P. Courvalin Unité des Agents Antibactériens, Institut Pasteur, 25–28, rue du Docteur Roux, 75724 Paris Cedex 15, France. Tel.: +33 45 68 83 20. Fax: +33 45 68 83 19. E-mail: firstname.lastname@example.org
Following the detection of glycopeptide-resistant enterococci (GRE) in 1986 and their subsequent global dissemination during the 1990s, many studies have attempted to identify the reservoirs and lines of resistance transmission as a basis for intervention. The eradication of reservoirs and the prevention of GRE spread is of major importance for two reasons: (i) the emergence of high-level glycopeptide resistance in invasive enterococcal clinical isolates that are already multiresistant, has left clinicians with therapeutic options that are only at the experimental stage; and (ii) the resistance genes may spread to more virulent bacterial species such as Staphylococcus aureus, Streptococcus pneumoniae and Clostridium difficile. VanA-type strains, resistant to high levels of both vancomycin and teicoplanin, are the most commonly encountered enterococci with acquired glycopeptide resistance in humans. A widespread VanA-type GRE reservoir was detected early in farm animals that were exposed to the glycopeptide growth-promoter avoparcin. Numerous studies have provided indirect evidence for the transfer of VanA-type GRE and their resistance determinants from animal reservoirs to humans. The data collected have expanded our understanding of the promiscuous nature of antibiotic resistance, and have provided the groundwork for logical decision-making with the objective of deterring the dissemination of resistant bacteria and of their resistance genes.
The first human enterococcal strains with acquired high-level glycopeptide resistance reported from the UK and France in 1988 [1,2] made headlines in medical journals. Uttley and colleagues described an outbreak of nosocomial infections and fecal colonization with high-level vancomycin-resistant enterococci among renal unit patients [1,3]. Leclercq and colleagues reported plasmid-mediated glycopeptide resistance in two fecal Enterococcus faecium strains from patients with leukemia . During the following decade, glycopeptide-resistant enterococci (GRE) rapidly emerged worldwide as an important nosocomial pathogen [3–13]. In 1999 more than 60% of E. faecium clinical isolates in the USA were reported to be resistant to vancomycin . Hospital data from the UK in 1999 revealed vancomycin resistance among 24% of clinical E. faecium isolates . However, the relatively high proportion of glycopeptide resistance in E. faecium in the UK is not representative of the situation in Europe according to a recent study that reported vancomycin resistance in 22 of 574 E. faecium strains (3.8%) from 27 countries .
From a clinical point of view, transferable high-level glycopeptide resistance in enterococci from humans is a cause of concern for two obvious reasons: (i) glycopeptide antibiotics, alone or in combination with aminoglycosides, are still considered the last-resort therapy for infections due to multiresistant Gram-positive bacteria, and (ii) glycopeptide resistance determinants have the potential to spread to more virulent bacteria. The transfer of glycopeptide resistance operons to various Gram-positive bacteria under laboratory conditions, as well as their heterospecific expression (Table 1), have made this risk obvious [17–20]. The appearance of these genes in clinical isolates of Cellulomonas (Oerskovia) turbata, Arcanobacterium (Corynebacterium) hemolyticum, Bacillus circulans and Streptococcus bovis underscores this concern.
Table 1. . Heterospecific expression of the vanA gene cluster
aVanco, vancomycin; bteico, teicoplanin. Modified from . The MIC of vancomycin against a Staphylococcus aureus harboring the vanA gene was 250–1000 mg/L .
VanA, which confers to high-level resistance to both vancomycin and teicoplanin, is the most common type of acquired glycopeptide resistance in human clinical enterococci. The epidemiology of VanA-type GRE is complex and has been the subject of recent reviews [24–32]. Several factors may have contributed to the striking global dissemination of this resistance among enterococci in humans. The putative transmission of VanA-type GRE and of their resistance determinants from avoparcin-associated animal GRE reservoirs to humans is one of these factors. This review summarizes and discusses our current knowledge on the possible transfer from animals to humans, and its potential impact on public health.
Evidence for animal to human transmission of antibiotic resistant bacteria and resistance determinants
Zoonoses are defined by the World Health Organization as ‘Diseases and infections which are naturally transmitted between vertebrate animals and man’; however, this definition has been debated . The term ‘anthropozoonoses’ refers to infections transmitted from lower animals to man, and among those are infections due to various enteropathogenic bacteria. The link between animal and human reservoirs of antibiotic-resistant bacteria is more obvious for pathogenic bacterial genera such as Salmonella and Campylobacter than for commensal species such as Escherichia coli and E. faecium. Infectious disease surveillance systems are based primarily on clinical infections that may reveal outbreaks of resistant pathogenic zoonotic bacteria, whereas the zoonotic transfer of commensals passes unnoticed. Hence, the first and most convincingly documented evidence for the transmission of antibiotic-resistant bacteria from animals to humans is, not unexpectedly, provided by enteropathogenic bacteria.
The possible danger of development of resistance in bacteria from farm animals became a concern in the early 1960s, following the description of transfer of drug resistance among Enterobacteriaceae  and the emergence of multiresistant strains of Salmonella in animals . In 1965 Anderson and Lewis described the possible transfer of multiple drug-resistant Salmonella enterica serotype typhimurium from calves to humans . The association between the use of antimicrobial drugs in animal feed and the emergence of resistance became a cause of concern at this time . H. Williams Smith wrote in 1968: ‘Ideally, only drugs that are unsuitable or not usually used for the treatment of diseases in animals and man, and that do not produce cross-resistance with ones that are, should be used in animal feed’. The scene was set for the Swann report on the use of antibiotics in animal husbandry and veterinary medicine . The committee tolerated the use of antibiotics in animal feed if these antibiotics (i) ‘have little or no application as therapeutic agents in man or animals and (ii) will not impair the efficacy of a prescribed therapeutic drug or drugs through the development of resistant strains of organisms’.
The early observations by Anderson and Lewis have been further substantiated by a number of studies. More recently, Mølbak et al reported a connection between the occurrence of S. enterica serotype typhimurium with reduced susceptibility to fluoroquinolones in a Danish pig herd and 25 human infections from which two patients died . Human infections due to fluoroquinolone-resistant Campylobacter have also been reported following use of these drugs in animals . Recently, the US Food and Drug Administration’s Center for Veterinary Medicine, proposed to withdraw the approval of fluoroquinolones for animal use due to the increase in fluoroquinolone resistance among Campylobacter species isolated from human patients associated with the use of fluoroquinolones in poultry since 1995 .
Levy and coworkers have provided experimental evidence for the possible transfer of antibiotic-resistant bacteria and of plasmid-encoded resistance determinants from animals to humans. A temperature-sensitive chloramphenicol resistance gene was used to trace a conjugative plasmid, pSL222-6 (also conferring resistance to tetracycline, sulfonamides and streptomycin), in E. coli strains derived from chickens. The resistances were subsequently transiently detected in fecal E. coli strains from humans at the farm, suggesting the transfer of resistant E. coli or of the conjugative plasmid from animal to human . Intra- and interspecies spread of animal antibiotic-resistant E. coli in a farm environment has been further documented . Interestingly, E. coli from one animal host was able to spread rapidly and to colonize the intestinal tract of other animals and humans in the absence of selective pressure.
Molecular evidence for animal to human transfer of resistance determinants
The evidence for transfer of resistance determinants from animal to human bacteria in vivo is strong when the selective pressure is only exerted in animals, as was the case for the streptothricin antibiotic nourseothricin in the former German Democratic Republic [44,45]. Nourseothricin replaced oxytetracycline in animal feed in 1983 and was used exclusively in animals. When its use was stopped in 1990, streptothricin resistance had spread from the fecal flora of nourseothricin-fed pigs to E. coli in the gut flora of farm personnel, their families, and healthy adults in the community. Similar E. coli strains were later isolated from humans with urinary tract infections, and the resistance determinant was eventually detected in clinical S. enterica and Shigella sonnei isolates [25,45]. Similarly, Chaslus-Dancla et al reported human clinical isolates of E. coli and S. enterica serotype typhimurium resistant to apramycin, an aminoglycoside restricted to veterinary therapy . The plasmid-borne apramycin resistance gene encoding a 3-N-aminoglycoside acetyltransferase type IV that also confers resistance to gentamicin, was initially detected only in animal strains. Recently a gene conferring resistance to streptogramin type A antibiotics and thus to quinupristin/dalfopristin, satG, has been detected in E. faecium isolates from animals and humans . Resistance was detected in human E. faecium before any clinical use of quinupristin/dalfopristin, suggesting animal to human spread of resistant bacteria or of their resistance determinants associated with the use of virginiamycin as a growth promoter. The emergence of resistance to type A streptogramins in animal enterococci has been associated with the use of virginiamycin in farm animals [48–50].
Detection of GRE reservoirs in the community and their association with the use of avoparcin in animal husbandry
The first human VanA-type GRE isolates were described in hospital patients associated with fecal colonization or infections [1–3]. Analysis of the first 100 GRE cases in New York hospitals during 1989–91 revealed two community-acquired urinary tract infections with GRE . This observation led the authors to suggest that GRE were spreading from hospitals to the community as has been shown for methicillin-resistant Staphylococcus aureus. However, community reservoirs of VanA-type GRE were soon discovered in several European countries, making the opposite direction of transfer equally likely.
Researchers in the UK and Germany presented the first evidence for a community reservoir of GRE in 1993 [51,52]. Bates and colleagues reported the ubiquitous presence of GRE in raw sewage, fresh and frozen chickens, and in fecal samples from various farm animals as well as patients from community practice. They suggested that the emergence of GRE in hospital patients was related to resistant organisms entering the food chain as a result of antibiotic use in animal husbandry . Klare and colleagues described the presence of high-level glycopeptide-resistant E. faecium in waste water from sewage treatment plants in different geographic parts of Germany, indicating a large environmental reservoir of the resistance determinant . Subsequent studies showed that the GRE strains from community reservoirs in the UK and Germany harbored the vanA gene cluster [53,54]. Further evidence for a large environmental reservoir of VanA-type GRE in Europe was provided by Torres and colleagues, who also described vanA enterococci in Spanish sewage . The German community reservoir also included humans, as demonstrated by the isolation of fecal VanA-type E. faecium in 12% of nonhospitalized humans in the Saxony-Anhalt federal county in the former German Democratic Republic, a region with a very limited human use of vancomycin at that time .
In 1995, German and Danish studies [54,57] revealed fecal colonization with VanA-type GRE in pigs and poultry exposed to the glycopeptide avoparcin as an animal growth promoter. Both Klare et al and Aarestrup showed a strong correlation between avoparcin and vancomycin MIC levels in E. faecium isolates as well as the presence of the vanA gene in glycopeptide-resistant strains, indicating cross-resistance to the two antibiotics [56,58]. Subsequently, a retrospective Danish cohort study provided strong evidence in favor of an association between the use of avoparcin and the occurrence of GRE in poultry farms and swine herds . Kruse et al confirmed these findings when examining Norwegian poultry farms and swine herds . The presence of fecal VanA GRE in chickens in Japan was also associated with the use of glycopeptides (e.g. avoparcin and orienticin) in animal husbandry . GRE were not detected in Swedish poultry farms and swine herds where avoparcin had not been used since 1986 . Similarly, no animal reservoirs of GRE were detected in the USA when chicken and turkey samples were examined . Avoparcin has not been licensed as a growth promoter in the USA and Canada. These data strongly suggest a causal relationship between the consumption of avoparcin and the occurrence of VanA-type GRE in farm animals.
The extent of the use of avoparcin in European animal husbandry compared with that of glycopeptides in humans is illustrated by a Danish report . ‘In Denmark, with a population of 5.2 million inhabitants, 24 kg of active vancomycin were used for humans in 1994. In comparison, 24 000 kg of active avoparcin was used for swine and broilers the same year. Thus, the amounts of active avoparcin used for livestock in 1994 exceeded the total use of vancomycin in the USA and in Europe in 1994.’
The association between the use of avoparcin use and the occurrence of VanA-type GRE has been further demonstrated by the decrease in the prevalence of GRE in animal husbandry following the ban of avoparcin in Denmark and Norway in 1995, in Germany in 1996, and in the remaining EU countries in 1997. Bager et al reported a significant decrease in the proportion of glycopeptide resistant strains of E. faecium in Danish chicken, from >80% in 1995 to <5% in 1998, when examining a randomly selected isolate from each flock for glycopeptide susceptibility . An Italian study reported a decrease in the prevalence of VanA-type GRE in poultry meat, 49 of 334 (15%) positive samples in March 1997 and 22 of 271 (8%) in October 1998, 18 months after avoparcin was banned . Klare et al isolated VanA-type GRE from 11 of 11 German poultry meat samples in 1994 and from 8 of 31 (26%) samples in 1997 after the withdrawal of avoparcin in January 1996 . However, the prevalence of VanA-type GRE in 1998 in poultry farms in Norway 3 years after the ban of avoparcin continued to be high . The proportion of glycopeptide-resistant enterococci as opposed to susceptible enterococci was not determined in this study. Environmental sampling on five Norwegian poultry farms 4 years after the ban of avoparcin revealed the persistence of VanA-type GRE, even after depopulation and the implementation of hygienic measures . GRE were not isolated from the hatchery providing day-old chicks to the farms, but within 3 weeks all flocks tested positive for GRE, indicating a persistent environmental reservoir for VanA GRE colonization in animals.
It thus appears that the ban of avoparcin endorsed by the EU probably led to a reduction in the proportion of GRE among enterococci in formerly exposed animals. However, the situation prior to the use of avoparcin has not been restored, and the prevalence of animals colonized with GRE at farms exposed to avoparcin is still very high in an environment without any apparent glycopeptide selection. The situation is analogous to that described by Smith in the UK after the ban of tetracycline as an animal growth promoter in 1971 . Four years later, the proportion of tetracycline-resistant E. coli isolated from the exposed pig population had slightly decreased, but the prevalence of pigs excreting the resistant organisms was still 100%.
Fecal VanA-type GRE colonization among outpatients and healthy community controls has been demonstrated in a number of European studies, varying from 0.1% to 28%[56,71–75]. Table 2 summarizes the data concerning human GRE colonization rates from various studies. The rates of GRE colonization in humans are low in Sweden and Norway. Torell et al detected a single VanA-type E. faecium carrier in a Swedish study of 670 outpatients . Norwegian surveys of human GRE colonization have been performed only in hospitals. Again, a single VanB-type (vancomycin-resistant, teicoplanin-susceptible) GRE was detected among 616 patients hospitalized in high-risk units , suggesting a very low frequency of GRE colonization in the general population. Van den Bogaard et al reported a 14% VanA-type GRE colonization rate among 117 (sub)urban residents of a farming district in The Netherlands using broth enrichment with vancomycin [74,75]. Van der Auwera et al found 11 VanA-type E. faecium carriers among 40 community-based controls in Belgium (28%) . In the same study performed in 1989–91, ‘in vivo enrichment’ of GRE was demonstrated by the recovery of VanA-type GRE in 14 of 22 (64%) healthy volunteers who received oral glycopeptides. This observation suggests an underestimation of human GRE colonization rates using traditional sampling strategies and emphasizes the importance of antibiotic selection in the occurrence of GRE, which has been confirmed in subsequent clinical, epidemiological and experimental studies [85,87,88]. Endtz et al detected VanA-type GRE in 12 of 624 (2%) hospitalized patients and in four of 200 (2%) patients from the community . Genotyping of the strains revealed a high degree of strain diversity that underscored the unlikelihood of hospital-engendered GRE endemicity and suggested that the origin of the VanA-type strains was within the community.
Table 2. . VanA-type GRE fecal colonization rates among humans in hospitals and in the community
aYear of sample collection; bNS, not specified.
The large differences among European countries of GRE colonization rates in humans in the community can be explained in several ways. The quantitative exposure to VanA-type GRE may be different because of differences in the use of avoparcin. Avoparcin was banned in Sweden in 1986. Its use in Norway was restricted to poultry and lasted for only 10 years . Differences in the microbiological methods used for the recovery of GRE have a strong impact on the results. Enrichment broth has been shown to significantly enhance the recovery of GRE compared with direct plating [89,90]. Cultural differences in food handling practices, as well as differences in meat consumption, may influence the level of transmission. It is also possible that the use of other animal growth promoters (i.e. tylosin) in several EU countries has coselected GRE due to linkage of the vanA gene cluster to other resistance determinants (i.e. erm genes). In fact, in the first GRE strains observed in humans in Europe, the macrolide and glycopeptide resistance genes were physically linked on plasmids [3,17]. Co-transfer of erythromycin and vancomycin resistance between enterococci was detected by in vitro filter conjugation experiments using the first GRE isolates from the UK as donors . Evidence for genetic linkage between macrolide resistance determinants (erm) and the vanA gene cluster has been described [17,91]. Of note is the high percentage of resistance to tylosin and erythromycin among GRE isolates from Dutch turkeys, turkey farmers, turkey slaughterers and suburban residents: 70%, 75%, 67% and 93% respectively . Thus, the human use of macrolides may have selected for intestinal colonization with enterococci harboring glycopeptide resistance determinants genetically linked to erm.
The widespread fecal carriage of VanA-type GRE by nonhospitalized humans in several European countries suggests an efficient transmission from large and/or multiple reservoirs. Following the ban of avoparcin in Germany, a decrease in the prevalence of human fecal GRE colonization was observed in healthy volunteers: from 12 of 100 (12%) in 1994, to 6 of 100 (6%) in 1996 and to 13 of 400 (3%) in 1997 . These figures indicate that interventions directed at animal GRE reservoirs also had a significant impact on the human nonhospital GRE reservoir. These observations also support an animal origin for the reservoir of human VanA-type GRE or vanA resistance determinants within the European community that may be acquired through the food chain via contaminated meat products. Enterococci can be transmitted to meat products despite high hygienic standards during the slaughtering process . Enterococcus faecium exhibits tolerance to extremities of pH, salinity and temperature; it has been shown to withstand heating to 65°C or 80°C, for 20 or 3 min, respectively , and to survive on dry surfaces for several days . Various studies report the recovery of VanA-type GRE from raw commercial meat products in various European countries [51,53,56,90,95]. Quednau et al were able to isolate VanA-type GRE from Danish, but not Swedish, meat products, which confirms an association between the use of avoparcin and the occurrence of GRE . GRE was not recovered from vegetarians in contrast to that recovered from meat-eaters in a Dutch study, which further implicates animal food products as the most plausible source of GRE colonization in nonhospitalized patients . Although Van den Braak et al did not confirm the difference in GRE colonization between vegetarians and meat-eaters in another Dutch study , this failure to confirm resulted from finding a single VanA-type E. faecium carrier among 276 meat-consuming controls.
Sources of GRE other than contaminated meat products or alternative modes of entry into the human community population have been proposed. A hospital origin for the genomically diverse VanA-type GRE colonizing 12% of healthy Germans was very unlikely since this observation was made in a geographic region with a highly restrictive use of vancomycin in humans . Enterococcus faecium is used in probiotics as well as in cheese fermentation and VanB-type GRE have been recovered from a probiotic products . However, the fact that VanA-type GRE was not found in food fermentation products in a large European study , nor in various USA probiotic products , have made this a less likely source. The detection of GRE in fecal samples from pet cats and dogs in Europe is related to the ingestion of contaminated food [98,99]. The likely transmission of VanA-type GRE between farm animals and humans may also apply to the GRE colonization of pet animals. To our knowledge, the possibility of transmission between pet animals and humans through direct or indirect contact has not been studied.
Farmers – the first ground for animal to human spread of GRE
If GRE were transferred from animals to humans, one would expect the rates of human colonization to vary according to the subject’s level of exposure to animals. Van den Bogaard et al and Stobberingh et al investigated the prevalence of VanA-type GRE colonization among turkeys, turkey farmers and turkey slaughterers [74,75]. The samples were collected in The Netherlands in 1996, the year the use of avoparcin was suspended in the EU. They reported colonization rates of 50% among 47 turkey flocks, 39% among 47 turkey farmers, 20% among 48 turkey slaughterers and 14% among samples from 117 area residents . The number of GRE relative to total enterococci in fecal samples was not significantly different between turkeys and turkey farmers . The high colonization rates of Dutch turkey farmers are comparable with those in a recent Norwegian study. In 1998, 3 years after the Norwegian ban of avoparcin, a total of 72 of 73 chicken flocks (99%) previously exposed to avoparcin and 13 of 73 farmers were still colonized with VanA-type GRE (18%) . The high prevalence of carriers among Norwegian farmers is in contrast to the absence of VanA-type GRE among Norwegian hospital patients . These observations clearly suggest a significant correlation between the level of exposure and the risk of human colonization with GRE. These data, in conjunction with the fact that VanA-type E. faecium in humans and animals are indistinguishable, as are vanA operons in animal and human GRE at farms exposed to avoparcin, imply transmission of GRE and/or their resistance determinants from animals to humans. Although one could imagine the reverse mode of transfer, the substantial and persistent animal and environmental reservoir at exposed farms to avoparcin [68,69] clearly supports the notion of transmission from animals to humans, by indirect or direct contact.
Comparison of GRE reservoirs in animals and humans at the strain level
GRE from animal and human sources have been molecularly characterized in a number of studies to analyze the genomic relatedness among strains. Ribotyping of 42 VanA-type GRE from farm animals, sewage and chicken from retail outlets by Bates and colleagues revealed polyclonality with 14 separate patterns . Two ribotypes of the nonhuman strains were also found among human clinical isolates. These findings suggested transmission lines between human, animal and environmental GRE reservoirs, although the discriminatory power of ribotyping can be questioned. Similar genomic diversity was observed among German VanA-type GRE from animal feed, sewage and nonhospitalized patients examined by pulsed-field gel electrophoresis (PFGE) and multilocus enzyme electrophoresis .
The extensive genomic heterogeneity of VanA-type GRE makes the likelihood of detecting genomically related strains among randomly selected animal and human strains extremely low. Examination of a large number of isolates from a single location would be necessary for an optimal assessment of strain-sharing among animals and humans. However, epidemiologically related and unrelated VanA-type GRE from humans and animals with indistinguishable or highly similar PFGE patterns have been found in several studies [74,75,100,101]. Van den Bogaard and colleagues  described VanA-type GRE that was indistinguishable in a Dutch farmer and his turkey flock. Examination of VanA-type GRE in Norwegian poultry and their farmers revealed a pair of animal and human strains that were closely related genetically . Both the Dutch and the Norwegian studies revealed indistinguishable vanA resistance determinants in genomically diverse but closely related epidemiologically animal and human strains, as shown by DNA sequencing and RFLP analysis of Tn1546 amplicons [74,75,100]. The molecular data obtained strongly suggest that animal GRE can colonize the human intestine. The observation of transient human intestinal colonization following separate ingestions of VanA-type E. faecium isolated from a pig and a chicken further substantiate the risk of transmission of GRE from animals to humans . This observation was recently extended and confirmed in a Danish randomized double-blind study. The ingestion of VanA GRE strains originating from food animals by human healthy volunteers resulted in temporary intestinal growth and colonization that was documented by molecular methods .
Host-specific genotypic or phenotypic markers would have been valuable for the analysis of interspecies bacterial transfer. Devriese and colleagues have shown host-specific sugar fermentation patterns among E. faecium strains from dogs and chickens [104,105]. Recently, amplified-fragment length polymorphism (AFLP) was used for the analysis of the genetic relatedness among VanA-type GRE isolates from hospital patients, nonhospitalized persons, pets and various farm animals . AFLP combines PCR and restriction analysis and allows the examination of the polymorphism among small restriction fragments. One of the aims of the study was to assess the contribution of animal husbandry GRE reservoirs to the occurrence of GRE in humans. Willems and colleagues disclosed four major AFLP genogroups (A–D), each sharing 65% of the restriction fragments. GRE genotypes that were specific for animal hosts were also detected in the fecal flora of their corresponding farmers, suggesting again the transmission from animals to humans.
Comparison of GRE reservoirs in animals and humans at the Tn1546 level
The vanA gene cluster, conferring inducible high-level vancomycin and teicoplanin resistance, is part of the nonconjugative Tn1546 transposon (Figure 1) that was originally described in human clinical E. faecium BM4147 . The vanA operon has been shown to reside on variably sized conjugative plasmids [3,17] and conjugative chromosomal elements , which appear to be responsible for the intra- and intercellular mobility of VanA-type resistance.
The epidemiology of Tn1546 has been extensively examined in a large number of genomically unrelated GRE isolated from different human and nonhuman reservoirs in various countries using different molecular techniques: PCR mapping [19,108–114], Southern hybridization [108,111,114–116], DNA sequencing [100,111,112,114,116,117] and long PCR RFLP analysis [100,114,116,118–121]. Polymorphism among Tn1546 elements results from the presence of IS1216V, IS3-like, IS1542, IS1251 and IS1476 in non-essential regions for expression of glycopeptide resistance [110–112,114–116,122]. Various deletions associated with these insertions have been described in the left orf1-orf2 region required for transposition as well as in the right accessory vanZ region (Figure 1). Apart from insertion elements and associated deletions Tn1546 elements seem to be highly conserved at the DNA sequence level. Extensive sequence analysis has demonstrated only six nucleotide substitutions in Tn1546 so far (Figure 1). The most prevalent is a G→T transition at position 8234 in the vanX gene, introducing an amino acid substitution (K→N). Both variations are present in human clinical strains. Interestingly, whereas the T-substitution is predominant in pig isolates [116,123], a conserved G in position 8234 has been observed in GRE from poultry and the farmers who deal with them [68,122]. The mutation at position 4847 (T→C, silent) in the vanS gene  is found in 57% of 70 GRE recovered from veal calves (Willems and Mevius, unpublished data). Two mutations, 7658 (T→C; V→A) in vanA and 9692 (C→T; P→L) in vanY, are found almost exclusively in human isolates from the USA . Finally, two mutations at position 1226–T→A in orf1 resulting in the introduction of a stop codon and the silent C→T mutation at position 6908 in vanH– were encountered in a veal calf  and in a human isolate  respectively.
The overall conclusion one can draw from these observations is that Tn1546 elements are well conserved at the sequence level and that two transposon subtypes, A1 and A2, are dominant and widely distributed in both human and animal GRE [111,112,114,116,118]. The A1 subtype is identical to the Tn1546 prototype . The A2 subtype has a 120-bp deletion at the left end associated with an IS1216V-IS3-like insertion and a G→T transition at position 8234 . The high degree of sequence identity of Tn1546 in genomically diverse GRE strains contrasts with the heterogeneity of the vanB gene clusters [117,124] and is consistent with a common pool of vanA resistance determinants among GRE reservoirs. Recent molecular analysis of Tn1546-like elements in human GRE suggests geographic clustering of transposon types associated with differences in meat consumption in European countries that indicate an animal origin of human vanA gene clusters in Europe .
Sequencing of the entire Tn1546 in 13 isolates revealed identical transposons in animal and human strains . Enterococcus faecium derived from human fecal matter and E. gallinarum derived from a veal calf harbored A1 subtype elements, whereas identical Tn1546 A2 subtype elements were found in a human blood culture and in a pig-derived E. faecium. These observations suggest that animal to human transfer of glycopeptide resistance can also occur at the gene level. The detection of indistinguishable Tn1546 elements in GRE in animals and farmers that are genetically diverse but closely linked epidemiologically strongly substantiate this hypothesis [74,75,100]. The larger GRE reservoir in animals in this setting implies that the transfer is most likely from animals to humans.
Although the glycopeptide resistance determinants from various organisms have been studied in detail [reviewed in 126], the origin of the vanA gene cluster in enterococci remains unknown. Potential reservoirs have been sought in lactic acid bacteria from the genera Lactobacillus, Leuconostoc, and Pediococcus that is intrinsically glycopeptide resistant [127–129], as well as in glycopeptide-producing organisms [130,131]. Glycopeptide resistance in lactic acid bacteria is associated with the synthesis of peptidoglycan precursors terminating in D-Ala-D-Lac and the expression of D-lactate dehydrogenase enzymes, as shown in VanA- and VanB-type GRE . However, molecular analysis demonstrated that the D-Ala-D-Lac ligases from lactic acid bacteria and those from VanA-type GRE constitute different phylogenetic groups [127–130]. Thus, these organisms are not a likely source for the vanA resistance determinants. Similar analysis of the resistance genes (vanH, vanA and vanX) from the glycopeptide-producing organisms Streptomyces toyocaensis and Amycolatopsis orientalis demonstrated approximately 60% amino acid identity in the deduced sequences of these genes compared with those in the vanA operon in BM4147 [130,131]. Southern hybridization of other glycopeptide-producing organisms, including A. coloradensis ssp. labeda (teicoplanin and avoparcin producer), demonstrated the ubiquitous presence of homologous clusters in these bacteria . In addition to the significant amino acid similarity noted above, the orientation and fine organization of the resistance genes in the glycopeptide producers are identical to those in Tn1546, suggesting a common ancestry. The evolutionary implications of these observations are significant with regard to the earlier report on antibiotic formulations contaminated with DNA from the organism used in the fermentative production . Webb and Davies examined a number of antibiotic preparations employed for animal and human use and found chromosomal DNA from producing organisms as well as identifiable macrolide, tetracycline and streptomycin resistance gene sequences. Some animal feed formulations contained nucleic acids in excess of 65 μg/g. Vancomycin preparations contained 16S ribosomal DNA from the glycopeptide-producing organism as shown by PCR amplification.
Taken together, these observations suggest glycopeptide preparations as a possible route for dissemination of glycopeptide resistance determinants in animal enterococci [130,131], in particular since antibiotics used as animal feed additives are poorly purified and since E. faecium is used as a probiotic. However, the sequence divergence makes a direct link between the glycopeptide resistance elements from the producers and those of enterococci unlikely. It does not rule out, however, ancient transfer and subsequent divergent evolution. Passage of the resistance determinants through a long chain of organisms that are related phylogenetically also represents a possible way for gene transfer from the producers to the enterococci. Candidate intermediates exist. Recently, the vanF glycopeptide resistance determinant of Paenibacillus popilliae has been described. The gene cluster has significantly higher sequence identity with the vanA operon than those of the glycopeptide-producing organisms [134,135]. Paenibacillus popilliae is a biopesticidal constituent used in the USA since the late 1930s, and the vanF gene was detected in a P. popilliae strain isolated in 1945. Furthermore, the percent amino acid identity (70%–80%) of the putative proteins encoded by the vanF operon with those from the vanA gene cluster suggests that certain links are still missing. The recent description of the vanG operon in human clinical GRE from Australia illustrates this point .
The US situation: Vancomycin as a human feed additive
The first human GRE in the USA were described in 1989 , and GRE were rapidly established as a widespread and frequent cause of nosocomial infections in the USA. From 1989 to 1993 the proportion of nosocomial enterococcal isolates resistant to vancomycin increased from 0.3% to 7.9%[136,137]. In 1999, over 60% of clinical E. faecium in the USA were reported to be resistant to vancomycin . Genotypic characterization revealed polyclonal [4,85,138–140] and clonal  nosocomial GRE outbreaks as well as evidence for interhospital transmission . The impressive spread of GRE within US hospitals contrasts with European hospitals where endemic nosocomial transfer of VanA-type GRE is still uncommon [15,16].
It is not clear how GRE were introduced into US hospitals. In contrast to Europe, no community GRE reservoir has been detected [63,86,142]. However, these studies have been limited in population size, and the direct plating technique used is less sensitive than selective enrichment broth [89,90]. Some observations suggest a community reservoir as a source of GRE in US hospitals. VRE colonization among patients within the first 24 h of admission to a hospital intensive care unit (ICU) in the USA has been described . Furthermore, analysis of the first 100 patients with GRE in New York hospitals revealed genomic heterogeneity among strains indicating multiple reservoirs . The latter observation is also consistent with an efficient transfer of the vanA gene cluster between different enterococci.
Avoparcin has not been licensed for use in animal husbandry in the USA. It has been argued that the impact of avoparcin-associated animal VanA-type GRE reservoirs on human GRE epidemiology is contradicted by the widespread occurrence of such strains in US hospitals where no community reservoirs have been detected [144,145]. The difference between GRE epidemiology in Europe and in the USA has been described as the American–European paradox . However, the extensive human use of vancomycin in US hospitals is in itself sufficient to explain the high proportion of GRE in this reservoir . Accordingly, the extensive use of avoparcin in European husbandry correlates with the large community reservoir of GRE . Vancomycin usage increased significantly throughout the 1980s in the USA, including a 10-fold increase in oral formulations from 1984 to 1990. The use of vancomycin in the USA is five times higher, based on population, as compared with the main European markets . For example, the amount of injectable vancomycin administered in a single 900-bed US hospital during 1991 (19 957 g) exceeded by 50% the amount of intravenous vancomycin used in Norway that year (13 344 g) [84,147,148]. Of note is the restricted use of oral vancomycin formulations in Europe, compared with its use in the USA. The impact of oral glycopeptides on the selection of fecal VanA-type GRE has been documented in human volunteers .
Introduction of the vanA gene cluster into a multiresistant enterococcal population, mostly E. faecium, might have contributed to the rapid increase in GRE infections and colonization among US hospital patients, in contrast to what has been observed in Europe. Characterization of 23 GRE isolates from the first 100 patients in New York City demonstrated ampicillin resistance in 22 isolates, as well as gentamicin and streptomycin high-level resistance (HLR) in 20 and 21 strains respectively . Similar resistance figures were detected in other early clinical GRE isolates from US hospitals . The recent European GRE study from 27 European countries revealed ampicillin resistance and gentamicin HLR in 51% and 23% of the E. faecium (n = 574), respectively, using NCCLS breakpoints for susceptibility testing . The 1999 data for ampicillin resistance and gentamicin HLR in clinical E. faecium isolates (n = 1263) in the USA were 80% and 70% respectively . ‘Pan-resistance’ (i.e. resistance to ampicillin, gentamicin, streptomycin and vancomycin) was found in 31% of clinical E. faecium isolates in the USA. Evidence for genetic linkage between transferable vanB glycopeptide and ampicillin resistance determinants has recently been reported for E. faecium. These observations support the hypothesis that coselection might have been an important factor in the emergence of nosocomial GRE infections in the USA.
Molecular analysis of the vanA operon in the first human VanA-type GRE strains from the north-eastern region of the USA indicated a dominant specific Tn1546 subtype with an IS1251 insertion in the vanS-vanH intergenic region . The IS1251 insertion seems to represent a unique genetic signature that can be used as a molecular marker in epidemiological studies and which has been shown to be present in other US VanA-type GRE strains [110,111,116]. Handwerger and colleagues demonstrated that the Tn1546::IS1251 gene cluster can reside on a mobile chromosomal element that was tentatively designated Tn5482. A role for Tn5482 in the spread of glycopeptide resistance in a New York City hospital has been suggested, although strict evidence for a chromosomal location was lacking . The Tn1546::IS1251 gene cluster was initially detected only in human isolates in the USA. However, identical IS1251-like insertions were recently documented in the vanS-vanH intergenic region in genomically diverse E. faecium isolates from Ireland (n = 2), Norway (n = 1) and the USA (n = 4), indicating trans-continental spread of glycopeptide resistance determinants . The Tn1546 elements in these strains all exhibited deletions in the orf1-orf2 region as well as a G→T transition at position 8234 in the vanX gene, which strongly suggest a common origin for the resistance determinants. To the best of our knowledge, the Tn1546::IS1251 gene cluster has not been detected in GRE from animals or food.
The Australian situation
Australian observations diverge from those made in the USA and most European countries and deserve specific comment. As in Europe, avoparcin has been widely used as a growth promoter for Australian farm animals [151,152]. The relative difference in the amounts of glycopeptides used in Australian farm animals and humans, respectively, is comparable with those described in Denmark . Between 1991 and 1993, >10 000 kg of avoparcin were used annually, compared with 193 kg of vancomycin used in the treatment of humans [151–153]. Until 1998, only 69 human GRE or clusters of strains have been described in Australian hospital patients [9,152]. The strains are polyclonal and widespread, with no clustering related to a specific health care institution. Apparently, the Australian situation resembles the GRE epidemiology in Europe. However, there seems to be interesting discrepancies. VanB was shown to be the most common (71%) clinical GRE type in Australia in 41 index isolates from 14 institutions in seven cities . Furthermore, a recent Australian study of fecal samples from 112 healthy volunteers in Melbourne revealed GRE colonization in only two (0.2%) specimens using Enterococcosel broth enrichment without vancomycin . Both samples contained VanB-type E. faecium. The apparent lack of VanA-type GRE colonization among healthy persons in the Australian community is in contrast to the results from several European countries also exposed to extensive use of avoparcin in animal farms. With regard to the already mentioned possibility of coselection of GRE through human macrolide usage, it is pertinent to stress that tylosin has been extensively used as an animal growth promoter in Australia . To our knowledge, neither the avoparcin-exposed farm environments nor commercial meat products have been examined in Australia for the presence of VanA-type GRE. Further speculation about VanA GRE epidemiology in Australia will be possible only when data concerning animal colonization and meat contamination have been collected.
Transfer of vanA resistance determinants in vitro and in vivo
When considering the spread of VanA-type GRE and glycopeptide resistance determinants between and within reservoirs, it is of considerable interest to examine the mobile genetic elements carrying the vanA gene clusters. The observation of identical or similar Tn1546 elements between genomically diverse enterococci from various reservoirs, in different enterococcal species, on several continents, as well as their heterospecific expression underlines the existence of efficient and promiscuous transfer systems for the vanA operon. The nonconjugative Tn1546 element, which is a member of the Tn3-like family of transposons, displays replicative transposition that allows intracellular mobility . Intercellular spread of Tn1546 has been associated with linkage to various plasmids [3,17,139,154,155] and chromosomal elements  that are self-transferable by conjugation.
Characterization of the genetic environment of Tn1546-elements has been performed in several studies. In their original characterization of Tn1546 in BM4147, Arthur and colleagues demonstrated heterogeneity in the insertion loci, suggesting that transposition plays an important role in dissemination of the vanA gene cluster . Southern hybridization of 35 Belgium GRE from 1989 demonstrated indistinguishable Tn1546-junction fragments in 20 of 35 (57%) isolates compared with pIP816, the vanA plasmid in BM4147, suggesting a wide spread dissemination of a single mobile element containing vanA. However, examination of various VanA-type GRE isolated from different countries between 1986 and 1992 by Miele and colleagues showed a polymorphism of considerable length in Tn1546 junction fragments, consistent with an early spread of the vanA gene cluster into various replicons . This observation emphasizes the importance of Tn1546 transposition into various conjugative elements as well as their horizontal dissemination in the global spread of VanA-type resistance. Comparison of sequences flanking Tn1546 elements in epidemiologically linked animal and human strains has not been performed to our knowledge.
Werner and colleagues have demonstrated large conjugative vanA plasmids (125–208 kbp) in German GRE that may appear as macrorestriction bands in PFGE, and accordingly, lead to the interpretation of a chromosomal localization of the vanA operon . Thus, hybridization of macrorestricted total DNA resolved by PFGE should be interpreted with caution. Evidence for a chromosomal localization of the vanA gene cluster should at least include linked hybridization to a housekeeping gene (i.e. rRNA gene).
The likelihood for intra- and interspecies transfer of vanA resistance determinants is dependent upon the host range and the efficiency of transfer of linked conjugative elements. Transfer frequencies, expressed in relation to the number of donor cells, vary from 10−4 to 10−8 between enterococci when using filter mating techniques [3,17,154]. The mobilization of Tn1546 by replicative transposition into enterococcal sex pheromone-response plasmids may increase conjugation frequencies considerably . In vitro transfer frequencies using filter-mating techniques may underestimate in vivo transfer potentials. Observations supporting this notion have been obtained in the gastrointestinal tract of gnotobiotic mice which demonstrated high frequency transfer of the vanA gene cluster between and within enterococci of human and animal origin [157 and unpublished results]. Interspecies transfer of the vanA gene cluster from enterococci to Streptococcus lactis, S. sanguis, S. pyogenes, L. monocytogenes and its heterospecific expression (Table 1) have been obtained under laboratory conditions . Detection of the vanA gene in clinical isolates of Cellulomonas (Oerskovia) turbata, Arcanobacterium (Corynebacterium) hemolyticum and Bacillus circulans confirm the relevance of these observations under natural conditions.
Transfer of the vanA operon to wild pathogenic bacteria such as S. aureus or S. pneumoniae could have catastrophic human effects. In vitro conjugative transfer of plasmids conferring antimicrobial resistance from E. faecalis to S. aureus, and vice versa, was demonstrated over 20 years ago [158,159]. The clinical importance of these observations is emphasized by the fact that enterococci and staphylococci share resistance determinants for various antimicrobial agents (reviewed in 160 and 161). The transfer of high-level vancomycin resistance from E. faecalis to S. aureus on agar plates as well as on mouse skin illustrates that this scenario may be just a matter of time and the presence of the vanA gene cluster on functional gene transfer-expression systems . GRE are isolated from the human perianal region, also a well-known habitat for S. aureus, when they are present in feces to levels above 10−4–10−5 CFU/g . And why are most S. aureus strains able to secrete a peptide that acts as a sex pheromone for E. faecalis strains carrying the plasmid pAM373 [162,163]? In order to answer this question, it would have been interesting to characterize the mobile genetic elements mediating transfer of high-level vancomycin resistance to S. aureus in the experiment by Noble et al. For obvious public health reasons such experiments should only be performed under strictly controlled laboratory conditions.
The avoparcin-associated VanA-type GRE reservoir and the potential impact on public health – evaluation of the existing evidence
Does the use of glycopeptides as animal growth promoters select for VanA-type GRE?
Several studies have provided strong arguments for a causal association between the use of avoparcin as an animal growth promoter and the occurrence of VanA-type GRE [53,54,57,59–61,68]. The time sequence and specificity criteria of this association were most convincingly demonstrated in the study by Bager et al. There was a significant decrease in glycopeptide resistance among animal enterococci after the ban of avoparcin as observed in poultry in Denmark . Although the rate of resistance declines after antibiotic suspension, observations at poultry farms exposed to avoparcin 3–4 years after the withdrawal of avoparcin suggest that VanA-type GRE may persist for many years [68,69].
Are GRE from animals able to colonize humans?
This state of affairs is very likely based upon the existence of clonally diverse VanA-type GRE colonizing the European community and the significant GRE reservoir observed in food animals as well as commercial meat products. The absence of host-specific molecular markers in enterococci precludes firm evidence for such transfer. However, the finding of GRE strains with similar or indistinguishable PFGE [74,75,100] or AFLP  patterns in animals and farmers that are closely related epidemiologically provides molecular evidence for this hypothesis. Berchieri’s heroic self-colonization experiment with animal GRE does not include a molecular analysis of strains and of their resistance determinants . However, the ability of VanA GRE strains from animal origins to colonize humans was recently confirmed in a Danish randomized double-blind study using molecular methods .
Do VanA resistance determinants from animal enterococci spread to human pathogens?
The observation of identical Tn1546 elements in genomically diverse animal and human clinical GRE illustrates the occurrence of a common vanA gene cluster reservoir . The detection of indistinguishable Tn1546 elements in GRE from animals and farmers that are closely related epidemiologically but that are genomically diverse, strongly suggests an interspecies transfer of the vanA gene cluster [74,75,100]. In vitro and in vivo transfer experiments between and within animal and human enterococci represent other lines of evidence supporting these observations [3,17,154,157]. The transfer from animals to humans seems to be the most likely in light of the extensive animal reservoir.
Do animal GRE cause human infections?
Das et al reported a case of community-acquired wound infection in a food-industry worker . The VanA-type E. faecalis isolated from the wound was most likely derived from the chicken carcasses in the factory and supports the possibility of transfer of infection to humans by direct or indirect contact. Again, the lack of host-specific molecular markers in enterococci precludes definitive evidence for an animal origin of this particular strain.
Does an animal pool of Tn1546 containing enterococci increase the probability of transfer of resistance to more virulent animal and human bacteria?
The global distribution of closely related Tn1546 elements in genomically diverse enterococci from various reservoirs, indicates the existence of highly efficient intraspecies-transfer systems for the vanA operon. Conjugation, which is the most likely mode of intercellular transmission of the vanA operon, requires physical contact between the donor and the recipient for transfer to occur. Transfer of the vanA gene cluster to more virulent genera such as staphylococci and streptococci has been demonstrated under laboratory conditions [17,18]. Human intestinal GRE colonization requires intimate contact with S. aureus in the perianal region. By reducing the pool of vanA gene clusters in human and animal enterococci the probability of Tn1546 transfer to wild pathogenic bacteria is also reduced.
The term ‘risk’ is often confused with probability and therefore used erroneously. Risk is defined as the probability that a certain event will take place, multiplied by the consequences arising if it happens. The consequences of high-level transferable glycopeptide resistance in S. aureus, S. pyogenes and S. pneumoniae can easily be imagined but would be awful to experience. The extensive prevalence of transferable vanA resistance determinants from enterococci in animal husbandry associated with the use of avoparcin has increased their availability for transfer to more virulent bacterial species.
The detection of substantial animal and human VanA-type GRE reservoirs within the community in Europe, and the elucidation of the most likely transmission lines between them has led to the implementation of measures to reduce the prevalence and spread of GRE. The EU ban on avoparcin usage was introduced to reduce human exposure to VanA-type GRE. Recent follow-up studies in Denmark, Germany and Italy have shown a reduction in the proportion of GRE among enterococci isolated from chicken and meat products, as well as the prevalence of fecal carriage among healthy humans in the community [65–67]. Thus, the EU ban seems to have had a certain impact. However, the prevalence of animals in farms exposed to avoparcin excreting VanA-type GRE has not declined, indicating a long-term reservoir of glycopeptide resistance determinants to which humans have been, and are still, regularly exposed [68,69]. This observation makes clear the stability of antibiotics in the environment and the sustained selection they exert [165–168]. This aspect of selection has not been extensively studied, although it is important as the use of antibiotics in animal feed is measured by tons and not by kilograms.
If we do not learn from history we are bound to repeat it. The core lesson of the GRE story is to encourage a policy of antibiotic usage that delays dissemination of resistant bacteria and resistance genes. The widespread occurrence of GRE in US hospitals is consistent with the nonprudent use of antibiotics in treating humans. The emergence of human VanA-type GRE in the community illustrates how the use of antimicrobial growth promoters in farm animals may influence human carriage of resistant bacteria. The scientific basis for the use of antibiotics in animal feed and its potential public health impact are still debated [144,169–172]. The Swann report  and the World Health Organization recommendations  accept the use of any antimicrobial agent for growth promotion in animals provided it is not used in human therapy or known to select for cross-resistance to antimicrobials used in human medicine.
During the last 30 years enormous progress has been made in the understanding of the genetics and biochemistry of resistance. Antimicrobial resistance is not a class phenomenon for two reasons: (1) the extension of the concept of cross-resistance to various drug classes; cross-resistance can occur, for example, by overlapping targets of antibiotics as shown by the decrease of activities of structurally unrelated macrolides, lincosamides and streptogramins (MLS) following synthesis of a ribosomal methylase . Drug efflux has recently been recognized as a common resistance mechanism in both Gram-positive and Gram-negative bacteria (reviewed in 175). The broad substrate transporters account for diminished susceptibility to β-lactams, aminoglycosides, tetracyclines, chloramphenicol, trimethoprim, MLS, sulfonamides, fluoroquinolones and others. (2) The increased occurrence of genetically linked and coexpressed resistance determinants: this is best illustrated by the integrons , first described in Gram-negative  and now also in Gram-positive bacteria . Genetic linkage and coexpression implies that the use of any antibiotic that is a substrate for one resistance mechanism will coselect for resistance to the others and thus maintain the entire gene set. With regard to transmission from animals to humans, it is pertinent that the genetic basis for multidrug-resistance (ampicillin, chloramphenicol, streptomycin, spectinomycin, sulfonamides, tetracycline) in Salmonella enterica serotype typhimurium DT104 is mediated by two linked integron structures [179–181].
Thus, there is no monospecific association between one class of antibiotics and a specific resistance mechanism. Since cross-resistance means cross-selection and linkage between resistance genes implies coselection, this prohibits de facto the use of any antimicrobial agent as a growth promoter. This constitutes a timely challenge to incorporate this concept for decision-making in issues concerning public health.
Arnfinn Sundsfjord was the recipient of a Senior International Fellowship (132736/300) from the Norwegian Research Council.