Diversity of Yersinia enterocolitica isolated from pigs in a French slaughterhouse over 2 years

Abstract The pig is one of the main reservoirs of Yersinia enterocolitica strains pathogenic to humans. A description of the Y. enterocolitica population in this reservoir, and accurate discriminatory techniques for typing isolates are needed for prevention, outbreak investigation, and surveillance. This study investigates the genetic diversity of pathogenic Y. enterocolitica isolates obtained from pig tonsils in a French pig slaughterhouse in 2009 (S1) and 2010 (S2). The use of Pulsed‐Field Gel Electrophoresis (PFGE) and MLVA as typing techniques was also compared and evaluated. First, a total of 167 isolates (12 of biotype 3 recovered during S1, and 155 of biotype 4 recovered during S1 and S2) were typed by PFGE using the XbaI enzyme. MLVA was then tested on all the biotype 3 isolates in addition to 70 selected biotype 4 isolates recovered over the 2 years. PFGE generated two specific XbaI‐PFGE profiles for biotype 3 isolates. Nine XbaI profiles were obtained for biotype 4, with a higher diversity (ID = 0.599) than biotype 3 (ID = 0.167). Two out of the nine XbaI profiles were reported during both surveys and at different months. MLVA improved the differentiation between isolates; the index of diversity reached 0.621 and 0.958, respectively, for biotype 3 (three MLVA types) and biotype 4 (32 MLVA types). The MLVA types for biotype 4 differed over the two surveys, but some isolates with different MLVA types were genetically closely related. This study provides an initial evaluation of the genetic diversity of Y. enterocolitica strains isolated from pigs in France. We show that some PFGE profiles are maintained in the pig production sector, and, through MLVA, that part of the Y. enterocolitica population remained genetically close over the two years. MLVA proved its effectiveness as a tool for investigating pathogenic Y. enterocolitica strains isolated from pigs.

Biotyping is used to evaluate the level of pathogenicity of Y. enterocolitica strains isolated from pigs, but it is insufficient to describe the diversity of pathogenic strains. It is necessary to characterize these strains more precisely within each biotype for the purposes of risk prevention, efficient outbreak investigation, and surveillance.
Different molecular typing techniques have been developed to more accurately describe Y. enterocolitica populations of porcine or human origin. The most commonly used of these techniques is Pulsed-Field Gel Electrophoresis (PFGE). Many studies have used this typing technique, and various restriction enzymes have been tested to generate digestion profiles. Many studies, for example, have used NotI and XbaI as restriction enzymes (Bonardi et al., 2014;Falcão, Falcão, Pitondo-Silva, Malaspina, & Brocchi, 2006;Fredriksson-Ahomaa, Stolle, & Stephan, 2007). However, limited diversity among biotypes 4 and 3 was observed at pig slaughterhouse level even between strains of different geographical origins, like Germany, Finland, New Zealand, and China (Fredriksson-Ahomaa et al., 2003;Gilpin et al., 2014;Liang et al., 2012). Another increasingly popular technique used to type Y. enterocolitica is the Multi-locus variable number tandem repeat analysis (MLVA). Several studies have reported a higher discriminatory power with this technique (Gierczynski, Golubov, Neubauer, Pham, & Rakin, 2007;Sihvonen et al., 2011).
In our study, we tested two typing techniques-PFGE and MLVA-on a collection of Y. enterocolitica isolates of porcine origin isolated in a French slaughterhouse during two consecutive years.
The aim was to evaluate the effectiveness of these two typing techniques to assess the diversity of the isolates and to evaluate their variation within the slaughterhouse over two years. Our study is the first one to describe the Y. enterocolitica population by both PFGE and MLVA in different years at the same slaughterhouse.

| Yersinia enterocolitica isolates
The Y. enterocolitica isolates considered in this study were collected during two surveys in the same French pig slaughterhouse. The latter is one of the largest slaughterhouses in France, with more than 1.5 million pigs slaughtered per year. The pigs slaughtered there have been supplied by the same farmers for many years.
The first survey (S1) was held from January to March 2009 (Fondrevez et al., 2010). This survey found 132 positive pigs (14.6%) out of the 900 pigs sampled in this slaughterhouse. The second survey (S2) was held from March 2010 to February 2011 (Fondrevez et al., 2014). This survey found 33 positive pigs (16.5%) out of the 200 pigs sampled in this slaughterhouse. The 200 pigs were sampled over 3 months (March, August, and November 2010) during this 1year survey (S2). We checked that the percentage of positive pigs in this slaughterhouse did not significantly change between the two surveys (χ 2 test, p = 0.511).
All the isolates were from pig tonsil swabs using the same bacteriological method as described previously (Fondrevez et al., 2014).
Biochemical assays were used to biotype the Y. enterocolitica isolates and were carried out as per the ISO10273, 2003 method. One isolate of each biotype detected from a given swab was selected for this study. Consequently, a total of 167 isolates were selected for further typing analysis.

| Pulsed-field gel electrophoresis
Pulsed-Field Gel Electrophoresis was carried out on all 167 Y. enterocolitica isolates included in this study. The Salmonella Braenderup H9812 strain was used as a reference size marker strain to allow comparison of the PFGE profiles from different gels.
Strains were sub-cultured on Plate Count Agar (PCA) at 30°C for 24 hr. The culture was suspended in TE buffer (0.01 M Tris-EDTA buffer, pH 8.0) and adjusted to an optical density (600 nm) of 1.5.
This suspension was then mixed with 1% agarose to make the plugs, which were incubated for 48 hr at 50°C in a lysis solution (Na 2 EDTA 0.5 M, pH9, N-lauryl-Sarcosyl 1%, proteinase K 1 mg/ml), and finally washed five times with TE buffer.
DNA was thereafter digested with 40 U of XbaI restriction enzyme (Roche, Boulogne-Billancourt, France) for 4 hr at 37°C. The electrophoresis conditions had an initial switch time of 1.5 s, with a final switch time of 18.0 s for 25 hr at 6.6 V. The gels were stained with GelRed TM Nucleic Acid (Biotium) and the restriction fragments were visualized under ultraviolet light. Tiff images from the GEL Doc 1000 Imaging System (Bio-Rad, Hercules, CA, USA) were imported into the ANSES laboratory's database on the PFGE patterns of Y. enterocolitica strains. Electrophoretic patterns were compared using BioNumerics® (Applied Maths, Sint-Martens-Latem, Belgium; version 7.6). Similarities between profiles were determined by calculating the Dice correlation coefficient-based on band positions-with a maximum position tolerance of 1% on the active zones (8.5%-96.5%). A dendrogram was constructed in order to reflect the similarities between strains in the matrix. Strains were clustered by the unweighted pair group method using the arithmetic mean (UPGMA) (Struelens, 1996). Simpson's index was determined as described by Hunter and Gaston (1988) to assess population diversity.
The two PCR products of each strain were diluted to 1/100 in sterile water, and run separately using capillary electrophoresis with an ABI 3130 DNA analyzer (Applied Biosystems, Foster City, CA, USA) with D (DS-30) fragment analysis chemistry according to the manufacturer's instructions. The Geneflo™ 625 ROX labeled (EurX, Gdańsk, Poland) was used as an internal size standard. Electrophoretic patterns were analyzed using BioNumerics 7.6 software (Applied Maths). To determine the correspondence between the allele size measured and the number of repeats, representative amplicons for all six VNTRs used in this study were subjected to sequence analysis using a BigDye Terminator v3.1 cycle sequencing kit (Applied Biosystems) with an ABI 3130 DNA analyzer (Applied Biosystems). An MLVA type was given for each combination of the six VNTRs and attributed to the isolates (Supporting Information Table S1). Simpson's index (ID) was determined to assess the diversity of the populations and the discriminatory power of each VNTR locus. A standard minimum spanning tree generated under BioNumerics 7.6 using the single and double locus variance priority rules was used to visualize the relationships between biotype 4 isolates.

| Comparison of the methods
The methods were compared using 80 isolates, including the 12 biotype 3 isolates (S1 survey), the 33 biotype 4 isolates obtained in 2010 (S2 survey) and 37 out of the 122 biotype 4 isolates obtained in 2009 (S1 survey). To avoid a sampling effect, the 37 biotype 4 isolates out of the 122 isolates available were selected according to two criteria: their XbaI-PFGE profiles and their prevalence in the population recovered in 2009. We validated this sampling because the population represented by the 37 isolates was not significantly different from the population represented by the 122 isolates (Fisher's test, p = 0.179).
Pulsed-Field Gel Electrophoresis and MLVA were compared according to two criteria: their discriminatory power estimated using Simpson's index (ID), and their concordance. This congruence was assessed using the adjusted Rand index (AR) which measures the overall agreement between two typing techniques (Hubert & Arabie, 1985) and the adjusted Wallace index (AW) which assesses the directional agreement by separately evaluating the concordance when each technique is used first (Severiano, Pinto, Ramirez, & Carrico, 2011). The adjusted Rand and the adjusted Wallace coefficients were determined using website https://www.comparingpartitions.

| Distribution of the isolates
The 167 isolates were distributed among two biotypes, with biotype 4 being the most prevalent (92.81% of the isolates). The other 12 isolates (7.18%) belonged to biotype 3, and were only recovered from the S1 survey. Interestingly, out of the 132 pigs detected positive for Y. enterocolitica, two carried both a biotype 4 and a biotype 3 isolate. The other 130 pigs were detected positive for only one biotype, either biotype 4 or biotype 3. No biotype 3 isolates were recovered from the S2 survey carried out on 200 pigs. This distribution of the biotypes from isolates in both surveys was not significantly different (Fisher's test; p = 0.126).

| Diversity using PFGE
A total of 11 XbaI-PFGE profiles were obtained from the 167 isolates and were coded X01-X11 (Table 1) . These profiles were only observed for this biotype and were genetically distant from biotype 4 on the dendrogram (<64.5% of genetic similarity; Figure 1). In addition, the X10 profile was very different from the X01 profile (59.5% of genetic similarity).
The biotype 4 isolates were closely related, sharing at least 88.7% of genetic similarity on the dendrogram ( Figure 1). Of the nine XbaI-PFGE profiles, X03 was the most prevalent (particularly during S1;

| D ISCUSS I ON
This study is the first to describe the Y. enterocolitica population among pigs slaughtered at the same slaughterhouse in two consecutive years and using two typing techniques. It highlights the relevance of the typing technique for monitoring the variation of this population over time.
Most of the isolates considered in this study were of biotype 4 (92.81%). The predominance of biotype 4, which is pathogenic to humans, is in accordance with many previous studies on pigs (Bonardi et al., 2014;Martinez et al., 2011;Martins et al., 2018;Rahikainen Ibañez et al., 2016;Van Damme, Habib, & Zutter, 2010). Biotypes 2 and 3 were also found in the slaughtered pigs examined. These biotypes are generally less common, with a prevalence below 10% (Fredriksson-Ahomaa et al., 2007;Martinez et al., 2011;Poljak et al., 2010). Although no biotype 2 isolates were detected in our study, some biotype 3 isolates were recovered during survey S1 in 2009. Statistical analysis indicated that F I G U R E 2 Distribution of the biotype 4 isolates (n = 70) in a minimum spanning tree according to their MLVA types (M01 to M35). To visualize the relationships between isolates, a standard minimum spanning tree (MST) was generated using BioNumerics software (ver. 7.6) with a categorical coefficient of similarity and single and double locus variance priority rules. Each circle represents a different MLVA type, its size being proportional to the number of strains belonging to that MLVA type. Branch thickness indicates how many loci are different in the MLVA types of the connected circles. Thick solid lines connect nodes that differ by one MLVA locus, thin solid lines connect nodes that differ by two or three MLVA loci and dashed lines connect nodes that differ by more than three MLVA loci. The halo surrounding the MLVA types groups together types belonging to the same complex. MLVA complexes were assigned if two neighboring types did not differ by more than three VNTR loci and if at least two types fulfilled this criterion. The XbaI-PFGE profiles (X01-X11) are also indicated near the corresponding MLVA type (M01-M35) the percentage of positive pigs and the distribution of isolates according to biotype did not differ significantly from one survey to another. The absence of biotype 3 in the second survey could be explained by the smaller number of pigs sampled (200) compared to the first survey (900).
To evaluate the genetic diversity of Y. enterocolitica isolated at slaughterhouse level in both surveys, we first typed the isolates by PFGE with restriction enzyme XbaI. The PFGE method has been widely used with different restriction enzymes to distinguish isolates of Y. enterocolitica. Isolates of biotypes 3 and 4 were differentiated using XbaI by Buchrieser, Weagant, and Kaspar (1994) and Najdenski, Iteman, and Carniel (1994). In our study, we also observed only two XbaI profiles among the 12 biotype 3 isolates. These profiles, X01 and X10, were only found for biotype 3 and clearly differed from biotype 4 profiles, with a genetic similarity less than 64.5%. A different finding was observed by Gilpin et al. (2014). When using ApaI or NotI enzymes alone or in combination, similar profiles were identified for biotypes 3, 2 and 4.
Our study revealed a low diversity of biotype 3 isolates(ID: 0.167). This lack of diversity may be explained either by the fact that we had only a few isolates belonging to this biotype, or because biotype 3 is generally less diverse. This low diversity for biotype 3 was In those studies as well as in ours the majority of biotype 3 isolates belonged to one common profile shared by, respectively, 75.0%; 53.1%, and 91.7% of the isolates (Gilpin et al., 2014;Liang et al., 2012;this study). The choice of the enzyme or combination of enzymes did not seem to significantly improve the power of PFGE to discriminate biotype 3 isolates. Moreover, the 439 strains tested in China by Liang et al. (2012) had profiles with 79% of similarity.
This supports the idea that PFGE encountered little genetic diversity among biotype 3 isolates and failed to subtype this biotype.
Interestingly, the two XbaI profiles recovered in our study shared only 59.5% of similarity. This low similarity may indicate that profiles were genetically distant from each other. Since only one isolate with the X01 profile had been recovered from the 12 biotype 3 isolates, we speculated that there may have been a recent emergence or disappearance of isolates having the X01 profile to the benefit of isolates having the X10 profile. Another hypothesis may be that X01 isolates were not as well-adapted as X10 isolates, and therefore remained at a low level of prevalence. In order to conclude on the biotype 3 isolates' genetic diversity and on the usefulness of PFGE for their typing, another typing method needed to be tested.
The 155 biotype 4 isolates in our study were grouped into nine XbaI profiles, giving an index of diversity of 0.605 over the two surveys.
Compared to biotype 3 isolates, biotype 4 isolates were better subtyped with the PFGE method. Also always superior to 0.500, the value of discriminatory index for biotype 4 may vary according to studies. The diversity of the biotype 4 population was similar for Fredriksson-Ahomaa, Korte, and Korkeala (2000) when they used the XbaI enzyme alone (ID = 0.594). Using NotI, three other studies obtained, respectively, an index of diversity of 0.564, 0.692 and 0.840 (Fredriksson-Ahomaa et al., 2000;Gilpin et al., 2014;Sihvonen et al., 2011). Since the same enzyme was used the difference in the ID should have reflected the genetic variability between the set of strains studied.
The PFGE discriminating power increased from 0.740 to 0.870 and 0.930 when Fredriksson-Ahomaa, Autio, and Korkeala (1999) used, respectively, NotI alone, NotI and ApaI or NotI, ApaI and XhoI to type biotype 4 strains. The use of a combination of enzymes may then increase the discriminating power of PFGE when biotype 4 isolates are studied. The choice of enzymes or their combination could therefore be an important parameter to take into account for an optimal discrimination of biotype 4 isolates. In our study, we noted that the biotype 4 population in the first survey (S1) was less diverse than that of the second survey (S2), despite the number of isolates tested during S1 being much larger (122 vs. 33). This may be due to the fact that samples from the first survey were concentrated over three months in 2009, while those from the second survey were spread out more over time, increasing the probability for the 33 isolates to have different profiles.
Although not identical, the profiles of biotype 4 isolates from both surveys exhibited only minor differences. The isolates with a high degree of similarity (88.7%) were then grouped, independently of the year of survey. The marked homogeneity of biotype 4 isolates had previously been observed with different sets of isolates and different enzymes (Bonardi et al., 2014;Falcão et al., 2006;Filetici, Anastasio, Pourshaban, & Fantasia, 2000).
In our study, PFGE divided biotype 4 isolates into two major groups representing 71% of all the tested isolates. Several analyses of the biotype 4 population with PFGE revealed the presence of one or two dominating profiles (Bonardi et al., 2014;Fredriksson-Ahomaa et al., 2000Gilpin et al., 2014). These profiles represented, respectively, 58% to 77% of the population studied. When considering the pulsotype, the major profiles were closely related (96.5% of similarity). We can therefore hypothesize that they shared common characteristics which promoted their persistence in the slaughtered pigs.
Moreover, the number and type of PFGE profiles differed from one year to the other. Thus, in our study, four XbaI-PFGE profiles identified in the first survey were no longer observed in the second survey, whereas three new XbaI-PFGE profiles were identified in the second survey. Therefore, the population of Y. enterocolitica must have varied over time, although XbaI profiles displayed minor deviations and were genetically closely related (sharing at least 88.7% of genetic similarity).
The genetic variation of Y. enterocolitica appeared to be uncorrelated to the year of isolation. Nevertheless, we observed that 76.
% of the biotype 4 isolates 4 belonged to two XbaI-PFGE profiles (X03 and X04) founded throughout the two surveys and at different months. Because the different profiles may compete and suffer mutual interference, the presence of the same profile over the years can be considered of interest. This situation had already been described previously. Fredriksson-Ahomaa, Meyer, Bonke, Stuber, and Wacheck (2010) recovered six NotI-ApaI-XhoI genotypes from the pigs slaughtered in the same slaughterhouse over more than one year. These genotypes came from 14 out of the 27 farms found positive to Y. enterocolitica. Another study concerning a retrospective analysis of clinical biotype 4 strains isolated between 2008 and 2010 indicated that some profiles persisted over several years (Martin, Cabanel, Lesoille, Menard, & Carniel, 2015). These observations suggest either that some biotype 4 genotypes were widely distributed and persisted for years, or that the PFGE method could not discriminate all strains. A clear predominance of major profiles may lead to an incorrect attribution of isolates, especially in case of possible outbreaks. As a matter of fact, the relationship between an isolate and an outbreak cannot be established if the isolate belongs to one of the major profiles.
In our study, the genetic variation of Y. enterocolitica biotypes 3 and 4 appeared to be quite limited when using the PFGE typing technique.
As several studies demonstrated MLVA's high discriminatory power (Gierczynski et al., 2007;Sihvonen et al., 2011) when applied to Y. enterocolitica from different sources. Alakurtti et al. (2016) reported that the discriminatory power of the loci varied from one country to another. Loci V4 and V9 are often reported to have a low discriminatory power compared to the other loci (Alakurtti et al., 2016;Sihvonen et al., 2011;Wang et al., 2012). In these studies, the discriminatory power value was between 0 and 0.60 for one of the two loci. Compared to MLVA data obtained from other countries, the present study indicated that all six VNTR loci showed a high discriminatory power with a value over 0.80. Nevertheless, by considering the IC95%, locus V4 was statistically less discriminatory than loci V2A and V5.
No MLVA types were observed to be common to both biotype 4 and 3 isolates in our study.  (Martin et al., 2015) and arguments about connections between the cases to be put forward. Because MLVA revealed a remarkably high genetic diversity of the pathogenic strains, the authors refuted the hypothesis of a single source of contamination or the expansion of a specific clone.
When using MLVA, we noted that the biotype 4 population found in the first survey was more diverse than that of the second survey, despite a similar number of tested isolates (37 vs. 33 Gierczynski et al. (2007) showed that the MLVA type remained unchanged after 20 serial passages of a strain in vitro. In vivo, the same MLVA type was recovered from humans after repeated isolations from one or more patients with yersiniosis (Gierczynski et al., 2007;Sihvonen et al., 2011) as well as in pigs on farms when sampled at a 6-month interval (Saraka et al., 2017

| CON CLUS ION
This study provides an initial evaluation of the genetic diversity of Y. enterocolitica strains isolated from pigs in France. The biotype 4 population is genetically more heterogeneous than the biotype 3 population.
With PFGE, we showed that some profiles were maintained in the pig production sector due to the presence of two XbaI-PFGE profiles in two consecutive years. With MLVA, we not only improved the differentiation between isolates, but also showed that clones recovered during both years may be genetically closely related. In addition, our study showed that MLVA successfully discriminated biotype 4 from biotype 3 isolates. MLVA, in combination or not with PFGE, has proved its effectiveness as a tool for investigating pathogenic Y. enterocolitica strains isolated from pigs and assessing the genetic diversity of this foodborne pathogen. Because typeability, reproducibility, and discriminatory power are key features in the evaluation of an epidemiological typing system, MLVA is a promising tool. Nevertheless, further studies are needed to improve our knowledge on how to establish a clear relationship between MLVA profiles and epidemiological data.

ACK N OWLED G M ENTS
We would like to thank Claire de Boisséson for her technical help with MLVA. We are grateful to Gilles Salvat for his help in obtaining financial support for a PhD. The PhD grant was funded by the Saint-Brieuc agglomeration and the Brittany region. This research constitutes a work package for a project funded by the Compte d'Affectation Spéciale "Développement agricole et rural" (Special Allocation Account for agricultural and rural development).

CO N FLI C T O F I NTE R E S T
The authors declare no conflicting interests.

E TH I C S S TATEM ENT
The research was done on isolates collected from tonsil swabs on pig carcases. No ethical approval for this study was needed.

DATA ACCE SS I B I LIT Y
Supporting Information Table S1 is included at the end of the manuscript in the section "Appendices". The authors adhere to all policies on sharing data and materials described in the guidelines for authors.