Intended for healthcare professionals

Clinical Review State of the Art Review

Antibiotic resistance in Enterobacteriaceae: mechanisms and clinical implications

BMJ 2016; 352 doi: (Published 08 February 2016) Cite this as: BMJ 2016;352:h6420
  1. Jon Iredell, professor/director1,
  2. Jeremy Brown, microbiology registrar2,
  3. Kaitlin Tagg, PhD candidate1
  1. 1Westmead Institute for Medical Research, University of Sydney and Marie Bashir Institute, Sydney, NSW, Australia
  2. 2Department of Microbiology, Institute of Clinical Pathology and Medical Research, Sydney
  1. Correspondence to: J Iredell jonathan.iredell{at}


Resistance of the Enterobacteriaceae to antibiotics, especially of the β lactam type, is increasingly dominated by the mobilization of continuously expressed single genes that encode efficient drug modifying enzymes. Strong and ubiquitous selection pressure has seemingly been accompanied by a shift from “natural” resistance, such as inducible chromosomal enzymes, membrane impermeability, and drug efflux, to the modern paradigm of mobile gene pools that largely determine the epidemiology of modern antibiotic resistance. In this way, antibiotic resistance is more available than ever before to organisms such as Escherichia coli and Klebsiella pneumoniae that are important causes of major sepsis. Modulation of the phenotype by host bacteria makes gene transmission less obvious and may in part explain why tracking and control of carbapenem resistance has been particularly problematic in the Enterobacteriaceae. This review discusses the underlying principles and clinical implications of the mobility and fixation of resistance genes and the exploitable opportunities and potential threats arising from apparent limitations on diversity in these mobile gene pools. It also provides some illustrative paradoxes and clinical corollaries, as well as a summary of future options.


Antibiotic resistance typically occurs within a few years of the introduction of a new antibiotic. Such resistance is not surprising because most modern antibiotics are derived directly or indirectly from microbial products. To mitigate this problem pre-existing resistance mechanisms may be identified in target pathogens even before the introduction of a new antibiotic.1 Bacterial DNA sequences in the human gut indicate the presence of proteins that are similar to important antibiotic resistance enzymes,2 and genes that are similar to “modern” antibiotic resistance genes are found in the environment and in samples dating back millions of years.3 4 This suggests an almost unlimited capacity within the global microbiome to resist any new antibiotic, which existed long before the evolution of modern humans (fig 1).


Fig 1 Timeline of bacterial evolution. The introduction of each antibiotic has been quickly followed by recognition of an adaptation (including a phenotype or mechanism, as marked on resistance timeline) to resist it. *Not in clinical use. AmpC=ampicillin hydrolyzing; CTX-M=cefotaximase, first identified in Munich; IMP=imipenemase; KPC=Klebsiella pneumoniae carbapenemase; MRSA=meticillin resistant Staphylococcus aureus; NDM=New Delhi metallo β lactamase; PRSA=penicillin resistant S aureus; TEM, SHV, and OXA=common and diverse groups of β lactam hydrolyzing enzymes; VRE=vancomycin resistant enterococci; VRSA=vancomycin resistant S aureus; VIM=Verona integron encoded metallo β lactamase

This review will focus on the management and control of antibiotic resistance in medically important Gram negative bacteria, particularly Escherichia coli and Klebsiella pneumoniae, which are the main agents implicated in severe sepsis and septic shock. It will examine the mechanisms by which resistance becomes established in bacterial populations and the basic principles underlying rational management, including antimicrobial stewardship and infection control surveillance. The review deals mainly with resistance to β lactam antibiotics because of the continuing clinical importance of these agents, particularly for the management of severe infection, and the emerging threat of carbapenem resistance in the Enterobacteriaceae (genera such as Escherichia, Klebsiella, Citrobacter, Serratia, Salmonella, Enterobacter, and Proteus). References to comparator bacteria illustrate important differences; table 1 provides a classification of common pathogenic Gram negative bacteria and the agents used to treat associated infections.

Table 1

Taxonomy and treatment of infection with common pathogenic Gram negative bacteria

View this table:

Sources and selection criteria

The references used in this review were identified through PubMed and Medline searches of articles published between 1980 and 2015 and through our personal libraries. Search terms included “mechanisms” and “resistance”, “ecology” and “resistance”, “horizontal gene transfer”, and “treatment” and “carbapenem resistance”. We prioritized early and definitive references, but cited recent high quality reviews when multiple references were relevant. We prioritized references relating to those antibiotic resistance mechanisms associated with E coli and K pneumoniae and those that are known to be transmissible and most pertinent to the widely used aminoglycoside and β lactam antibiotics, particularly third generation cephalosporin and carbapenem antibiotics.

Incidence and prevalence

Antibiotic resistance in human isolates

The 2014 World Health Organization report on global resistance provided a snapshot of Africa, the Americas, the eastern Mediterranean, Europe, South East Asia, and the western Pacific.5 The carriage of extended spectrum β lactamases (ESBLs; see Glossary) confers resistance to many penicillins and cephalosporins (table 2) and often co-occurs with mechanisms that confer resistance to other types of antibiotics. The prevalence of ESBL-type resistance in both E coli and K pneumoniae varies widely across countries (fig 2) and is presumably related to factors such as antibiotic availability and restriction, waste and water management, and the general standard of living and healthcare.5 6 7 8

Table 2

Selected important β lactamases in the Enterobacteriaceae

View this table:

Fig 2 Antibiotic resistance by region. Increasing rates of extended spectrum β lactamase (ESBL) carriage in Klebsiella pneumoniae (KP) and Escherichia coli (EC) and high rates of fluoroquinolone resistance are seen in all regions, whereas amikacin resistance remains relatively rare. Data were collated from the SMART studies of intra-abdominal and urinary tract infection.9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 Fluoroquinolone and aminoglycoside resistance data for the Asia-Pacific region come from Mendes and colleagues27

Some parts of the world report that 60% or more of E coli and K pneumoniae are resistant to important β lactam hospital antibiotics, such as third generation cephalosporins (for example, cefotaxime), and travelers regularly import resistance into countries with lower prevalence.28 29 30 31 The use of β lactam and aminoglycoside antibiotics (such as gentamicin) has long been widespread and resistance has been managed with carbapenems (such as meropenem and doripenem) and fluoroquinolones (such as ciprofloxacin and levofloxacin).

Fluoroquinolone resistance is now also present in at least half of reported clinical isolates of E coli from many locations around the world however,32 and some European, South East Asian, and eastern Mediterranean regions report similar prevalence rates for carbapenem resistance.5 The co-occurrence of resistance to extended spectrum β lactams, carbapenems, aminoglycosides, and fluoroquinolones is well described.

Even in countries with strict controls on antibiotic use and where antibiotic resistance is uncommon, resistance seems to be rising. In Canada and Australia, the prevalence of ESBL phenotypes in E coli has roughly doubled from around only 3% five years ago.33 34 35 Ciprofloxacin resistance in clinical isolates of E coli increased from 5.4% in 2010 to 6.9% in 2012 in Australia, and from 21% to 27% between 2007 and 2011 in Canada, where fluoroquinolone usage has been somewhat higher than in Australia.33 34 35

Subgroups with high prevalence of resistance

“New” antibiotic resistance threats are often first recognized in travelers returning to countries with well developed health systems and laboratory diagnostics,36 and screening of those returning after healthcare exposure is widely recommended.29 37 Returning Dutch and Swedish international travelers were reported to have ESBL colonisation rates of 24%, from a very low baseline in those countries, and 10-fold higher colonisation rates of E coli that contained ESBL were reported in asymptomatic military personnel stationed in Afghanistan compared with those stationed in the United States.38 39 40 Long term residents of aged care facilities, who have frequent exposure to healthcare and antibiotics and experience common cross-transmission of bacteria, also tend to have a high prevalence of antibiotic resistance.41 42 43 44 45

Much of the available antibiotic resistance data are from clinical (typically, hospital) contexts, in which antibiotic exposure and cross-transmission are expected to be increased, especially in resource limited settings. However, prevalence varies between populations in a region. For example, recent studies of urine samples from asymptomatic patients report no carbapenem resistance and that only 20% of E coli and less than 10% of K pneumoniae had ESBL-type resistance in rural settings near Delhi,46 where high levels of antibiotic resistance are commonly reported in clinical isolates. This has implications for the design of surveillance and it implies that antimicrobial stewardship and infection control initiatives still have much to offer.

Antibiotic resistance in isolates from environmental and agricultural sources

Antimicrobials in animal husbandry and the effluent and waste from hospitals and factories are probably important drivers of resistance.47 The feces of humans and animals contaminate the environment, and drinking and environmental water supplies may harbor highly resistant E coli in both resource poor and rich countries.48 49 50 51 52 53 54 Antibiotic resistant human pathogens are also common in food and in food chain animals,55 56 57 58 59 60 61 and household pets may carry similar multi-resistant isolates to humans.62 63 Wild animals are often affected,64 particularly scavengers such as seagulls, which are important vectors of antibiotic resistance including in known human pathogens.65 66 67

Defining antibiotic resistance

The designation of a pathogen as antibiotic susceptible or resistant is a key role of the diagnostic microbiology laboratory. This is done primarily by defining the minimum inhibitory concentration (MIC; see Glossary) at which bacterial growth is inhibited under standardized conditions in vitro.68 Consensus “breakpoints” define susceptible, resistant, and (sometimes) intermediate MIC ranges for specific bacteria. They are based on whether the MIC can be reliably achieved in a patient with regard to drug dosing (pharmacokinetics) and mode of action (pharmacodynamics) and could therefore be reasonably expected to result in therapeutic success.69 These decisions may be further informed by the determination of an “epidemiological cut-off” (see Glossary), which describes the normal distribution of the MIC within a population.70 71

A resistant MIC predicts antibiotic failure but a susceptible MIC is no guarantee of success. Therapeutic failure despite in vitro susceptibility can be caused by reduced antibiotic penetration or activity (in sites such as abscesses or beyond the blood-brain barrier) and sometimes by high local concentrations of a hydrolyzing enzyme produced by dense populations of bacteria (the “inoculum effect”; see Glossary).71 In addition, bacteria are generally most vulnerable to antibiotics when they are rapidly growing and metabolizing (for example, in the diagnostic laboratory) because most antibiotics target these processes. Unfortunately, however, the growth stage of an organism in a clinically important infection may differ greatly from that in which antibiotic susceptibility is determined in vitro.

Bacterial biofilms (see Glossary) are common in the clinical environment on abiotic surfaces including catheters and implanted prostheses, as well as infected natural surfaces such as bone, cartilage, and heart valves. Biofilms often contain populations in growth phases that make them less vulnerable to antibiotics that target growth and remodeling processes. This disparity between the susceptibility of vegetative organisms and those in biofilms often results in relapse of infection a few days after antibiotics are stopped, typically after a good initial response (fig 3).72 73


Fig 3 Differences in growth states and antibiotic susceptibility. Bacteria in a biofilm state (blue bacteria) growing on a prosthetic hip joint (A) release occasional rapidly growing organisms (yellow) into the blood (B). Antimicrobial susceptibility testing of these isolates in ideal growth conditions (C) may overestimate the susceptibility of organisms in the biofilm. This is because rapidly growing bacteria are easily killed by standard therapy (yellow curve, centre) whereas bacteria in less susceptible growth states may survive initial therapy (blue curve, centre). The shape of these curves and the associations between them vary with bacterial and drug types

Bacterial adaption in Gram positive and Gram negative bacteria

Clinical outcomes have been most closely related to the MIC, rather than the minimum bactericidal concentration (MBC) beyond which viable cells cannot be recovered in standardized protocols with moderate inocula (see Glossary), but adaptive change occurs at all levels of exposure that do not result in bacterial death. There may be great differences between MIC and MBC, which vary with organism, drug, and growth conditions as different populations arise among the progeny of a single organism to suit changed conditions.74 75 Acquired antibiotic resistance thus results from normal adaptive capacities that are a balance of compromises or evolutionary trade-offs,76 77 78 and that do not necessarily require acquisition of new genetic material.

Bacterial strategies for coping with antibiotic selection pressure (see Glossary) include target modification, drug exclusion or expulsion, and drug modification. Table 3 and fig 4 compare these strategies for some of the important and common antibiotic classes used for medically important Gram negative and positive bacteria. The Gram stain highlights important biological differences that make adaptive strategies more predictable for given organism-drug combinations. Gram positive organisms have a simpler peptidoglycan-rich cell wall than Gram negative bacteria, and their lifestyles are often more externalised (for example, Staphylococcus aureus persists on dry surfaces). By contrast, Gram negative bacteria typically have a lifestyle in which liquid phase motility, the management of permeable channels through their hydrophobic outer envelope, and the exchange of information (including genetic information) with near neighbors may be much more important.

Table 3

Different resistance mechanisms for three important antibiotic classes

View this table:

Fig 4 Important mechanisms of antibiotic resistance in the Enterobacteriaceae. Porin deficiencies or alterations (A) reduce antibiotic access and efflux pumps (B) may actively transport antibiotics out of the cell. β lactamases (C) acting in the periplasmic space hydrolyse β lactam antibiotics and thereby prevent disruption of the cell wall. Intracellular (for example, aminoglycoside modifying) enzymes (D) alter antibiotics. 16S rRNA methylases (E) prevent aminoglycoside binding. Mutations in targeted DNA gyrase and topoisomerase IV genes (F) render fluoroquinolones ineffective. Pentapeptide Qnr proteins (G) prevent fluoroquinolones from effectively binding to DNA gyrase through target mimicry

Role of the outer envelope in Gram negative bacteria

Gram negative bacteria that tolerate an environment in which external toxins are abundant and osmotic pressures variable (such as Acinetobacter baumannii, Pseudomonas aeruginosa) are usually much less permeable than members of the Enterobacteriaceae that operate competitively in dense communities in the relatively protected environment of the mammalian gut. Transmembrane pumps are important for coping with drugs that act intracellularly (such as aminoglycosides, quinolones, tetracyclines, and macrolides) but provide little protection against drugs such as β lactam antibiotics, which act in the periplasmic space between the inner cytoplasmic (plasma) membrane and the hydrophobic outer membrane.79

A range of relatively non-specific pores (porins) in the hydrophobic outer membrane are the main barrier to β lactam antibiotics in Gram negative bacteria (fig 4). The outer membrane of environmentally hardy A baumannii is around seven times less permeable to carbapenem and cephalosporin antibiotics than that of P aeruginosa and up to 100 times less permeable than E coli, largely due to differences in the type and number of outer membrane porins.80 Such differences have a predictable effect on antibiotic resistance strategies (fig 5).80 81 In general, Gram negative organisms best adapted to survive in hospital drains or on equipment (Acinetobacter, Pseudomonas, and even Enterobacter spp) are most likely to combine reduced outer membrane permeability and periplasmic hydrolysis with efflux systems82 to produce strong resistance phenotypes (fig 5).


Fig 5 Variable resistance strategies in Gram negative bacteria. Organisms living in toxic or occasionally hyperosmolar environments often have mechanisms for restricting antibiotic influx and exporting antibiotics. By contrast, those that live in a relatively less noxious and more stable environment in the gut may be more adapted to compete by acquiring genetic material from neighbors. The antibiotic resistance mechanisms in a given bacterial type can be predicted to some extent by its lifestyle

Intrinsic mechanisms of drug modification

Drug modification is a major mechanism of resistance. Chromosomally encoded penicillin hydrolyzing (ampicillin hydrolyzing, AmpC) enzymes have general structural similarities to the ubiquitous penicillin binding proteins that have an important role in cell wall remodeling. These enzymes are commonly found in medically important Gram negative bacteria in the orders Enterobacteriales and Pseudomonadales. Chromosomal AmpC enzymes are variably induced on exposure to β lactam antibiotics, such as cephalosporins, and other agents such as aztreonam. AmpC enzymes are less active against some modified penicillins such as oxacillin, but are also less effectively inhibited by the classic β lactamase inhibitors (such as clavulanate).

Strong AmpC inducing antibiotics that are also good AmpC substrates (such as ampicillin, first generation cephalosporins) are clearly ineffective in vitro and therefore create no confusion when tested in the diagnostic laboratory. By contrast, weak inducers that are relatively poor substrates (such as third generation cephalosporins) can appear effective in vitro but occasionally fail in vivo due to selection of derepressed mutants (see Glossary) that significantly overproduce the enzyme compared with wild-type organisms (see ECOFF in Glossary).

Chromosomal AmpC induction is complex and variable,83 but the important clinical point is that derepression-type mutations that lead to stable high level expression of AmpC-type enzymes are not unusual and can lead to clinical failure.84 Consequently, relevant authorities urge vigilance for the development of resistance to cephalosporins while treating infections caused by Enterobacter, Serratia, and Citrobacter spp in particular. These bacteria are often reported by microbiology laboratories as resistant despite in vitro susceptibility, on the presumption that this mechanism is present.85

Acquired mechanisms for drug modification

A range of genes encoding drug modifying traits have emerged into the mobile gene pool, expressed from constitutive promoters (see Glossary) at similar and apparently optimized levels, and many move easily between bacteria like E coli and K pneumoniae.86 Despite the broad protection that accrues from other mechanisms such as drug exclusion or specific removal from the cell, even environmentally adapted bacteria like Acinetobacter and Pseudomonas spp commonly acquire additional drug modifying traits in this way, implying that this is an efficient adaptive strategy.

The means of horizontal transfer of genetic traits varies with bacterial lifestyle. For example, marine organisms such as Vibrio cholerae are likely to receive advantageous traits from formally packaged bacterial viruses (transduction by bacteriophages).87 88 89 However, organisms that live in the human gastrointestinal tract may be more likely to directly take up free DNA (transformation) or plasmids, including self transmitting (conjugative) plasmids (see Glossary) (fig 4).90

Resistance genes are made available in this gene pool after being mobilized from the chromosome of a range of organisms through the “copy and paste” replication mechanisms of small genetic elements such as transposons, integrons, and insertion sequences (see Glossary: “mobile genetic elements”).91 92 93 The subsequent transfer of these gene packages into efficient vehicles such as conjugative plasmids allows their dissemination into pathogenic strains, provided the donor and recipient are “ecologically linked” through a shared habitat (such as the gastrointestinal tract) or a chain of organisms that link the donor to the final organism (fig 6).94 95


Fig 6 Ecology of antibiotic resistance. The processes of horizontal gene transfer—transduction, transformation, and conjugation—mobilize genes between different environments. Rare gene capture events mediated by small mobile genetic elements mobilize the gene (for example, from chromosome to plasmid or conjugative transposon; 1), which may in turn transfer to a new bacterial cell (1B). Packaging and transfer of a resistance gene by bacteriophages (transduction; 2) is probably very rare. Uptake of free DNA by competent bacteria (transformation; 3) may also occur. Under antibiotic pressure, resistant organisms have a selective advantage and become part of the resistance gene pool (A). They may then spread as a successful clone (B) from which conjugative plasmids may be disseminated widely (C). Small mobile elements may move by formal mechanisms or by simple recombination between multi-resistance regions (D) in plasmids or chromosomes. Rare mutations that arise in existing resistance genes can generate new variants with activity against many different antibiotics (E)96

The “capture” of the original gene in this way is rare and its subsequent mobilization on to a suitable vehicle such as a plasmid equally so. The ultimate success of a gene in the mobile gene pool relates to its capacity for transfer to other successful mobile vehicles and the access of these vehicles into successful host bacteria.71 97 98 This enhanced mobility (for example, within a conjugative plasmid) is crucial to the success of a resistance gene.99 Mutations that subsequently occur within the resistance gene may not be competitive enough to persist.100 Alternatively, they may give rise to more successful variants—for example, the widespread blaTEM and blaCTX-M gene families seem to have evolved in situ after their initial mobilization to confer a broad and diverse range of advantageous phenotypes.101 102

The rarity of successful gene capture and mobilization events constitutes an ecological bottleneck (fig 6) that results in a relatively small pool of genes that can be acquired and transmitted to and between human pathogens.94 103 As a result, some resistance genes are globally disseminated whereas others are never seen more than once in a clinical setting. For example, genes encoding resistance to third generation cephalosporins ( or aminoglycosides are diverse and numerous,104 but few are commonplace,105 and this enables genotypic tests with high predictive values.106 107

The development of multi-resistance regions

Resistance to several antibiotics can be carried on individual resistance plasmids or on the chromosome. The accumulation of resistance genes around an initial insertion event in a region of DNA that acts as a “founder element” creates dynamic and diverse multi-resistance regions in chromosomes and plasmids.91 There are two important corollaries of the accretion of resistance genes into multi-resistance regions, both of which promote their ecological success. The first is that the increasing prevalence of common genetic sequences may increase the ease with which new resistance genes are incorporated—for example, by homologous recombination (see Glossary) or as a result of their enrichment with gene capture systems (see Glossary). The second is that each multi-resistance region has pluripotent resistance potential, so that exposure to one drug selects for resistance to many.98

For example, gentamicin resistance genes are commonly found with the ESBL gene blaCTX-M-15 on plasmids in E coli,108 and treatment with gentamicin will therefore often (co-)select organisms that are resistant to both gentamicin and extended spectrum β lactams. This makes the choice of an alternative antibiotic to minimize selection for a particular gene or phenotype increasingly difficult, but it is essential to understand these associations to aid effective antimicrobial stewardship.

Implications for infection management and control

Variable expression of resistance traits

Porins, efflux pumps, and inducible AmpC enzymes, are actively regulated for optimal efficiency, but horizontally acquired genes are generally constitutively expressed from promoters within the insertion sequence or gene capture system itself.86 Acquired aminoglycoside modifying enzymes produce a clinically important aminoglycoside resistant phenotype as reliably in E coli as in P aeruginosa, but this is not so for other resistance traits. For example, variants of the quinolone resistance trait QnrB are common on antibiotic resistance plasmids in E coli but do not produce a clinically significant quinolone resistant phenotype alone.109 This “conditional” phenotype is commonly associated with enzymes that hydrolyze β lactams, particularly, carbapenem antibiotics, and is an important consideration for most contemporary antibiotic prescribers.

Most major β lactamases such as the CTX-M-type ESBL enzymes hydrolyze their primary targets (for example, cefotaxime) so well as to result in clinically important antibiotic resistance without the need for additional mechanisms that remove the antibiotic from the bacterial cell or restrict antibiotic entry to the cell. Consequently, they are predictably associated with a clinically significant level of antibiotic resistance even in highly antibiotic permeable species such as E coli. MIC variations all exceed the susceptibility breakpoint but the variations in MIC normally go unreported by diagnostic laboratories, the organisms being described simply as either “susceptible” or “resistant.”

This dichotomous reporting creates confusion in the case of carbapenemase producing Enterobacteriaceae. For example, in E coli and K pneumoniae that produce the metallo β lactamase (MBL) enzyme IMP-4, the MIC for carbapenem antibiotics is commonly below the breakpoint value, except in the presence of an augmenting factor (such as a porin defect).110 111 Similarly, the OXA-24 enzyme is associated with marked carbapenem resistance in low permeability A baumannii but not with levels of antibiotic resistance in E coli that would suggest a risk of treatment failure.112 By contrast, the “K pneumoniae carbapenemase” (KPC) enzyme is almost always linked to marked carbapenem resistance, and this is possibly because it is most commonly found in K pneumoniae strains in which this phenotype is augmented (see below).

In addition, structural similarities between carbapenems and other β lactam antibiotics mean that many carbapenems are susceptible to attack by ESBL and AmpC enzymes, although not very efficiently and usually not resulting in clinically relevant increases in carbapenem MICs that would be expected to be associated with treatment failure. This means that although carbapenemases generally result in much higher carbapenem MICs in the same host strain, ESBL and AmpC enzymes expressed in a porin deficient host may be more common causes of clinically significant carbapenem resistance where carbapenemases are rare.

The plasticity of the “accessory genome”

Bacteria that frequently share genes often do so through the exchange of plasmids, which provide access to the enormous genetic potential within the microbiome in places such as the human gastrointestinal tract.2 113 114 115 116 In addition to common small non-mobilizable plasmids, this “accessory genome” often includes several different large, transmissible, low copy number plasmids of 60-200 kb in each bacterial cell, even in wild animals without particular antibiotic resistance.117 This accessory genome thus comprises 10% or more of the total genome of species such as E coli and K pneumoniae. Many of these plasmids are conjugative (self transmissible) or mobilizable (with the help of a conjugative plasmid). Conjugative plasmids can quickly convert life threatening bacteremic sepsis from being antibiotic susceptible to resistant after only one or two antibiotic doses by direct acquisition of the plasmid by the pathogen or by expansion of subpopulation(s) in which that plasmid resides.118

Discussion of the relatedness of plasmids and of more specific approaches to define relationships between plasmids119 is beyond the scope of this review but the plasmid replicon type (see Glossary) can be regarded as a key determinant of both (in)compatibility and of host range (that is, the capacity for a plasmid to become stably established in a given bacterial population). Mutual incompatibility of plasmids that share the same replication system has been used for decades as a convenient typing scheme, and polymerase chain reaction (PCR) based replicon typing is a popular surrogate.120 Plasmid stability and “addiction” systems (see Glossary) that ensure persistence of the plasmid in a bacterial population are linked to the replicon itself, and the presence of specific plasmid addiction systems can generally be predicted by replicon type (see “Ecological fixation of antibiotic resistance” section).

The natural plasmid complement of a given bacterial cell does not necessarily contain antibiotic resistance genes, but these may be acquired from any DNA locus (including another plasmid) as part of a mobile genetic element that has originally “captured” the gene from elsewhere. The original source of a gene in the mobile pool is not always known, however, and may be distant in time and environmental context from that of the bacterial population in which it is first recognized.

Successful genes, plasmids and clones


The most successful of the ESBL genes, blaCTX-M-15, seems to have been captured in its original form from Kluyvera ascorbata, a member of the Enterobacteriaceae that is rarely pathogenic in humans, by a mobile genetic element,121 122 along with a small amount of other genetic material. It is efficiently expressed by a promoter that resides within the element that initially captured it, in this case the insertion sequence ISEcp1. There are probably many reasons for its global spread,101 because it is found on different plasmid types and on the chromosome in pathogenic E coli strains.123 124 125 However, its success is due at least in part to an association with highly successful uropathogenic E coli (such as sequence type (ST)131) subclones that also harbor chromosomal fluoroquinolone resistance,124 126 127 128 as well as its presence on IncF-type plasmids that are particularly common in E coli.129 130 The development of a PCR based assay for rapid detection of E coli subtype ST131, for example, has added value because of the strong associations between known pathogenic strains and subtypes and resistance determinants that are tracked for the purposes of treatment and infection control.131

K pneumoniae ST258 and the KPC carbapenemase

Another example is that of K pneumoniae ST258 and the KPC carbapenemase.132 The strong epidemiological association between gene and bacterium may relate to plasmids that are relatively Klebsiella specific (for example, InFIIK-type) on which the KPC resistance gene is commonly found. It is important to note that pathogenic K pneumoniae subtypes that carry the KPC gene133 134 135 136 have often not only lost a functional OmpK35 “matrix” porin but also have a potentially important variation in the OmpK36 outer membrane “osmo” porin that is expected to augment the resistance phenotype.134 137 E coli carrying the KPC gene are much less resistant than K pneumoniae ST258 with the same gene when both are present together.134 KPC gene transmission between strains, species, and patients may therefore go undetected by phenotypic screening methods, as discussed later, although the spread of KPC probably relates primarily to its successful association with the Klebsiella strains in which the carbapenem resistance phenotype is marked.

Mobility within and between common pathogenic species such as K pneumoniae and E coli is also seen for transmissible carbapenem resistance traits linked to outbreaks in many countries.138 139 140 However, the carbapenem resistant phenotype is disproportionately over-represented in K pneumoniae compared with E coli,141 perhaps by up to 20-fold.142 Relevant factors may include the relative ease with which reduced permeability is tolerated in the Klebsiellae and the ecological connectivity to soil and other environments from which “new” resistance genes may emerge.

Ecological fixation of antibiotic resistance

An important question is whether the persistence of resistant bacteria in the environment is due to low level antibiotic contamination, non-antibiotic selection, the stability of the resistance genes and transfer elements, or a combination of these factors.143 Antibiotics in environments that are associated with humans (rivers, wastewater treatment plants, hospitals, aquaculture, farms), even at low concentrations, select for resistant organisms (and associated elements: plasmids, mobile genetic elements, genes),144 145 and a range of determinants of success are to some extent common to all such adaptations in prokaryotic and eukaryotic systems.146

Resistant organisms clearly persist asymptomatically in the microbiome without antibiotic selection, but travelers have been shown to clear these organisms over a few months after returning to a country with low resistance rates.147 148 In settings where the prevalence of resistance is generally high, continued fecal carriage is more likely,149 and some strains seem to be more likely to persist than others.150 Thus, the likelihood of resistant organisms (or plasmids) being simply replaced by antibiotic susceptible organisms (or plasmids) of the same type and thereby diluted into the local microbiota is low when the background prevalence of antibiotic resistance or of organisms adapted to human colonization is high.

Important unknowns include the extent to which bacteria and plasmids lose redundant genes or promote delivery of strongly selected genes to more secure genomic locations—for example, well adapted plasmids or the chromosome. There may be critical thresholds relating to fitness costs (see Glossary), the background reservoir of diverse non-resistant plasmids, and the content of the resistance gene pool, beyond which a return to antibiotic susceptibility is unlikely even if antibiotic usage stopped completely.151

Reduced bacterial “fitness” is often described as a disadvantage of resistance plasmid carriage, but measures of “fitness” are highly context specific and are often defined in terms of growth rates in optimal conditions in vitro. Under these conditions the carriage of resistance plasmids may impose little fitness cost on a bacterial host or a cost that is quickly ameliorated and fitness even enhanced.152 153 154 As indicated above, large, low copy, conjugative resistance plasmids stabilize themselves in bacterial populations by ensuring effective partitioning at cell division and by directly poisoning populations from which the plasmid is lost (plasmid “addiction”).155 These addiction systems are common in acquired genetic elements in bacteria and are widely recognized in conjugative plasmids.

An example of the type of addiction systems found in these large plasmids is the combination of a stable toxin and a more labile or shorter acting antitoxin, so that cell death occurs if the plasmid is lost and the antitoxin can no longer be produced.156 Even small resistance plasmids without systems to ensure partitioning of plasmids into each new bacterial cell after division (or to kill off any new bacterial cells in which this fails) seem to co-evolve with bacterial populations to ensure persistence in the absence of antibiotic selection.157 158

It may therefore be helpful to consider the entire population of a single bacterial strain type as a genetic ecosystem, within which certain genetic niches are occupied by plasmids representing mutually exclusive examples of each various type. The main ecological parameters to consider in this paradigm are:

  • Host range (the capacity to become established in different bacterial populations) of conjugative elements such as plasmids

  • The mutual associations and incompatibilities of these conjugative elements

  • The advantages for host bacteria that are associated with the acquisition and loss of these conjugative elements

  • The capacity of host bacteria to ameliorate any associated fitness costs

  • The capacity of host bacteria to exchange and recombine genetic material

In the case of the typical large conjugative and addictive antibiotic resistance plasmids, they are relatively fixed in bacterial accessory genomes. The dynamics of this system are key determinants of the epidemiology of modern transmissible antibiotic resistance in Gram negative bacteria, particularly the Enterobacteriaceae. The general concept of genetic ecology could be applied equally to other mobile genetic elements and the ecological constraints on their dissemination.

Surveillance guidelines

Despite the variation in antibiotic resistance that may be associated with host strain context, phenotypic screening for acquired resistance is often highly effective, especially in conjunction with genetic methods, when the transmissibility of the resistance trait is limited. When the transmissible resistance trait is widely disseminated among different species and the association between its presence and the particular phenotype breaks down, direct detection of the transmissible trait is needed.

Detection and sampling of transmissible carbapenem resistance

Currently the most problematic area seems to be that of detection of transmissible carbapenem resistance traits in the Enterobacteriaceae. It is important to distinguish between the detection of carbapenemase producing Enterobacteriaceae (CPE) that contain a carbapenem hydrolyzing enzyme and the carbapenem resistant Enterobacteriaceae (CRE) that have a non-susceptible MIC when considering this.

Existing guidelines for clinicians and laboratories on how best to efficiently detect and track transmissible carbapenemases (tables 4 and 5) are based on limited data and vary significantly.85 159 160 161 162 Screening of a large patient group is time consuming and costly. Targeted screening of at risk people is a compromise that detects those most likely to carry resistant organisms and who are most at risk of adverse consequences. Available guidelines emphasize the need for screening contacts in the outbreak setting but vary with regard to screening of all high risk patients (Australian) or even all admissions (European Society of Clinical Microbiology and Infectious Diseases; ESCMID).

Table 4

Guidelines for surveillance and testing: patient screening and identification

View this table:
Table 5

Guidelines for carbapenemase surveillance and testing: isolate screening and identification

View this table:

One pragmatic approach is simply to screen only for organisms that are evidently antibiotic resistant in vitro (such as KPC producing K pneumoniae, OXA-23 producing A baumannii) as would normally be done for non-transmissible resistance in other bacteria. This means that a transmissible trait will go undetected in organisms in which it produces only a low level of antibiotic resistance (such as IMP -producing K pneumoniae and KPC producing E coli), allowing a reservoir of transmissible resistance to go undetected. Such traits may then emerge unpredictably in different strains that have high MICs and may not respond to antibiotic treatment. Some authorities therefore contend that an acquired carbapenemase should always be reported, even in the context of a susceptible MIC.163 Clearly, we need more detailed genetic epidemiological data for risk-benefit calculations that best inform cost effective screening policies.

The apparently limited diversity of the shared gene pool results in a high negative predictive value for targeting genetic traits that are known to be successful, but this approach will obviously not detect novel mechanisms.106 107 Currently, major laboratory guidelines advise phenotypic investigation and confirmation of isolates that reach MIC thresholds associated with therapeutic failure.85 159

Detection of carbapenemase genes in the Enterobacteriaceae may be improved by lowering the MIC threshold of concern (for example, using a lower carbapenem concentration in growth media), by using surrogate phenotypes (for example, using cefotaxime resistance to detect MBL-type carbapenemases), or by using associated traits (such as co-transmitted aminoglycoside resistance). The ideal associated traits for screening purposes are closely genetically linked and consistently expressed. For example, high level amikacin resistance is unusual in most countries but is commonly associated with plasmid borne blaNDM.164 Such resistance should be regarded as a reason to test for blaNDM in an organism such as E coli in which the carbapenem resistant phenotype may not be very striking.

Unfortunately, associated traits are too often absent from strains of interest or too often present in other strains to be reliable. Traditional susceptibility testing to screen for organisms that are resistant or nearly resistant to a particular antibiotic may help reduce the number of candidate organisms that need to be subjected to further testing by more direct (such as genetic) methods and a combined approach is probably optimal. Testing for specific hydrolytic capacity is also a useful direct method and has the obvious advantage of detecting the phenotype caused by a novel gene, but it may be less sensitive than specific nucleic acid detection for known genes.165 166

Monitoring the mobile resistance gene pool

When considering an outbreak of antibiotic resistant bacteria it is crucial to determine the mechanism(s) of dissemination. At one extreme is a resistance mutation in Mycobacterium tuberculosis that is attributable to a single nucleotide change, which creates a new clonal variant that subsequently spreads. The transmission chain is effectively tracked by comparative genomic analyses (the highest resolution of which is to determine the entire DNA sequence of each genome) because exchange of resistance genes is relatively unimportant.167 A phenotype from a gene that tends to be linked to a specific bacterial type (such as KPC in ST258 K pneumoniae) can also be informatively studied by whole genome sequencing to obtain a transmission chain and to detect potential emerging bacterial types.168 169

At the other extreme is highly mobile plasmid borne resistance with a broad host range among the Enterobacteriaceae, in which the pattern of mobility determines the epidemiology of transmission. The adaptive capacity of bacteria is greatly enhanced by ready access to the large gene pool. For example, CMY-2 AmpC-like β lactamases constitutively (permanently) expressed from plasmids have become the dominant local source of an AmpC phenotype in E coli.106 170 171 Once accessible to human pathogens through the shared mobile gene pool, dissemination of resistance trait(s) can proceed rapidly. Sharing of genetic traits is expected to be less common between distantly related organisms (such as Acinetobacter and Escherichia spp) than between members of the Enterobacteriaceae, which share many plasmid types.172 Nevertheless, the plasmid host range (the range of bacterial types capable of receiving and supporting a given plasmid; see above) may include completely different classes of bacteria,173 and plasmids play a vital role in connecting diverse species and ecosystems and in disseminating antibiotic resistance traits widely.174

The role of the bacterial host in increasing the prevalence of antibiotic resistance in a population, especially if that host is already successful in a particular ecological space for reasons other than antibiotic resistance, was discussed above for E coli ST131 with blaCTX-M-15 and K pneumoniae ST258 with blaKPC.136 Horizontal plasmid transfer may also be enhanced when specific bacterial populations expand in the context of gut inflammation or antibiotic treatment.175 176 A complete epidemiological picture of a mobile resistance trait would ideally contain detailed information about:

  • The genetic context of the resistance gene(s) of interest (associated genes and mobile genetic elements, such as insertion sequences and transposons; whether the trait is likely to be present as a single gene or within a complex genetic unit)

  • The vehicle(s) in which the resistance trait is present (for example, plasmid type and host range or whether it is only present on the chromosome)

  • The prevalence of permissive strains (bacteria that could receive this gene or its vehicle and augment the phenotype through features such as porin variation(s)).

Microbiota and gene pools should be monitored within hospital contexts and also within the community. It is important to understand the interactions between these groups, especially dilution of resistance in the gut microbiota by incoming “healthy” microbes, and the policy implications for antimicrobial stewardship and infection control (see “Detection and sampling of transmissible carbapenem resistance” section).

The relations between specific vehicles of the resistance gene pool (especially conjugative plasmids) and the bacterial subtypes in which they are detected in antimicrobial resistance surveys (see “Antibiotic resistance in human isolates” section above) may be an important epidemiological determinant that is as yet poorly understood.

The role of antibiotic stewardship

Drug companies have turned away from unprofitable antibiotic development, even as international calls for action highlight the need for new antibiotics and for antimicrobial stewardship strategies “to preserve the integrity and effectiveness of the existing antimicrobial armamentarium.”5 177 Unfortunately, less than 25% of blood cultures are positive even in severe sepsis,178 and timely information is often unavailable. Reliable and robust point of care diagnostics for sepsis and antibiotic resistance are essential tools to develop.

Antibiotic resistance rates may fall with reduced prescribing,179 and the potential for appropriate antimicrobial stewardship to reduce antimicrobial resistance and improve individual patient outcomes is obvious.180 181 182 183 However, assumptions that “narrow spectrum” antibiotics (defined in terms of medically important bacteria) have less impact on the microbiota have not been well tested and some of these assumptions may be incorrect.

It is commonly thought that “narrower spectrum” third generation cephalosporins are preferable to “broader spectrum” carbapenems or even piperacillin-tazobactam in clinical situations in which either would be equally effective. However, data relating to the ecological impact on the microbiota suggest that third generation cephalosporins and cefepime are more likely to result in subsequent infection with P aeruginosa, ESBL producing Enterobacteriaceae, and multiply resistant S aureus and Clostridium difficile.184 185 186 It is currently unclear exactly what “appropriate” antibiotic prescribing is, and translational research is urgently needed in this area.187 Indeed, it has been argued that an aggressive antibiotic based curative policy promotes the resistance it aims to avoid.188 Our approach to antibiotic stewardship probably needs to be much more sophisticated.

The paradox of selective gut decontamination

Selective decontamination of the digestive tract uses non-absorbable antibiotics in the oropharynx and gut (selective oral decontamination) plus four days of intravenous antibiotics (usually the third generation cephalosporin, cefotaxime). This approach results in overall reduced rates of antibiotic usage, and the reported absolute overall mortality benefit of 3-6% exceeds that of other widely accepted medical interventions (such as urgent angioplasty for myocardial infarction).189 190 However, it is not widely practised outside the areas in which it is championed because of concerns about the development of antibiotic resistance in countries where the background prevalence rate of antibiotic resistance is high.191 It may be that the additive effect of the intravenous antibiotic is the least important component,192 and the use of a drug such as piperacillin-tazobactam might not have the adverse ecological effect that is expected from cefotaxime.

Emerging treatments

Management of ESBL infection has been well reviewed including the contentious issue of extended spectrum penicillin-β lactamase inhibitors (such as piperacillin-tazobactam) instead of carbapenems.193 Consequently, we will focus again on CRE as the most challenging clinical problem.

Non-carbapenem based regimens are traditionally preferred for treating infections with CRE.194 However, many carbapenemase producing isolates without an adjunctive resistance mechanism have carbapenem MICs that are only one or two dilutions above the clinical breakpoint, and are achievable in vivo. Observational studies of treatment of infections with CRE have looked at a heterogeneous mix of organisms and patients, and carbapenem containing regimens have been shown to confer a mortality benefit in the context of a MIC that is not conspicuously resistant.195 196 Extended infusion regimens may restore efficacy even when the in vitro MIC is slightly above the resistance breakpoint,197 and dual carbapenem therapy may be considered for KPC producing Enterobacteriaceae using ertapenem as an additional (sacrificial) substrate to saturate or overwhelm the capacity of the KPC enzyme to hydrolyze the principal therapeutic carbapenem.198 199 200 The benefits of dual β lactam therapy warrant review,201 whether they accrue from synergistic action at a single target site or from targeting different penicillin binding proteins (as for enterococcal therapy with ampicillin and cefotaxime or ceftriaxone).202

Newer antibiotics

New antibiotic classes for which resistance is expected to be slow or difficult to develop are welcome arrivals.203 Agents that suppress virulence characteristics,204 205 206 rather than kill the microbe, should also select less strongly for resistance and are attractive candidates for co-administration with current antimicrobials,207 208 209 210 211 although none is immediately available.


Tigecycline is a glycylcycline antibiotic in the tetracycline class that was approved in 2005 for intra-abdominal infections as well as complicated skin and skin structure infections. However, pharmacodynamic and pharmacokinetic properties make it less suitable as monotherapy for serious intra-abdominal infection and hospital acquired pneumonia.212 213 214 Resistance is not uncommon among the Enterobacteriaceae, especially in species such as Proteus spp, which may exhibit carbapenem resistance on acquisition of a carbapenemase gene: a UK study of CRE showed less than half to be susceptible to tigecycline.214


Avibactam is a novel non-β lactam β lactamase inhibitor with activity against common KPC and “classic” ESBL enzymes (Ambler class A), AmpC enzymes (Ambler class C), and OXA-48 (Ambler class D), but not MBLs (Ambler class B). Ceftazidime-avibactam has performed well in phase II trials of intra-abdominal infection (with metronidazole) and urinary tract infection,215 and the addition of avibactam to aztreonam has promise against MBLs such as NDM-1 (New Delhi metallo β lactamase 1), in which ESBL enzymes that would otherwise simply hydrolyze the aztreonam are common.216


Eravacycline, a novel fluorocycline antibiotic, withstands usual tetracycline resistance mechanisms, has broad activity against enteric Gram negative pathogens including multi-resistant isolates and anaerobes, and may be as safe and effective as ertapenem in complicated intra-abdominal infection.217 218


Plazomicin is a novel aminoglycoside that is resistant to modification by most currently described transferases and acetylases but is ineffective in the presence of the 16S rRNA methylases that are commonly acquired with NDM.219 220 Other potential novel agents and inhibitors are discussed elsewhere.221 222

Antimicrobial peptides

Antimicrobial peptides are generally short amino acids that kill bacteria by multiple mechanisms, including membrane pore formation,223 224 and they are synergistic with traditional antibiotics.225 Although the fact that these peptides have multiple targets theoretically protects against the development of resistance, they are part of the innate immunity of plants and animals and multiple resistance mechanisms have been recognized.226 Improvements in bioavailability and stability, and reduced toxicity, are needed to make them part of the deliverable antimicrobial armamentarium of the future.223

Older antibiotics

Colistimethate sodium

Colistimethate sodium (colistin), first described in 1954, interacts with lipopolysaccharide in the bacterial outer membrane. It is bactericidal in a concentration dependent manner and early establishment of adequate tissue levels by giving “loading doses” may increase efficacy and reduce nephrotoxicity (which affects 10-30% of recipients and may be related to total cumulative dose).227 228 This drug may be comparable to tigecycline and carbapenems,195 with which it is used in combination (or with aminoglycosides) as well as in monotherapy for infections caused by CRE.228 However, dosing is difficult and clinically apparent resistance increases with drug exposure.228 229 New compounds with similar bactericidal activity may be less toxic but show cross resistance with colistimethate sodium,230 231 and a plasmid borne resistance trait has recently been recognized.232


Fosfomycin acts by inhibiting peptidoglycan synthesis in the cytosol.233 The antibiotic is well tolerated and bactericidal levels are maintained in the urine for 72 hours after a single dose.234 Intravenous fosfomycin has also been used successfully in critically ill patients, although almost always in combination with another active agent.235 236 Resistance due to decreased uptake, target site, or drug modification may emerge during treatment.233

Non-antibiotic treatments

Lytic bacteriophages (phages)

Lytic bacteriophages (phages) are bacterial viruses that were discovered before penicillin 237 238 239 and developed in Soviet Russia after the second world war while antibiotics and antibiotic resistance developed in the West.240 They are often highly target specific and synergistic with antibiotics,241 and they may be valuable in antibiotic resistant and biofilm-type infections, infecting target bacteria and disappearing as they are consumed.242 Empiric treatment usually requires a phage “cocktail” to overcome resistance, but problems associated with resistance, targeting, immunogenicity, and diffusion must be resolved before they can be used routinely.241 243 244 Intrinsic bacterial defense systems such as the CRISPR (clustered regularly interspaced short palindromic repeats)-Cas9 attack complex, which ordinarily functions as an adaptive immune mechanism to combat bacteriophages and foreign DNA, can be re-engineered to specifically target resistance, virulence, or other autologous DNA sequences, including with phage delivery.245 246 247


Bacteriotherapy (fecal microbiota transplantation) to restore normal gut microflora is experiencing a resurgence of interest.248 249 250 251 252 This interest has been driven largely by the increased incidence of severe C difficile associated diarrhea and colitis,253 254 and by the recognition of opportunities for intervention in other diseases.255 Certain antibiotics such as third generation cephalosporins have long been known to be associated with colonization and infection by opportunistic and antibiotic resistant pathogens including C difficile.186 Differential colonization effects are evident within 48 hours of admission to intensive care,185 and these antibiotics may be associated with less effective spontaneous recovery of phylum level balance within gut microflora compared with those without such associations.184 A better understanding of the effect of probiotics and antibiotics on the gut microflora is therefore essential.256 Increased understanding of the role of microbial production of, and competition for, key substrates and apparent successes in manipulating these dynamics with limited bacterial combinations to ameliorate dysbiosis hint at more elegant solutions than whole fecal microbiota transplantation in the future.257 258 259

Glossary and acronyms

  • Antibiotic selection pressure: The pressure to adapt that is exerted on bacterial populations by antibiotics that threaten their survival

  • β lactams and β lactamases: β lactams are the most widely used type of antibiotic (penicillins, cephalosporins, and related compounds such as carbapenems), especially in hospitals. Recognition of increasingly widespread resistance to the early drugs was followed by chemical modification of the penicillins and the introduction of new classes such as cephalosporins (first, second, third, and later “generations”), cephamycins (often grouped with second generation cephalosporins), monobactams (such as aztreonam), and carbapenems (such as meropenem). The third generation cephalosporins (such as cefotaxime, ceftriaxone, ceftazidime) are the archetypal “extended spectrum β lactam” antibiotics. The functional Ambler classification divides β lactam hydrolyzing enzymes into class A (such as extended spectrum β lactamases), class C (AmpC enzymes), and class D (OXA) serine proteases and the Ambler class B metalloenzymes (such as NDM). By convention, β lactamase enzymes are described in upper case (for example, CTX-M-15) and their genes in appended subscript to an italicized lower case bla prefix (for example, blaCTX-M-15)

  • Biofilms: Bacterial growth style of an extracellular matrix plus organisms (often multiple species) that display multiple phenotypes and growth characteristics

  • Breakpoint/MIC/MBC: Breakpoints are used to define antibiotic susceptibility and resistance in clinical microbiology. The lowest antibiotic concentration at which bacterial growth is inhibited in optimal conditions in vitro is the minimal inhibitory concentration (MIC) and that at which no viable cells can be recovered is the minimal bactericidal concentration (MBC). If the MIC for a given drug-bacterial combination does not exceed the predefined standardized susceptibility breakpoint, the bacteria are deemed to be susceptible.

  • Conjugative plasmid: Large (>60 kbp) replicative extrachromosomal element (usually circular) that encodes the capacity to be transmitted to receptive bacterial cells

  • Constitutive promoter: A region of DNA upstream of a gene that induces expression of that gene at a constant level; this is a common feature of mobile genetic elements that have “captured” a resistance gene (see Gene capture)

  • CRE and CPE: The terms CRE (carbapenem resistant Enterobacteriaceae) and CPE (carbapenemase producing Enterobacteriaceae) are not synonymous. Production of carbapenemases may not result in clinically relevant carbapenem resistance. Similarly, β lactamases not regarded as typical carbapenemases may produce carbapenem resistant phenotypes in certain settings

  • CTX-M: Most common class of ESBL enzymes, recognized for cefotaxime (CTX) hydrolysis and first isolated in Munich

  • Derepressed mutant: A cell containing a mutation that decreases the repression of gene expression; this is important in certain Enterobacteriaceae that produce AmpC enzymes

  • ESBL (extended spectrum β lactamases): The classic ESBLs are Ambler class A enzymes capable of hydrolyzing third generation cephalosporin antibiotics. The term is sometimes more loosely applied to any enzyme that hydrolyzes third generation cephalosporins including members of all other classes of β lactamases even, arguably, those known for hydrolyzing carbapenems, the carbapenemases

  • Epidemiological cut-off (ECOFF): The upper limit of the normal distribution of MICs in a given bacterial population (see breakpoint/MIC/MBC)

  • Fitness costs: The general adverse consequences for growth and survival that may result from a mutation or adaptation to a specific circumstance (such as the acquisition, loss, or alteration of an important gene or genes)

  • Gene capture: Certain genes seem to have been excised from their original genetic locus by genetic elements that promote excision from the chromosome or by packaging in a way that results in the mobilization of that gene into a transmissible gene pool and, usually, efficient expression of the protein encoded by that gene. See also “Mobile genetic elements”

  • Homologous recombination: Nucleotide sequences are exchanged between highly similar or identical DNA molecules, sometimes introducing new unrelated DNA sequences that are flanked by these regions of highly similar DNA

  • Inocula and the inoculum effect: The inoculum is the starting amount of viable bacteria in any growth environment; the effectiveness of many antibiotics varies with the inoculum, with larger inocula having seemingly higher MICs than smaller inocula. This “inoculum effect” is the result of a high density of bacteria producing extracellular antibiotic resistance proteins (such as hydrolyzing enzymes) that concentrate locally to destroy more antibiotic than individual cells or smaller inocula could. In the laboratory testing (unlike in vivo), the inoculum for susceptibility testing is standardized, partly for this reason

  • KPC: the Klebsiella pneumoniae carbapenemase is found in particular K pneumoniae strains and is generally associated with high levels of antibiotic resistance in those strains

  • MBL/metallo β lactamases: The Ambler class B family of cation (usually zinc) dependent enzymes (such as IMP, imipenemase; NDM, New Delhi metallo β lactamase) hydrolyze most β lactam-type molecules, including cephamycins (second generation cephalosporins) and carbapenems (such as meropenem) but usually not monobactams (such as aztreonam)

  • Microbiota: The entire population of microorganisms within a given ecological system, such as the gut or the ocean

  • Microbiome: The entire microbe associated genetic content within a given ecological system, including genes and non-gene elements

  • Mobile genetic elements: Transposons and insertion sequences and similar small genetic units capable of transferring DNA such as antibiotic resistance genes

  • Plasmid host range: The diversity of bacterial types that a given plasmid can enter and stably reproduce within; this is therefore a key determinant of plasmid epidemiology

  • Plasmid replicon: The reproductive unit, typically referring to those genes specifically dedicated to plasmid replication; specific differences are used to define plasmid (replicon) types and incompatibility groups

  • Plasmid incompatibility: The mutual intolerance of similar plasmids in a given cell, predictable largely on the basis of replicon type, and commonly used to group plasmids (for example, IncF, IncI)

  • Plasmid addiction systems: Bacterial toxicity upon plasmid loss after cell division (sometimes called post-segregational killing, PSK) is commonly a result of long acting residual toxins that are more stable than plasmid provided antitoxins (sometimes called toxin-antitoxin, TA, systems). Plasmid addiction systems are widespread in conjugative plasmids and are relatively predictable for a given replication (incompatibility) type; they are important in stabilizing antibiotic resistance traits in the accessory genome of bacteria

  • Porins: Protein channels in the outer membrane that permit transport of solutes, often by passive diffusion, and constitute an important gateway to the cell from the outside


Antibiotic resistance is a natural adaptive process in bacteria that pre-dates the evolution of modern humans but that may have been accelerated by the omnipresence of antibacterial compounds in medicine, agriculture, and the environment.260 Public and clinical policy paradoxes in our approach to antibiotic stewardship and the prophylaxis of infection require a more sophisticated understanding of the effects of antibiotics on microbial populations and interactive microbial systems.

Current knowledge makes it easier to understand why carbapenem resistance in enteric bacteria still occasionally surprises us, given the complexities of co-selection, the phenotypically silent spread of mobile traits that encode it, and the variability of the bacterial factors that augment it. For now, our knowledge of the apparently restricted diversity in transmissible antibiotic resistance gene pools can be exploited for diagnostic and screening purposes. However, it highlights the possibility that mobile gene pools in major pathogens such as E coli and K pneumoniae may be being driven consistently toward loss of more diverse antibiotic susceptible elements, while the capacity to shed antibiotic resistance genes from the gene pool remains undefined. This in turn suggests the potential for an “ecological tipping point” in the mobile gene pool,151 beyond which point the usual plasmids available to be acquired by our gut bacteria, even in the absence of any specific selection, may only be antibiotic resistance plasmids of limited diversity. It emphasizes the urgent need to better understand the pools of mobile resistance genes and their inter-relationships in host bacterial populations.


Cite this as: BMJ 2015;351:h6420


  • Thanks to Mikala Wang for critical reading of the manuscript.

  • Contributors: JI, JB, and KT reviewed the literature, developed the tables and figures, and wrote the article. JI is guarantor.

  • Competing interests: We have read and understood BMJ policy on declaration of interests and declare the following interests: none.

  • Provenance and peer review: Commissioned; externally peer reviewed.

  • The BMJ did not request patient input on this article when it was commissioned.


View Abstract