serve to understand in more depth the mechanisms leading to the emergence of resistance and to develop novel strategies impeding this emergence. One such concept is the MSW, which comprises a concentration range in which resistant bacteria can be selectable under an antimicrobial selective force (Drlica, 2001). This concept also includes the pharmacodynamic knowledge. Another concept, which will not be discussed here, is combination therapy, which aims at obtaining synergistic effects by administering two or more antibiotics at the same time. A crucial way of limiting the spread of antibiotic resistance is by the reduction of antibiotic use. This is illustrated by the correlation between antibiotic use and the frequency of antibiotic resistance (discussed in the next paragraph).

Antimicrobial use and resistance in ecological studies

For many years, ecological studies have shown a link between the frequency of resistance and the use of antimicrobials. This has been observed both in hospital and in community environments. Moreover, the same correlation has been seen in agriculture, in particular in the cattle industry (Levy, 1997; Shryock & Richwine, 2010). The observed increase in the incidence of resistance is an expected result. By contrast, when antimicrobial use is reduced or abolished, a decrease of resistance prevalence may occur, but is less obvious or not achieved (Enne et al., 2001; Sundqvist et al., 2010; Andersson & Hughes, 2011).

Since the introduction of antimicrobials into medical practice, the increase in the frequency of resistance has become a common collateral effect of their use. First anecdotal reports were followed by a huge body of evidence for a vicious circle in which the introduction of a new antimicrobial agent is followed by the emergence of resistance. The subsequently needed development of new compounds that are not affected by previous resistance mechanisms will lead, sooner or later, to the selection of new resistances. This has been well illustrated on several occasions (Seppälä et al., 1997) and can be particularly well documented with some microorganisms such as Staphylococcus aureus (Mackenzie et al., 2007; Anonymous, 2008). A supranational multicentre study has demonstrated that European countries with a higher consumption of antimicrobials in hospitals have a higher incidence of methicillin-resistant S. aureus (MRSA) (Mackenzie et al., 2007). When the consumption of different antibiotics was monitored, it was found that the linkage was stronger with expanded-spectrum cephalosporins than with other classes of antibiotics. In some instances, a time-series analysis is required to observe a relationship between consumption and resistance. For instance, a single hospital study showed that MRSA increase was concomitant with the increased use of different antimicrobials, but the correlation between MRSA incidence and the consumption of macrolides, cephalosporins and quinolones varied over time (Monnet et al., 2004).

A reciprocal conclusion from supranational multicentre ecological studies is that countries with high resistance figures normally have a high consumption of all (or nearly all) monitored antimicrobials. This was illustrated in the ARMed study performed in the Mediterranean area with different sentinel bacteria, including S. aureus and Escherichia coli (Borg et al., 2008, 2010). In the same vein, vancomycin-resistant enterococci (Hsueh et al., 2005b), extended-spectrum β-lactamase (ESBL)-producing Enterobacteriaceae (Meyer et al., 2010), carbapenemase-producing Gram negative bacteria (Furtado et al., 2010), multiresistant Pseudomonas aeruginosa (Polk et al., 2004; Hsueh et al., 2005a; Weng et al., 2011) and Acinetobacter baumannii (Hsueh et al., 2005a; Meyer et al., 2009) have also been linked to an increased use of antimicrobials in single institutions. Remarkably, risk factor analysis has shown that there is also a correlation between antimicrobial use and colonization or infection with these organisms (Rodríguez-Baño & Pascual, 2008; Benenson et al., 2009; Gasink et al., 2009; Lepelletier et al., 2010).

Conversely, a decrease of antimicrobial use and subsequent reduction of resistance have been less frequently reported, perhaps due to difficulties in ascertaining such effects and due to a potentially longer observation time needed. Early experiences with some resistant pathogens such as ESBL-producing Enterobaceriaceae appeared to be promising (Meyer et al., 1993; Lee et al., 2004). More recently, however, with the exception of some experiences in countries with low frequencies of resistance levels and low colonization pressure (i.e. number of patients or individuals colonized with a resistant microorganism), it has been difficult to curtail ESBL-producing organisms (Meyer et al., 2009; Tängdén et al., 2011). This can be attributed, in part, to multiresistance and potential selection with different antimicrobials (Morosini et al., 2006; Cassier et al., 2011). For this reason, cycling or rotation strategies (i.e. scheduled rotation of one class of antibiotic with one or more different classes exhibiting comparable spectra of activity, at the level of the hospital wards) are not currently recommended because of a potential accumulation of mutations or resistance genes on mobile genetic elements (Brown & Nathwani, 2005; van Loon et al., 2005). Other strategies such as antibiotic mixing need to be validated (Bal et al., 2010; Masterton et al., 2010). This strategy involves antibiotic rotation at the level of individual patients (e.g. patients consecutively receive a carbapenenem, a cephalosporin, a quinolone and so on in a cyclical manner).

Global reduction of antimicrobial use does not always result in a decline of resistance, either. For instance, a nonreverting trimethoprim/sulphonamide resistance has been observed in E. coli despite a clear decrease in the use of these antibiotics in a community setting (Enne et al., 2001; Sundqvist et al., 2010). This lack of reversion to susceptibility can be due to the absence of a significant fitness cost of the resistance traits or to their permanent fixation (e.g. on an integron) in the bacterial population (Cantón et al., 2003a; Sundqvist et al., 2011). Antibiotic stewardship programmes have been in part designed to try and curtail resistances once they have arisen (Tamma & Cosgrove, 2011).

Antibiotic pharmacokinetics and prevention of emergence of resistance: antibiotic concentration gradients, mutant prevention concentration (MPC) and MSW

As stated above, the definition of antimicrobial resistance both at the clinical and at the microbiological level uses a concentration-dependent criterion, based on the MIC. At concentrations higher than the MIC, the susceptible population should in principle be inhibited, whereas a minority of variants (resistant mutants) harbouring resistance mechanisms will not be inhibited. Nevertheless, these resistant variants will be inhibited by a higher antibiotic concentration (i.e. the MIC of the resistant variants). Antibiotic concentrations surpassing the MIC of a susceptible population, but below that inhibiting the resistant variants constitute the selection window. This window depends on the pharmacokinetics of the drug and on the formation of antibiotic concentration gradients in the human body. The concept of antibiotic gradients in antibiotic selection was introduced by Baquero & Negri (1997) and was originally developed with two in vitro models, one of which used Streptococcus pneumoniae expressing different penicillin susceptibility levels due to distinct penicillin-binding protein (PBP) arrangements (Negri et al., 1994). The other model used a collection of isogenic E. coli isolates with different ESBL variants, conferring variable levels of cefotaxime and ceftazidime resistance (Blázquez et al., 2000). In both models, selection of resistant bacteria only occurred within a certain window of concentrations, but neither above nor (presumably) below these concentrations. Antibiotic selective gradients ensured the selection of bacteria with very small differences in MIC values (Baquero et al., 1998). In vivo confirmation of this principle of concentration-dependent selection was obtained in animal models and is supported by mathematical modelling (Negri et al., 2000). It has been proposed that the practical application of this concept aiming at prevention of resistance might be achieved by establishing a MPC, derived from in vitro susceptibility testing (Drlica, 2001; Zhao & Drlica, 2008). The MPC, which is above the MIC, is defined as the concentration that restricts the emergence of first-step resistant mutants within a susceptible population. Thus, once the antibiotic concentration surpasses the MPC at the site of infection, the emergence of resistance is expected to be limited (Cantón et al., 2006). The MSW (Drlica, 2001) refers to the range of concentrations between the MIC and the MPC (see Fig. 2 and Glossary).

Figure 2

MSW and MPC. The figure illustrates three different situations where an antibiotic is administered. Curves represent the pharmacokinetics (concentration over time) of an antimicrobial agent and squared boxes represent the bacterial population. (A) The pharmacokinetic curve is below the MIC; thus, no selection of a resistant mutant subpopulation within the wild-type population is expected, but see text for a discussion of the possible selection of resistant mutants at subinhibitory concentrations. (B) The pharmacokinetic curve is mainly within the MSW; therefore, the resistant mutant subpopulation within the wild-type population can be selectable. (C) The pharmacokinetic curve surpasses the MPC; thus, the susceptible bacteria are inhibited and selection of a resistant mutant subpopulation is potentially avoided. This is an idyllic, desired outcome as resistant mutants may still be amplified eventually, even when microorganisms are within the MSW.

Unlike MIC testing, which typically uses an inoculum size of approximately 104–105 CFU mL−1, the calculation of the MPC needs a large inoculum (approximately 109–1010 CFU mL−1). This high inoculum is chosen to ensure the presence of first-step resistant mutants within the susceptible bacterial population. Practical calculation of MPC values requires a series of agar plates with increasing antibiotic concentrations; the minimum concentration inhibiting all bacterial growth defines the MPC. An initial view of the MPC did not consider the emergence of resistant bacteria as a dynamic phenomenon over time. MPC should also be established with new PK/PD knowledge (Croisier et al., 2004; Tam et al., 2005, 2007; Mouton et al., 2011), which is considered one of the most important limitations. In addition, some bacterial cells, when applied to agar plates at the MPC, may survive even though they may not be visualized as colonies. These bacterial cells, which do not grow, are persister cells (not mutants). Persisters are dormant cells that form stochastically in microbial populations and are tolerant to antibiotics. When the antibiotic is removed after the challenge, persisters are found to be susceptible to the antibiotics (Lewis, 2010) (see the section Emergence of antimicrobial resistance in biofilms and the contribution of hypermutators and persisters for a more extensive discussion).

Initial calculations of MPC values were developed for quinolones and microorganisms that accumulated resistant mutants in a step-wise manner under appropriate antibiotic pressure, including Mycobacterium tuberculosis, S. aureus and S. pneumoniae (Dong et al., 1999, 2000; Blondeau et al., 2001). Consequently, high antimicrobial doses ensuring antimicrobial concentrations above the MPC at the infection site were suggested for respiratory tract infections (Cantón et al., 2006). In addition, combination therapy approaches were justified, particularly when the frequency of resistant mutants under selection pressure with a single drug was high, as in pulmonary tuberculosis or in ventilator-associated pneumonia due to P. aeruginosa (Zhanel et al., 2006). Double mutants rarely emerge under the selective pressure of two antimicrobials.

The MPC concept can be used with bacteriostatic and bactericidal agents and has been calculated for several antimicrobials, including fluoroquinolones, β-lactams, glycopeptides, lipopeptides, macrolides, oxazolidinones, tetracyclines and ansamycins, among others. Table 1 shows the MPC values for different antimicrobials and microorganisms. Theoretically, compounds killing bacteria rapidly (below the MPC) might have advantages over those antimicrobial agents exhibiting lower killing rates (Firsov et al., 2004; Chung et al., 2006). Antimicrobials that kill more rapidly than the time required for outgrowth of the resistant mutants are particularly advantageous. In this case, the MPC value might even overestimate the concentration needed to restrict mutant development. Although the MPC concept was initially used for first-step mutants, it can also be applied to second-step mutants and higher-order mutants. MPC values for such evolved mutants are higher than those for first-step mutants and can be calculated once first-step mutants have emerged. The limits to the clinical application of the MCP concept are imposed by a potential toxicity encountered at high antimicrobial doses. Furthermore, it is difficult to avoid the selection of resistance in commensal bacteria, which can subsequently transfer resistance traits to pathogens.

View this table:
Table 1

MPC values of different antibiotics against different organisms

MicroorganismAntibioticMIC50 (mg L−1)MPC50 (mg L−1)Reference
Pseudomonas aeruginosaCiprofloxacin0.122Cantón et al. (2003b)
Levofloxacin0.258
Ceftazidime232
Pseudomonas aeruginosaImipenem232Credito et al. (2010)
Meropenem0.58
Doripenem0.54
Escherichia coliNalidixic acid1.532Hansen et al. (2006)
Ciprofloxacin0.0120.3
Escherichia coliImipenem0.250.5Credito et al. (2010)
Meropenem0.030.06
Doripenem0.030.125
Streptococcus pneumoniaeLevofloxacin12Homma et al. (2007)
Moxifloxacin0.1250.5
Staphylococcus aureusCiprofloxacin0.34Zhao et al. (2003)
Levofloxacin0.122
  • * MIC and MPC values were obtained with Streptococcus pneumoniae ATCC 49619.

  • MIC and MPC values were obtained with Staphylococcus aureus NR450.

  • Values were obtained on Luria–Bertani agar plates (Cantón et al., 2003b); Mueller–Hinton agar plates (Hansen et al., 2006); trypticase soy agar plates (Credito et al., 2010); TSA-DHB agar (tryptic soy agar with 5% defibrinated horse blood) (Homma et al., 2007); or GL agar plates [0.3% Casamino acids, 0.3% yeast extract, 0.1 M NaCl, 0.2% sodium lactate, 0.1% glycerol and 1.5% agar (pH 7.8)] (Zhao et al., 2003). Unless indicated otherwise, the MIC and MPC values are expressed as MIC50 and MPC50 values.

Other pharmacokinetic-based indices used in strategies to decrease the selection of antibiotic resistance

If antibiotic regimens are to be designed not only for clinical efficacy, but also for suppression of resistance emergence (Mouton et al., 2011), then more refined strategies may ultimately be required. They are presented in Supporting Information and include the consideration of the time in which a bacterial population is within the MSW (TMSW). This parameter was studied in a mouse thigh bacterial infection model with a fluoroquinolone used in veterinary medicine (marbofloxacin); selection of resistant bacteria occurred after infection with a low (105 CFU) or a high (107 CFU) E. coli inoculum (Ferran et al., 2009). As observed in a previous in vitro model, the emergence of resistance was more frequent when both the initial bacterial inoculum size and TMSW increased (Ferran et al., 2007). In the case of a high inoculum, the net number of resistant mutants increased over time, due to a higher probability of enrichment for resistant mutants during treatment, particularly if antibiotic concentrations were below the MPC (Ferran et al., 2007). From a practical point of view, when using antimicrobials in a clinical setting, there is a need for a prompt antibiotic treatment, which minimizes the increase in the inoculum size, as well as for antibiotic regimens precluding a prolonged period of time within the MSW. Thus, the duration of antimicrobial therapy should be as short as possible. Mathematical models support this recommendation (Levin & Udekwu, 2010).

The MPC concept and the MSW hypothesis can be applied to resistance associated with horizontal gene transfer

The MPC concept and the MSW hypothesis have been built around mutational resistance. Several such studies have focused on fluoroquinolones and topoisomerase mutants of S. pneumoniae, S. aureus, E. coli and M. tuberculosis (Dong et al., 1999, 2000; Blondeau et al., 2001; Hansen & Blondeau, 2005; Cui et al., 2006; Olofsson et al., 2006; Firsov et al., 2008). Further studies have been performed with P. aeruginosa, which frequently displays fluoroquinolone resistance due to topoisomerase mutations and overexpression of efflux pumps (Hansen et al., 2006; Zhanel et al., 2006; Pasquali & Manfreda, 2007; Plasencia et al., 2007). However, both the MPC concept and the MSW hypothesis can be applied to resistance mechanisms involving horizontal gene transfer and recombination.

The acquisition of resistance mechanisms by horizontal gene transfer is a dynamic process. Once a susceptible population acquires a resistance gene, by the acquisition of plasmids, transposable elements or phages, the recipient cells transmit the resistance gene to their descendants and furthermore may act as donors to susceptible cells. In the absence of a fitness cost, both processes rapidly enrich for a resistant population and this can occur even in the absence of antimicrobials (García-Migura et al., 2007; Pallecchi et al., 2008; Prelog et al., 2009). Nevertheless, when the pertinent antimicrobial agent is present at a concentration equal to or higher than the MIC, a selection process will occur as long as its concentration kills or inhibits the remaining susceptible population and does not affect the emerging population carrying the resistance trait. In S. pneumoniae with different PBPs (transmissible by transformation) and in E. coli expressing different plasmid-borne β-lactamases (TEM-ESBL), resistant bacteria were obtained over a short concentration range, but not when the antibiotic concentration surpassed the MIC for the population harbouring the acquired resistance genes (Negri et al., 1994; Baquero & Negri, 1997; Blázquez et al., 2000).

The implication of the MPC in acquired resistance has also been studied in Enterobacteriaceae harbouring either a qnr class gene, the aac(6′)-Ib-cr gene or the qepA gene, all of which belong to plasmid-mediated quinolone resistance (PMQR) genes and lead to low-level resistance to these antibiotics (Table 2) (Rodríguez-Martínez et al., 2007; Briales et al., 2011; Luo et al., 2011). In the absence of other mechanisms affecting fluoroquinolone susceptibility, the presence of a qnr class gene increases the MIC of fluoroquinolones between four- and 128-fold, although the MIC values remain below the susceptible breakpoints. It has been suggested that the expression of Qnr proteins facilitates the selection of high-level quinolone-resistant mutants. In vitro models have shown that the qnrA gene increases the MPC against fluoroquinolones in E. coli and Klebsiella pneumoniae (Rodríguez-Martínez et al., 2007). Moreover, the MPC of ciprofloxacin was dramatically increased when PMQR genes were present together with chromosomal gyrA and/or parC (topoisomerase) mutations in Salmonella Typhimurium (Luo et al., 2011). Thus, a low-level resistance mechanism could not only lead to an increase in the MIC values, but also accelerate the generation of high-level resistance. In other words, the presence of PMQR genes increases the probability of isolates to be within the MSW and this appears to facilitate the selection of resistant mutants (Rodríguez-Martínez et al., 2007; Luo et al., 2011).

View this table:
Table 2

Effects of different low-level PMQR mechanisms (qnrA, qnrB, qnrB4, qnrS1, aac(6′)-Ib-cr and qepA) on MIC and MPC values of ciprofloxacin in Escherichia coli and Salmonella Typhimurium with and without chromosomal mutations affecting quinolone resistance (gyrA and parC)

MicroorganismsResistance mechanismCiprofloxacin
PMQRMutation in QRDRsMIC (mg L−1)MPC (mg L−1)
Escherichia coli 0.0021
gyrA0.1254
qnrA0.1258
qnrAgyrA0.516
qnrB0.1252
qnrBgyrA0.58
qnrS10.1254
qnrS1gyrA18
Salmonella Typhimurium0.0150.125
parC0.0150.125
gyrA0.252–4
parC+gyrA464
aac(6′)-Ib-cr0.061
aac(6′)-Ib-crparC0.061
aac(6′)-Ib-crgyrA18–16
aac(6′)-Ib-crparC+gyrA16128
qepA0.1252
qepAparC0.1252
qepAgyrA232
qepAparC+gyrA16128
qnrB40.54
qnrB4parC0.54
qnrB4gyrA1–216–32
qnrB4parC+gyrA864
qnrS10.52
qnrS1parC0.52
qnrS1gyrA18–16
qnrS1parC+gyrA864
  • * PMQR, plasmid-mediated quinolone resistance mechanisms associated with (1) qnr genes encoding Qnr proteins protecting topoisomerases from the action of the fluoroquinolones; (2) aac(6′)-Ib-cr gene encoding an acetylase that modifies the amino group of the piperazin ring of the fluoroquinolones; and (3) qepA gene encoding an efflux pump affecting fluoroquinolones.

  • QRDRs, quinolone resistance-determining regions in the topoisomerases, including those from gyrA and parC.

  • Data from Briales et al. (2011).

  • § Data from Luo et al. (2011).

  • Gene encoding an aminoglycoside acetyltransferase conferring reduced susceptibility to ciprofloxacin by N-acetylation.

In conclusion, dosing and regimen schemes avoiding low-level resistance mechanisms are easier to implement than those avoiding high-level resistance mechanisms. Nevertheless, as low-level resistant microorganisms can readily enter the MSW for high-level resistance, antibiotic regimens should ideally be designed to prevent high-level resistance as well.

Undesirable effects of subinhibitory antibiotic concentrations

Low concentrations of certain antibiotics, including fluoroquinolones and β-lactams, have been reported to fuel mutagenesis and to increase the risk for emergence of resistance (Couce & Blázquez, 2009). These antibiotics directly induce the expression of error-prone DNA polymerases or increase the formation of reactive oxygen species (ROS) inducing the SOS response, for example in Enterobacteriaceae and P. aeruginosa (Blázquez et al., 2006; Kohanski et al., 2010). These phenomena have also been associated with transient hypermutation (Blázquez et al., 2003), which increases the probability of bacteria to become resistant in the presence of an antimicrobial agent despite being below the concentration range of the MSW. Conversely, in vitro inactivation of the recA gene, which is required for the induction of the SOS response, reduces mutagenicity and counteracts the effect of sublethal antimicrobial concentrations (Thi et al., 2011). Furthermore, subinhibitory concentrations of ciprofloxacin stimulate genetic recombination in E. coli, which may facilitate the incorporation of genes after horizontal transfer and thereby enriches for resistant mutants (López et al., 2007; López & Blázquez, 2009).

It can be speculated that antibiotic-resistant mutants might present a fitness gain in the presence of sublethal antibiotic concentrations (Andersson & Hughes, 2011). Importantly, subinhibitory antibiotic concentrations might also contribute to the emergence of resistance via gene duplication, according to the gene-duplication-amplification model (Andersson & Hughes, 2009; Andersson, 2011) mentioned before. The frequency of tandem gene duplications may be thousands of times higher than the frequency of spontaneous mutations. Survival by gene amplification of an otherwise susceptible strain might also provide a possibility of acquiring ‘mature’ resistance genes by horizontal gene transfer during a period when the bacteria coexist with resistance donors.

Subinhibitory concentrations promoting resistance can also be relevant in faecal microbiota, particularly for fluoroquinolones, which are partially excreted through the biliary tract. This has been observed in human volunteers treated orally with ciprofloxacin; ciprofloxacin-resistant bacteria emerged mainly when the local concentrations of the antibiotic were lower than the MIC (Fantin et al., 2009). This also explains the higher rates of ciprofloxacin resistance in faecal Enterobacteriaceae recovered from cancer or leukaemia patients exposed to fluoroquinolones during neutropenia or in patients with hepatic cirrhosis receiving fluoroquinolone prophylaxis (Carratalà et al., 1996; Aparicio et al., 1999).

These observations have practical consequences for antimicrobial treatment. Dose regimens avoiding subinhibitory concentrations should be ensured, particularly during the first part of the antimicrobial treatment. This can justify a high loading dose in the case of those antimicrobials for which distribution into the infection site may be decreased by serum protein binding or because of the physicochemical characteristics of the compound.

Emergence of antimicrobial resistance in biofilms and the contribution of hypermutators and persisters

The MPC concept and the MSW hypothesis have been developed for a planktonic mode of bacterial growth and not for biofilms (formed by sessile bacteria), which are a typical mode of bacterial growth in chronic infections. Biofilms are consortia of bacteria embedded in a self-produced polymer matrix consisting of polysaccharides, proteins and DNA (Donlan & Costerton, 2002; Hall-Stoodley et al., 2004; Speziale et al., 2008; Høiby et al., 2010; López et al., 2010). At least three different mechanisms (of which only the first is specific for biofilms) can influence antimicrobial susceptibility: (1) the presence of an extracellular matrix affecting diffusion of antibiotics into embedded bacterial cells; (2) lesions in the mismatch repair system (MMR) or in the DNA oxidative repair system (GO), resulting in hypermutators; and (3) emergence of persistent bacterial cells (Driffield et al., 2008; Yang et al., 2008; Mulcahy et al., 2010).

Alterations in DNA repair genes can affect the MMR system, which involves the mutS, mutL and uvrD genes, or the GO system, which involves the mutT, mutY and mutM genes (Oliver et al., 2002; Chopra et al., 2003; Smania et al., 2004; Mandsberg et al., 2009). These mutations may be induced by oxidative stress under biofilm conditions via the production of endogenous ROS. Thus, a deficient antioxidant system enhances hypermutability (Hassett et al., 1999; Mai-Prochnow et al., 2008).

The high cell density in biofilms, as compared with that of a planktonic mode of growth, increases the absolute numbers of resistant mutants that can be selectable under antibiotic pressure. Furthermore, horizontal gene exchange is enhanced in biofilms. Whereas the mutation frequency is usually stable over time in bacteria, an increased value of this parameter has been demonstrated in bacteria growing in biofilms (Molin & Tolker-Nielsen, 2003; Driffield et al., 2008; García-Castillo et al., 2011). So-called hypermutators (=mutators) have a spontaneous mutation rate that is 10–1000 times higher than that of a wild-type strain. They are frequently recovered from chronic infections, particularly those involving the respiratory tract (i.e. from patients with cystic fibrosis, bronchiectasis or chronic obstructive pulmonary disease) where biofilm formation is a feature of colonization or infection and where the selection of resistant mutants is high (Oliver et al., 2000; Maciá et al., 2005; Martínez-Solano et al., 2008; Valderrey et al., 2010). Genetic adaptation of P. aeruginosa in cystic fibrosis is undoubtedly a clear example of this situation. Loss-of-function mutations in DNA repair genes play an important role in the adaptive evolution of P. aeruginosa, which includes the development of antibiotic resistance among other important phenotypic changes such as mucoidy or loss of the quorum-sensing machinery (Ciofu et al., 2010; Oliver & Mena, 2010). In conclusion, the presence of a high proportion of hypermutators in a bacterial population inevitably leads to antimicrobial multiresistance, which in part accounts for a significant number of treatment failures observed in this type of infection (Maciá et al., 2005; Rodríguez-Rojas & Blázquez, 2009).

The fact that cystic fibrosis lung infections are recalcitrant appears to be related to yet another phenomenon, i.e. the formation of persister cells. The frequency with which cells enter the persister state can increase by mutation (Mulcahy et al., 2010). As persisters are insensitive to antimicrobial action, the unresponsiveness of patients to regular antibiotic treatment could be explained not only by hypermutators causing resistance mutations, but also by persisters, whose numbers, in turn, can be increased by hypermutability.

Units of selection in antibiotic resistance: towards multilayered epidemiology of antibiotic resistance

In the preceding paragraphs, it was implicitly assumed that the selection of antibiotic resistance occurs because of the selection of resistant bacterial cells. In recent years – as highlighted by other papers of this Thematic Issue – it has become evident that, below the cellular level, selection can operate on genes, operons and other multigene modules, genomic islands and mobile genetic elements such as integrons, integrative-conjugative elements, transposons and plasmids; these genetic entities might constitute separate objects for evolutionary trends (Baquero, 2004). At the supracellular level, clones, clonal complexes, quasi-species, species, communities and, in general, integrated microbiotic ensembles might also constitute units of selection for antibiotic resistance. Units of selection find connectivity at the subcellular level via horizontal gene transfer and at the supracellular level by interactions between cells and supracellular entities. Each of these units of selection should be taken into account in surveillance of resistance. In a multilayered epidemiological approach, the changes in the frequency of each unit of selection involved in antibiotic resistance should be monitored. The detection of emerging high-risk associations between them might help predict future trends in antibiotic resistance (Baquero, 2011).

Concluding remarks

Antibiotic use and emergence of resistance are undoubtedly linked. However, to some extent, emergence of resistance can be avoided or at least diminished with appropriate antimicrobial regimens. The application of the MPC concept and the MSW hypothesis may contribute to this objective although they have several limitations. Originally, they were built around mutational resistance and did not consider the acquisition of resistant genes. In addition, initial views did not focus on the MPC as a dynamic phenomenon over time, which should be established with new PK/PD knowledge. Toxicity of antibiotics can preclude the administration of doses that would be sufficiently high to eliminate all resistant forms of bacteria. To avoid the selection of resistant bacteria, in some situations, combination therapy approaches can be justified. Table 3 summarizes the challenges and objectives for an optimal antibiotic treatment that limits the emergence of resistance as much as possible. Regimens for antimicrobials should be designed not only to attain clinical efficacy, but also to minimize the emergence and spread of resistance.

View this table:
Table 3

Challenges and objectives for optimal antibiotic treatments avoiding the emergence of resistance according with the MPC concept and the MSW hypothesis

ChallengeObjective
Early treatmentTo avoid an increase in inoculum size and in absolute numbers of the mutant subpopulation that can be selectable during antimicrobial treatment To avoid the establishment of isolates within biofilm structures, which increase the risk for persistence, hypermutability and resistance development
Attainment of an antimicrobial concentration at the site of infection at or beyond the MPCTo inhibit both the susceptible population and the resistant subpopulation (and, if possible, to inhibit the resistant population when this has been selected)
Avoidance of an antimicrobial treatment resulting in antimicrobial concentrations at the site of infection within the MSWTo decrease the possibility of selection of the resistant subpopulation within the MSW
Shorten the time of antimicrobial concentrations within the MSWTo limit the time of mutant subpopulation selection
Avoidance of subinhibitory antimicrobial concentrations at the site of infectionTo avoid the increase in resistance development due to the potential emergence of (transient) hypermutators

Glossary: terminology used

View this table:

Supporting Information

Additional Supporting Information may be found in the online version of this article:

Appendix S1. Additional parameters used in strategies aiming at decreased selection of antibiotic resistance.

Acknowledgements

The content of the manuscript was obtained in part from research EU-funded projects (LSHM-CT-2003-503335 and HEALTH-F3-2008-223031). We thank Dr Fernando Baquero and Dr Dieter Haas for critical suggestions to improve the manuscript. The authors whish also to evoke the memory of their friend Dr Cristina Negri who did a notable work in the field of selective concentrations and antimicrobial resistance.

Footnotes

  • Editor: Dieter Haas

References

View Abstract