Introduction

Poultry and poultry products, including meat and eggs, have long been recognised as an important source of food-borne infections caused by Salmonella enterica (1,2). The global increase in human infections with serovar Enteritidis observed in the late 1980's and early 1990's (3) was almost entirely attributable to the presence of this organism within the poultry production industry worldwide. However, the implementation of national monitoring programmes, together with control measures including vaccination, has resulted in recent years in a reduction in cases of human salmonellosis associated with the consumption of poultry and egg products in the UK (4).
The Scottish Salmonella Reference Laboratory (SSRL) receives all strains of S. enterica isolated from cases of human infection from hospital laboratories. Regional veterinary laboratories, public analysts and water authorities submit isolates of animal, food and environmental origin. All isolates are fully serotyped, phage typed (where applicable), and tested for resistance to 15 antimi-
crobial agents. A selection of isolates is further typed by plasmid profile analysis and pulsed-field gel electrophoresis. Results are reported to the Scottish Centre for Infection & Environmental Health (SCIEH).
In the early summer 2002, a number of isolates of S. enterica serotype Java (S. Java) were examined as part of an exercise to build up a database of plasmid profile and pulsed-field gel electrophoresis (PFGE) data for the top ten serotypes isolated from humans in Scotland at that time (see table 1). Although numbers of isolations of this serotype were relatively small, a small increase was apparent (figure 1). Considerable diversity was observed in plasmid profile, antibiogram, and PFGE pattern, but some isolates from sporadic human cases were found to possess a PFGE pattern indistinguishable from that in some strains of Java isolated from poultry meat. This observation prompted further investigation of other human and poultry isolates of S. Java held in the culture collection at SSRL, and monitoring of all new isolations of this serovar.

Table 1
The top ten ranking serotypes of S. enterica isolated from human infections in Scotland in 2001-02

Laboratory studies

A total of 59 strains of S. Java were included in the study. Thirty isolates originated from 29 human cases between June 1994 and November 2002. All human isolates from 2001 and 2002 were examined while clones isolated previously were selected based on resistance to antibiotics. The remaining 29 isolates were all isolated from poultry meat or poultry skin. Poultry isolates were submitted to SSRL by commercial poultry companies after recovery during routine in-house sampling. Although information on these isolates is incomplete, it is thought that thirteen of them had been recovered from poultry meat originating in the Netherlands, and sampled on 6 different occasions between July 1997 and September 2002.
Plasmid profiling was carried out as previously described (2). PFGE was carried out using a standardised protocol which has been implemented in a number of European reference laboratories under the European Union funded Salm-gene project (see article pp46-50). This was done to allow ease of comparison with other countries. Briefly, chromosomal DNA plugs were digested with the restriction endonuclease XbaI, and subjected to electrophoresis using a CHEF DR-II® apparatus (BioRad, USA). Running conditions were 1% agarose in 0.5x TBE, 6 V/cm for 22 hours, initial pulse 2 seconds, final pulse 64 seconds. Analysis of PFGE patterns was carried out using Phoretix 1-D Advanced® (v 5.0) software (Nonlinear Dynamics, England). Antimicrobial resistance was determined by growth on solid media containing antibiotics at breakpoint concentrations (ampicillin 50 mg/ml, cefotaxime 1mg/ml, chloramphenicol 20mg/ml, ciprofloxacin (high level) 0.5mg/ml, ciprofloxacin (low level) 0.12 mg/ml, furazolidone 20mg/ml, gentamicin 20mg/ml, kanamycin 20mg/ml, nalidixic acid 40mg/ml, netilmicin 20mg/ml, spectinomycin 100mg/ml, streptomycin 20mg/ml, sulphamethoxazole 100mg/ml, tetracycline 10mg/ml, trimethoprim 2mg/ml).

Results and discussion

When comparing typing data for isolates of S. Java, the initial impression was that we were dealing with a highly diverse organism (see table 2). All but 13 of the isolates examined for plasmid content contained at least one plasmid and up to five different plasmids. It was also noted that isolates recovered from poultry meat sampled at the same time did not all possess the same plasmid profile. These isolates contained between 3 and 5 plasmids in different combinations. While these differences may at first seem to complicate matters, it should be noted that variation in plasmid profile in salmonellae with close epidemiological connections has been reported previously (2,5).
The results from antibiotic resistance screening also revealed wide variability. All isolates from poultry products were multiresistant (resistant to 3 or more antibiotics) while 11 human isolates were fully sensitive, one was resistant to two antibiotics and 18 were multiresistant. Antimicrobial resistance data from the SSRL database on all human isolates of S. Java confirmed that the number of multiresistant strains has been on the increase since the mid-1990s (figure 1).

Table 2
Summary of typing results for isolates of S. enterica Java recovered from humans and poultry products in Scotland

Pulsed-field gel electrophoresis has been demonstrated to be a valuable tool in the investigation of outbreaks of salmonellosis, and in the identification of sources of infection. Digestion of chromosomal DNA with XbaI provided a good range of fragment sizes and easily discernible patterns under these running conditions (figure 2). Comparison of fragment patterns by the Dice coefficient and using UPGMA (unweighted pair-group method using arithmetic averages) clustering generated two major clusters (Figure 3). Cluster A comprised 10 of the 29 human isolates examined, the earliest from 1999, and 28 of the 30 poultry isolates. All isolates from poultry imported from Holland belonged to this cluster. PFGE patterns within this cluster were relatively homogeneous, 30 isolates had an identical fragment pattern (designated JavX1) with the remaining variants differing by a small number of bands. Cluster B comprised 19 human isolates and 2 poultry isolates of undetermined origin. This cluster was much more heterogeneous with only 3 patterns being represented by more than a single isolate.

It has recently been reported that a particular clone of S. Java has become predominant in poultry production in
Germany in the latter half of the 1990s (6), and this can be distinguished by its XbaI PFGE profile (X8). Preliminary comparison strongly suggested that this pattern was the same as the JavX1 pattern reported in this study. Moreover, a similar degree of variability in plasmid profile and antibiogram was described in the German study, and the authors had identified isolates from Belgium and Holland as belonging to the same clone.
Seventeen of the thirty isolates recovered from poultry were recovered from meat of unknown origin. The latest available report on the isolation of salmonella from livestock in the United Kingdom records only a single isolation of S. Java from cattle, sheep, pigs or poultry, including the statutory monitoring of breeding flocks and hatcheries, in the period from the beginning of 1997 to the end of 2001 (7). This isolation was made from ducks or geese in 2000. From the same report, no isolations of S. Java were made from animal feedstuffs during 2000 or 2001. It would therefore seem unlikely that the poultry meat infected with S. Java originated in the United Kingdom.
The potential for poultry meat to act as a vehicle for multiresistant strains is a matter of concern for public health. In recent years, the number of cases of human infection caused by multiresistant Java strains with the type A PFGE profile has increased in Scotland (one in 1999, one in 2000, two in 2001 and five in 2002). In particular, the presence of resistance to quinolone antimicrobials, at high as well as low levels of resistance, is important. Fluoroquinolone antibiotics such as ciprofloxacin are the drugs of choice in cases of invasive salmonellosis in humans. Cross-resistance between quinolones and fluoroquinolones is well documented (8). In a recent Danish study, an increased mortality rate was observed in patients in a two-year period following infection with S. Typhimurium when the infecting strain was resistant to quinolones (9). From the Netherlands, it has been reported that levels of infection with S. Java have increased in poultry from 2% of all isolates prior to 1996 to 40% in 2001, and the clone responsible has been demonstrated as the X8/JavX1 PFGE type (10). Resistance to flumequin, a quinolone antibiotic currently licensed for use in livestock in the EU, has increased in S. Java in Holland from 3% between 1996-99 to 20% between 2000-02.
Plasmid profiling and antibiogram typing have previously given valuable information on the source of salmonella infections. However, in this case PFGE proved to be invaluable in building an association between cases of human infection and isolates from poultry meat. The evidence from PFGE analysis, together with the epidemiological information available for human infections, and the origins of poultry meat, strongly implicate imported poultry meat as an important source of S. Java infections in humans in Scotland. Despite the recent reports on the high levels of S. Java infections in poultry flocks in Germany and the Netherlands (6,10), no reports of a similar increase in human cases have appeared and there have been suggestions that the current poultry associated clone may be of reduced virulence in man. In this study, we demonstrate that infection in humans, although relatively rare, does occur. The same multiresistant clone found in poultry in Germany and the Netherlands has been responsible for a significant proportion of human cases of S. Java infection in Scotland, particularly since 2000. All new cases of S. Java in the human population in Scotland will continue to be monitored.
The application of standardised protocols to facilitate the investigation of international outbreaks has previously been advocated (11). This approach was invaluable in allowing the direct comparison of PFGE typing results with those of colleagues in Germany and the Netherlands.

Acknowledgements