Evolution and epidemiology of extended-spectrum β-lactamases (ESBLs) and ESBL-producing microorganisms
Corresponding author and reprint requests: M. Gniadkowski, Sera & Vaccines Central Research Laboratory, ul. Chelmska 30/34, 00–725 Warsaw, Poland
Tel: +48 22 851 46 70
Fax: +48 22 841 29 49
The rapid and irrepressible increase in antimicrobial resistance of pathogenic bacteria that has been observed over the last two decades is widely accepted to be one of the major problems of human medicine today. Several aspects of this situation are especially worrying. There are resistance mechanisms that eliminate the use of last-choice antibiotics in the treatment of various kinds of infection. Many resistance mechanisms that emerge and spread in bacterial populations are those of wide activity spectra, which compromise all or a majority of drugs belonging to a given therapeutic group. Some mechanisms of great clinical importance require specific detection procedures, as they may not confer clear resistance in vitro on the basis of the interpretive criteria used in standard susceptibility testing. Finally, multiple mechanisms affecting the same and/or different groups of antimicrobials coexist and are even co-selected in more and more strains of pathogenic bacteria. The variety of β-lactamases with wide spectra of substrate specificity illustrates very well all the phenomena mentioned above. Being able to hydrolyze the majority of β-lactams that are currently in use, together they constitute the most important resistance mechanism of Gram-negative rods. Three major groups of these enzymes are usually distinguished, class C cephalosporinases (AmpC), extended-spectrum β-lactamases (ESBLs) and different types of β-lactamases with carbapenemase activity, of which the so-called class B metallo-β-lactamases (MBLs) are of the greatest concern. This review is focused on various aspects of the evolution and epidemiology of ESBLs; it does not cover the problems of ESBL detection and clinical relevance of infections caused by ESBL-producing organisms.
A general definition of extended-spectrum β-lactamases (ESBLs) consists of several elements . ESBLs are active site serine Ambler's class A or class D β-lactamases [2,3], which are able to hydrolyze oxyimino-β-lactam compounds at a rate that is equal to or higher than 10% of that for benzylpenicillin. They demonstrate good activity against broad-spectrum penicillins and cephalosporins, and are inhibited by site-directed β-lactamase inhibitors . In the β-lactamase functional classification scheme by Bush, Jacoby and Medeiros, ESBLs are located in two subgroups of group 2, namely subgroups 2be (extended-spectrum β-lactamases; Ambler's class A enzymes) and 2d (cloxacillin-hydrolyzing β-lactamases; Ambler's class D ESBLs) .
For several reasons, ESBLs pose a serious clinical problem. The first of these arises directly from their wide substrate specificity, which in general includes almost all penicillins, cephalosporins (except cephamycins) and monobactams [1,4]. The second is the fact that ESBL producers frequently appear susceptible in vitro to some ESBL substrates [5–7], which, among other reasons, is due to high quantitative differences in the activity of certain ESBL variants against particular compounds [1,4,8–11]. However, because such drugs were often found to be non-efficacious in vivo [12–16], it is broadly accepted today that, aside from those identified in urinary tract infections, ESBL-producing isolates should, by definition, be reported as resistant to all ESBL substrates . This creates a real challenge for clinical microbiology laboratories, which should precisely detect the ESBL phenotype [6,17–24]. Finally, the third important aspect of the presence of ESBLs results from their complex and dynamic evolution and epidemiology.
ESBLs are usually described as acquired β-lactamases that are encoded mostly by plasmid-located genes [4,25]. As such, they are a very recent evolutionary development; the first clinical isolates expressing acquired ESBLs were identified in Germany in 1983 [26,27]. In 1985 the first nosocomial outbreaks caused by ESBL producers started to occur in France , and then in the USA at the end of the 1980s and the beginning of the 1990s [8,13,29–31]. It is, however, noteworthy that some species-specific β-lactamases are biochemically indistinguishable from ESBLs and are classified into the Bush subgroup 2be too , such as OXY enzymes of Klebsiella oxytoca[32,33] or CME β-lactamases of Chryseobacterium meningosepticum[34,35]. Moreover, certain species-specific Bush subgroup 2e cephalosporinases , e.g. CepA of Bacteroides fragilis, CblA of Bacteroides uniformis or L2 of Stenotrophomonas maltophilia[38,39], differ from the 2be enzymes only by their rather poor activity against penicillins. Both the 2be and 2e species-specific β-lactamases reveal clear structural homologies to some types of classical ESBLs [34,35,40–44]. All of these data indicate that the model of a class A β-lactamase with a substrate-binding cavity broad enough to interact with oxyimino-β-lactams is not, in fact, so recent in bacterial evolution, and that some ESBLs could inherit their principal characteristic from species-specific ancestral β-lactamases.
The number of ESBL variants identified has been constantly growing since 1983, demonstrating how rapid ESBL evolution has been. More than 100 different natural ESBL variants are known at present, and the list seems to be far from complete (http://www.lahey.org/studies/webt.htm) [42–45]. They have been classified into nine distinct structural/evolutionary families based on comparisons of their deduced amino acid sequences (families TEM, SHV, CTX-M, PER, VEB, GES, TLA, BES and OXA). Major families are formed by TEM and SHV enzymes, which include the first ESBL variants identified [26–28,46,47], and which have remained the most prevalent ESBL types, in terms of both variant numbers and numbers of producer strain isolations (http://www.lahey.org/studies/webt.htm). OXA enzymes, which are the only ESBLs of class D , represent another relatively prevalent ESBL family (http://www.lahey.org/studies/webt.htm). TEM, SHV and OXA ESBLs are structural derivatives of acquired, so-called broad-spectrum β-lactamases, such as TEM-1, TEM-2 and SHV-1 (all Bush subgroup 2b), and OXA-2 and OXA-10 (subgroup 2d), respectively (http://www.lahey.org/studies/webt.htm) [1,4,25,48]. Their deeper evolutionary origins are unclear, with the exception of SHV enzymes, which most probably evolved from the K2 variant of Klebsiella pneumoniae species-specific β-lactamase . The genealogy of the remaining ESBL families is, in general, more mysterious; however, as mentioned above, amino acid sequence comparisons have revealed similarities of some of them to certain species-specific β-lactamases from Bush subgroups 2be and 2e. These homologies range from 30–40% up to 99%, which indicates common origins of the discerned clusters of enzymes and, in the case of a very high homology value, even the direct descent of an ESBL from a species-specific enzyme. A good example is the CTX-M family, which has been rapidly growing in recent years (http://www.lahey.org/studies/webt.htm) ; this groups ESBLs related to β-lactamases of Kluyvera ascorbata (99% homology with the CTX-M-2 variant), Klebsiella oxytoca, Citrobacter diversus, Proteus vulgaris and Serratia fonticola[40,41,51]. On the other hand, the minor ESBL families PER, VEB and TLA have been found to be distantly related to each other and also to Bacteroides fragilis CepA, Bacteroides uniformis CblA and Chryseobacterium meningosepticum CME β-lactamases [34,35,42,43]. All these observations lead to the conclusion that ESBL activity is demonstrated by enzymes with substantial diversity in terms of structure and evolutionary origins.
Most of what we know about ESBL evolution comes from studies of TEM and SHV enzymes, which, as mentioned above, are mainly structural mutants of TEM-1, TEM-2 and SHV-1 penicillinases, respectively. The prevalence of these β-lactamases in nosocomial Enterobacteriaceae populations, reflected, for example, by an ampicillin resistance rate of about 50% in Escherichia coli, is one of the key factors responsible for the wide spread of TEM and SHV ESBLs around the world. Comparisons of nucleotide sequences of TEM and SHV ESBL genes with those coding for their parental penicillinases have revealed numerous point mutations, which cause amino acid substitutions in a β-lactamase polypeptide. Many of these substitutions have been found to affect enzyme structure and activity in different ways, and were thoroughly discussed by Knox in his review . The most important are spectrum-extending mutations (position 164 in TEMs, 179 in SHVs and 238 in both), which, by enlarging the β-lactam-binding site, make enough room for enzyme interactions with compounds possessing bulky oxyimino side-chains. Causing serious structural alterations, these mutations diminish the overall enzymatic activity of a β-lactamase when compared to parental enzymes [48,52]. Other substitutions (104 in TEMs and 240 in TEMs and SHVs) enhance β-lactamase interactions with oxyimino side-chains of specific compounds (ceftazidime and aztreonam) . There is a mutation (237 in TEMs) which modulates β-lactamase activity by reducing it against certain β-lactams (ceftazidime and aztreonam) while increasing it towards others (cefotaxime and cephalothin) . Another substitution (182 in TEMs) serves as an intragenic suppressor that compensates for β-lactamase structural defects resulting from spectrum-extending substitutions . In some TEM and SHV enzymes, ESBL-type mutations have been identified together with those that determine β-lactamase resistance to β-lactam inhibitors, inhibitor resistance-type substitutions (positions 69, 130, 275, 276) [52,55–57]. Finally, not only substitutions but also amino acid insertions or deletions have been found either to be responsible for or to possibly contribute to ESBL activity [58,59]. What is noteworthy is that a combination of mutations with defined roles does not have to result in a simply additive effect on β-lactamase activity. TEM-52 has been shown to hydrolyze moxalactam and cefotetan, which is unlikely be attributable to any of its substitutions alone (104, 182, 238) , whereas in SHV-10 the IR-specific substitution (130) is strongly detrimental to ESBL activity .
Among all the changes identified in TEM and SHV ESBLs, there are also single amino acid substitutions, which are considered to be neutral (e.g. position 21 in TEMs ); however, it is possible that some of these need a specific context of other mutations to reveal their cryptic role [53,54]. Neutral or nearly neutral mutations do not undergo strong, if any, selective constraints, and therefore their repeated selection under strictly defined pressure is less likely than that of mutations which turn out to be profitable in a given environment. In general, this is the case with silent mutations too, which can also be found in ESBL genes when compared to each other. Some of the neutral or nearly neutral substitutions and at least the majority of silent mutations observed were inherited by TEM and SHV ESBL genes from different variants of their parental penicillinase genes [62–65] and constitute specific markers of diverse genealogic lineages of ESBL genes. Analysis of these mutations is very useful in evolutionary as well as epidemiologic investigations within particular ESBL families [64,66]. This has been demonstrated, for example, in studies of genes coding for TEM-12 and TEM-26 ESBLs (blaTEM-12 and blaTEM-26) in clinical isolates from British, French and US hospitals. Specific distribution of silent mutations in these genes has excluded the possibility of the transfer of producer strains from the USA to the UK and France, and clearly indicates events of convergent evolution [67–70].
Some ESBL variants have evolved and may still evolve directly from broad-spectrum β-lactamases, due to single spectrum-extending mutations [47,71–73]; however, the vast majority of them have emerged by the stepwise acquisition of different mutations by pre-existing ESBLs. This process was very well documented by the work of Bradford et al., who identified two ESBL genes differing by one nucleotide substitution in a single Klebsiella pneumoniae isolate. The blaTEM-10 gene probably appeared as a result of duplication of blaTEM-12, which was followed by the acquisition of an additional mutation . The comparison of genes coding for different TEM ESBL variants in Polish hospitals strongly suggests that, in a sequence of single genetic changes, blaTEM-48 gave rise to blaTEM-47 and blaTEM-49, and then blaTEM-68 evolved from blaTEM-47[57,66].
The selection pressure that drives ESBL evolution has usually been attributed to the intense use of oxyimino-cephalosporins, mainly ‘third-generation’ cephalosporins . This was well documented in a study conducted in 1993–94 in a hospital in Cleveland, where the frequency of ESBL-producing Klebsiella pneumoniae isolates was positively correlated with the use of ceftazidime and negatively with that of piperacillin–tazobactam, to which the isolates were susceptible . There also exists an interesting observation from Japan, where in the 1990s ESBL-producing organisms were sporadically identified and usually expressed the so-called Toho β-lactamases from the CTX-M family [74–77]. It is possible that the low incidence of ESBLs in Japan has been at least partially due to the intense use of cephamycins and carbapenems in this country . As ESBLs often demonstrate selective preferences regarding different oxyimino-β-lactams, selection of a particular enzyme variant in a given center has been frequently attributed to the specific profile of antibiotic use. This was, for example, the case with a high prevalence of the TEM-26 ceftazidime-hydrolyzing β-lactamase in several US hospitals in 1988–92, into which ceftazidime was introduced in high amounts at that time [8,13,29]. In 1996–97, the CTX-M-3 cefotaxime-hydrolyzing ESBL was found to predominate in a hospital in Warsaw, where the consumption of ceftriaxone and cefotaxime exceeded that of ceftazidime by around four times . However, such a correlation has not always been observed, as is the case in a study performed in a Chicago hospital in 1990–91, in which the TEM-10 ceftazidime-hydrolyzing β-lactamase was identified as the most prevalent ESBL variant. Its high prevalence coincided with the introduction of ceftazidime to the hospital, but because the use of cefotaxime was much higher at that time, a greater selective potential of ceftazidime was suggested .
In recent papers by Blázquez et al., a wider view of the problem has been proposed, according to which ESBLs evolve under the constant or fluctuating pressure of various β-lactam antibiotics, including diverse oxyimino-compounds as well as penicillins and early-generation cephalosporins [53,79]. The authors postulate that such conditions may be responsible for the selection of TEM enzymes carrying the modulating substitution at position 237, which converts ESBL variants of highly ‘asymmetric’ substrate preferences into those with much more balanced activity [53,80]. This hypothesis probably contains also one of the main explanations for the fact that only a fraction of all possible spectrum-extending mutations have been selected under natural conditions when compared to laboratory mutant studies [81–84]. Structural defects resulting from some spectrum-extending mutations appear to diminish β-lactamase activity too far against, for example, aminopenicillins . It could also be suggested that pressure of the use of non-oxyimino-β-lactams may be responsible for the selection of mutation combinations that further extend ESBL activity, as in TEM-52 towards moxalactam , or for the selection of so-called complex mutant enzymes, which combine ESBL- and IR-type mutations [55–57]. Out of three such enzymes identified to date under natural conditions, two variants, TEM-50 and TEM-68, have been found to retain both activities at significant levels [56,57].
The concept of multi-β-lactam pressure does not explain all the phenomena that are observed in ESBL evolution and epidemiology. ESBLs characterised by highly selective substrate preferences are being selected [50,68,80] and may remain prevalent at the level of a single medical center or on a wider geographic scale [78,85,86]. Moreover, as Blázquez et al. point out, enzyme variants with ‘rare’ spectrum-extending mutations may also be sporadically selected. By way of example, it could be the case that the mutation in position 179, which, although never found in natural TEM ESBLs, is identified in some SHVs [79,87–89]. This may be due to reduced competition that occurs in specific niches of complex nosocomial environments  or unique profiles of antibiotic use in particular hospitals or hospital wards. The ‘modulating tendency’ of multi-β-lactam pressure may also be limited by the fact that ESBLs often appear efficacious in vivo against their weaker substrates [12–15]. Nevertheless, the multi-β-lactam pressure hypothesis seems to be of great importance.
The strong selective pressure of the use of β-lactam drugs exerted on ESBL producer strains operates not only on ESBL gene coding regions, but also on their promoters, copy number and other genes. Such changes may significantly affect the β-lactam resistance of strains by increasing its level, expanding it for non-ESBL substrates, and/or compromising the activity of β-lactamase inhibitors. Several reports have documented the selection of strains that hyperproduced ESBLs due to various genetic changes such as promoter up-mutations, insertion of transposable elements close to the promoter region, or multiplication of the ESBL gene copy number [90–93]. Other work has demonstrated the emergence of strains expressing two ESBLs of different types [68,85,86,94,95]. In Argentina and Poland, where cefotaxime-hydrolyzing CTX-M ESBLs are widespread, the ‘double-ESBL’ isolates producing such an enzyme and an ESBL with a much stronger activity against ceftazidime have been found [85,86,95]. Owing to mutations in genes coding for porin proteins, ESBL producer strains frequently show reduced permeability for antibiotics, and this is usually manifested by cefoxitin resistance in Klebsiella pneumoniae strains [96–98]. In recent years, ESBLs have become more and more prevalent in species characterised by inducible class C cephalosporinases (AmpC), such as Enterobacter spp., Citrobacter freundii or Serratia marcescens, which frequently segregate mutants with high-level constitutive production of AmpC enzymes [4,48,78,95,99,100]. Strains that are both ESBL producers and AmpC derepressed mutants have been already identified in several studies [86,101,102], despite the masking of the ESBL phenotype by the inhibitor resistance of AmpC β-lactamases.
ESBL-producing isolates are frequently resistant to other kinds of antimicrobials, including aminoglycosides, quinolones and co-trimoxazole [13,28,31,78,89,103]. What is especially frustrating is that genes coding for ESBLs and, for example, aminoglycoside-modifying enzymes (AMEs) often reside within the same conjugative plasmids and therefore are transmitted together from one strain to another [78,104,105]. This means that resistance to two different kinds of drugs may be co-selected by the use of either one. Bell and Turnidge demonstrated that, in 1998–99 in Western Pacific countries and South Africa, the frequency of resistance in ESBL-producing Enterobacter cloacae to gentamicin, ciprofloxacin and co-trimoxazole could reach values of approximately 65%, 60% and 20%, respectively .
ESBL epidemiology should be considered at different levels, namely the level of a single patient, of a single medical institution, and on a wider geographic scale, and at each of these it strongly depends on evolutionary phenomena that occur in ESBL-producing strains. A very informative example of such an overlap between evolutionary and epidemiologic problems in the course of infection of a single patient was reported by Rasheed et al., who followed the 3-month progression of an Escherichia coli infection of a child with several bacteremia episodes. Analysis of multiple isolates of the organism which were collected over time revealed a very complex view of both the linear and radiative evolution of an originally single bacterial strain, which coincided with intense antimicrobial therapy. It involved such events as acquisition of the SHV-1-encoding gene, hyperproduction of the enzyme, emergence of SHV-8 ESBL by a single point mutation in blaSHV-1, porin alterations, and acquisition of resistance to other antimicrobials . On the other hand, studies by Marchandin et al. have demonstrated the interspecies dissemination of an ESBL gene-carrying plasmid in a single multibacterial infection/colonisation case [106,107]. In one of these, the horizontal transfer of a plasmid with the TEM-24-encoding gene was observed between strains of Enterobacter aerogenes, Escherichia coli, Proteus mirabilis and Pseudomonas aeruginosa.
ESBLs have been found in a wide range of Gram-negative rods; however, the vast majority of strains expressing these enzymes belong to the family Enterobacteriaceae [4,108–110]. Numerous species of this family have already been identified as hosts of ESBL expression, including organisms of the so-called Enteric Group 137, with as yet no clear taxonomic position . Klebsiella pneumoniae seems to remain the major ESBL producer; this is reflected both by its usually outstanding contribution to all ESBL-expressing isolates identified in medical centers, and by the high incidence of ESBL production in nosocomial populations of the species [4,108,110,112]. Another very important organism is Escherichia coli, which has a remarkably high representation among ESBL producers circulating in hospital environments [4,108,112]. As mentioned above, ESBLs have become common over time among species with inducible AmpC β-lactamases; this is well illustrated by a recent Argentine survey, in which approximately 14% of all inducible AmpC strains studied were found to express ESBLs . What is also of note is the growing incidence of ESBLs in Salmonella spp.; sporadic isolates of ESBL-producing salmonellae or nosocomial outbreaks caused by these organisms have been already identified in numerous countries of Latin America, Africa, Europe and Asia [113–115]. Non-Enterobacteriaceae ESBL producers are much rarer, with Pseudomonas aeruginosa being the most important among them . This organism is a frequent host of OXA-type ESBL expression [73,117,118]; however, Pseudomonas aeruginosa isolates producing TEM, SHV and PER ESBLs have also been identified in several countries [106,119–123]. PER β-lactamases have also been found in Acinetobacter spp. [124,125] and in Alcaligenes faecalis. Finally, TEM-17 has been detected in Capnocytophaga ochracea and SHV-12 in Burkholderia cepacia.
ESBL-producing organisms are usually selected in hospitals [4,108,129]; however, a number of reports from the USA and France have documented the possibility of their emergence in nursing home facilities [121,130,131]. Recent studies on the appearance of Salmonella spp. strains with the acquired CMY-2 AmpC enzyme in livestock [132,133] have also demonstrated the danger of selection of ESBL producers in such environments. The frequency of ESBL producers in a hospital depends on many factors and may vary greatly; in a 1988–90 survey conducted in 12 French centers, the incidence of ESBL-producing isolates in Klebsiella pneumoniae populations ranged from 0% to approximately 50% . The distribution of ESBL-expressing strains within a hospital is similar to that of other drug-resistant organisms, and they are usually more prevalent in wards in which patients with numerous infection risk factors are hospitalised. These are mainly intensive care units (ICUs) [14,99,102,112,134,135], surgical wards , pediatrics and neonatology wards [57,92,93,112], rehabilitation units , oncologic wards [8,112] and chronic-care facilities [29,136].
An ESBL variant (ESBL producer strain) may appear in a center due to de novo selection, which may result in a novel type of enzyme [17,57,60,89,130] or in one that has been previously identified in another institution (convergent evolution) [67–69,137]. It may also be imported by a patient from another center, even though that may be located in another city or country [31,57,138–141]. Once selected, the ESBL variant may spread in the center by different means, including clonal dissemination of the producer strain [14,92,93,142,143] or horizontal transmission of the ESBL gene-carrying plasmid among non-related strains [29,144]. A growing number of ESBL outbreaks have recently been attributed to concurrent plasmid transfer and clonal spread events [78,102,134]. In some studies, the same ESBL-encoding gene has been identified in different plasmids present in bacterial strains in a hospital, which suggests either convergent evolution of the gene in a single center or horizontal gene transfer among different plasmids [31,78,131]. Indeed, some ESBL-encoding genes are located within transposons or integrons, which strongly facilitates such a transfer [42,45,70,118,145].
ESBL outbreaks may be large. In an approximately 500-bed general hospital in New York, 165 patients were infected or colonised by TEM-26-producing Klebsiella pneumoniae over a 19-month period at the beginning of the 1990s [13,30]. In 1993–94, 148 ESBL-producing Klebsiella pneumoniae isolates were recovered from patients of an approximately 900-bed hospital in Cleveland . An approximately 1000-bed hospital in Barcelona reported a figure of 145 patients infected or colonised with ESBL-producing Klebsiella pneumoniae between May 1993 and June 1995 . ESBL outbreaks have also often occurred, as was revealed in a study performed by Babini et al. on ESBL-producing klebsiellae from ICUs of 13 European hospitals in 1994 and 1997–98. Sixty-five per cent of the 1994 isolates and 69% of the 1997–98 isolates represented clonal outbreak strains, and these were identified in nine of the ICUs involved in the analysis . The lower digestive tract of colonised patients has been recognised as the major source of ESBL-producing organisms , and their cross-transmission among patients has been attributed to hands of medical personnel [16,92]. Some ESBL outbreaks have been revealed to result from contamination of various diagnostics supplies, such as thermometers , or gel used in ultrasonography .
An ESBL gene-carrying plasmid, an ESBL-producing strain and even an ESBL outbreak may persist in a medical center over a prolonged period. A family of related plasmid molecules containing blaSHV-5 was identified in Enterobacteriaceae isolates collected in a hospital in Albany between 1993 and 1999 . In the aforementioned study of klebsiellae from ICUs of 13 European centers, nine ESBL-producing strains were observed both in 1994 and 1997–98; moreover, three of these caused permanent outbreaks . Recently, Coque et al. have reported results of an analysis of all ESBL-expressing Klebsiella pneumoniae isolates collected in a hospital in Madrid over a period of 11 years (1989–99). They have found that the blaTEM-69-carrying plasmid was present in that population between 1995 and 1999, and that two different ESBL-producing strains circulated in the hospital for 2 and 5 years, respectively. Interestingly, this study has also documented the opposite phenomenon, which is a rapid turnover of ESBL producers in a medical center. The two strains mentioned above were the only ones out of 28 ESBL-expressing Klebsiella pneumoniae strains identified that persisted for longer periods in the hospital .
Dynamic evolution is one of the main reasons why numerous ESBL variants and producer strains exist in a center at the same time. By the end of 1988, five different TEM ESBLs (TEM-5, TEM-8, TEM-12, TEM-16 and TEM-24) had been identified in the Klebsiella pneumoniae population in a hospital in Clermont-Ferrand . Three ESBL variants of different families (TEM-10, TEM-26 and SHV-7) coexisted in Enterobacteriaceae strains of a hospital in New York in 1993 . In 1996–97, in all, 22 CTX-M-3 or SHV-type ESBL-producing strains of seven Enterobacteriaceae species (identified as RAPD types) were distinguished in a hospital in Warsaw . But probably the highest diversity of ESBLs and ESBL producers ever described in a single center was observed by Essack et al. in the 1996 Klebsiella pneumoniae population from a hospital in Durban. Nine distinct strains were discerned among 13 ESBL-producing Klebsiella pneumoniae isolates collected at the time, and these were found to express many combinations out of three TEM and six SHV β-lactamases, including different ESBL variants (TEM-53, TEM-63, SHV-2, SHV-5, SHV-20, SHV-22) .
One of the most intriguing issues of wider-scale ESBL epidemiology is the frequency of ESBL producers in larger geographic regions. This is very difficult to estimate, for a variety of reasons; one of these arises from high variations in ESBL incidence among different hospitals, which means that the selection of centers for survey studies strongly influences the mean data obtained . It should also be considered that, due to the dynamics of ESBL epidemiology, survey results might describe a transient situation, characteristic for a given region only at a certain point in time . However, such studies have been undertaken and give us some insights into the problem concerning the global spread of ESBLs. Recent data from the SENTRY Antimicrobial Surveillance Program, performed on Klebsiella pneumoniae, Escherichia coli, Proteus mirabilis and Salmonella spp. isolates collected in 1997–99 from all over the world, showed that ESBL frequency in Klebsiella pneumoniae may account for about 45% in Latin America, 25% in the Western Pacific, 23% in Europe and 8% in the USA . The very high incidence of ESBL production in Latin America has also been observed in a SENTRY study for this region, in which 40.5% of Klebsiella pneumoniae isolates collected were found to be ESBL positive . In a 1994 survey carried out on klebsiellae from 35 ICUs in 10 countries of western and southern Europe, ESBLs were identified in 22.8% of all isolates analyzed and in 28.6% of Klebsiella pneumoniae isolates . Similar data, with ESBL incidence standing at 25.4% of all klebsiellae, was obtained in a follow-up study in 1997–98 .
On a single-country scale, the frequency of ESBL production has been investigated in France since the mid-1980s, and the results of two out of several surveys performed in the country will be discussed here. In a 1988–90 study of β-lactam resistance in 12 university hospitals, ESBL producers comprised 13.3% of Klebsiella pneumoniae isolates and 1.5% of all Enterobacteriaceae isolates; however, it was this study in which the aforementioned high variations between hospitals were observed . In 1998, a similar survey was performed involving 14 hospitals from 11 regions of France, and ESBLs were detected in 9.5% of Klebsiella pneumoniae isolates and 3.2% of all Enterobacteriaceae isolates. Strikingly, the highest incidence of ESBLs was observed in Enterobacter aerogenes (53.5%); however, this was due to the intense multicenter spread of a single clone of this organism, which may have been a transient epidemiologic phenomenon . In 1999, an ESBL survey was performed in 28 hospitals in South Korea and revealed a frequency of 18.1% ESBL producers among Klebsiella pneumoniae isolates .
It was observed in the 1980s that the same ESBL variant might be present in several or even many different medical institutions. Such a spread of an ESBL variant can occur due to the dissemination of a single clone of an ESBL-producing organism in numerous centers of a region, which seems to strongly indicate frequent occurrences of strain transmission from one center to another. This was well documented in France, with an SHV-4-producing Klebsiella pneumoniae clone being found in 14 hospitals by the end of 1988 , and the previously mentioned TEM-24-producing Enterobacter aerogenes clone spreading into 21 hospitals by 1997 . Recently, a single clone of CTX-M-type ESBL-producing Salmonella typhimurium has been identified in Russia, Hungary and Greece . In other cases, the presence of a single ESBL variant in different centers may be attributed mostly to the dissemination of an ESBL gene-carrying plasmid; however, this also suggests an important role for the transmission of producer strains between institutions. For example, plasmid transfer was revealed to be responsible for the spread of TEM-3 in France, identified in strains of 10 different species in 26 hospitals by 1989 [155,156], and of CTX-M-3 in Poland, found in eight species in 15 hospitals by 2000 . Finally, convergent evolution may also lead to a wide geographic representation of an ESBL variant [67–70], and this has probably been a major factor in the ‘international career’ of SHV-5, despite documented cases of its importation to one country from another . SHV-5 has been found in almost every country in which ESBLs have ever been studied down to the molecular level [57,64,95,97,134,148,157–165], and in some of these, e.g. in Poland, its gene was identified in a wide variety of plasmids in bacterial isolates from numerous hospitals [57,78,135]. The described phenomena, alone or in combination with each other, may lead to the high prevalence of certain ESBL variants in particular regions or countries. The French 1998 ESBL survey revealed that TEM-24 was expressed by 51.9%, and TEM-3 by 35.4%, of all ESBL-producing isolates included in the study, which means that these two enzymes are mainly responsible for the clear predominance of TEMs among all ESBLs that can be currently observed in France .
Less than 20 years after the first identification, ESBL-producing Enterobacteriaceae have become common microorganisms in medical institutions of many countries and larger geographic regions. Together with methicillin-resistant staphylococci, vancomycin-resistant enterococci and some other types of pathogens, they now make a major contribution to the serious clinical problem created by nosocomial infections with very limited therapeutic options. Various factors of different kinds facilitate their spread, including the diversity of parental β-lactamases in ESBL evolution, the worldwide prevalence of their specific types, the relative ease of emergence of an ESBL, the ‘mobility’ of ESBL genes and, finally, the strong selective pressure of antibiotic use. The very rapid evolution of ESBLs seems to run in several directions, such as increased activity, its modulation or further extension, and, together with other evolutionary changes observed in ESBL producer strains, it is leading to an increase in the level and an expansion of the spectrum of their resistance phenotypes. In recent years, ESBLs have greatly widened the range of their host species, and the growing incidence of ESBLs in Salmonella spp. demonstrates the danger of their spread among pathogens circulating in livestock and the community. ESBL outbreaks are usually complicated and difficult to eradicate, and frequently evolve into complex endemic situations characterised by a high variety of producer strains expressing multiple β-lactamase combinations at a single center. The insufficiently monitored movement of infected or colonised patients between different medical institutions often causes a very wide geographic spread of particular ESBL variants and their producer strains. All these factors taken together illustrate that ESBL epidemiology today is very dynamic, and the prediction of its consequences in the future is difficult.
The author would like to thank Waleria Hryniewicz, Ewa Sadowy and Andrew Hazlewood for critical reading of the manuscript.