Yersinia enterocolitica (YE) is an enteropathogenic bacterium transmitted via food or water and may cause sporadic infections as well as foodborne outbreaks of yersiniosis [1–5]. The symptoms of yersiniosis range from mild diarrhea to severe clinical manifestations and post-infectious complications such as reactive arthritis, myocarditis, glomerulonephritis, and erythema nodosum [6]. Y. enterocolitica can be divided into six biotypes, of which biotypes 1B and 2-5 are known to be pathogenic to humans.

At present, pulsed-field gel electrophoresis (PFGE) is commonly used to discriminate between Y. enterocolitica strains. However, there are no standard PFGE procedures or databases similar to those, e.g., for Escherichia coli O157:H7, Salmonella, and Shigella standardized by PulseNet [7]. Most of the restriction enzymes used in PFGE for Y. enterocolitica produce patterns with a high number of bands that are not ideal for analysis. Furthermore, the global homogeneity of the pulsotypes among Y. enterocolitica 4/O:3 is high and different pulsotypes often display only minor differences [8–11]. However, the discriminatory power of PFGE has been improved by using more than one restriction enzyme [12].

Most bacterial genomes contain repeats of DNA sequences called "variable-number tandem repeats" (VNTR). These VNTR regions can be applied in the PCR-based subtyping of strains by multilocus variable-number tandem-repeat analysis (MLVA). MLVA is increasingly used for typing, surveillance and epidemiological investigations of pathogenic bacteria [13]. A study investigating the development of an MLVA subtyping method to be used for Y. enterocolitica 4/O:3, based on six loci, was reported recently [14].

Although yersiniosis is seldom treated with antimicrobials, medication may be required, for example in the case of immuno-compromised patients. Y. enterocolitica is a known ß-lactamase producer and thus is resistant to ß-lactam antibiotics such as ampicillin, carbenicillin, penicillin, and first-generation cephalosporins [15–20]. In recent studies done in Switzerland, the USA, Germany, and Austria, Y. enterocolitica strains have shown high susceptibility to antimicrobials other than ß-lactams [21–24]. However, multiresistant Y. enterocolitica strains have also been reported, e.g., from Spain and Brazil [16, 25, 26]. The antimicrobial resistance of Y. enterocolitica has not been monitored regularly in Finland although the surveillance of antimicrobial resistance would be useful for epidemiological studies. Over 20 years ago, 186 Finnish Y. enterocolitica strains were studied and found to be resistant only to ampicillin and susceptible to ceftriaxone, tetracycline, sulpha-trimethoprim, and ciprofloxacin [27].

The aim of the present study was to determine how MLVA using fluorescently labeled primers and fragment analysis compares to PFGE in its discriminatory power with regard to the sporadic and outbreak-related strains of YE bio/serotypes 4/O:3. We included traditional antimicrobial susceptibility testing in our study to see whether it provides additional information for the genotypic analysis concerning, e.g., the geographical source of infection. We therefore used MLVA and PFGE to type 104 sporadic and outbreak-associated Y. enterocolitica strains and determined the sensitivity of the strains to 12 antimicrobial agents. In addition, we studied the genetic basis of the antimicrobial resistance that was detected.

In our study, MLVA typing using fluorescently labeled primers and fragment analysis was shown to be a high-resolution discriminatory method for epidemiological investigations of Y. enterocolitica. In the present study, the discriminatory power of MLVA was 99.9% while that of NotI PFGE was 87.9%. Our results were in agreement to those obtained by Gierczyński and colleagues [14] who demonstrated that the used MLVA markers are highly discriminatory and added the evidence that this method can successfully be applied for the outbreak strains of Y. enterocolitica ssp. palearctica biotypes 2 and 4. In the present study, only the VNTR loci V2A, V4 and V5 were found in six BT 1A strains tested with the MLVA method (data not shown). Another MLVA method designed using Y. enterocolitica ssp. enterocolitica strain 8081 whole genome and with four loci was introduced recently [28]. The method showed potential for the epidemiological investigation for YE biotype 1A strains with DI of 87% and worked also for six tested BT 2 and BT4 strains [28]. The discriminatory power of PFGE can be improved by using more than one restriction enzyme. For instance, the discriminatory index of 74% achieved with NotI PFGE was increased to 93% by using further characterization with ApaI and XhoI enzymes of 128 YE 4/O:3 strains [29]. However, both the time required and the costs of PFGE rise considerably when several restriction enzymes are used. The amount of working time needed for the PFGE protocol with one enzyme is two to three days, MLVA using fragment analysis can be done in one day.

In December 2003, authorities from the city of Kotka, Finland reported an outbreak of gastroenteritis. Investigations revealed that it was caused by Y. enterocolitica 4/O:3 [30]. Approximately 30 people fell ill; 12 patients had culture-confirmed, multiresistant YE 4/O:3 infection. Three of them had appendectomies before the disease was recognized as yersiniosis. Most of the patients had abdominal pain (94%), fever (78%), and diarrhea (72%). Most of the patients had eaten in the same cafeteria in the Port of Kotka between November 25 and December 15, 2003. However, the contaminated food source was not discovered, as the food served in that cafeteria was no longer available for microbiological investigations. In addition, a cohort study among cafeteria users did not show a significant association between any food and illness. During a microbiological sampling of the cafeteria's kitchen a month later, in January 2004, hygienists noticed some shortcomings in food handling and hygiene practices that increased the possibility of cross-contamination in the cafeteria. While no YE 4/O:3 strains were found in the specimens collected from the cafeteria, YE biotype 1A strains were isolated from iceberg lettuce imported from Spain and from domestic carrots. Unfortunately, the antimicrobial susceptibilities of these strains are not known. At the time of the outbreak in Kotka, there were around 20 confirmed YE 4/O:3 cases in other locations in Finland, mainly in the Turku area. The cases were suspected to be linked with the larger outbreak, but no epidemiological evidence for this was found.

MLVA played a key role in confirming that the cases which occurred in the city of Kotka in 2003 belonged to a single outbreak: 12 isolates representing the Kotka outbreak were clonal by MLVA, and differed distinctly from those of epidemiologically unrelated strains that shared the same PFGE pulsotype. Another suspicion of outbreak was refuted by MLVA: six 1-year-old children had been infected in 2006 by YE 4/O:3 strains that shared the same PFGE pulsotype (5NotI_ye a). Interviews, however, revealed no epidemiological connection between the cases. All of these strains which shared the same PFGE pulsotype were found to be of different types in MLVA. We also detected some evidence that the MLVA method can be as useful with YE 2/O:9 outbreaks as it was with YE 4/O:3. In a household outbreak in 2009, a mother and two children had YE 2/O:9 strains found to be identical in MLVA (data not shown here). MLVA also identified identical YE 2/O:9 strains in a school/day care center outbreak that occurred in Finland in 2010 (data not shown here).

Support was obtained for genetic stability among sporadic cases, since two MLVA-typed strains were isolated twice from the same patient at intervals of 7 or 19 days. In both cases, the MLVA and PFGE types remained identical. Similar observations of the stability of the Y. enterocolitica MLVA markers' loci in vivo had also been reported earlier [14]. Genetic events will eventually alter the MLVA patterns, but the rate of alteration is not known. However, previous studies confirmed that the MLVA type remained the same after as many as 20 serial passages of colony plating [14].

Our previous case-control study revealed that travel abroad was a risk factor for Y. enterocolitica infection in Finland [31]. In the present study, we found a statistically significant association between the antimicrobial multiresistance of YE strains and travel. The results indicate that a considerable number of multiresistant Y. enterocolitica infections are actually imported, and that domestic Y. enterocolitica strains in Finland are rather susceptible to antimicrobials. For instance, all of the nalidixic acid-resistant strains were isolated from patients who had been infected while on vacation in Spain or Brazil, countries where multiresistant Y. enterocolitica strains have been described previously [16, 25, 26]. The multiresistant strains belonged to certain PFGE pulsotypes, which were not found among susceptible strains. This is perhaps due to the DNA of the resistance plasmid. The MLVA types were so varied that no hint of the origin of the strains could be obtained on that basis.

In the outbreak that occurred in Kotka, the patients had not been abroad before falling ill. However, the antimicrobial multiresistance of the outbreak strain nevertheless suggests that the strain originated from abroad. Spanish iceberg lettuce, at least, had been used in the cafeteria. In 2005 Salmonella enterica serotype Typhimurium, with a resistance profile identical to that detected now for the Y. enterocolitica outbreak strain, was isolated in an outbreak situation in Finland and traced to iceberg lettuce imported from Spain [32].

The resistance of Y. enterocolitica to NAL is based on point mutations in the fluoroquinolone resistance-determining regions of gyrA [26, 33]. In our study, the strains resistant to NAL had amino acid changes stemming from point mutations in the gyrA gene: i.e., either Ser83Arg, Asp87Tyr, or Asp87Asn. Two of these mutations are identical to those reported previously for fluoroquinolone-resistant Y. enterocolitica strains [33].

Conjugation experiments confirmed that in Y. enterocolitica, the antibiotic resistance to CHL, STR, and SUL, at least, is encoded on a large conjugative plasmid and can easily be transferred to a susceptible Y. enterocolitica strain. Conjugative plasmids that carry antibiotic resistance genes have been isolated from a variety of clinical strains, but reports of this for Y. enterocolitica are rare. Hundreds of different antibiotic resistance cassettes have been identified as residing on mobile resistance integrons [34]; owing to the cassette nature of the resistance genes, they can easily change the resistance repertoire. In fact, one of the outbreak strains in our study had altered antimicrobial resistance and lacked resistance to TET. A study on the persistence of TET-resistant E. coli in colonic microbiota observed that three out of 13 strains lost TET resistance during intestinal colonization [35].

MLVA was less labor-intensive than PFGE and the results were easier to analyze, especially because they were independent of subjective interpretation. PFGE can still be useful for surveillance of the sources and transmission routes of sporadic Y. enterocolitica strains in future. However, for outbreak investigations, MLVA offers a powerful tool for the discrimination of Y. enterocolitica strains. More sporadic and outbreak Y. enterocolitica strains should be subjected to MLVA typing in order to determine whether this method could be considered for use as a new gold standard for outbreak investigations of Y. enterocolitica. This study revealed that multiresistant Y. enterocolitica strains do appear in Finland, but that the multiresistance was mainly associated with travel. All three nalidixic acid resistant strains were associated with travel to Spain or Brazil. Interestingly, all outbreak strains studied here were also multiresistant. Thus, traditional susceptibility testing provides additional information useful for genetic typing methods in epidemiological investigations.