Evolution and spread of antibiotic resistance

Transcription

1 Journal of Internal Medicine 2002; 252: REVIEW Evolution and spread of antibiotic resistance B. HENRIQUES NORMARK & S. NORMARK From the Swedish Institute of Infectious Disease Control and the Microbiology and Tumor Biology Center, Karolinska Institutet, Stockholm, Sweden Abstract. Normark BH, Normark S (Swedish Institute for Infectious Disease Control and Karolinska Institutet, Stockholm, Sweden). Evolution and spread of antibiotic resistance (Review Article). J Intern Med 2002; 252: Antibiotic resistance is a clinical and socioeconomical problem that is here to stay. Resistance can be natural or acquired. Some bacterial species, such as Pseudomonas aeruginosa, show a high intrinsic resistance to a number of antibiotics whereas others are normally highly antibiotic susceptible such as group A streptococci. Acquired resistance evolve via genetic alterations in the microbes own genome or by horizontal transfer of resistance genes located on various types of mobile DNA elements. Mutation frequencies to resistance can vary dramatically depending on the mechanism of resistance and whether or not the organism exhibits a mutator phenotype. Resistance usually has a biological cost for the microorganism, but compensatory mutations accumulate rapidly that abolish this fitness cost, explaining why many types of resistances may never disappear in a bacterial population. Resistance frequently occurs stepwise making it important to identify organisms with low level resistance that otherwise may constitute the genetic platform for development of higher resistance levels. Selfreplicating plasmids, prophages, transposons, integrons and resistance islands all represent DNA elements that frequently carry resistance genes into sensitive organisms. These elements add DNA to the microbe and utilize site-specific recombinases/integrases for their integration into the genome. However, resistance may also be created by homologous recombination events creating mosaic genes where each piece of the gene may come from a different microbe. The selection with antibiotics have informed us much about the various genetic mechanisms that are responsible for microbial evolution. Keywords: antibiotic resistance, compensation, cost, evolution, mechanisms, reversion. Introduction The antibiotics we have at our disposal make use of the major differences that exist between prokaryotic and eukaryotic cells. Thus, vancomycin and b-lactam antibiotics interfere at different levels with the synthesis of the bacterial cell wall (peptidoglycan, murein). Quinolones inhibit bacterial enzymes involved in the replication of DNA. Rifampicin inhibits transcription by binding to bacterial RNApolymerase. Macrolides, tetracyclines and aminoglycosides inhibit 70S-ribosomal function thereby inhibiting protein synthesis. Sulphonamide and trimetoprim inhibit biosynthesis of tetrahydrofolic acid and also interfere with DNA-synthesis. The above categories of antibiotics have been in existence for a long time. The oxazolidinones represent the first truly new class of antibacterial Ó 2002 Blackwell Science Ltd 91

2 92 B. HENRIQUES NORMARK & S. NORMARK agents to reach the marketplace in several decades. They have a unique mechanism of action involving inhibition of the initiation step of protein synthesis and are not cross-resistant to other classes of antibiotics. This drug is currently reserved for the treatment of antibiotic-resistant Gram-positive infections [1]. Many of the existing antibiotics target genes that are essential for the microbe. With the developments in genetics and functional genomics clever screens can now be made for genes that are essential for a microbe during different growth conditions. Of particular interest are those genes that are essential for bacterial growth within a host but not in vitro, as they probably encode potential drug targets that have never been possible to screen for before [2 4]. In recent years a wealth of data have been accumulated about the mechanisms by which different bacteria are able to cause disease [5, 6]. Development of drugs that interfere with bacterial virulence and growth in vivo rather than bacterial growth in vitro has for a long time been one of the goals of researchers working on bacterial pathogenesis. Interference with bacterial secretion of virulence proteins or with bacterial attachment and colonization are two potential targets for novel drugs. By genomic comparisons it is also possible to identify potential targets present in all bacteria in contrast to genes present only in a subset of species. The absence of the target in the human genome is at least an indication that the inhibitor will not interfere with essential human functions [7]. Thus, the tools are there to develop novel antibiotics that might be more selective than existing antibiotics. For example, narrow spectrum drugs specific against Helicobacter pylori or Chlamydia pneumoniae are probably essential to avoid unnecessary selection for resistance in other organisms if large human populations are to be treated to prevent stomach cancer or myocardial infarction. There are, however, no examples of antibacterial agents against which bacteria have not been able to develop resistance, with the possible exception of certain antibacterial peptides that are naturally made by the innate immune system [8]. Therefore, although there is hope for novel classes of antibiotics to be developed, bacteria will certainly evolve resistance against them. It therefore makes sense to do everything possible to prolong the life span of existing antibiotics. By knowing how resistance evolves and spreads in a population steps can be introduced to prevent or at least delay the spread. In many instances bacterial resistance of a given type has already occurred within a given species. In such cases we can only attempt to prevent further selection and spread in the human population and the transmission to other bacterial species. In a few but important cases resistance has not yet evolved in a particular bacterial species, i.e. penicillin resistance in group A streptococci. One genetic event anywhere in the world giving rise to a penicillin-resistant group A streptococcus will have serious consequences in the treatment of many infections caused by this organism. One should be aware of that we only have to go years back to a situation where there was no penicillin resistance amongst pneumococci, and pathogenic Neisseria worldwide. Antibiotic resistance in bacteria is a result of classical Darwinian selection. In this review we will focus on the underlying genetic mechanisms for antibiotic resistance and the parameters involved in the spread of resistance. Physiological resistance to antibiotics The biofilm example Resistance can be physiological meaning that resistance is only expressed during certain growth conditions. Perhaps the most discussed type of physiological resistance is that seen in bacterial biofilms. Opportunistic pathogenic bacteria such as Pseudomonas aeruginosa can adopt a sessile biofilm lifestyle, seen for example in lung infections in individuals with cystic fibrosis. Bacteria growing in biofilms are notoriously difficult to eradicate with antibiotic treatment. The actual mechanism behind this physiological resistance is far from clear. It has been argued that antibiotics may have difficulties in penetrating the organized matrix that surround bacteria in biofilms. Also, it may be that certain genetic systems are activated during biofilm mode of growth, inducing physiological resistance, that is not activated during free-living (planctonic) mode of growth [9]. It was however, recently argued that bacteria growing in biofilms or in a free living planctonic form show the same degree of antibiotic susceptibility [10]. One reason why bacteria in biofilms are difficult to treat could be that these organisms are in a stationary phase characterized by a balanced state of growth and death. It has been

3 REVIEW: EVOLUTION AND SPREAD OF ANTIBIOTIC RESISTANCE 93 known for years that bactericidal antibiotics such as b-lactams given to a bacterial culture usually do not sterilize the culture. There are always a low number of viable bacteria remaining. These are called persisters. In the stationary phase and during biofilm mode of growth, such persisters might occur in larger numbers. The actual mechanism by which these persisters are not killed by the drug is not known, but suggest that many antibiotics are not directly killing the microbe but triggering a programme within the organism itself that leads to death. This problem will be discussed later in this review when we discuss antibiotic tolerance. The location of the infection The location of the microbe during infection may, in some instances, prevent the drug from reaching appropriate concentrations where the microbe is growing. It has, for example, recently been shown in a murine model that Escherichia coli expressing so called Type I fimbriae, may be internalized by uroepithelial cells during a bladder infection [11]. It is therefore possible that recurrent UTI in females by E. coli, despite antibiotic treatment, could be the result of an outgrowth from a small number of intracellularly located organisms surviving the treatment. Intrinsic antibiotic resistance Intrinsic resistance means that each member of an entire bacterial species is resistant without any additional genetic alteration. For example, mycoplasma is always resistant to b-lactam antibiotics as they lack peptidoglycan as a cell wall component. Likewise, many enteric bacterial species including P. aeruginosa, exhibit a very low susceptibility to hydrophobic antibiotics like macrolides, because hydrophobic antibiotics have difficulties penetrating the outer membrane of these organisms. Even hydrophilic antibiotics can be partially restricted from penetrating the outer membrane as they have to utilize water-filled protein channels, porins, in the outer membrane [12, 13]. Pseudomonas aeruginosa exemplifies an organism with a high intrinsic resistance Pseudomonas aeruginosa is particular. It has been demonstrated that the high intrinsic resistance in this organism arises from the combination of unusually restricted outer-membrane permeability and other natural resistance mechanisms such as energy-dependent multidrug efflux pumps and chromosomally encoded b-lactamase [14]. Pseudomonas aeruginosa produces a number of broadly specific multidrug efflux systems, including MexAB- OprM and MexXY-OprM [15]. In addition, these and two additional tripartite efflux systems, MexCD-OprJ and MexEF-OprN, promote acquired multidrug resistance as a result of mutational hyperexpression of the efflux genes [16, 17]. In addition to antibiotics, these pumps promote export of numerous dyes, detergents, inhibitors, disinfectants and organic solvents. The efflux pump proteins are highly homologous and consist of a cytoplasmic membrane-associated transporter, an outer membrane channel-forming protein and a periplasmic membrane linker protein (Fig. 1) [18]. Homologues of these systems are present in Stenotrophomonas maltophilia, Burkholderia cepacia, Burkholderia pseudomallei where they also play a role in export of and resistance to multiple antimicrobial agents and/or organic solvents [19]. Efflux pumps are present in most bacteria although many do not provide intrinsic resistance and only reveal a resistance phenotype when overexpressed by mutations. Although the natural function of multidrug efflux systems is largely unknown, their contribution to intrinsic antibiotic resistance and acquired resis- Non-functioning state Channel Transporter Efflux pump Outer membrane Linker Inner membrane Functioning state Channel Transporter Antibiotic Outer membrane Linker Inner mem brane Fig. 1 The figure attempts to illustrate the structure of a tripartite efflux pump in Gram-negative bacteria. The three components involve a transporter in the inner membrane, a channel protein in the outer membrane and a periplasmic linker protein. When the pump is in a functional state the linker has been suggested to fold back on itself allowing the transporter to be brought in close association with the channel protein allowing for drug efflux across both membranes from the cytosol directly into the medium. Modified from [18]. The drug specificity for different efflux pumps may differ. Resistance is frequently caused by an increased synthesis of proteins constituting the pump. This often occurs by mutations abolishing a transcriptional repressor.

4 94 B. HENRIQUES NORMARK & S. NORMARK tance as well as their conservation in a number of important human pathogens make them logical targets for therapeutic intervention. In organisms with a high level of natural resistance, such as P. aeruginosa, acquired resistance to most classes of antibiotics can readily arise by mutations causing further increase in resistance. Thus, mutations can occur in Pseudomonas leading to an overproduction of the chromosomal AmpC b-lactamase leading to an increased resistance to virtually all b-lactam antibiotics, except the carbapenems [20]. Likewise, efflux pumps can be overexpressed by mutations. In P. aeruginosa the mexr and nfxb genes act as repressors for the mexab-oprm and mexcd-oprj efflux pumps, respectively [21, 22]. Mutations in these repressor genes result in the overexpression of efflux pump proteins and increased resistance to those antibiotics that act as substrates for the respective pump. Pseudomonas aeruginosa possesses a number of porin proteins in the outer membrane. OprD is a porin involved in the transport of basic amino acids. This porin also provides the main entry site for carbapenem antibiotics like imipenem and meropenem and consequently mutations resulting in a decreased production of OprD mediates carbapenem resistance in Pseudomonas [23, 24]. Acquired antibiotic resistance, a multitude of mechanisms Acquired resistance to antibiotics occurs either by mutations (point mutations, deletions, inversions, insertions, etc. within the bacterial genome) or by horizontal gene transfer. For each class of antibiotics there are usually a number of mechanisms that can cause resistance. These mechanisms may also differ depending on the bacterial species and its genetic make-up. The main mechanisms of resistance include (i) decreased uptake of the drug (i.e. mutated porins), (ii) increased export (i.e. up-regulated efflux pumps), (iii) inactivation or modification of the drug target (mutations in ribosomal proteins, penicillin binding proteins, methylation or mutation of ribosomal RNA, etc.), (iv) introduction of a new drug resistant target (i.e. horizontal acquisition of meca in methicillin resistance) or other by-pass mechanisms (i.e. d-ala-d-lactate synthetase in enterococcal vancomycin resistance or a novel dihydrofolate reductase in trimetoprim resistance), (v) hydrolysis of the antibiotic (b-lactamase), (vi) modification of the antibiotic (i.e. aminoglycoside modifying enzymes). Resistance through mutations occurs at different frequencies depending on the mechanisms of resistance The chromosomal AmpC b-lactamase example The mutation frequency leading to resistance can differ markedly depending on the mechanism leading to resistance. Thus, if an inactivation of a gene can lead to resistance, the frequency of resistance development is much higher than if the gene has to be specifically altered to code for a resistant protein with its normal activity maintained. Chromosomally mediated b-lactam resistance in enteric opportunists and in P. aeruginosa provides a good example of markedly differing mutation frequencies between related bacterial species. In organisms such as Enterobacter cloacae and Citrobacter freundii chromosomal resistance to b-lactams occur at a high mutation frequency, whilst the frequency is much lower for an organism such as E. coli. These three organisms encode the same type of chromosomal AmpC b-lactamase. In E. coli the ampc gene is produced constitutively at low levels. Resistant mutants overproducing the enzyme occur in vitro at the low frequency of about 10 )9. These resistant mutants carry specific point mutations or insertions that increase the transcription of the ampc gene [25 27]. They can also be caused by attenuator mutations allowing an increased transcriptional readthrough into ampc or by ampc gene amplifications (Fig. 2) [28, 29]. In E. cloacae and C. freundii the chromosomal AmpC b-lactamase is also expressed at low levels but is inducible by b-lactams. Mutants producing high constitutive levels of AmpC b-lactamase in these inducible organisms occur at a very high frequency, 10 )6 )10 )7. These mutants, in most cases, contain loss of function mutations in the ampd gene [30, 31]. Although E. coli also carry an ampd locus, inactivation has no consequence for its own ampc-transcription [30, 32]. Unlike E. coli, a regulator, AmpR, controls the expression of the ampc gene in Gramnegative organisms producing an inducible AmpC b-lactamase (Fig. 3) [33, 34]. During normal growth, in the absence of b-lactams, the AmpR regulator acts as a repressor preventing ampc tran-

5 REVIEW: EVOLUTION AND SPREAD OF ANTIBIOTIC RESISTANCE 95 Genetic events leading to Amp C β-lactamase overproduction in Escherichia coli 1) Up-promotor mutations 2) Attenuator mutations P A ampc gene 3) IS-elements creating a strong hybrid promotor 4) Gene amplification Fig. 2 Illustration of the various mutational events that may lead to b-lactamase overproduction in organisms such as Escherichia coli that normally produces low constitutive levels of chromosomal AmpC-b-lactamase. (i) Up-promotor mutations are quite rare, but can increase the ability for RNA-polymerase to initiate transcription. The first such mutation to be discovered, ampa, contained a 1-bp insertion between the conserved )10 and )35 regions of the promoter. (ii) There is a weak transcriptional terminator in the 5 -untranslated region. So-called attenuator mutations abolishing this terminator structure result in a fivefold increase of b-lactamase expression. Integration of insertion sequence IS2 into the ampc-promoter has been shown to result in a hybrid promotor that is considerably stronger than the wildtype promotor. (iii) Finally, duplication of DNA segments encompassing the ampc gene can occur via recombinations between short direct repeats. Once a duplication has occurred it can readily be amplified to multiple copies resulting in b-lactamase hyperproduction because of the many ampc copies. scription [35]. During such conditions AmpR is interacting with the natural cytosolic precursor for the peptidoglycan, UDP-muramyl pentapeptide [36]. This interaction prevents AmpR from activating the ampc gene during normal growth. In the presence of b-lactams, there is an increased degradation of peptidoglycan derived muropeptides. These muropeptides are taken up into the bacterial cytosol via the AmpG transporter and displace UDP-muramyl pentapeptide from AmpR, converting AmpR into an activator turning on ampc-transcription. Hence, the AmpC b-lactamase becomes inducible by b-lactams [37, 38]. In all Gram-negative bacteria, recycled muropeptides are reused for new peptidoglycan synthesis. A key step in the reutilization process is the cleavage of the stem peptide from the recycled muropeptide by the cytosolic AmpD amidase (Fig. 3) [39]. In organisms expressing an inducible AmpC b- lactamase, mutations in ampd result in a cytosolic accumulation of recycled muropeptides that constantly keeps AmpR in an active form, allowing for a constitutive hyperproduction of AmpC b-lactamase even in the absence of b-lactams [40, 41]. Thus, although both E. coli and E. cloacae possess a chromosomal AmpC b-lactamase gene and a virtually identical murein recycling system, mutations in the recycling system will only cause resistance in the species that express an AmpR regulatory protein. The differences in mutation frequencies explain why therapeutic failure, caused by the selection of a resistant mutant, may occur with b-lactam treatment of an initially sensitive E. cloacae or C. freundii infection but not an E. coli infection. Besides E. cloacae and C. freundii a number of other Gram-negative organisms are expressing an inducible AmpC b-lactamase, notably P. aeruginosa, Yersinia enterocolitica and Serratia marscecens. They can all acquire b-lactam resistance through ampd mutations [42]. Plasmids have been identified carrying chromosomal ampc b-lactamase [43]. If such plasmids also carry an ampr gene or if the chromosomes of the recipient organism carry that gene, an extremely high b-lactamase production can be obtained from ampd-mutations, because of the high copy number of plasmid-borne ampc. Resistance in Mycobacterium tuberculosis is caused by chromosomal mutations Certain bacterial species such as Mycobacterium tuberculosis have an extremely low capability to exchange DNA with its surrounding. Thus, develop-

6 96 B. HENRIQUES NORMARK & S. NORMARK AmpC β -lactamase induction and murein recycling in Enterobacter cloacae Murein (peptidoglycan) Crosslinking Murein turnover β -lactam AmpC anh-mur NAc-tripeptide (tetrapeptide) PBP AmpG Inner membrane AmpR p AmpR β -lactamase induction AmpC + UDP AmpD Biosynthesis Recycling Fig. 3 The intimate connection between b-lactamase induction, murein turnover and recycling of murein fragments in Gram-negative bacteria with an inducible AmpC b-lactamase. Turnover products from the murein (peptidoglycan) are generated by the action of murein hydrolases. The turnover product 1,6-anhydroMurNAc tripeptide is taken up into the cell via the AmpG transporter. In the cytosol this muropeptide is degraded by the cytosolic AmpD amidase releasing the tripeptide, l-ala-dap-lys, that can be directly reutilized for murein biosynthesis. In the regulation of the inducible AmpC b-lactamase the transcriptional regulator AmpR is negatively controlled by components in the biosynthetic pathway, such as UDP-MurNAc-pentapeptide. When murein turnover is increased, as believed to happen upon b-lactam treatment, more recycled muropeptides are taken up into the cell allowing for an increase of cytosolic 1,6-anhydroMurNAc tripeptide even in the presence of AmpD amidase. This recycled muropeptide is believed to release UDP-MurNAc-pentapeptide from AmpR allowing this regulator to act as a transcriptional activator for the ampc gene. In ampd mutants 1,6-anhydroMurNAc tripeptide is not degraded in the cytosol allowing for a high concentration of this muropeptide also in the absence of b-lactams. In such mutants AmpR is constantly active causing AmpC-b-lactamase hyperproduction even in the absence of b-lactam inducer. ment of resistance in this species depends entirely on genetic alterations within the genome [44]. Mycobacterium tuberculosis and other members of the M. tuberculosis complex use several strategies to naturally resist the action of many antimicrobial agents. First, the mycobacterial cell is surrounded by a specialized, highly hydrophobic cell wall that results in decreased permeability to many compounds. Active efflux pumps and degrading or inactivating enzymes are also present and provide intrinsic resistance to many agents. Resistance of M. tuberculosis to antimycobacterial drugs is the consequence of spontaneous mutations in genes that encode either the target of the drug or enzymes that are involved in drug activation. Resistance-associated point mutations, deletions or insertions have been

7 REVIEW: EVOLUTION AND SPREAD OF ANTIBIOTIC RESISTANCE 97 described for all first-line drugs (isoniazid, rifampin, pyrazinamide, ethambutol and streptomycin) and for several second-line and newer drugs (ethionamide fluoroquinolones, macrolides, nitroimidazopyrans) [45]. Multidrug resistance, defined as resistance at least to isoniazid and rifampin, develops by sequential acquisition of mutations at different loci, usually because of inappropriate patient treatment [46]. Stepwise increase in antibiotic resistance, the fluoroquinolone example Single step mutations leading to full resistance occur, but are clinically not as important as resistance that is gradually built up by a series of successive mutations. Resistance to fluoroquinolones is a good example. Such resistance is built up by spontaneous mutations in target genes gyra, gyrb, parc and pare, and genetic events leading to an increased expression of efflux pumps, such as mutations abolishing the expression of a transcriptional repressor for the pump [47 51]. As different fluoroquinolones have different affinities for the target proteins, mutations in those genes may cause different types of resistance spectra. Once this multistep process to fluoroquinolone resistance involves efflux pumps resistance may also develop to other antibiotics if they are substrates for the up-regulated pump [52]. In this pathway from sensitivity to resistance, the initial mutation may have a small effect on fluoroquinolone resistance and may be overlooked. However, this first step mutation may prove extremely important for the level of resistance obtained by the second mutation [53]. Dosage regimens, pharmaco-dynamic properties of the drug and the total antibiotic load are important parameters in the development of multistep resistance. In every treated individual the time period, where MIC is higher than the MIC of the original strain but lower than the first type of mutation leading to increased resistance, must be kept as short as possible to avoid selection. The recommendation from a theoretical standpoint is to give the highest dosage possible to those that are treated. It is likely that a higher dosage given during a shorter treatment time will give a lower selection for multistep resistance than treatment with the same total amount of antibiotic given over a longer time period. We are currently witnessing the development of multistep resistance to both penicillin and fluoroquinolones in Streptococcus pneumoniae. Clearly, steps have to be taken to slow down this development. Efficiency of DNA-repair affect mutation frequencies to resistance, the mutator example Mutation frequencies may differ between bacterial species and also within the same species, depending on the efficiency of error correcting DNA repair systems as well as the efficiency of error prone DNA repair systems. Of particular interest are bacteria with a so-called mutator phenotype [54]. Such strains exhibit a much higher mutation frequency to resistance against a number of antibiotics, than normal for that species [55, 56]. It was recently shown that the mutation frequency to antibiotic resistance in H. pylori is generally higher than for enterobacteria, and in a recent study as many as 33% of H. pylori strains exhibited a mutator phenotype [57]. The nature of this high mutation frequency is not known. Mutations in the mismatch repair gene muts, that in enteric bacteria give a mutator phenotype, had no effect on mutation rate in H. pylori. It remains to be shown whether mutator strains within a bacterial population are those that drive resistance development caused by mutations. Fitness cost and compensatory mutations Mutations to resistance frequently occur in genes that fulfil essential functions in the cell. These mutations therefore have to generate alterations in a protein preserving its natural function, whilst reducing or abolishing its interaction with the antibiotic. This will reduce the number of potential alterations that are compatible with growth of the organism. Most mutations resulting in antibiotic resistance will at the same time have an effect on the normal function of the target protein. This has been clearly demonstrated both for ribosomal protein mutations causing aminoglycoside resistance as well as mutations in elongation factor G and GyrA, resulting in fusidic acid and quinolone resistance, respectively [58]. Fitness cost resulting from antibiotic resistance has mainly been studied in Salmonella typhimurium and Staphylococcus auerus, but it most likely represents a general phenomenon [59]. For example, resistance to clarithromycin in H. pylori was recently associated with a biological cost. In a

8 98 B. HENRIQUES NORMARK & S. NORMARK murine model the mutant was outcompeted by its sensitive parental strain in the stomach [57]. Importantly, fitness cost can be compensated for by additional mutations usually but not necessarily affecting the target protein. In an in vivo recycling experiment in mice, compensatory mutations arose very quickly resulting in resistant organisms that could not be outcompeted when infected together with the sensitive parental strain [60]. Interestingly, different types of compensatory mutations were selected in vivo during infection as compared with in vitro suggesting that environmental conditions prevailing in the host may affect the mutational spectrum [61]. Compensatory mutations in the presence of the resistance mutations may affect the active site for a target protein, such that it now operates as efficiently as the wild type protein. However, compensatory mutations are likely to work only in the context of the resistance mutation. Thus, compensatory mutations will prevent mutational reversion to sensitivity [62]. In any bacterial population, emergence, spread and stability of antibiotic resistance will be determined by several factors including the volume of drug use, the rate of formation of resistant mutants, the biological cost of resistance and finally the rate and extent of genetic compensation of the fitness cost. An antibiotic resistant organism that has no fitness cost by being resistant will behave as any susceptible bacterial clone within a population that is not under antibiotic selection. For antibiotic susceptible pneumococci we know that specific clones can rapidly emerge in a population without any obvious selective forces [63]. It can remain abundant in the population for several years, but are often replaced by other clones with time. Hence, even fully compensated antibiotic resistant strains are likely to decrease in number with time in the absence of a selective pressure. However, they are not likely to disappear completely and will rapidly expand in the population if antibiotic selection pressure is exerted. Horizontal gene transfer and antibiotic resistance Most bacterial genomes that have been sequenced contain a large proportion of DNA that have been relatively recently acquired from other sources. This horizontally acquired DNA usually encodes functions that are of selective advantage to the organism such as antibiotic resistance, virulence and biodegradation pathways. There are a number of different DNA elements described transferring antibiotic resistance: self replicating plasmids (that can be self transmissible by conjugation or brought into the cell by tranformation or transduction), prophages, transposons, integrons and resistance islands. Resistance plasmids, transposons and integrons The so-called R-factors, found in enteric bacteria, were the first examples of horizontally transferred antibiotic resistance [64]. R-factors are quite complex plasmids carrying not only antibiotic resistance genes, but also transfer functions allowing a rapid transmission within a population of related Gramnegative organisms. The resistance genes on these plasmids are either carried on plasmid integrated transposons or are inserted in an integron [65, 66 70]. Transposons are mobile genetic DNA elements that encode a site-specific transposase allowing sitespecific insertions and excisions. Bacteria contain a large number of transposable elements that can be categorized in four major groups according to their mechanisms of transposition. These are: class I: insertion sequences (IS) and compound transposons (with IS sequences at their termini) which usually require only one protein for transposition to occur (e.g. Tn10); class II: complex transposons and insertion sequences with short inverted repeats in which transposition is replicative and requires two gene products (e.g. Tn3); class III: transposable bacteriophage (e.g. Mu). The fourth group consists of the transposons and IS of variable mechanism, which do not fall into the above classes (e.g. Tn7) [71, 72] (Fig. 4). An integron consists of a site-specific integrase and the corresponding DNA target sequence [73]. Resistance gene cassettes, carrying a corresponding target sequence, are integrated one by one into such target sites mediated by the activity of the integrase (Fig. 5). Integrons are frequently included in transposons [74, 75]. Transposons and integrons have evolved as a mean for microorganisms to change more rapidly than possible by only mutations [76]. Transposons and integrons are frequently carried on plasmids, but can also have a chromosomal location.

9 REVIEW: EVOLUTION AND SPREAD OF ANTIBIOTIC RESISTANCE 99 IR1 IRr mer tnp Tn kb aada 1 sul1 R-det Tn ,2 kb Tn9 homolog IS1b cata 1 IS1a Fig. 4 Illustration of a self-transmissible and multiple resistant R-plasmid (NR1). The inserts represent mobile transposable elements. Note that a transposon may integrate into other transposons. Reproduced with permission (Liebert et al., Microbiology and Molecular Biology Reviews 63: 507). fin O rep A t r a NR 1 (R 100) 94.5 kb ori T Tn kb IS10 R teta,r IS10 L In Gram-positive bacteria, certain transposons, termed conjugative transposons, can be transferred from one organism to another without being carried as part of a plasmid [77]. The vanb locus of Enterococcus faecium has been localized to a Tn961- like conjugative transposon that itself was integrated in the chromosome within a larger transferable element that also contains a mutated pbp5 gene encoding high-level resistance to ampicillin. This finding helps explain why many vancomycin-resistant enterococci also are resistant to ampicillin [78]. Plasmids may carry resistance to several antibiotics. In a hospital setting, selection by one antibiotic may result in a selection for resistance to other antibiotics carried by the same plasmid. Plasmids carrying antibiotic resistance genes often encode resistance to heavy metals and detergents. Selection pressure exerted by the latter type of compounds may therefore also select for antibiotic resistance. The recently published complete genome of a multidrug resistant Salmonella enterica serovar Typhi gives an illustrative example of the evolution of multiple antibiotic resistance as a consequence of repeated DNA-insertion events [79]. This strain has a large conjugative plasmid that carries 18 genes involved in resistance to a large number of antimicrobial agents and heavy metals. Several intact and degenerate integrases and transposases are found on this plasmid. Resistance has been acquired by a number of successive IS-mediated genetic events. Thus, a chloramphenicol resistance cassette flanked by IS1 has been integrated into a tetracycline resistance transposon. In the chloramphenicol cassette a mercury resistance operon cassette

10 100 B. HENRIQUES NORMARK & S. NORMARK Integrase P P Integron Target site flanked by IS4321 is found. Finally, a cassette encoding b-lactam resistance, sulphonamide resistance and streptomycin resistance, and flanked by IS26, has finally found its way into the mercury resistance cassette. Variations on this theme are enormous and are found both in Gram-negative and Gram-positive bacteria. Extended spectrum b-lactamases Resistance genes carried by plasmids are also subjected to genetic alterations by mutations and possibly also by recombination. One example is the very common TEM b-lactamase encoded by the bla gene. This particular gene is widespread amongst enterics and usually carried on the Tn 3 transposon [80]. Haemophilus influenzae acquired TEM b-lactamase around 1974 [81] and pathogenic Neisseria around 1976 [82]. The TEM b-lactamase is efficiently hydrolysing b-lactams such as benzylpenicillin and ampicillin, but has a poor ability to hydrolyse third generation cephalosporins. However, variants of this enzyme have evolved with time, exhibiting an extended spectrum also including the third generation cephalosporins [83]. A number of alterations are present in these TEM variants that slightly alter the active site to also allow hydrolysis of antibiotics such as ceftazidime. Extended spectrum b-lactamase have not only evolved from TEM but also from SHV type of b-lactamases [84]. Plasmid mediated resistance in Gram-positive bacteria Resistance gene l Resistance gene ll Fig. 5 Schematic representation of a class 1 integron. DNA cassettes, often carrying antibiotic resistance determinants, are inserted one after another by a site-specific recombination mechanism involving specific attachment sites and an integrase. Plasmid mediated resistance is also very common amongst Gram-positive organisms. Today it is taken for granted that S. aureus should be resistant to benzyl penicillin because of the production of a b-lactamase from the bla gene, that is often carried on plasmids, frequently carrying other resistance genes, that can be transduced by phages to susceptible recipients [85]. The same bla gene is also found in enterococci. It is therefore remarkable that penicillin resistance in S. pneumoniae to this date has never been caused by b-lactamase expression and that penicillin resistance in group A streptococci has yet to be reported. There are no obvious explanations to this. Plasmids are rarely observed in pneumococci. So potential replication of staphylococcal plasmids in the pneumococcus might be deficient. Also, pneumococci prefer to take up linear DNA by transformation. No transducing bacteriophages are known for pneumococci. It could also be that restriction enzymes efficiently degrade DNA originating from staphylococci. Resistance islands, the meca example The meca gene in methicillinresistant Staphylococcus aureus is present on a large (about 50 kb) large DNA-element that is inserted into the bacterial chromosome. This so-called resistance island (also called SCCmec) encodes proteins with homologies to recombinases/integrases, and it has been shown that these enzymes may catalyse excition as well as integration of the mec-resistance island [86]. The stepwise build up of penicillin resistance in S. pneumoniae has occurred by multiple horizontal gene transfer events Instead of acquiring b-lactamase S. pneumoniae has evolved penicillin resistance in a much more elaborate and time consuming manner, explaining the relatively late appearance of pneumococcal penicillin resistance despite heavy selection. The so-called penicillin binding proteins (PBPs) represent targets for b-lactam antibiotics. The larger membrane bound PBPs catalyse the transpeptidation reaction required to cross-link the stem peptides in the peptidoglycan. Some of these PBPs also have transglycosidase activity required to covalently bind the disaccharide units, from the building block, together. Penicillin resistance in pneumococci has been generated by a series of horizontal DNA transfer events

11 REVIEW: EVOLUTION AND SPREAD OF ANTIBIOTIC RESISTANCE 101 allowing extensive remodelling of one or more of the large molecular weight penicillin binding proteins [87]. The resulting mosaic genes encode penicillin binding proteins with kinetic alteration in the interaction with penicillin, whilst preserving the recognition for the natural stem peptides. Penicillinresistant isolates of S. pneumoniae may contain mosaic genes encoding low-affinity forms of PBP2x, PBP2b and PBP1a. Low affinity variants of PBP2a and PBP1b also occur [88]. These low affinity PBP variants have evolved through interspecies gene transfer from commensal streptococci [89]. However, horizontal gene transfer between different penicillin-resistant pneumococcal strains have most probably contributed to the stepwise increase in penicillin resistance. Compensatory mutations reducing the fitness cost of these scrambled low affinity PBP variants have likely occurred as well fine tuning the kinetic properties of the altered PBP towards natural substrates [90]. Pathogenic Neisseria have also evolved penicillin resistance by rebuilding their PBP targets through horizontal gene transfer [91]. Both pneumococci and pathogenic Neisseria are readily transformable and exist in an environment containing related nonpathogenic commensals, providing the source of pbp-gene fragment building blocks. Penicillin-resistant pneumococcal clones have been shown to spread globally Spread of antibiotic resistance within a hospital, within a country or worldwide is usually accomplished by specific clones with a high transmission ability. These clones can be identified through different genetic finger printing methods of which multilocus sequencing (MLST) is the most appropriate method for international surveillance studies. In this method, a number of so-called house keeping genes are sequenced from each isolate and mutational differences are identified. The closer relationship there is between two isolates the fewer are the differences in the house keeping genes [92]. If the organism is not undergoing any type of horizontal gene transfer the sequence differences between two isolates belonging to the same species reflect the time that has elapsed since they departed from the same bacterium. Extensive horizontal gene transfer is a confounding factor in these analyses, as genetic divergence appears much faster in such bacterial populations. Having data banks with MLST profiles for large number of resistant bacteria worldwide greatly facilitates identification of particularly problematic international clones [93]. Using molecular typing methods specific penicillinresistant and/or multi-resistant pneumococcal clones have been traced and found to spread globally [94]. In most studies clones with intermediate or high penicillin resistance are more frequent amongst isolates from nasopharynx of healthy carriers, and from local infected sites such as the middle ear in otitis media as compared to invasive isolates from blood or liquor [95]. Also, penicillin resistance is more frequent amongst serotypes associated with a good ability to colonize the upper respiratory tract. Such organisms presumably have a high transmission capacity and live for long times in a microecosystem containing commensal relatives allowing for extensive antibiotic selection and interspecies genetic exchange. Other pneumococcal serotypes, such as serotypes 1 and 3, are common in invasive disease but are rarely found in the nasopharynx of healthy carriers. Penicillin resistance is rare amongst isolates of these two serotypes. Such isolates presumably have a high attack rate but may have a low capacity to colonize and grow in the upper respiratory tract. Development of penicillin-resistant and/or multiresistant pneumococci that have evolved a high transmission rate as well as a high attack rate, would be a serious treat to the treatment of invasive pneumococcal disease. Most epidemiological studies show no increased mortality in invasive disease caused by penicillin-resistant pneumococci compared with penicillin-susceptible isolates. If this is because of successful treatment with other antibiotics or with penicillin at a higher dosage or if this reflects a lower virulence for penicillin-resistant clones in invasive disease is not known. Prevention of clonal spread of penicillin-resistant pneumococci includes intervention at the step of transmission (isolation, hand washing, etc.) and at the step of selection by reducing the antibiotic load. Vaccination targeting particular resistant clones might also be a valid approach. The conjugated 7-valent vaccine against S. pneumoniae is given protection against most of the serotypes associated with penicillin resistance [96]. Immunization against these serotypes might introduce selective bottle necks for resistant pneumococcal clones if most children are being vaccinated. However, such

12 102 B. HENRIQUES NORMARK & S. NORMARK a strategy may also allow for an expansion of novel resistant clones of serotypes not covered by the vaccine. How b-lactams and vancomycin kill S. pneumoniae and the phenomenon of antibiotic tolerance b-lactams and vancomycin are lytic antibiotics meaning that bacteria normally lyse as a consequence of antibiotic action. In S. pneumoniae a laboratory mutant of strain R6 was isolated several years ago that was not lysed by penicillin [97]. This mutant had lost the ability to produce the major autolysin LytA. LytA is an l-alanine-muramyl amidase that releases stem peptides from the peptidoglycan. A lyta mutant is not only nonlytic after treatment with penicillin, but also after vancomycin treatment. Moreover, the mutant no longer lysed spontaneously upon reaching the stationary phase, in contrast to the parental strain. LytA is produced throughout the pneumococcal growth cycle and there is no induction of the gene when lytic antibiotics are added. Consequently, the LytA amidase must somehow be prevented from acting on its substrate during active growth. Lytic antibiotics trigger a response in the pneumococcus resulting in the enzymatic activation of LytA and self-afflicted death through lysis [98]. Although a lyta mutant is not lysed by penicillin and vancomycin, these agents remain able to kill the organism, but less efficiently. In pneumococci, lytic antibiotics therefore trigger a lytic death dependent on LytA as well as a nonlytic death that operates through an unknown killing system. A series of penicillin and/or vancomycin tolerant mutants were recently identified and characterized in the laboratory strain R6 by Tuomanen s group [99, 100]. These mutants still produced normal amounts of LytA but were neither lysed nor killed by lytic antibiotics suggesting that they were defective in both death processes. A two component system vncr/vncs was identified in this search where inactivation of the VncS histidine kinase sensor resulted in complete tolerance whereas inactivation of the corresponding VncR response regulator had no measurable phenotype. The VncS sensor has therefore been suggested to be part of a signal transduction system leading to antibiotic induced death. This antibiotic triggered death system also contains a small secreted peptide, Pep27 as well as an ABC transporter [101]. Clinical isolates of pneumococci showing a reduced lysis to lytic antibiotics occur at a high frequency, although they produce apparently normal levels of LytA. Many of these isolates show a reduced killing by penicillin and/or vancomycin, but are eventually killed by the drugs [102]. Although these isolates have been termed antibiotic tolerant, there are to our knowledge no clinical isolates reported that show a complete tolerance to lytic antibiotics, defined as no significant killing at 10 MIC. In contrast to the laboratory strain R6, most clinical isolates of S. pneumoniae are encapsulated. We found that pneumococci of certain capsular serotypes appeared less lytic to penicillin and/or vancomycin than pneumococci of other serotypes. Notably, isolates of capsular serotypes 2, 4, 9V were in general less lytic than pneumococci of serotype 3. A mutant defective in capsular production was isolated in strain T4. Interestingly, this nonencapsulated mutant was significantly more lytic than the corresponding type 4 capsulated parental strain both after penicillin treatment and after growth to stationary phase (Henriques et al., unpublished data). Hence, it appears that capsular polysaccharides of certain chemistries are able to inhibit the LytA mediated lytic response. The underlying mechanism for this is not known. It has been shown, however, that several capsular types can be covalently attached to the peptidoglycan one exeption being the type 3 polysaccharide [103]. Our hypothesis is that capsular polysaccharides by binding to peptidoglycan may sterically interfere with the ability of LytA to reach its target. Variations in capsular chemistry, degree of peptidoglycan association level for capsular polysaccharide, variation in the level of capsular production and variations in the peptidoglycan structure may all be factors affecting the lytic response of encapsulated clinical isolates of S. pneumoniae. Reverting resistant bacteria to sensitivity With the detailed knowledge at hand for a number of resistance mechanisms there should be possibilities to revert existing resistance. The first useful example of this approach has been the combination of b-lactams with b-lactamase inhibitors. Vancomycin resistance is particularly interesting in this respect as this drug is widely recognized as the last

13 REVIEW: EVOLUTION AND SPREAD OF ANTIBIOTIC RESISTANCE 103 line of defence in many hospital-acquired bacterial infections. This drug exerts its antibacterial activity by binding to d-ala-d-ala termini of peptidoglycan precursors preventing these from being incorporated into the growing wall. Enterococci resistant to vancomycin carry the vana or vanb gene cluster that allow some of the peptidoglycan precursors to terminate with d-ala-d-lac, which preclude binding to vancomycin. Recently, small molecules have been generated that are capable of a selective hydrolysis of the altered peptidoglycan precursors, causing a significant reduction in MIC to vancomycin [104]. Reversal of meticillin resistance in S. aureus would be particularly important because of the increasing number of cases with meticillin-resistant bacteria. It has recently been found that meticillin resistance requires not only the meca gene encoding the novel penicillin binding protein PBP2A, but also the transglycosylase domain of the native PBP2 protein. Hence, genetic inactivation of this domain in PBP2 prevented meticillin resistance. Hence, a combination of methicillin and novel transglycosidase inhibitors might be a way to combat meticillin-resistant staphylococci [105]. In S. pneumoniae penicillin resistance was recently shown to be reverted by genetically inactivating the murmn operon encoding enzymes required for the formation of branched stem peptides in the peptidoglycan. Interestingly, pneumococcal murmn mutants, unable to produce branched stem peptides, became hypersensitive to lytic antibiotics and also lysed faster than the wild type when exposed to such antibiotics. Tolerance to lysis induced by lytic antibiotics may therefore require the presence of branched stem peptides in the pneumococcal peptidoglycan. This work shows that there exists an important link between antibiotic resistance and antibiotic tolerance in S. pneumoniae [106, 107]. Again, inhibitors against enzymes such as MurM and MurN might be useful in combination with b-lactams against penicillin-resistant pneumococci. Efflux pumps are also interesting targets for new inhibitors. To identify potential efflux pump inhibitors, the growth inhibitory activity of synthetic compounds and natural product extracts was assayed using strains of P. aeruginosa overexpressing each of the three efflux pumps: MexAB-OprM, MexCD-OprJ and MexEF-OprN, that were grown with or without subinhibitory concentrations of levofloxacin. One broad-spectrum efflux pump inhibitor was found to decrease intrinsic resistance of P. aeruginosa to fluoroquinolones significantly. The same inhibitor also reversed acquired fluoroquinolone resistance in Pseudomonas mutants overexpressing efflux pumps [108]. References 1 Evans GA. The oxazolidinones. Curr Infect Dis Rep 2002; 4: Heithoff DM, Conner CP, Hanna PC, Julio SM, Henschel U, Mahan MJ. Bacterial infection as assessed by in vivo gene expression. Proc Natl Acad Sci USA 1997; 94: Mahan MJ, Tobias JW, Slauch JM, Hanna PC, Collier RJ, Mekalanos JJ. Antibiotic-based selection for bacterial genes that are specifically induced during infection of a host. Proc Natl Acad Sci USA 1995; 92: Slauch JM, Mahan MJ, Mekalanos JJ. In vivo expression technology for selection of bacterial genes specifically induced in host tissues. Meth Enzymol 1994; 235: Sansonetti PJ. Microbes and microbial toxins: paradigms for microbial mucosal interactions III. Shigellosis: from symptoms to molecular pathogenesis. Am J Physiol Gastrointest Liver Physiol 2001; 280: G Stebbins CE, Galan JE. Structural mimicry in bacterial virulence. Nature 2001; 412: Arigoni F, Talabot F, Peitsch M et al. A genome-based approach for the identification of essential bacterial genes. Nat Biotechnol 1998; 16: Zasloff M. Antimicrobial peptides of multicellular organisms. Nature 2002; 415: Whiteley M, Bangera MG, Bumgarner RE et al. Gene expression in Pseudomonas aeruginosa biofilms. Nature, 2001; 413: Spoering AL, Lewis L. Biofilms and planktonic cells of Pseudomonas aeruginosa have similar resistance to killing by antimicrobials. J Bacteriol 2001; 183: Mulvey MA, Schilling JD, Hultgren SJ. Establishment of a persistent Escherichia coli reservoir during the acute phase of a bladder infection. Infect Immun 2001; 69: Nikaido H. Crossing the envelope: how cephalosporins reach their targets. Clin Microbiol Infect 2000; 6 (Suppl. 3): Nikaido H. Preventing drug access to targets: cell surface permeability barriers and active efflux in bacteria. Semin Cell Dev Biol 2001; 12: Hancock RE, Speert DP. Antibiotic resistance in Pseudomonas aeruginosa: mechanisms and impact on treatment. Drug Resist Updat 2000; 3: Li XZ, Nikaido H, Poole K. Role of mexa-mexb-oprm in antibiotic efflux in Pseudomonas aeruginosa. Antimicrob Agents Chemother 1995; 39: Gotoh N, Tsujimoto H, Tsuda M et al. Characterization of the MexC-MexD-OprJ multidrug efflux system in DeltamexAmexB-oprM mutants of Pseudomonas aeruginosa. Antimicrob Agents Chemother 1998; 42: Kohler T, Michea-Hamzehpour M, Henze U, Gotoh N, Curty LK, Pechere JC. Characterization of MexE-MexF-OprN, a positively regulated multidrug efflux system of Pseudomonas aeruginosa. Mol Microbiol 1997; 23: