G.S. Simonsen, Department of Medical Microbiology, University and University Hospital of Tromsø, Tromsø, Norway.
Avoparcin was used as a feed additive in Norwegian broiler and turkey production from 1986 until 1995. It was banned due to the selection of VanA-type vancomycin-resistant enterococci (VRE) in animal husbandry and to reduce the potential for human exposure to VRE. The aim of the present study was to investigate the prevalence of VRE carriage in Norwegian poultry farmers and their poultry three years after avoparcin was banned. Corresponding faecal samples from poultry and humans on farms where avoparcin had previously been used (exposed farms, n = 73) and farms where avoparcin had never been used (unexposed farms, n = 74) were analysed for the presence of VRE. For each farm, one sample from the poultry house and one sample from the farmer were obtained. VRE were isolated from 72 of 73 (99%) and eight of 74 (11%) poultry samples from exposed and unexposed farms, respectively. VRE were isolated from 13 of 73 (18%) and one of 74 (1%) farmer samples from exposed and unexposed farms, respectively. All VRE isolates were highly resistant to vancomycin and possessed the vanA gene, as shown by PCR. The high prevalence of VRE is in accordance with previous Norwegian studies, and shows a remarkable stability of the VanA resistance determinant in an apparently non-selective environment.
Vancomycin-resistant enterococci (VRE) were first detected in the UK and France in 1986 ( Leclercq et al. 1988 ; Uttley et al. 1988 ) and have become an important cause of nosocomial infections worldwide. From Norwegian hospitals, however, only sporadic cases of VRE infection or colonization have been reported ( Harthug et al. 1998 ).
In 1993, Bates reported isolation of VanA-type VRE from non-human sources in the UK including faeces from various farm animals and sewage, suggesting a community VRE reservoir ( Bates et al. 1993 ). Furthermore, a German study documented the occurrence of VanA-type resistant Enterococcus faecium in the waste-water from sewage treatment plants ( Klare et al. 1993 ). In 1995, results from Danish and German studies showed faecal VRE colonization in pigs and poultry as well as in humans in the community, suggesting that the use of avoparcin as a feed additive selected for the occurrence of VanA-type VRE ( Aarestrup 1995; Klare et al. 1995a , b).
Avoparcin, a glycopeptide cross-resistant to vancomycin, has been extensively used as a growth-promoting feed additive in many countries excluding the USA and Canada ( Donnelly et al. 1996 ). Avoparcin was first licensed in the European Community (EC) in 1975 ( Donnelly et al. 1996 ). In 1986, avoparcin was introduced for use in the Norwegian poultry industry for growth promotion, and to prevent necrotic enteritis among broiler chickens and turkeys. In order to reduce the human exposure to VanA-type VRE, the use of avoparcin was temporarily banned in Norway and Denmark in 1995, and in Germany in 1996. In 1997, the use of avoparcin was suspended in all EU countries.
Transmission of VRE of animal origin to humans through the food chain has been proposed as the most likely connection between animal VRE reservoirs and humans in the community. Contact between animals and humans at avoparcin-exposed farms has also been shown to be associated with human VRE colonization ( Kruse and Rørvik 1996; Van den Bogaard et al. 1997 ; Simonsen et al. 1998a ; Stobberingh and Van den Bogaard 1999). Recent data indicate a decrease in the proportion of VRE among enterococci isolated from broiler chickens in Denmark ( Bager et al. 1999 ) and a reduced prevalence of VRE in poultry meat samples and among humans in Germany ( Klare et al. 1999 ) after the selective pressure of avoparcin was removed. The aim of the present study was to investigate the prevalence of VRE carriage in Norwegian poultry farmers and their poultry three years after avoparcin was banned, and to compare the results with previous Norwegian VRE studies ( Simonsen et al. 1998b ; Kruse et al. 1999 ). VRE isolates were subjected to susceptibility testing of relevant human clinical antimicrobials and antimicrobial feed additives previously used in Norwegian poultry production. Limited molecular analysis was performed to evaluate possible transmission of VRE strains and/or VanA resistance determinants between human and animal VRE reservoirs.
Materials and methods
Selection of poultry farms
All producers of broiler chickens, turkeys, layer chickens or pullets (young hen, parent of layer chickens) delivering poultry products to five different distributors in the eastern part of Norway (n = 390) were invited to take part in the study. One hundred and fifty-four producers volunteered to participate, and 147 were included in the study after submitting one faecal sample from a poultry flock on the farm and one faecal sample from a person in daily contact with the poultry. The exposed group comprised 73 farms producing broiler chickens and/or turkeys where avoparcin had previously been used. The unexposed controls were 74 farms producing layer chickens or pullets where avoparcin had never been used. The collection of human samples was approved by The Regional Committee for Medical Research Ethics, University of Tromsø, Norway. Sampling was conducted during the summer of 1998, three years after avoparcin was banned in Norway.
Collection of faecal samples
The participating farmers received sterile plastic tubes and a written instruction on how to perform the faecal sample collection. Fresh faecal material collected from the centre of the floor or from underneath the pens in the poultry house was placed in a plastic tube. A stool sample from the participating person was placed in another plastic tube. The human and poultry samples were sent by mail without delay to the University Hospital of Tromsø and the National Veterinary Institute in Oslo, respectively.
All samples were frozen at − 70 °C prior to analysis. The isolation of VRE from poultry samples was performed using a direct method only. A 5 g portion from each faecal sample was dissolved in 45 ml peptone water (Difco). From the suspended material, 100 µl were plated onto Slanetz and Bartley's (S & B) enterococcus agar plates (Oxoid) ( Slanetz and Bartley 1957) with and without vancomycin 50 mg l−1 (Sigma). S & B plates without vancomycin were included to confirm growth of enterococci in the samples. After incubation at 37 °C for 48 h, typical enterococcal colonies from the vancomycin-containing agar plates were subcultured on blood agar plates (Difco). After overnight incubation at 37 °C, one colony from each positive sample was purified for further analysis.
Isolation of VRE from human samples
The isolation of VRE from human samples was performed using enrichment broth. Faecal material was solubilized in sterile 0·9% NaCl, mixed with equal volumes of brain heart infusion (BHI) broth (Oxoid) with 20% glycerol, and frozen at – 70 °C. At the time of analysis, 10 µl were inoculated into Enterococcosel enrichment broth (Beckton Dickinson) with natriumazid 0·15 g l−1 (Merck) and incubated at 37 °C for 48 h. Samples with a positive bile-esculine reaction were subsequently plated onto Cephalexin Arabinose Aztreonam agar plates without (CAA) and with vancomycin 8 mg l−1 (Abbot) (CAAV) ( Ford et al. 1994 ). CAA plates were included to confirm growth of enterococci in the samples. From each CAAV plate with growth, one colony was subcultured for further analysis.
Presumptive identification of Enterococcus spp. was based on typical colony morphology and Gram stain, absence of catalase production and a positive pyrrolidonyl arylamidase test (Difco). Further identification to species level was performed by the RAPID ID32 Strep test kit (bioMérieux, Marcy l'Etoile, France) and assays for pigment production (bioMérieux) and motility (Difco). The species identities of Ent. faecium, Ent. faecalis, Ent. gallinarum and Ent. casseliflavus were confirmed by species-specific PCRs targeting the ddl genes, the vanC1 gene and the vanC2/3 gene, respectively ( Dutka Malen et al. 1995 ; Patel et al. 1997 ). Isolates identified by the RAPID ID32 Strep test kit as Ent. durans/Ent. hirae were separated into sucrose-fermenting Ent. hirae and non-fermenting Ent. durans ( Facklam and Collins 1989). Quality control strains for species identification are listed in Table 1.
Table 1. Reference strains used for quality control purposes
Specificity; enterococcosel enrichment broth
Specificity; enterococcosel enrichment broth
Sensitivity; CAA agar and S & B agar, ddl PCR, antimicrobial susceptibility testing
The susceptibility to vancomycin was determined to confirm high-level vancomycin resistance. Bacitracin susceptibility was examined because zincbacitracin was the only antimicrobial feed additive used in Norwegian poultry production (mainly for layer chickens) after the avoparcin ban was introduced. Finally, ampicillin and aminoglycoside susceptibilities were evaluated, as these are the drugs most commonly used for the treatment of serious human enterococcal infections in Norway.
Minimal inhibitory concentrations (MIC) of vancomycin, ampicillin and bacitracin were determined using the Etest (AB Biodisk, Solna, Sweden) according to the instructions of the manufacturer. The isolates were categorized as vancomycin susceptible (MIC ≤ 4 mg l−1), intermediately susceptible (MIC 8 or 16 mg l−1) or resistant (MIC ≥ 32 mg l−1), and ampicillin susceptible (MIC ≤ 8 mg l−1) or resistant (MIC ≥ 16 mg l−1), according to National Committee of Clinical Laboratory Standards (NCCLS) guidelines (M100-S9, January 1999). No NCCLS-defined breakpoints are available for bacitracin, thus, MIC values are reported without categorization. All isolates were screened for the presence of high level resistance (HLR) to aminoglycosides using NeoSensitabs™ discs containing gentamicin 250 mg, kanamycin 500 mg and streptomycin 500 mg, according to the instructions of the manufacturer (Rosco, Taastrup, Denmark). The resistance genotype of vancomycin-resistant isolates was confirmed by vanA PCR ( Simonsen et al. 1998a ). Quality control strains for antimicrobial susceptibility testing are listed in Table 1.
Molecular characterization of VRE isolates
Limited molecular analysis was performed to evaluate possible transmission of VRE and/or VanA resistance determinants between human and animal VRE reservoirs. Corresponding human and poultry VRE isolates were analysed by pulsed-field gel electrophoresis (PFGE) as described by Dahl ( Dahl et al. 1999a ), with the addition of mutanolysin 40 U ml−1 (Sigma) in the lysis step and proteinase K 50 µg ml−1 (Promega) to eliminate nuclease activity. PFGE patterns were compared visually according to the criteria of Tenover ( Tenover et al. 1995 ). Restriction analysis of vanX amplicons using DdeI (New England Biolabs, Beverly, MA, USA) to differentiate between G and T at position 8·234, was performed on all isolates ( Jensen 1998).
The relative risk (RR) with Taylor series 95% confidence interval (CI) was calculated using Epi Info version 6·04b (CDC, Atlanta, GA, USA and WHO Geneva, Switzerland). P<0·05 was considered significant.
VRE in faecal samples from poultry and farmers
The results are summarized in Table 2. VRE were isolated from 72 of 73 (99%) poultry faecal samples from exposed farms, and from eight of 74 (11%) poultry faecal samples from unexposed farms (RR: 9·1, 95% CI 4·7–17·7). Analysis of human faecal samples revealed VRE carriage in 13 of 73 (18%) farmers on exposed farms, and in one of 74 (1%) farmers on unexposed farms (RR: 13·2, 95% CI 1·8–98·2). For all exposed farms with VRE-positive farmers, VRE were isolated from the poultry samples as well. For one unexposed farm, the farmer was colonized with VRE while no VRE were isolated from the poultry sample. As only one VRE isolate from each positive sample was further analysed, the total number of VRE included 80 poultry- and 14 farmer-derived isolates.
Table 2. Number of VRE-positive and VRE-negative corresponding faecal samples from poultry and farmers on 147 Norwegian poultry farms
Among the poultry VRE isolates, 67 were identified as Ent. faecium (84%), nine as Ent. hirae (11%), two as Ent. durans (2·5%) and two as Ent. casseliflavus (2·5%). Nine of 14 human VRE isolates were identified as Ent. faecium (64%) and the remaining five as Ent. hirae (36%). For the 13 farms where VRE were isolated from corresponding poultry and human samples, all poultry isolates were identified as Ent. faecium whereas the human isolates consisted of nine Ent. faecium and four Ent. hirae.
Antimicrobial susceptibility testing and vanA PCR
All VRE isolates expressed high-level vancomycin resistance (MIC ≥ 256 mg l−1) and possessed the vanA gene as shown by PCR. The isolates were susceptible to ampicillin and did not express high-level resistance to aminoglycosides. The bacitracin MIC range was from 0·38 mg l−1 to ≥256 mg l−1 with a median value of 4 mg l−1, and 25% of the isolates had MIC ≥256 mg l−1.
Related PFGE patterns were not detected in the 13 pairs of human and poultry isolates from the same farms ( Fig. 1). All vanX amplicons from both poultry and human VRE isolates were of the ‘poultry type’, with G at position 8234 of the vanA gene cluster ( Fig. 2).
This study demonstrates a high prevalence of VRE on Norwegian poultry farms three years after the use of avoparcin was banned. These findings are consistent with the results of a study of Norwegian poultry conducted during 1995–97 ( Kruse et al. 1999 ). Thus, there has been no detectable decrease in the prevalence of VRE in poultry, despite the discontinued use of avoparcin since June 1995. These results are in contrast to a recent study which indicated a decreased prevalence of VRE in Danish poultry after avoparcin was banned, as the proportion of Ent. faecium isolates resistant to vancomycin was reduced from >80% in 1995 to <5% in 1998 ( Bager et al. 1999 ). However, the results in the Danish study are not directly comparable with the present data. The methodology used for monitoring antimicrobial resistance in Denmark (DANMAP) does not select for VRE, and only a single, randomly selected enterococcal isolate from each farm is subjected to antimicrobial susceptibility testing ( Aarestrup et al. 1998 ). Klare et al. in Germany isolated VRE from eight of 31 (26%) poultry meat samples at the end of 1997 ( Klare et al. 1999 ), suggesting a decreased prevalence of VRE compared with the results from 1994 when all of 11 poultry meat samples analysed were VRE-positive. Again, divergent conclusions are reached but the results cannot be compared directly due to differences in the materials studied and the methodology used.
The continuing high prevalence of VRE on Norwegian poultry farms previously using avoparcin is surprising. The poultry industry is based on an ‘all-in all-out’ production, and to preclude direct transmission of bacteria from flock to flock, the houses are supposed to be cleaned, disinfected and left empty for 10–30 days before a new flock of chickens is introduced. However, recently presented data show that VRE survive in the Norwegian poultry houses between each cycle and subsequently colonize the chickens after arrival on the farm ( Borgen et al. 2000 ).
Possible mechanisms for the persistence of resistance determinants include direct selective pressure, or physical linkage of resistance elements to genes encoding other traits that are selected for in the environment ( Marshall et al. 1990 ; Salyers and Amabile-Cuevas 1997 and references therein). Residuals of avoparcin or unidentified glycopeptide analogue producers might be present in the farm environment and represent a selective pressure. Linked selection by use of other antimicrobial feed additives seems unlikely. Avoparcin was the only antimicrobial feed additive used in Norway in substantial quantities between 1986 and 1995, and alternative feed additives such as virginiamycin and tylosin did not replace avoparcin as they did in other European countries until the EU introduced further restrictions in 1999. A small amount of zincbacitracin was used in Norwegian poultry production after avoparcin was banned in June 1995, 64 and 27 kg active substance in 1996 and 1997, respectively (Norwegian Agricultural Inspection Service). The observed median bacitracin value of 4 mg l−1 as well as the wide susceptibility range do not support the hypothesis of co-linked selection, but may rather reflect the natural bacitracin MIC range among enterococci. However, ionophore coccidiostats have been extensively used in the Norwegian poultry industry both before and after avoparcin was banned. Ionophore feed additives for poultry have a simultaneous antimicrobial and coccidiostatic effect, but little is known about resistance mechanisms and co-selective capability of these compounds ( Anon. 1997). Resistance to metallic ions or disinfectants might select for VRE, and large conjugative elements harbouring the vanA gene cluster could carry genes encoding advantageous properties for survival in the farm environment ( Salyers and Amabile-Cuevas 1997). Recent studies in one of our laboratories have shown a high in vitro stability of plasmid-located VanA resistance determinants in serial transfer studies without selection pressure, as well as high in vivo transfer frequencies of the vanA gene cluster ( Dahl et al. 1999b ; Johnsen et al. 1999 ). These findings are consistent with the observed persistence of VanA resistance determinants in the farm environment.
Limited data are available on the prevalence of VRE carriage among humans in avoparcin exposed vs non-exposed agricultural communities. Kruse et al. ( 1999) reported six VRE carriers out of 10 avoparcin-exposed farmers, and no carriers among 16 unexposed farmers, in Norway using a direct selective plating method. Van den Bogaard et al. reported VRE carriage rates of 39% among 47 farmers and 20% among 48 slaughterers exposed to avoparcin-fed turkeys in the Netherlands using a VRE enrichment procedure ( Van den Bogaard et al. 1997 ). The present study confirms this strong association between work at avoparcin-exposed farms and VRE carriage, with a carrier rate of 18% among 73 poultry farmers working on farms previously using avoparcin.
European studies of healthy volunteers in the community have shown VRE carrier rates in the range of 2–14% ( Endtz et al. 1997 ; Van den Bogaard et al. 1997 ). VRE colonization among humans in the community has not been investigated in Norway, but a survey of 616 Norwegian hospital patients in 1997 using the same enrichment method as in the present study did not reveal any VanA-type VRE carriers ( Simonsen et al. 1998b ). The apparent lack of VRE carriers outside the agricultural community in Norway as opposed to Germany ( Klare et al. 1995a ) and the Netherlands ( Endtz et al. 1997 ) may be explained by the extent of previous avoparcin usage. Different food handling practices in these countries may also have an influence on the total number of VRE carriers. Alternatively, the extensive use of feed additives such as tylosin and virginiamycin in continental Europe may have created co-linkage between the vanA gene cluster and resistance determinants to antibiotics commonly used by humans.
Whether humans on previously avoparcin-exposed farms contract enterococcal vancomycin resistance from their poultry, or they constitute a separate VRE reservoir as a direct result of avoparcin exposure, is difficult to discern. No VRE carriers were detected among 19 Norwegian animal feed mill workers exposed to avoparcin (data not shown). Dutch ( Van den Bogaard et al. 1997 ; Stobberingh et al. 1999 ) and Norwegian ( Simonsen et al. 1998a ) studies have reported identical or highly similar PFGE patterns, and identical structure of the vanA-possessing Tn1546 transposon, in VRE isolates from animals and humans on the same farms. In the present study, no PFGE pattern-related isolates were detected from corresponding human and animal samples. However, as only a single isolate from each sample was analysed, the possibility of VRE transmission between animals and humans cannot be ruled out. Identical vanX amplicon digests with the so-called poultry nucleotide G at position 8234 in all isolates in this study suggest vanA gene cluster transfer between human and poultry reservoirs.
So far, the Norwegian agricultural VRE reservoir has not been associated with enterococcal vancomycin resistance in Norwegian human clinical isolates. This may be explained by the low prevalence of infections with methicillin-resistant Staphylococcus aureus (MRSA) and the subsequent low usage of vancomycin in Norwegian hospitals. The general susceptibility of Norwegian agricultural VRE to ampicillin and aminoglycosides, which are the drugs of choice for invasive enterococcal infections in humans, should be noted. As the number of MRSA infections recorded by the Norwegian Nationwide System for Notification of Infectious Diseases is now increasing, the significance of the agricultural VRE reservoir may change. Further studies of this reservoir are warranted both with regard to public health concerns and the elucidation of molecular mechanisms of the persistence of VanA resistance determinants in an apparently non-selective environment.
The authors thank all the participating poultry farmers and their organizations for valuable collaboration and help in the collection of samples, Gerhard Schaller for practical organization of sampling, and Bjørg Haldorsen and Hanne Tharaldsen for excellent laboratory assistance. This study was funded by the Norwegian Board of Health, the Norwegian Research Council, and the Odd Berg Foundation.