ABSTRACT
Sporadic and epidemiologically linked Yersinia enterocolitica strains (n = 379) isolated from fecal samples from human patients, tonsil or fecal samples from pigs collected at slaughterhouses, and pork samples collected at meat stores were genotyped using multiple-locus variable-number tandem-repeat analysis (MLVA) with six loci, i.e., V2A, V4, V5, V6, V7, and V9. In total, 312 different MLVA types were found. Similar types were detected (i) in fecal samples collected from human patients over 2 to 3 consecutive years, (ii) in samples from humans and pigs, and (iii) in samples from pigs that originated from the same farms. Among porcine strains, we found farm-specific MLVA profiles. Variations in the numbers of tandem repeats from one to four for variable-number tandem-repeat (VNTR) loci V2A, V5, V6, and V7 were observed within a farm. MLVA was applicable for serotypes O:3, O:5,27, and O:9 and appeared to be a highly discriminating tool for distinguishing sporadic and outbreak-related strains. With long-term use, interpretation of the results became more challenging due to variations in more-discriminating loci, as was observed for strains originating from pig farms. Additionally, we encountered unexpectedly short V2A VNTR fragments and sequenced them. According to the sequencing results, updated guidelines for interpreting V2A VNTR results were prepared.
INTRODUCTION
Yersinia enterocolitica is a food-borne pathogen that causes human yersiniosis. Yersiniosis is the third most commonly reported food-borne zoonosis (1) and often occurs in young children (2). The most frequent symptom is gastroenteritis with diarrhea, but additional sequelae such as reactive arthritis and erythema nodosum may occur, especially in adults. Y. enterocolitica is a heterogeneous species that is divided into several biotypes and serotypes (2). Bioserotypes 1B/O:8, 2/O:5,27, 2/O:9, 3/O:3, and 4/O:3 are associated commonly with human disease. Of these, bioserotype 4/O:3 strains have been responsible for yersiniosis cases in Europe, the United States, Canada, and Japan (3–6). Recently, bioserotype 2/O:9 infections have been on the rise (7–9). Besides humans, bioserotype 4/O:3 frequently is isolated from samples of pig origin (2, 10–13).
Pulsed-field gel electrophoresis (PFGE) has been one of the most commonly used genotyping methods for epidemiological studies of Y. enterocolitica (14–17), but more-sophisticated and automated methods are needed and recently have been applied. Multiple-locus variable-number tandem-repeat analysis (MLVA) has been developed for several pathogenic bacterial species (18, 19). The method is based on finding variable-number tandem repeats (VNTRs) in the genome and multiplying these fragments in a PCR. Different genotypes are distinguished by the number of repeats in each locus (18). The application developed for Y. enterocolitica 4/O:3 uses six VNTR loci, i.e., V2A, V4, V5, V6, V7, and V9 (20). MLVA also has proved successful with Y. enterocolitica serotype O:9 (7, 8, 21, 22). A separate MLVA application has been developed for Y. enterocolitica biotype 1A (23).
MLVA has proved to be a promising tool for Y. enterocolitica outbreak investigations, due to its high discriminatory power (7). However, the number of strains studied has been limited. Therefore, MLVA should undergo further testing with several sporadic and epidemiologically linked strains, to define its suitability for practical use. In this work, we used a previously developed MLVA typing method to investigate a collection of Y. enterocolitica strains from both sporadic and epidemiologically linked origins that had been acquired over 12 years, between 1995 and 2007. The strains originated from Finland, Germany, England, and Russia and were isolated from human patients as well as from pig farms, meat stores, and slaughterhouses. The discriminatory power, advantages, limitations, interpretation, and use of MLVA in epidemiological studies of Y. enterocolitica were evaluated. Among strains from the same epidemiological origin, we commonly observed instability in loci V2A, V5, V6, and V7. Lastly, our observation of some unexpectedly short V2A PCR products was further studied.
MATERIALS AND METHODS
Y. enterocolitica strains and isolation of DNA.For this study, we chose a total of 379 strains of Y. enterocolitica from Finland (n = 288), Germany (n = 46), England (n = 34), and Russia (n = 11). The strains were isolated between 1995 and 2007 and originated from clinical samples from human patients (n = 150), tonsil and fecal samples from pigs collected at slaughterhouses (n = 183), and pork samples collected at meat stores (n = 46). Of these, 128 samples collected at slaughterhouses were traced to have originated from 35 pig farms. We included both sporadic (n = 278) and epidemiologically linked (n = 101) strains. The latter strains were considered epidemiologically linked when they originated from pigs from the same farm. The strains belonged to serotypes O:3 (n = 363), O:9 (n = 9), and O:5,27 (n = 7). We previously divided all serotype O:3 strains into 81 genotypes using PFGE with the enzyme NotI, and we divided 206 of those strains into 105 genotypes using the additional enzymes ApaI and XhoI (16, 17, 24–26). The strains were grown on tryptic soy agar (Difco, Lawrence, KS) at 28°C for 18 to 24 h, and the DNA was extracted using guanidium thiocyanate (27).
Multiple-locus variable-number tandem-repeat analysis.The MLVA was based on the method described by Gierczynski et al. (20), with modification of the PCR run conditions (7). The forward primers for VNTR loci V2A, V4, V5, V6, V7, and V9 were fluorescently labeled with 6FAM, VIC, NED, and PET dyes (Applied Biosystems, Foster City, CA) and divided into two PCRs. One microliter of the template was added to the PCR mixture containing a 0.2 mM deoxynucleoside triphosphate mixture (Thermo Fisher Scientific, Vantaa, Finland) and 0.5 U of DyNAzyme II polymerase (Thermo Fisher Scientific), in a total reaction volume of 25 μl. A 1.5-μl volume of 10-fold diluted PCR products was denatured with 10 μl of HiDi formamide (Applied Biosystems) mixed with 0.3 μl of the GeneScan 500 LIZ internal standard (Applied Biosystems), boiled for 3 min, and placed on ice. The sizes of the PCR products were determined by capillary electrophoresis with an ABI Prism 310 genetic analyzer (Applied Biosystems). An injection time of 5 s and an injection voltage of 15 kV were used for electrophoresis. The run time was 28 min, the voltage was 15 kV, and the run temperature was set at 60°C. Certain V2A VNTR fragments of unexpected size (less than 263 bp) were sequenced using a BigDye Terminator v3.1 cycle sequencing kit (Applied Biosystems) with an ABI 3730xl DNA analyzer (Applied Biosystems).
Data analysis.The MLVA data were collected with GeneScan software (Applied Biosystems) and analyzed with BioNumerics v. 5.1 (Applied Maths NV, Sint-Martens-Latem, Belgium). The tandem-repeat number for each locus was saved as character data type in the BioNumerics software. The clustering of the MLVA results was calculated using Euclidian distances. To compare the discriminatory abilities of the VNTR loci, Simpson's discriminatory indexes were calculated according to the method described by Hunter and Gaston (28). The chi-square test was used to compare MLVA and PFGE results.
RESULTS
Discrimination of strains.Among the 379 strains examined, we detected 312 different MLVA types. A dendrogram of the results appears in Appendix S1 in the supplemental material. After excluding strains with the same epidemiological source, we detected 262 different MLVA types among 278 strains. The Simpson discriminatory index for the sporadic strains was 0.999. All of the MLVA types in the present study were discovered to be country specific, as none of the types was detected in several countries.
Groups of similar MLVA types were detected among strains originating from the same source, i.e., humans or pigs (see Appendix S1 in the supplemental material). Within several MLVA profiles, the PFGE results obtained by using three enzymes aggregated accordingly. However, a few strains with similar MLVA genotypes revealed diverse PFGE results. Interestingly, we observed in the MLVA results a significant (P < 0.05) association between PFGE type NA/AA/HA and repeat numbers of two in locus V4 and three in locus V9 (see Appendix S1 in the supplemental material). Finally, MLVA showed higher discriminatory power than the standard method PFGE, even with three enzymes. Among the 206 serotype O:3 strains that were divided into 105 genotypes using PFGE with the enzymes NotI, ApaI, and XhoI, we detected a total of 193 different MLVA types.
Typing of human and porcine strains.We observed similar MLVA types between strains isolated from (i) human fecal samples, (ii) human patients and pigs, and (iii) pigs originating from the same farm but also from two farms. The same MLVA types were found in human clinical isolates from Finland. Evidence of epidemics limited to certain provinces was observed between 1996 and 1999 (Table 1). Six groups of closely related MLVA types that included 19 human Y. enterocolitica cases were limited to one province. Additionally, four groups of similar MLVA profiles that were detected over a period of 3 years were found in two or three provinces within the country. Among strains from humans and pigs, similar MLVA results were found for human fecal samples and samples originating from pigs (Table 2).
Similar MLVA results for Yersinia enterocolitica strains isolated from clinical fecal samples from human patients in certain provinces of Finland
Similar MLVA results for Yersinia enterocolitica strains isolated from pig and human sources
A farm-specific MLVA type was found for 17 farms (49% of all farms studied). Two separate farm-specific MLVA profiles were discovered for 13 (37%) farms, whereas three, four, and five different MLVA profiles were found for two farms, three farms, and one farm, respectively. Interestingly, all English pig farms that were included in this study had more than one MLVA type. Similar MLVA types were found for the pig farms GB2 and GB4, GB4 and GB5, FI5 and FI6, FI12 and FI13, FI17 and FI18, and RU1 and RU2, which were located in England, Finland, and Russia (Table 3). Vast variations in distances were observed for the geographical positions of the farms with similar MLVA types, ranging from 10 km between farms FI5 and FI6 to 300 km between farms RU1 and RU2. Samples from pigs that originated from farms FI5 and FI6 and also from farms FI12 and FI13 were collected at the same time and at the same slaughterhouse, whereas samples originating from the other farms were collected separately.
Similar MLVA results detected on two farms
Variations in V2A, V5, V6, and V7 VNTR loci.Within the MLVA results for strains that originated from the same pig farms, variations in tandem-repeat numbers were noted commonly for the loci V2A, V5, V6, and V7. Among the strains from 15 farms, we observed variations in one to three of these loci (Table 4). However, such variations were not observed for loci V4 and V9. The discriminatory index values for tandem-repeat loci V2A, V4, V5, V6, V7, and V9 were 91.1%, 53.7%, 83.3%, 84.9%, 82.1%, and 46.0%, respectively. For one strain, the PCR product for VNTR locus V7 primers remained absent, despite several repeated reactions.
Variations detected for different VNTR loci of Yersinia enterocolitica strains originating from the same pig farms
Unexpectedly short V2A VNTR fragments.For 100 strains, the V2A VNTR fragments obtained were shorter than expected, being less than 263 bp. However, the lengths of these exceptionally short fragments varied in 6-bp intervals, with results of 232, 238, 244, 250, 256, and 262 bp. The short V2A fragments of all lengths were sequenced in order to clarify our results. The real repeat numbers of these fragments, as confirmed by sequencing, were 2, 3, 4, 5, 6, and 7 when the VNTR fragment lengths were 232, 238, 244, 250, 256, and 262 bp, respectively (Table 5). These results indicate an updated interpretation of the results for the V2A VNTR locus.
Distribution of numbers of tandem repeats in the V2A VNTR locus
DISCUSSION
We typed a large number of Y. enterocolitica strains from human and porcine sources and from four different countries by using MLVA with six VNTR loci. The method showed a high discriminatory capacity, since 312 different MLVA types were found among 379 Y. enterocolitica strains. MLVA found 45 genotypes among 62 Y. enterocolitica 4/O:3 strains in a previous study (20) and 77 types among 88 Y. enterocolitica strains in another study (7). When strains with the same epidemiological sources were excluded, we detected 262 MLVA types among 278 strains, whereas 21 Y. enterocolitica strains with no epidemiological link were previously discriminated into 20 genotypes using MLVA (20).
Interestingly, with randomly chosen clinical strains isolated from human fecal samples in Finland, we observed similar MLVA types over a period of several years. Some of these strains were isolated from the same provinces, indicating unidentified geographically limited epidemics, while others were distributed in different provinces within the country but still seemed to be cases with apparently common sources. Most Y. enterocolitica infections are sporadic, and outbreaks are considered uncommon (2). Epidemics caused by Y. enterocolitica seem to exist, however, although many of them have remained unidentified. The occurrence of the same MLVA types in clinical samples collected over several years indicates long-term persistence and the availability of the same infectious sources.
Similar genotypes of strains that were isolated from humans and pigs in corresponding time periods were discovered with MLVA, suggesting a link between these two sources. Strains isolated from humans and porcine sources showed similar PFGE types previously (16, 17). As MLVA has markedly greater discriminatory capacity than PFGE, our results confirm that edible pig offal is an important source in the transmission of Y. enterocolitica between pigs and humans. However, some strains that originated from human clinical samples were isolated before the corresponding MLVA types were obtained from porcine samples. As certain Y. enterocolitica strains seem to persist on pig farms (29), the common MLVA types found among pigs and humans likely existed on the farms before the isolation from human clinical samples included in this study.
Among strains from pig origins, similar MLVA types were detected mainly from pigs originating from the same farms. However, some MLVA types were found on several farms. Y. enterocolitica strains can be transmitted from one farm to another with infected pigs (29). After pigs from different origins are mixed at the farms, different MLVA types are incorporated and spread rapidly in the pig population within each unit. Previous contacts and animal sales between the farms might have enabled the spread of similar MLVA types at these farms. Interestingly, similar MLVA types found at several farms were isolated only within one country. Of these, two MLVA profiles were obtained from samples collected on the same day from the same slaughterhouse, and the sampled pigs might have been exposed to contamination at the slaughterhouse instead of being infected at the farms.
Among the six VNTR loci used in this study, the discriminatory power of VNTR locus V2A was the highest, whereas VNTR loci V4 and V9 showed the lowest discrimination. This finding is in agreement with those of a previous study (7). Regional differences in the discriminatory abilities of different loci seem to exist, as among Chinese strains; locus V5 exhibited the highest discriminatory power, and the least variation occurred in loci V4 and V2A (22). Besides serotypes O:3 and O:9, MLVA proved successful for serotype O:5,27 strains in this study. The 16 strains of serotypes O:5,27 and O:9 were divided into 10 MLVA types. MLVA grouped these strains according to their serotypes and farms of origin. All serotype O:5,27 and O:9 strains in this study originated from samples from pigs from England. These serotypes are particularly common among English pigs, whereas serotype O:3 predominates among pigs from other European countries (13).
Among MLVA results obtained from the same pig farms, variations were detected commonly for four (V2A, V5, V6, and V7) of the six VNTR loci. Allelic variation of a single VNTR locus was reported recently for Mycoplasma pneumoniae (30). We detected such variations in several VNTR loci in Y. enterocolitica. Both insertions and deletions of tandem repeats were observed. Variations were evident in strains isolated from pigs originating from the same farms. Since the same strains are assumed to persist on farms (29), it is unsurprising that mutations occur in these strains over time and lead to variations in VNTRs. The generated variations in one or more VNTRs might cause the MLVA results for identical clones to appear different (19). In addition, environmental stress can affect the stability of tandem repeats (31). Most of the mutations in VNTR units were shown to be single-repeat differences (32), but multiple-repeat mutations also occur. Of these, two-repeat mutations are more common than three-repeat mutations, which are more common than four-repeat mutations. However, we found no variations in the farm-specific results for VNTR loci V4 and V9, which have VNTR unit lengths of 7 and 12 bp, respectively. These less-discriminating loci thus are longer than the other loci, all of which have a VNTR unit length of six base pairs. It seems that the more-discriminating loci occasionally might be even hypervariable and might be too discriminating for long-term use. This should be taken into consideration when interpreting results, especially those obtained from pig farms. Differences in the VNTR numbers of one or several loci can, in the long run, occur within strains from a single epidemiological origin such as a pig farm.
Interestingly, PFGE type NA/AA/HA (26) was significantly associated with MLVA results showing repeat numbers of two and three for VNTR loci V4 and V9, respectively (see Appendix S1 in the supplemental material). This was the most clearly observed congruity between the results of these two methods in the present study. The strains with this genotype originated from Finland, England, and Russia and were of both human and pig origins. All of these strains belonged to bioserotype 4/O:3.
Compared with the amplicon lengths and corresponding tandem-repeat numbers reported in previous studies (7, 20), V2A primers produced PCR products that were unexpectedly short, as their lengths were only 232 to 262 bp. The minimal length of a V2A fragment has been reported to be 263 bp (20). However, the sequencing results of our study showed that these 232- to 262-bp fragments contained two to seven VNTR units. According to the previously published Y. enterocolitica subsp. palearctica 4/O:3 Y11 genome (33), we also noticed that the actual number of repeats was nine when the size of the V2A amplicon was 274 bp. In previous studies, the repeat number of nine equaled lengths of 282 and 310 bp for V2A fragments (7, 20). Based on this finding, we prepared an updated table to interpret the results for the V2A VNTR locus (Table 5).
Performing MLVA was fast. With the protocol used, the MLVA results were ready within 24 h. MLVA results can be analyzed with automated software, which reduces the possibilities for human error. MLVA results can be compared between laboratories more easily than obsolescent gel images, but this requires interlaboratory consistency in MLVA typing and universal guidelines for interpretation of the results. However, precise guidelines are difficult to develop due to the varying durations of outbreaks and the different environments surrounding the pathogens (34). The MLVA method has high discriminatory power, evidently higher than that of PFGE, as several previous studies have noted (7, 22, 35). MLVA appears to be a practical tool for genotyping Y. enterocolitica, especially in short-term epidemiological studies. Among strains that originate from pigs, MLVA distinguishes the strains according to their farms of origin, but variations in loci V2A, V5, V6, and V7 are common. Reliable comparisons of results between laboratories require uniform interpretation of the results.
ACKNOWLEDGMENTS
This study was performed at the Centre of Excellence in Microbial Food Safety Research, Academy of Finland (grants 118602 and 141140), and was partly funded by the Ministry of Agriculture and Forestry of Finland (grant 2849/502/2008).
Maarit Penttilä, Erika Pitkänen, and Anu Seppänen are acknowledged for their technical assistance.
We declare that we have no competing interests.
FOOTNOTES
- Received 15 March 2013.
- Returned for modification 12 April 2013.
- Accepted 22 April 2013.
- Accepted manuscript posted online 1 May 2013.
Supplemental material for this article may be found at http://dx.doi.org/10.1128/JCM.00710-13.
- Copyright © 2013, American Society for Microbiology. All Rights Reserved.