Selection and evolution of resistance to antimicrobial drugs

Authors

  • Diarmaid Hughes

    Corresponding author
    1. Department of Medical Biochemistry and Microbiology, Biomedical Center, Uppsala University, Uppsala, Sweden
    • Address correspondence to: Diarmaid Hughes, Department of Medical Biochemistry and Microbiology, Biomedical Center, Uppsala University, Box 582, Uppsala, Sweden. E-mail: diarmaid.hughes@imbim.uu.se

    Search for more papers by this author

Abstract

The overuse and misuse of antibiotics over many years has selected a high frequency of resistance among medically important bacterial pathogens. The evolution of resistance is complex, frequently involving multiple genetic alterations that minimize biological fitness costs and/or increase the resistance level. Resistance is selected at very low drug concentrations, such as found widely distributed in the environment, and this selects for resistant mutants with a high fitness. Once resistance with high fitness is established in a community it is very difficult to reduce its frequency. Addressing the problem of resistance is essential if we are to ensure a future where we can continue to enjoy effective medical control of bacterial infections. This will require several actions including the discovery and development of novel antibiotics, the creation of a continuous pipeline of drug discovery, and the implementation of effective global antibiotic stewardship to reduce the misuse of antibiotics and their release into the environment. © 2014 IUBMB Life, 66(8):521–529, 2014

Introduction

Antibiotics are the medical wonder of our age. They make possible the safe application of invasive medical procedures where the probability of associated bacterial infections is very high, they prolong the lifespan of the elderly and the immunocompromised, and they reduce morbidity and mortality from common community-acquired infections such as bacterial pneumonia [1-4]. The widespread perception of antibiotics as wonder drugs is also the seed to their destruction as an effective medical tool. As wonder drugs they have been grossly overused and abused [5-8]. In human therapy, they are often prescribed for nonbacterial infections, and in many parts of the world they are self-prescribed and sold with little or no medical oversight. In addition, they are added as growth promoters in agricultural animal feed, added to the water in aquaculture, or sprayed onto orchards. A class of drug that was introduced to medicine to save human lives is now mostly used outside of human medicine. The abuse of antibiotics over the past 70 years has exposed global bacterial populations to a selection pressure to adapt. In abusing antibiotics, humanity has unwittingly conducted an uncontrolled experiment in bacterial evolution on a global scale with potentially serious negative consequences for our continued ability to enjoy the benefits of medical knowledge. The result is that today antibiotic-resistance, and frequently multidrug-resistance, is widespread among bacterial pathogens.

Resistance compromises our ability to deliver high quality medical care in both the community and the hospital environment. Effective antibiotic therapy is essential for many advanced medical procedures (e.g., heart transplants or hip replacements) and for the treatment of cancer patients and others with suppressed immune systems such as the elderly and those with AIDS or those undergoing chemotherapy. In addition, effective antibiotics are essential for treatments requiring the insertion of catheters, a procedure associated with a high risk of bacterial infection. Resistance to antibiotics among human pathogens increases the risks of treatment failure due to the application of an ineffective antibiotic, it also increases the risk of mortality by increasing the time from an initial diagnosis to an effective therapy, and it increases morbidity by increasing the use of more toxic antibiotics as replacements for those rendered ineffective by resistance [9, 10] and imposes an extra healthcare cost and productivity loss that in the EU was estimated as costing at least €1.5 billion in 2007 (Technical Report: The bacterial challenge: time to react (2009) is available on the website of the European Medicines Agency, http://www.ema.europa.eu/docs/en_GB/document_library/Report/2009/11/WC500008770.pdf). The increasing realization that antibiotic resistance imposes a financial burden on society, in addition to exacerbating personal issues of morbidity and mortality, is a strong motivation for governments and other stakeholders to adopt measures to address the problems caused by antibiotic resistance.

However, before heading bravely into the future and implementing a raft of measures to deal with a current problem it could be useful to review how we got to this position. This review will examine how bacteria experience and respond to the selective pressure of antibiotic exposure. In particular, the review will discuss evidence that:

  1. antibiotic usage is a driving force for resistance frequency;
  2. resistance once established is likely to persist in the population;
  3. resistance frequently develops by multistep genetic alterations (MDR can also occur in a single step associated with the acquisition of a mobile genetic element or overexpression of an efflux system);
  4. fitness is an important issue in the fixation of resistant mutants;
  5. selection occurs even at very low antibiotic concentrations suggesting that selection of resistance can occur in the wider environment;
  6. weak antibiotic selection selects for high fitness.

Finally, the review will outline where we are headed and discuss actions that need to be considered if we are to continue to enjoy the medical benefits of effective antibiotics.

Actions Have Consequences

Antimicrobial Resistance Frequency Is Increasing in Europe

There is reliable information on antibiotic usage and resistance frequencies in most European countries, publically available on website of the European Centre for Disease Prevention and Control, ECDC (www.ecdc.europa.eu). The information, contained in the Antimicrobial Resistance Interactive database (EARS-Net, www.ecdc.europa.eu/en/healthtopics/antimicrobial_resistance/database), covers the most important human pathogens and antibiotics, and is updated annually. As an example, the proportion of fluoroquinolone-resistant E. coli isolates has increased significantly throughout Europe in the decade from 2002 to 2012, the most recent year for which data is available (Fig. 1). Overall, the data on EARS-Net show that there is significant variation in the volume of antibiotic usage between European countries, in the use of broad versus narrow spectrum antibiotics, and in the frequencies of resistant pathogens. Although in some cases resistance frequencies show a decrease or no change from one year to the next, in general the trend across Europe is for an increasing proportion of clinical isolates collected over the past decade to exhibit antibiotic resistance or multidrug-resistance.

Figure 1.

Fluoroquinolone resistant E. coli in Europe. Increase in the proportion of fluoroquinolone resistant (R+I) E. coli isolates in Europe from 2002 to 2012. From the website of the European Centre for Disease Prevention and Control (http://www.ecdc.europa.eu/en/healthtopics/antimicrobial_resistance/database/). [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Why Is the Frequency of Resistance Increasing?

The logical suspicion is that the total amount of antibiotic that interacts selectively with bacteria is the major driving force pushing the global frequency of resistance upward. However, the question is complex and any answer is tentative because we are currently observing an uncontrolled experiment, in real time, where cause and effect are difficult to establish. However, if we accept the hypothesis that it is the relative level of exposure to antibiotics which correlates positively with selection of resistant clinical isolates, then we must address at least two related questions: (i) where does this exposure leading to resistance selection occur geographically; and (ii) how does exposure lead to the creation of antibiotic-resistant human pathogens?

Bacterial exposure to antibiotics not only occurs in the human body during therapy but it also occurs in the wider environment. A large fraction of the antibiotics used in therapy are excreted in urine in an active form [11], in addition to which antibiotics are used in very large quantities in veterinary medicine, agriculture, and aquaculture [12, 13]. This raises the question of which geographical locations are most important for selecting the resistant pathogens that cause problems for clinical therapy. For some long-lasting or chronic infections, such as Mycobacterium tuberculosis causing tuberculosis or Pseudomonas aeruginosa causing lung infections in cystic fibrosis patients, there is good evidence that resistance can arise by mutation during the course of therapy [14-16]. One of the best examples demonstrating the stochastic development of antibiotic resistance mutations in patients while undergoing drug therapy is the development of macrolide resistance in Mycobacterium avium septicemia treated by monotherapy [17]. This study analyzed blood isolates of M. avium taken from 38 patients before and after the development of clarithromycin resistance. Point mutations associated with resistance were identified in all resistant relapse isolates but in none of the susceptible pretreatment isolates. In vitro investigation revealed the same point mutations as observed in vivo [17]. However, for most combinations of bacteria and antibiotic, where infections are acute and therapy relatively short-term, there is very little data showing that resistance arises during therapy, although there are a few well-documented examples [10, 18]. In the vast majority of acute infections whenever resistance is detected it is usually assumed to have pre-existed therapy, leaving open the question of how, when, and where resistant strains arise. Within Europe, where very good geographical data are available (www.ecdc.europa.eu/en/healthtopics/antimicrobial_resistance/database), there is a positive correlation between high total levels of antibiotic consumption in a country and high proportions of resistant bacteria. These data also show a positive correlation between the use of broad-spectrum antibiotics and higher proportions of resistant bacteria. The tentative conclusion is that the quantity and quality of antibiotics used in medicine in a country is an important factor in predicting the proportion of resistant bacteria in that country, with the important caveat that this does not take into account antibiotics used outside of human medicine. An important implication of this correlation between total antibiotic usage in human medicine and frequency of resistance in different countries is that it suggests that local policies on antibiotic usage could in principle be used to influence the local frequency of resistance.

Acquired antibiotic resistance arises by two different biological processes: (i) mutation within the existing genome that reduces antibiotic susceptibility and (ii) acquisition of horizontally transferred genetic material that introduces a novel gene conferring reduced susceptibility to an antibiotic [19]. For example, resistance to the first-line drug rifampicin in M. tuberculosis arises exclusively by mutation (there is little or no horizontal genetic transfer, HGT, into this bacterial species). In contrast, resistance to β-lactam antibiotics in E. coli is almost exclusively by HGT of plasmids carrying a gene encoding a β-lactamase enzyme. Resistance to fluoroquinolones in E. coli is typically a mixture of the two mechanisms: chromosomal mutations in target and efflux-regulator genes are nearly always present, but often in combination with a plasmid encoding a gene that further reduces susceptibility [20]. Each mechanism of resistance is more or less relevant to the clinical situation depending on the particular combination of bacterial species and antibiotic.

Can Local Actions Mitigate or Minimize the Resistance Problem?

One can compare neighboring countries with significantly different levels of antibiotic consumption for evidence that local policy can affect resistance frequency. The ECDC (www.ecdc.europa.eu) issues annual reports on the antibiotic consumption (in defined daily doses, DDD, per 1,000 inhabitants per day: DID) for 29 European member countries (Fig. 2). The data show that annual consumption differs by a factor 3.1 between the highest in Greece (35.1 DID) and the lowest in the Netherlands (11.4 DID). Huge differences in consumption between countries have also recently been reported for eastern Europe, ranging from 15.3 DID in Armenia up to 42.3 DID in Turkey [21]. It is especially instructive to compare the Netherlands with its immediate geographical neighbor, Belgium. These countries cover similar geographical areas, have similarly sized populations, and share an open border allowing people to travel freely between the countries, but have separate healthcare systems and implement significantly different antibiotic-usage policies. Belgium has a significantly greater per capita consumption of antibiotics (29.0 DID) than the Netherlands (11.4) and this is associated with a significantly higher proportion of resistant bacterial isolates. For example, the fluoroquinolone resistance data from 2011 recorded for E. coli, K. pneumoniae, and P. aeruginosa. Belgium consumed 2.7 DID fluoroquinolones versus 0.8 DID in the Netherlands. Fluoroquinolone resistance for these three species was recorded as 22%, 15%, and 21%, respectively, in Belgium, versus 14%, 7%, and 7% for the same species in the Netherlands. Similar differences are found for β-lactams, macrolides, and aminoglycosides. A reasonable conclusion is that locally enforced restrictive antibiotic usage practices can be beneficial and associated with locally lower frequencies of resistance.

Figure 2.

Differences in antibiotic consumption in different countries of Europe. Geographical distribution of antimicrobial consumption of antibacterials for systemic use in the community (primary care sector) in Europe, reporting year 2011. From the website of the European Centre for Disease Prevention and Control (http://www.ecdc.europa.eu/en/healthtopics/antimicrobial_resistance/database/). [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Antibiotic Resistance has a Fitness Cost

One reason why restrictive usage of antibiotics might slow the rate of increase in the frequency of resistant isolates in a geographical area is that resistant bacteria often have a lower biological fitness than susceptible strains. Experimental studies have shown that the occurrence or acquisition of antibiotic resistance is often associated, at least initially, with a reduction in bacterial fitness [22-25]. Relative fitness can be defined as a reduction in growth rate in vitro or in vivo or in terms of colonization, transmissibility, or virulence. There are several reasons why a correlation between resistance and reduced fitness is expected. One is that target mutations that reduce the antibiotic affinity frequently reduce the activity of essential bacterial enzymes. Because antibiotics disrupt essential activities in bacteria (cell wall building, protein synthesis, RNA transcription, DNA replication) the specific molecular targets of antibiotics are usually essential and conserved and mutational alterations usually reduce their specific activity with negative effects on bacterial fitness. However, a large part of the clinical resistance problem is caused by the acquisition of plasmids expressing genes for enzymes, such as β-lactamases, that reduce the effectiveness of antibiotics without altering the essential target. The evidence suggests that resistance due to HGT, or alteration of regulation of the resistance mechanism, is frequently associated with reduced fitness, at least initially, but that these costs can be quickly ameliorated by compensatory evolution [26, 27]. The frequent correlation between antibiotic resistance and fitness costs has raised the question whether restrictive usage could be used to as a lever to reduce the frequency of resistant pathogens. The logic of the suggestion is that a high frequency of an antibiotic-resistant pathogen with low-fitness can only be maintained in the population by the selective pressure of antibiotics.

Resistance Persists

The possibility that restricting antibiotic usage might result in a reduction in the frequency of resistant bacteria has been tested in several studies made in closed environments such as hospitals and in open community environments [28]. Interventions in closed systems like hospitals generally show a high level of success and reduction in resistance frequency can be observed within weeks to months [29-31]. In addition to possible fitness costs an important factor contributing to the success of interventions in hospitals is a dilution effect where the incoming population reduces the frequency of resistance within the hospital [32-34]. Interventions in community settings have not been so successful. A 97% decrease in the consumption of trimethoprim/sulfamethoxazole in the UK between 1991 and 1999 did not result in a reduction in sulfamethoxazole resistance [35]. In two other studies, in Finland and Iceland, restriction was associated with a fall in the subsequent frequency of resistance but the possibility that clonal shifts rather than fitness costs were the cause, was not addressed or ruled out [36, 37]. A recent prospective study made in Sweden [38] failed to show any major effect of restricting trimethoprim for a 2-year period in a particular Swedish county. The study was controlled with regard to compliance, information on previous and concurrent resistance frequencies and antibiotic usage, as well as having an equivalent set of data from a comparator county in Sweden. The main conclusion was that even with very good compliance in restricting use of one class of antibiotic (80% reduction in use for the 2-year period), the reduction in the frequency of resistance was marginal and it would require a much longer intervention to achieve even a small reduction. The authors speculated that confounding factors inhibiting a greater reduction in resistance frequency were possibly a low biological fitness cost of the resistance, and possible genetic co-selection by the antibiotics used as a replacement for the one that was restricted [38]. The overall conclusion from these studies is that once antibiotic resistance is established in a bacterial population it may prove to be very difficult to reverse the process by a policy of restrictive usage [28].

Fitness is Important in the Clinical Selection of Resistance

The evidence from in vitro experiments is that antibiotic resistance almost always comes with a fitness cost. However, these experiments also show that the magnitude of the fitness cost can vary depending on the particular genetic alteration, and in some cases the cost is too small to be measured suggesting that it may be entirely absent. This raises the question of whether relative fitness costs play any role in the clinical selection of resistant variants. This is an important question to answer but is difficult to address in a meaningful manner, partly because clinical isolates with resistance phenotypes often have very complex differences in genotype, making it difficult to make comparisons and definitively link cause with effect. However, the question has been successfully addressed in M. tuberculosis by applying in vitro information on the relative fitness of different streptomycin resistance mutations (differences in in vitro growth rates associated with different mutant alleles) with the relative frequencies of these same alleles among clinical tuberculosis isolates. The analysis is possible in M. tuberculosis because the species is relatively uniform genetically, with little or no influence from HGT. The results were clear and striking [39]. Streptomycin resistant mutants selected in vitro in M. smegmatis (a genetic model for Mycobacteria) and in M. tuberculosis each carried one of several different single amino acid substitutions at codon 42 in ribosomal protein S12, each allele arising spontaneously at similar frequencies. In contrast, when streptomycin-resistant clinical isolates of M. tuberculosis were examined the result was very different. Of 90 resistant isolated tested, 89 carried one particular allele: Lys42 substituted with Arg42 [39]. This mutant allele has no measureable fitness cost in vitro. This result suggests that at least for streptomycin resistance in M. tuberculosis there is a strong bias in favor of a low-cost resistance mutation, and that therefore biological fitness is an important selected parameter in the development of antibiotic resistance.

Evolution of Resistance is Often a Multistep Process

The data on the selection of streptomycin resistance above [39], and similar data on other aminoglycosides and macrolides [40, 41], show that single nucleotide mutations are sufficient to cause a high-level resistance phenotype. This is the classical view of resistance development: that resistance occurs because of a single event: either the occurrence of a mutation or the acquisition of a resistance gene by HGT. However, while the example of Lys42Arg causing streptomycin resistance in M. tuberculosis is a good illustration of one change causing clinically relevant resistance, there are good reasons for thinking that in many cases resistance development is a multistep evolutionary process (Fig. 3). Two examples are discussed below.

Figure 3.

Multistep evolution of antibiotic resistance. A general model illustrating evolutionary paths for a lineage under antibiotic selection. Resistance may arise without loss of fitness but more often will involve a loss of fitness and the possibility of additional mutations to restore fitness without loss of resistance. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Genetic Compensation of Fitness Costs of Resistance

There are many examples of in vitro antibiotic resistance evolution where the high fitness cost of a primary resistance mutation has been shown to be genetically compensated by the acquisition of secondary mutations in the same or other genes [23, 25, 42, 43]. However, it is more difficult to test whether that compensatory evolution occurs during clinical resistance selection. Once again, M. tuberculosis is the source of some convincing examples of fitness compensatory evolution. Kanamycin resistance in M. tuberculosis arises by the occurrence of any one of several different point mutations in the gene for 16S rRNA (A1408G, C1409U, G1491U, and the double mutation C1409A/G1491U). When the frequencies of each mutant allele were measured in clinical isolates, one allele, A1408G, was most frequent being found in 95% of isolates. This suggested that A1408G was probably a low cost mutation. Each of the mutant alleles was genetically reconstructed in an isogenic wild-type background and their relative fitness and resistance levels measured [44]. This analysis confirmed that the most frequent allele was indeed a relatively low cost allele that conferred a very high level of resistance (minimal inhibitory concentration [MIC] >1,024), explaining its high frequency among resistant clinical isolates. Especially interesting was the phenotype of double mutation allele that was isolated at a higher frequency (C1409A/G1491U, 1.2%) than a single mutation allele (G1491U, 0.6%). When the phenotypes of the reconstructed strains were measured the rare single mutation was found to confer a moderate level of resistance (MIC 64) but to have a very high fitness cost (12% per generation). In contrast, the double mutant also conferred a moderate level of resistance (MIC 16) but was essentially cost free (0.1% per generation). This seems to be a genuine example of fitness compensatory evolution occurring during the development of clinical resistance.

Another example where fitness compensatory evolution probably occurs clinically is during the development of rifampicin resistance in M. tuberculosis. Rifampicin is a first line antibiotic in the treatment of M. tuberculosis but in vitro selection in many species have shown that high-level resistance arises as a result of individual point mutation in rpoB, encoding the β-subunit of RNA polymerase, the molecular target of the antibiotic. Recently, whole genome sequencing of M. tuberculosis isolates has shown a strong correlation between rifampicin resistance and the presence of additional mutations in genes for different subunits of RNA polymerase [45, 46]. Genetic analysis is difficult in M. tuberculosis so it could only be speculated that the presence of additional mutations might be an example of fitness compensatory evolution. Support for the compensatory hypothesis came from a genetic analysis made in Salmonella enterica [47, 48]. They showed that rifampicin resistance arises by the selection of any one of several mutations in a small region of rpoB and that most of these mutations have a significant fitness cost. When mutant strains were evolved in vitro they rapidly acquired secondary mutations that were shown by genetic reconstruction to be fitness compensatory mutations. These compensatory mutations occurred in the genes for each of the major subunits of RNA polymerase, and tellingly they mapped in very similar locations to the additional mutations discovered in rifampicin-resistant M. tuberculosis isolates. These data strongly suggests that the development of resistance to rifampicin in M. tuberculosis very frequently involves the selection of second-site fitness compensatory mutations.

The data presented above concerning in vitro competition assays and the relative clinical prevalence of specific antibiotic resistance mutations strongly suggest that at least some antibiotic resistance mutations impose a fitness cost on M. tuberculosis. Given that M. tuberculosis grows very slowly one could ask what is the nature of that fitness cost: is it growth rate or some other parameter? A recent population-scale analysis of natural populations of M. tuberculosis has found evidence that genes involved in transport and in the metabolism of inorganic ions are under strong purifying selection [49]. The authors speculate that the hostile within-host environment imposes strict demands on M. tuberculosis physiology, and that this associates a substantial fitness cost with new mutations, accounting for the purifying selection. The targets of antibiotics for which fitness costs have been demonstrated in vitro are in the ribosome and the RNA polymerase. Alterations in these targets will very likely result in pleiotropic effects on bacterial physiology in addition to their influence on drug susceptibility. Accordingly, while it is possible that reduced relative growth rate per se is the relevant in vivo fitness cost of antibiotic resistance, it is also possible that growth-rate-neutral alterations in cell physiology associated with the resistance mutations cause the actual fitness cost in the patient. Regardless, the current data suggest that measurements of growth competitiveness in vitro are a good surrogate measure for the relative fitness of M. tuberculosis in vivo.

When One Change is Insufficient to Confer Clinical Resistance

Another situation in which multistep evolution occurs is when a single genetic change is insufficient to confer a clinically relevant level of resistance. This is the situation with regards to fluoroquinolone resistance and E. coli [24, 50]. There is no single genetic alteration that is known to confer resistance in E. coli. Individual target mutations, efflux regulator mutations, or plasmids expressing resistance genes, all result in an MIC that is at best 5–10-fold below the clinical breakpoint. Resistant clinical isolates always carry multiple resistance mutations and frequently also carry resistance plasmids. A typical fluoroquinolone-resistant E. coli isolate will have four to six and possibly more genetic alterations directly linked with increasing the MIC [50, 51].

Selection Occurs Way Below Mic

Antibiotic resistant mutants will be selectively enriched in a drug concentration range referred to as the mutant selective window (MSW). The MSW has been defined as the antibiotic concentration range between the MIC of a susceptible strain and the mutant preventive concentration (MPC) of the most-resistant single-step mutant [52]. In other words, at concentrations above the MIC, concentrations that prevent the growth of the susceptible strain, mutants with reduced susceptibility will be selected. However, this traditional model does not take into account what happens at concentrations below the MIC of the susceptible strain. Sub-MIC concentrations occur at the beginning and end of therapy due to the pharmacodynamics of drugs. Sub-MIC concentrations will also occur during therapy due to poor drug penetration at certain body sites, poor patient compliance, or inadequate dosing regimen that allows concentrations to go below MIC, and will occur very frequently when antibiotics are diluted in the wider environment. It has been speculated that low concentrations of antibiotics might play a role in the selection of clinical resistance [53, 54] but it is only recently that experimental tests of the effects of antibiotics at sub-MIC on the selection of resistance have been made [55, 56]. In these studies the relative competitive fitness of isogenic mutants carrying common resistance mutations or genes was measured at a range of antibiotic concentrations. The major finding of both studies was that mutants were in every case tested selected at antibiotic concentrations below the MIC of the susceptible strain. The logic of sub-MIC antibiotic selection is illustrated in Fig. 4. In the absence of any antibiotic the susceptible strain typically has a competitive advantage (because of the fitness cost of the resistance determinant). However, at a certain concentration the inhibitory effect of the antibiotic on the susceptible strain will balance the fitness cost of the resistance determinant. This is the minimal selective concentration (MSC), above which the resistant strain has a selective advantage. In the concentration range between the MSC and the MIC both strains can grow but the resistant strain has a competitive advantage and will be enriched. In the traditional selective window, above the MIC of the susceptible strain, only the resistant mutant can grow.

Figure 4.

Antibiotic selective windows theory. Between the MIC of the susceptible strain, MICsusc, and the minimal selective concentration (MSC) there is a sub-MIC selective window where both susceptible and resistant mutants can grow but resistant mutants are selectively enriched by the antibiotic. Reproduced from Ref. [55]. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

There are several important conclusions that came out of these studies. One is that sub-MIC selection occurred with all antibiotics tested. A second is that sub-MIC also selected de novo resistance [55]. A third is that the concentration range for sub-MIC selections went as low as more than 200-fold below the MIC of the susceptible strain [55]. Thus for the most common mutation associated with ciprofloxacin resistance, GyrA Ser83Leu, the MSC was 1/230 MIC and 0.1 ng/mL. This concentration is below the ciprofloxacin concentration found in many natural environments [57-59] and suggests that sub-MIC selection of resistance might be a relevant and widespread phenomenon in nature. This returns us to the question of whether resistance is primarily selected and enriched by antibiotic contamination in the wider environment, prior to becoming as a problem in human infections.

Weak Selection Selects Low-Cost (High-Fitness) Mutants

The significance of sub-MIC selection is twofold. One is that it vastly extends the geographical area where antibiotic resistance can be selected, to include soils, lakes, rivers, seas, as well as compartments in the human body undergoing antibiotic therapy, and less susceptible members of the human microbiome. However, a second and potentially even more important consequence of sub-MIC selection is that it intrinsically selects for low-cost resistance. The reason for this connection is apparent in Fig. 4. Increasing concentrations of antibiotic progressively reduce the relative fitness of a susceptible strain. In general the relative magnitude of the reduction in fitness (e.g., growth rate) caused by an antibiotic acting on a susceptible strain will be proportional to the concentration of the antibiotic. For any pair of susceptible and resistant strains there will be some antibiotic concentration at which their fitness difference is zero. Accordingly, if the intrinsic fitness difference between the strains, in the absence of antibiotic, is very small it will only require the presence of a relatively low antibiotic concentration to equalize their fitness. Any antibiotic concentration above this value (the MSC, minimal selective concentration) will further suppress the relative fitness of the susceptible strain and confer an advantage on the mutant strain. This is the situation for GyrA Ser83Leu which is selected at 1/230 the MIC of the susceptible wild-type [55]. Its low fitness cost may explain why it is so prevalent among resistant clinical isolates [50, 51]. Concentrations of ciprofloxacin found widely in natural environments will be sufficiently high to give a growth advantage to mutant variants carrying the Ser83Leu resistance mutation. The implication is that if clinical resistance evolves via sub-MIC selection then it is likely to be associated with very low fitness costs and will not be amenable to reversal by restrictive usage policies that are imposed after resistance has been established in the community.

Summary and Conclusions: What Can or Should We Do to Address the Problems?

This review summarizes some of what we know about the biology of antibiotic resistance selection and points out that there are still many gaps in our knowledge. What is clear at this stage is that a high frequency of resistance, once established in the community for a particular combination of pathogen and antibiotic, is very unlikely to be reduced by the subsequent introduction of restrictive usage policies. It will already be too late, the reason being that resistance will most likely already have been selected for a very low fitness cost.

Human medicine, relying heavily on effective antibiotics, faces a potential crisis in the coming years unless solutions can be found. Here are some obvious suggestions for actions to address the situation.

  1. Restrict the use of antibiotics, and their release into the wider environment, to reduce selection pressure. As argued above, restrictive use after resistance is established is unlikely to reduce the frequency of resistance, but it might reduce the rate at which the frequency of resistance to existing antibiotics increases. This will be important to buy time while research and development to find novel antibiotics continues. In addition, if and when novel antibiotics are introduced to the clinic, having such restrictive measures in place from the beginning will be important to reduce resistance selection and prolong the effective lifespan of the new drugs.
  2. Develop improved therapies, including combination therapies with existing drugs, that are effective against current resistant strains. This action may buy some time while new antibiotics are researched but it is very unlikely to provide a long-term solution.
  3. Develop improved diagnostics. These should be available at the point of care and should include rapid tests for susceptibility to support the use of the most appropriate antibiotic therapy, and so that narrow spectrum antibiotic therapy becomes a more attractive and widely used option. This could have the dual benefit of improving the economics of narrow spectrum drug development and also reduce the selection pressure for resistance by minimizing the exposure experienced by nontargeted bacteria.
  4. Develop new antibiotic drugs that are active against current resistant strains. Included here could be not only classical antibiotics but also adjuvants that enhance the effects of antibiotics, and antivirulence drugs that may also prove effecting in infection control. The discovery and development of novel drugs is the only realistic long-term option to maintain effective infection control in medicine. However, its success will depend not only on scientific discovery and development but also on establishing a viable economic model to ensure a continuous pipeline of drugs, and an appreciation of the value of antibiotic stewardship to ensure that new drugs remain effective for a long time.
  5. Improve the quality of hygiene measures, both at the hospital and in the community, including reducing the release of antibiotics into the environment.

The political realization that effective healthcare faces a potential crisis because of antibiotic resistance has recently resulted in major initiatives being taken in both Europe and the USA to tackle the problem. The European Union via its Innovative Medicines Initiative has launched the New Drugs for Bad Bugs (ND4BB) programme in 2012 to tackle a range of issues associated with antibiotics and antibiotic-resistance [60]. The aim is to combat antibiotic resistance by tackling the scientific, regulatory, and business challenges that hamper the development of new antibiotics. One of the programmes in ND4BB, ENABLE, is charged with advancing the development of novel antibiotics against Gram-negative bacteria and with identifying promising antimicrobial candidates for testing in clinical trials. The overall aim of ND4BB is to encourage public-private collaborations to improve early-stage antibacterial drug discovery and help make a sustainable impact on the future of antibacterial drug discovery. The reasonable assumption is that resistance will ultimately limit the effective useful lifespan of any new antibiotics introduced to the clinic. The long-term solution is to develop a continuous and sustained drug discovery pipeline and to put in place antibiotic stewardship policies that ensure maximum medical benefit from these investments.

Acknowledgements

D.H. acknowledges support from Vetenskapsrådet (Swedish Science Council), SSF (Swedish Strategic Science Foundation), Vinnova (Swedish Innovation Science), and the Knut and Alice Wallenberg Foundation (RiboCore Project).

Ancillary