Antibiotic-resistant bacteria that are difficult or impossible to treat are becoming increasingly common and are causing a global health crisis. Antibiotic resistance is encoded by several genes, many of which can transfer between bacteria. New resistance mechanisms are constantly being described, and new genes and vectors of transmission are identified on a regular basis. This article reviews recent advances in our understanding of the mechanisms by which bacteria are either intrinsically resistant or acquire resistance to antibiotics, including the prevention of access to drug targets, changes in the structure and protection of antibiotic targets and the direct modification or inactivation of antibiotics.
Antibiotic resistance is a global health emergency.
Resistance mechanisms exist for all current antibiotics, and few new drugs are in development.
Resistance can occur via a reduction in the intracellular concentration of drug, by target site alteration or protection and by the direct inactivation of antibiotics.
The mobilization of resistance genes into pathogens is making the treatment of severe infections challenging owing to a lack of effective antibiotics.
Our understanding of the molecular mechanisms of resistance has recently increased as a result of advances in systems biology, genomics and structural biology.
New knowledge about antibiotic resistance should be used to inform the design of novel therapeutic agents that might not be subject to, or can circumvent, mechanisms of resistance.
Antibiotics underpin modern medicine; their use has reduced childhood mortality and increased life expectancy, and they are crucial for invasive surgery and treatments such as chemotherapy. However, the number of infections caused by multidrug-resistant bacteria is increasing globally, and the spectre of untreatable infections is becoming a reality. The most recent World Economic Forum Global Risks reports have listed antibiotic resistance as one of the greatest threats to human health1,2,3. It is estimated that in Europe 25,000 people die each year as a result of multidrug-resistant bacterial infections and that this costs the European Union economy €1.5 billion annually1. In the United States more than 2 million people are infected with antibiotic-resistant bacteria annually, with 23,000 deaths as a direct result4,5. In addition to increased resistance to existing agents, there is a lack of new antibiotics in development. The word antibiotic has become synonymous with 'antibacterial drug': therefore, in this article the term antibiotic has been used throughout.
Bacteria can be intrinsically resistant to certain antibiotics but can also acquire resistance to antibiotics via mutations in chromosomal genes and by horizontal gene transfer. The intrinsic resistance of a bacterial species to a particular antibiotic is the ability to resist the action of that antibiotic as a result of inherent structural or functional characteristics (Fig. 1). The simplest example of intrinsic resistance in an individual species results from the absence of a susceptible target of a specific antibiotic; for example, the biocide triclosan has broad efficacy against Gram-positive bacteria and many Gram-negative bacteria, but it is unable to inhibit growth of members of the Gram-negative genus Pseudomonas. Although this was initially thought to be due to active efflux6, it has more recently been shown that it is instead due to the carriage of an insensitive allele of fabI that encodes an additional enoyl-ACP reductase enzyme — the target for triclosan in sensitive species7. A second example relates to the lipopeptide daptomycin (first approved for clinical use in 2003), which is active against Gram-positive bacteria but is not effective against Gram-negative bacteria. This is due to an intrinsic difference in the composition of the cytoplasmic membrane; Gram-negative bacteria have a lower proportion of anionic phospholipids in the cytoplasmic membrane than do Gram-positive bacteria, which reduces the efficiency of the Ca2+-mediated insertion of daptomycin into the cytoplasmic membrane that is required for its antibacterial activity8. The intrinsic resistance of some Gram-negative bacteria to many compounds is due to an inability of these agents to cross the outer membrane: for example, the glycopeptide antibiotic vancomycin inhibits peptidoglycan crosslinking by binding to target d-Ala-d-Ala peptides but is only normally effective in Gram-positive bacteria as, in Gram-negative organisms, it cannot cross the outer membrane and access these peptides in the periplasm9.
Recent studies have led to the identification of many genes that are responsible for intrinsic resistance to antibiotics of different classes, including β-lactams, fluoroquinolones and aminoglycosides. This was achieved using high-throughput screens of high-density genome mutant libraries that were created by targeted insertion or random transposon mutagenesis in bacteria such as Staphylococcus aureus, Escherichia coli and Pseudomonas aeruginosa10,11. Screening these libraries for antibiotic susceptibility has identified possible novel drug combinations in which one agent can inhibit an intrinsic resistance mechanism and so increase the spectrum of activity of other antibiotics beyond normal target species. For example, analysis of the susceptibility phenotypes resulting from inactivation of all non-essential E. coli genes identified putative targets, including thioredoxin (TrxA), thioredoxin reductase (TrxB), FabI, RecQ, SapC and the d-Ala-d-Ala carboxypeptidase DacA, which, if inhibited, can greatly promote the activity of existing drugs, including rifampicin, triclosan, nitrofurantoin, aminoglycosides and some β-lactams11. Understanding the genetic basis of intrinsic bacterial resistance, and hence the spectrum of activity of an antibiotic, can therefore guide the development of new combinations of agents with improved or expanded activity against target species. Various studies have identified in vitro synergies between unconventional combinations of antibiotics that can be used to target particularly problematic pathogens such as Acinetobacter baumannii and Neisseria gonorrhoeae12,13,14. Combining information from studies seeking synergy with those that use genetic screens to identify interactions between biochemical pathways could extend the arsenal of antibiotics and enable the use of existing drugs against species that are thought to be intrinsically resistant.
In addition to intrinsic resistance, bacteria can acquire or develop resistance to antibiotics. This can be mediated by several mechanisms, which fall into three main groups: first, those that minimize the intracellular concentrations of the antibiotic as a result of poor penetration into the bacterium or of antibiotic efflux; second, those that modify the antibiotic target by genetic mutation or post-translational modification of the target; and third, those that inactivate the antibiotic by hydrolysis or modification.
Each of these mechanisms has been reviewed over the past decade15,16,17. Therefore, this Review provides an update of the latest research for each type of antibiotic resistance mechanism and puts it into global context in terms of prevalence, the biological impact on the bacterium and the potential impact on clinical treatment.
Prevention of access to target
Reduced permeability. Compared with Gram-positive species, Gram-negative bacteria are intrinsically less permeable to many antibiotics as their outer membrane forms a permeability barrier18,19. Hydrophilic antibiotics cross the outer membrane by diffusing through outer-membrane porin proteins. In most Enterobacteriaceae, the major porins, such as the outer-membrane proteins OmpF and OmpC of E. coli, are thought to function as non-specific channels; previous evidence that suggested drug-binding sites were present within these channels now seems to be incorrect18,20. Therefore, reducing the permeability of the outer membrane and limiting antibiotic entry into the bacterial cell is achieved by the downregulation of porins or by the replacement of porins with more-selective channels. This well-established mechanism of intrinsic antibiotic resistance in Gram-negative bacteria has been reviewed previously (see, for example, Ref. 21). However, recent data have shown that in Enterobacteriaceae, Pseudomonas spp. and Acinetobacter spp., reductions in porin expression significantly contribute to resistance to newer drugs such as carbapenems and cephalosporins, to which resistance is usually mediated by enzymatic degradation21,22,23,24,25. For example, clinically relevant resistance to carbapenems in Enterobacteriaceae can occur in the absence of carbapenemase production if mutations reduce porin production or if mutant porin alleles are present22,26. The selective pressure exerted by carbapenems to favour the emergence of mutations in porin genes, as well as in genes that regulate porin expression, has been shown by the rapid accumulation of mutations in these genes in E. coli and Enterobacter spp. after carbapenem exposure23,27,28. In addition, isolates of Klebsiella pneumoniae that express porin variants have been associated with clonal lineages that have caused global outbreaks of infection24,29.
Increased efflux. Bacterial efflux pumps actively transport many antibiotics out of the cell and are major contributors to the intrinsic resistance of Gram-negative bacteria to many of the drugs that can be used to treat Gram-positive bacterial infections. When overexpressed, efflux pumps can also confer high levels of resistance to previously clinically useful antibiotics. Some efflux pumps have narrow substrate specificity (for example, the Tet pumps), but many transport a wide range of structurally dissimilar substrates and are known as multidrug resistance (MDR) efflux pumps.
There are well-studied examples of MDR efflux pumps that are present in all bacteria, and new pumps that export antibiotics continue to be described. In the past 2 years, these have included MdeA in Streptococcus mutans, FuaABC in Stenotrophomonas maltophilia, KexD in K. pneumoniae and LmrS in S. aureus30,31,32,33. Although all bacteria carry multiple genes that encode MDR efflux pumps on their chromosomes, some have been mobilized: that is, they have been mobilized onto plasmids that can transfer between bacteria. Genes coding for a novel tripartite resistance nodulation division (RND) pump were recently found to be carried on an IncH1 plasmid that was isolated from a Citrobacter freundii strain that also carried the gene for the antibiotic-targeting enzyme New Delhi metallo-β-lactamase 1 (NDM1)34. This is a worrying development as it shows that this resistance mechanism is transmissible and could be rapidly disseminated to other clinically relevant pathogens.
The RND family of MDR efflux pumps is found in Gram-negative bacteria and is the best characterized of the clinically relevant MDR efflux transporters. When overexpressed, RND pumps confer clinically relevant levels of MDR and export an extremely wide range of substrates35. Well-studied examples include the multidrug efflux pump AcrB in E. coli and MexB in P. aeruginosa. RND pumps, such as AcrB, are homotrimers that reside in the inner membrane and form a tripartite complex with a periplasmic adaptor protein, such as AcrA or MexA, and an outer-membrane channel, such as TolC or OprM. Substantial progress has been made in understanding the structure and function of these efflux pumps (reviewed in Refs 36, 37). Briefly, co-crystallization of E. coli AcrB in complex with substrates, along with computational studies, has shown that AcrB has two distinct binding pockets (distal and proximal). These pockets can accommodate substrates of different sizes and properties, which explains how the pumps can transport and provide resistance to such a broad range of antibiotics19,38,39,40,41,42. Further understanding of substrate binding has been gained from the detection of mutations occurring within the binding pockets; for example, a substitution at residue 288, within the distal binding pocket of AcrB, was found in clinical isolates of Salmonella enterica subsp. enterica serovar Typhimurium selected during antibiotic therapy. The substitution altered substrate binding and conferred antibiotic resistance, including to ciprofloxacin (J.M.A.B, and L.J.V.P, unpublished observations).
A structure of the whole tripartite system of E. coli AcrAB–TolC based on data produced with cryo-electron microscopy of an artificial assembly of mutant proteins has more recently been published, thereby addressing several outstanding questions about the pump structure and function43. Du and colleagues showed that the stoichiometry of the pump is 3/6/3 (AcrB/AcrA/TolC), which differs from previous models44 but is in agreement with structures for RND pumps, including CusBA in other Gram-negative species, and with data regarding the stoichiometry of periplasmic adapter proteins45,46,47,48.
Bacteria that overexpress efflux pumps, including Enterobacteriaceae, P. aeruginosa and S. aureus, have been isolated from patients since the 1990s49,50,51. Understanding the mechanism of overexpression is important as this is a common mechanism of resistance in clinical isolates, and thorough knowledge of the mechanism could enable the design of novel therapeutics to prevent the production of efflux-pump proteins. The transcription of genes that encode efflux pumps is controlled by local regulators, which are encoded alongside the efflux-pump genes, and by global regulators, which have broader biological functions (Fig. 2a). Global regulators include transcription factors of the AraC–XylS family, such as MtrA, which enhances transcription of mtrCDE in N. gonorrhoeae52, and RamA, which activates transcription of acrAB–tolC in Salmonella spp.53. Many of the AraC–XylS family of regulators are encoded alongside a repressor of the multiple antibiotic resistance protein (MarR) family, including marR, which is encoded with marA in E. coli, and ramR, which is encoded with ramA in S. Typhimurium54,55. Overexpression of the AraC–XylS transcription factors follows loss of repression by the corresponding MarR family protein56,57,58. The increased expression of these transcription factors, including MarA, SoxS, RamA or Rob in Enterobacteriaceae, confers MDR as a result of the increased expression of efflux pumps and repression of porin proteins53,59.
The high-level expression of efflux genes seen in multidrug-resistant bacteria is often due to mutation in the regulatory network controlling efflux-pump expression (Fig. 2b,c). These mutations can be within a local repressor, a global transcription factor or intergenic sites that alter the expression of pump genes or their regulators60,61,62,63,64,65,66. Mutations can alter promoter activity; a recent example is the detection of a single-base-pair mutation in the consensus −10 sequence upstream of mtrC in N. gonorrhoeae that forms a new promoter with stronger activity, causing constitutive overexpression of the efflux pump and multidrug resistance54.
Increased expression of efflux pumps can also occur as a result of induction in response to environmental signals and in conditions in which their function is required. For example, the acrAB genes in E. coli and Salmonella spp. are induced by small molecules that would be encountered during infection, such as indole and bile57,67,68,69, whereas expression of MtrCDE of N. gonorrhoeae and NorA of S. aureus is responsive to iron limitation, which is a common hallmark of the host environment70,71. The most common mechanism of induction of efflux pump gene expression is the direct binding of a molecule to a transcriptional repressor protein, which decreases binding of the repressor to its target DNA (Fig. 2d). The structures of several regulatory repressor proteins have recently been solved, improving our understanding of this mechanism of resistance. For example, the structure of the Mycobacterium tuberculosis TetR family transcriptional repressor Rv1219c has been recently solved; this protein controls expression of the ABC family transporter RV1217c–Rv1218c, which is responsible for the export of isoniazid and rifampicin72, and the structure revealed a large multidrug-binding pocket73. Similarly, the structure of RamR, the transcriptional repressor of the ramA gene, has been solved in complex with several ligands74. Inducing compounds reduced the binding affinity of RamR to DNA, increasing ramA expression. Understanding the molecular basis of induction of efflux could allow the development of chemical modulators to prevent efflux de-repression, and these modulators could be used in conjunction with antibiotics in novel therapeutic strategies.
Changes in antibiotic targets by mutation
Most antibiotics specifically bind to their targets with high affinity, thus preventing the normal activity of the target. Changes to the target structure that prevent efficient antibiotic binding, but that still enable the target to carry out its normal function, can confer resistance (Fig. 3). During the course of infection there are often large and diverse populations of pathogens, and if a single point mutation in the gene encoding an antibiotic target can confer resistance to the antibiotic, strains with this mutation can then proliferate. The genes that encode the targets of some antibiotics exist in multiple copies; for example, linezolid (the first oxazolidinone antibiotic to be developed, which entered clinical use just over 10 years ago) targets the 23S rRNA ribosomal subunit of Gram-positive bacteria, which is encoded by multiple, identical copies of its gene. Clinical use of linezolid has selected for resistance in S. pneumoniae and S. aureus by mutation in one of these copies, followed by recombination at high frequency between homologous alleles, which rapidly produces a population weighted in favour of carriage of the mutant allele75,76,77.
Transformation — that is, uptake of DNA from the environment — can confer antibiotic resistance by target protein modification through the formation of 'mosaic' genes. The archetypal example of this is penicillin resistance in S. pneumoniae, which is conferred by mosaic penicillin-binding protein (pbp) genes encoding penicillin-insensitive enzymes. These mosaic alleles have arisen by recombination with DNA from the closely related species Streptococcus mitis. Mosaicism in the penA gene (which encodes a PBP) in N. gonorrhoeae has also been linked with high-level resistance to extended-spectrum cephalosporins. Infections with N. gonorrhoeae are currently one of the biggest challenges in clinical microbiology owing to the emergence of pan-resistant isolates and the resulting loss of the utility of ceftriaxone as a therapeutic option for these isolates78.
Another example of a target change is acquisition of a gene homologous to the original target, such as in methicillin-resistant S. aureus (MRSA), in which methicillin resistance is conferred by acquisition of the staphylococcal cassette chromosome mec (SCCmec) element. This carries the mecA gene, which encodes the β-lactam-insensitive protein PBP2a; this protein enables cell wall biosynthesis to occur despite the native PBP being inhibited in the presence of antibiotic79. Many SCCmec elements have been identified in different Staphylococcus species, and there is evidence that the mecA allele has been mobilized several times80. The presence of a mecA-containing SCCmec element was thought to be the defining characteristic of MRSA. However, a divergent mecA allele, named mecC, has recently been identified in isolates from livestock and humans in the United Kingdom, Denmark and Ireland80,81,82. The mecC gene shares 70% nucleotide identity to mecA and is carried on a type XI SCCmec element. The presence of mecC MRSA presents a potential problem for diagnosis as it is not detected by standard PCR assays for mecA or by agglutination assays for PBP2a83. The antibiotic-resistance phenotype of mecC MRSA also differs slightly from mecA MRSA: isolates carrying mecC are more susceptible than mecA isolates to oxacillin but retain resistance to cefoxitin32,84,85.
Modification (and protection) of targets
Protection by modification of the target can also be an effective means of antibiotic resistance that does not require a mutational change in the genes encoding the target molecules (Fig. 3). In recent years, protection of targets has been found to be a clinically relevant mechanism of resistance for several important antibiotics; for example, the erythromycin ribosome methylase (erm) family of genes methylate 16S rRNA and alter the drug-binding site, thus preventing the binding of macrolides, lincosamines and streptogramins73. Another recently identified example is the chloramphenicol–florfenicol resistance (cfr) methyltransferase, which specifically methylates A2503 in the 23S rRNA; this confers resistance to a wide range of drugs that have targets near this site, including phenicols, pleuromutilins, streptogramins, lincosamides and oxazolidonones (including linezolid)86. Initially isolated in 1997 from a bovine staphylococcal isolate, the cfr gene was subsequently isolated from animal and human isolates of Gram-positive and Gram-negative bacteria, including S. aureus and E. coli87. The erm and cfr genes are both often carried on plasmids, which function as vectors to drive their wide dissemination77,88. The aminoglycoside antibiotics are protein synthesis inhibitors that function by binding to the ribosome. One mechanism of resistance to aminoglycosides is modification of the target ribosome by methylation. This was not previously thought to be a clinically relevant mechanism of resistance, but the enzymes that are responsible have recently been detected in several bacterial pathogens; for example, the armA gene, which encodes a methyltransferase, has been found in clinical isolates of Enterobacteriacae throughout North America, Europe and India, and rmt genes, which encode another methyltransferase, have been found in North America, Central and South America, and India89,90.
The qnr families of quinolone resistance genes have also been found on plasmids in various pathogens. The qnr genes encode pentapeptide repeat proteins (PRPs), which bind to and protect topoisomerase IV and DNA gyrase from the lethal action of quinolones. A recent model of the mechanism of action of PRPs is based on structural data of QnrB1, which suggests that PRPs interact with topoisomerase–quinolone complexes after drug binding and promote release of the quinolone91. This rescues the topoisomerase and enables it to complete its normal activity and re-ligate DNA, thus preventing the release of double-stranded DNA breaks, which normally result from quinolone–topoisomerase interaction91.
The polymyxin antibiotics, polymyxin B and polymyxin E (also known as colistin), are cyclic antimicrobial peptides with long, hydrophobic tails that target Gram-negative bacteria92,93. This specificity is based on their binding to lipopolysaccharide (LPS); antibacterial activity is conferred by the hydrophobic chain, which disrupts both cell membranes72,73. Owing to a lack of alternative therapies in recent years, colistin has become widely used in the treatment of infections by multidrug-resistant P. aeruginosa, Acinetobacter spp. and Enterobacteriaceae and, as a consequence, polymyxin resistance has developed. This resistance is often associated with changes in the expression of regulators affecting LPS production, which result in alterations in the target and reduce binding of the drug. For example, mutations that result in overexpression of pmrC, the product of which modifies lipid A, have been detected in the genes encoding the two-component regulatory system PmrAB92,94,95,96. These mutations cause the addition of phosphoethanolamine to lipid A and, as a result, colistin binding is reduced through lowering of the negative charge of the LPS95. Mutations in the genes encoding the PhoPQ two-component system, or its regulators, can also confer colistin resistance via increased expression of the PmrAB system97,98. This is a common resistance mechanism in K. pneumoniae99.
Daptomycin specifically targets anionic phospholipids in the cytoplasmic membrane of Gram-positive bacteria and, in the presence of calcium ions, it inserts into the membrane, causing depolarization and the loss of intracellular contents8,10. In S. aureus, daptomycin resistance can occur as a result of point mutations in mprF (which encodes the multiple peptide resistance factor, a protein that decorates anionic phospholipid phosphatidylglycerol with l-Lys), resulting in remodelling of the phospholipid content of the membrane; this in turn alters the membrane charge and phospholipid composition, thus reducing binding of daptomycin100. One important use of daptomycin has been to treat serious infections caused by enterococci, particularly Enterococcus faecium. However, resistance can emerge during therapy; although the mechanisms by which enterococci can resist daptomycin are not fully understood, recent evidence has identified several genes and pathways that correlate strongly with daptomycin susceptibility101,102. Among these, mutations within the genes encoding the LiaFSR regulatory system, which coordinates membrane stress response, and the YycFGHIJ system, which controls cell wall homeostasis, have been found in multiple isolates that show daptomycin non-susceptibility. Some isolates contain mutations affecting both systems, which shows that they are not mutually exclusive as resistance mechanisms103. Disruption of the LiaFSR system alters phospholipid content; this results in the redistribution of cardiolipin-rich domains in the membrane, which are the primary targets for daptomycin. The removal of primary binding sites for daptomycin from the septum, which is the primary cellular target in enterococci, is thought to confer daptomycin resistance; this is a novel example of how spatial reorganization of the target can confer resistance104.
Direct modification of antibiotics
As well as preventing antibiotics from entering the cell or altering their targets, bacteria can destroy or modify antibiotics, thus resisting their action.
Inactivation of antibiotics by hydrolysis. The enzyme-catalysed modification of antibiotics is a major mechanism of antibiotic resistance that has been relevant since the first use of antibiotics, with the discovery of penicillinase (a β-lactamase), in 1940 (Ref. 105) (Fig. 4). Thousands of enzymes have since been identified that can degrade and modify antibiotics of different classes, including β-lactams, aminoglycosides, phenicols and macrolides. There are also subclasses of enzymes that can degrade different antibiotics within the same class; for example the β-lactam antibiotics, such as penicillins, cephalosporins, clavams, carbapenems and monobactams, are hydrolysed by a diverse range of β-lactamases106,107,108,109. The expansion of antibiotic classes to include derivatives that have improved properties has been reflected by the emergence of hydrolytic enzymes that have altered spectra of activity. The early β-lactamases, which were active against first-generation β-lactams, were followed by extended-spectrum β-lactamases (ESBLs) that have activity against oxyimino-cephalosporins110. The carriage of diverse ESBLs and carbapenemases, including the IMP (imipenemase), VIM (Verona integron encoded metallo β-lactamase), K. pneumoniae carbapenemase (KPC), OXA (oxacillinase) and NDM enzymes in Gram-negative bacteria such as K. pneumoniae, E. coli, P. aeruginosa and A. baumannii, has underpinned the emergence of isolates that are resistant to all β-lactam antibiotics; this has serious implications for the treatment of severe infections, particularly in hospital patients108,110,111. In some cases, the expansion of resistant bacterial clones has been responsible for increased rates of resistance, whereas in others, plasmids that carry resistance genes have spread widely. Irrespective of the mechanism, the common feature has been the rapid spread and expansion of resistance in a relatively short period of time.
There are hundreds of variants of the CTX-M genes (which encode ESBLs that are notable for their greater activity against cefotaxime than other oxyimino-β-lactams). These genes are classified into different subgroups, which have resulted from the escape of genes from the chromosome of Kluyvera species found in soil112. This escape has been facilitated by insertion sequences, notably ISEcp1, followed by transfer on conjugative plasmids into many bacterial species112. The CTX-M14 and CTX-M15 enzymes have become the most widely isolated ESBLs worldwide113,114, especially in cephalosporin-resistant E. coli and K. pneumoniae isolates. CTX-M15-producing K. pneumoniae isolates are mainly a nosocomial problem, whereas CTX-M15-producing E. coli strains are disseminated in patients in the community115. Plasmids that belong to the IncFII incompatibility group, which is found widely in Gram-negative pathogens, have been particularly associated with the spread of blaCTX-M15, in some cases with particular clonal lineages (for example, the ST131-H30 clone of E. coli)115.The blaCTX-M14 gene is often associated with the IncK plasmid pCT in human, animal and environmental isolates of Enterobacteriaceae116,117,118.
As a result of the increased numbers of bacteria carrying ESBL genes, the clinical use of carbapenem antibiotics has increased over the past decade. In turn, this has been associated with increasing numbers of clinical isolates carrying β-lactamases with carbapenem-hydrolysing activity (known as carbapenemases)119,120,121. Carbapenemases mainly include members of the class A, B and D β-lactamases119,120,121. The hallmark of these enzymes is the ability to inactivate a broad range of β-lactams, including carbapenems and extended-spectrum cephalosporins. Although first identified on the chromosomes of single species, many carbapenemases are now plasmid-mediated and have been reported in Enterobacteriaceae, P. aeruginosa and A. baumannii121.
The spread of carbapenemases has occurred in different ways, as exemplified by the kpc and ndm genes. The serine carbapenemase KPC was first identified in K. pneumoniae in 1996 (Ref. 122) and has since been described in several Enterobacteriaceae123. The kpc gene is plasmid-borne and is associated with a dominant clone of KPC-producing K. pneumoniae, ST258, which is found worldwide124. This gene is often carried on the pKP-Qil plasmid or on closely related variants125, and there are several variants of the kpc gene that encode proteins that can be differentiated by single-amino-acid substitutions (although most retain similar activity); KPC2- and KPC3-producing strains have been responsible for outbreaks in the United States, Greece, Israel and the United Kingdom126,127. Since its first description in India in 2009 (Ref. 128), the NDM carbapenemase has become one of the most widespread carbapenemases and is found in Gram-negative pathogens including A. baumannii, K. pneumoniae and E. coli throughout the world. NDM confers resistance to all β-lactams except aztreonam. The ndm genes are often found on broad-host-range conjugative plasmids belonging to several incompatibility or replicon types, including IncA, IncC, IncF, IncHI1 and IncL–IncM129,130,131, and in conjunction with other antibiotic-resistance genes132. Interestingly, the spread of NDM has been underpinned by the extreme mobility of the blaNDM gene itself, which is mediated by an ISAba125 element upstream of the ndm gene rather than by expansion of an epidemic clone of bacteria or plasmid, as seen with the kpc genes107. NDM-producing isolates have been found in many species: the genes are located both on plasmids and on the host chromosome and able to move between the two at high frequency107,133. Initially, infections with NDM-carrying bacteria were epidemiologically linked to the Indian subcontinent, Balkan states and Middle East107. However, recent reports have identified outbreaks that have occurred and been sustained in other parts of the world106,134.
Inactivation of antibiotic by transfer of a chemical group. The addition of chemical groups to vulnerable sites on the antibiotic molecule by bacterial enzymes causes antibiotic resistance by preventing the antibiotic from binding to its target protein as a result of steric hindrance (Fig. 4). Various different chemical groups can be transferred, including acyl, phosphate, nucleotidyl and ribitoyl groups, and the enzymes that are responsible form a large and diverse family of antibiotic-resistance enzymes135.
Aminoglycoside antibiotics are particularly susceptible to modification as they tend to be large molecules with many exposed hydroxyl and amide groups. Aminoglycoside-modifying enzymes confer high levels of resistance to the antibiotic (or antibiotics) that they modify. There are three main classes of aminoglycoside-modifying enzymes: acetyltransferases, phosphotransferases and nucleotidyltransferases. These classes are evolutionarily diverse and vary in the aminoglycosides that they can modify and in the part of the molecule that is modified136. However, recent molecular dynamic studies have suggested that all three are able to bind to aminoglycosides as their active sites mimic the target environment of the ribosomal binding cleft137. The substrate specificity is due to protein flexibility and the size of the active site, in combination with the modulation of the thermodynamic properties of the enzyme when in complex with bound substrate136. A worrying recent development is the discovery of a novel genomic island in Campylobacter coli isolated from broiler chickens in China: this genomic island encodes six aminoglycoside-modifying enzymes, including members of all three classes, and confers resistance to several aminoglycoside antibiotics that are used in the treatment of Campylobacter infections, including gentamicin138.
Another recent discovery is the widespread presence of a family of rifamycin-resistance genes in Actinomycetes and pathogenic bacteria. These genes were discovered by identifying a common sequence element (rif-associated-element (RAE)) that is present in various genera of Actinomycetes upstream of rifamycin-resistance genes. A bioinformatic analysis of sequences deposited in GenBank identified the presence of the RAE in a wide range of species and also identified a novel rifamycin phosphotransferase that is associated with the RAE. This use of phylogenetics and functional assays has identified a previously unknown reservoir of rifamycin-resistance genes139.
Summary and conclusions
As a result of the widespread use of antibiotics in human medicine — as well as in animal treatment, horticulture, beekeeping, anti-fouling paints (used in the marine and oil industries) and laboratories carrying out genetic manipulation — the evolutionary pressure for the emergence of antibiotic resistance is great. Antibiotic resistance is a natural phenomenon, and bacteria have been evolving to resist the action of natural antibacterial products for billions of years (Box 1). Although the ability of bacteria to become resistant to antibiotics has long been appreciated, our knowledge of the remarkable diversity of mechanisms involved has increased greatly in recent years. Advances in genomics, systems biology and structural biology have dissected many of the precise events underpinning resistance and will continue to provide greater understanding. This information, if used properly, should aid the discovery and development of new agents that can circumvent or neutralize existing resistance mechanisms. Indeed, increased understanding of resistance has also provided new targets for discovery; for example, inhibitors of MarA, which regulates the expression of genes that encode for MDR efflux pumps, have been identified and have been shown to limit the severity of E. coli infection in a mouse urinary tract infection model140.
Studies of resistance development and mechanisms of resistance must be a mandatory requirement at an early stage of drug development; such studies will enable academic institutions and industry to work together. We now have the ability to rapidly evaluate the potential for the emergence of resistance to novel drugs, identify where and when this might occur and determine the mechanisms responsible. The early identification of naturally occurring resistance mechanisms and targets that can accommodate numerous structural changes should lead to the discontinuation of the development of agents that are likely to fail in the clinic as a result of resistance. Resources could instead be focused on agents that are less likely to drop out of the pipeline for microbiological reasons. Knowledge about how and when resistance occurs and potential synergies with combinations of agents will also facilitate the development of dosing regimens that can help to minimize the emergence of resistance to current and new antibiotics, enabling these drugs to be used to best effect. This is crucial in the short term as new agents are not likely to enter widespread clinical practice in the immediate future. The challenge for the field is now to make best use of the available technologies, information and expertise to ensure the impact of resistance is fully accounted for in the urgent development of next-generation antibacterial drugs.
The authors would like to acknowledge the Medical Research Council (MRC), Biotechnology and Biological Sciences Research Council (BBSRC), Royal Society, Department for Environment, Food and Rural Affairs (DEFRA) and the National Institute for Health Research (NIHR) for funding their research.
- Enoyl-ACP reductase
An enzyme key in the production of fatty acids which is the target for triclosan.
A natural or semi-synthetic fatty acid-linked peptide chain that targets the cell membrane (for example, daptomycin).
A natural or semi-synthetic amino sugar-linked peptide chain that targets terminal D-Ala-D-Ala dipeptides (for example, vancomycin).
An important class of antibiotics, members of which contain a β-lactam ring and inhibit peptidoglycan synthesis by covalent binding to the active-site Ser of penicillin-binding proteins. β-lactam subclasses include carbapenems, cephalosporins, penicillins, monobactams and clavams.
Synthetic compounds that target topoisomerases. Examples include nalidixic acid and ciprofloxacin.
Natural or semi-synthetic amino sugars that target translation by binding to the 30S subunit of the ribosome. Examples include gentamicin, tobramycin, streptomycin and kanamycin.
- Extended-spectrum β-lactamases
(ESBLs). β-lactamase enzymes that are able to hydrolyse extended-spectrum oxyimino cephalosporins.