Abstract
Three large clusters of Salmonella Typhimurium infections in Denmark in 2008 and 2009 were defined by multilocus variable number of tandem repeat analysis (MLVA). One of these proved to be the hereto largest Danish cluster of salmonellosis with 1446 cases. Two smaller clusters with a total of 197 and 89 cases, respectively, were seen concurrently. These clusters shared epidemiological characteristics such as age distribution, geography, and time. To investigate the possible genetic relationship between the cluster strains, these were further characterized by phage typing, pulsed-field gel electrophoresis, and Optical Mapping. Although the MLVA method proved robust and well-performing in detecting and defining clusters, the employment of a second typing method detected an additional fourth cluster among the isolates. The cluster strains were stable throughout the almost 2-year period, even though we detected changes in three of five MLVA loci in a small fraction of isolates. These changes were mainly due to the gain or loss of single repeats. Optical Mapping of the large cluster strain indicated no increased content of virulence genes; however, Optical Mapping did reveal a large insert, a probable prophage, in the main cluster. This probable prophage may give the cluster strain a competitive advantage. The molecular methods employed suggested that the four clusters represented four distinct strains, although they seemed to be epidemiologically linked and shared genotypic characteristics.
Introduction
The use of molecular typing methods is often a central part of the detection and investigation of foodborne outbreaks of disease. Cases related to the mentioned outbreaks in the United States were detected and defined by pulsed-field gel electrophoresis (PFGE), which is the most widely used method for epidemiological investigations of Salmonella and other foodborne bacterial pathogens. In recent years, multilocus variable number of tandem repeat analysis (MLVA) is increasingly used in Europe as the primary method for Salmonella Typhimurium outbreak detection (Lindstedt et al., 2004; Torpdahl et al., 2007; Bruun et al., 2009). For example, a pork-related outbreak in Scandinavia (Bruun et al., 2009) was mainly defined and investigated by the use of MLVA. Generally, the molecular typing methods chosen for outbreak detection should have the ability to cluster epidemiologically related isolates and distinguish these from nonrelated isolates. This requires a fairly good discriminatory power, but at the same time the detected markers should be sufficiently stable to ensure that it is still possible to identify all epidemiologically related isolates in, for example, a long-lasting outbreak, where some genetic divergence would be expected. The interpretation of molecular typing results is usually improved by the use of more than one method of targeting different genetic markers.
In Denmark, national outbreaks of Salmonella typically include <50 cases (Torpdahl et al., 2007). In spring 2008, some unusual clusters of Salmonella Typhimurium were detected and investigated by the use of MLVA, PFGE, and phage typing. In the following months, the number of human infections with isolates belonging to the main cluster rose to a total of 1446 cases. This is the largest Salmonella outbreak described in Denmark, which has a population of 5.5 million people, and this outbreak is comparable to the multistate outbreaks described in the United States. Two other clusters with a total of 197 and 89 confirmed cases in 2008–2009, respectively, were seen concurrently. Epidemiological characteristics were shared across clusters and it was a hypothesis in the outbreak investigations that these clusters were related epidemiologically, and possibly caused by the same sources and/or were present in the same reservoir (Ethelberg et al., 2008). In spite of careful investigations, including analysis of samples from food and animals and the use of numerous epidemiological tools, the sources of infection have not been determined for any of these outbreaks (Ethelberg et al., 2008). The cluster strains share some common characteristics and these call for further investigations into the possible biological relationship between these cluster strains. Further, these large and prolonged outbreaks give a unique opportunity to study the stability of the applied molecular markers and how these should be interpreted and used for case definitions in outbreak situations.
The objectives of this study were to characterize the cluster strains and elucidate their possible genetic relationship. Further objectives are to describe and discuss the challenges of using case definitions, based on the molecular typing, of patient isolates in situations with unknown epidemiological links.
Materials and Methods
Isolates
Salmonella Typhimurium strains were derived from the Danish laboratory-based surveillance system of human gastrointestinal infections in 2008–2009. Bacterial pathogens were isolated from human fecal samples either at the regional clinical laboratories or at the diagnostic laboratory at Statens Serum Institut. Salmonella isolates were serotyped according to the Kaufman-White scheme (Grimont and Weill, 2007). As is standard procedure in Denmark, confirmed Salmonella Typhimurium isolates were subjected to MLVA typing for outbreak and cluster detection. In addition, isolates (or a subset of isolates) were further characterized by phage typing, antimicrobial resistance profiling, PFGE, and Optical Mapping.
Multilocus variable number of tandem repeat analysis
The MLVA method developed by Lindstedt et al. (2004) was performed as previously described (Torpdahl et al., 2007; Larsson et al., 2009). A minimum spanning tree was calculated in BioNumerics 6.01 (Applied Maths, Sint-Martens-Latem, Belgium) based on MLVA profiles of all strains MLVA typed in Denmark to date.
Phage typing
Phage typing of Danish Salmonella Typhimurium isolates was performed at the National Food Institute, Technical University of Denmark (DTU Food), in accordance with international standards (Callow, 1959; Anderson et al., 1977).
Antimicrobial resistance
Susceptibility to a standard panel of antimicrobial agents was determined as described previously (Torpdahl et al., 2007).
Pulsed-field gel electrophoresis
PFGE was carried out according to the PulseNet protocol using the XbaI restriction enzyme (Ribot et al., 2006). Restriction patterns were compared by the use of BioNumerics 6.01.
Optical Mapping
The Optical Mapping was performed by OpGen (Gaithersburg, Maryland) as previously described (Zhou et al., 2004). In short, the bacterial cells were embedded in low melting agarose and lysed within the plugs. High-molecular-weight DNA was recovered and loaded onto an optical chip device with a positively charged surface. The genomic DNA was immobilized as single molecules in channels on the chip during the restriction cutting with endonuclease NcoI, ensuring that the fragments were now placed in the same order as they appeared in the genome. The genomic DNA was stained with a fluorescent dye and scanned to measure the size and order of restriction fragments. Finally, the optical map was created by overlapping the numeral fragment maps.
Results
Clusters of salmonellosis
Onset for clusters 1, 2, and 3 was registered in spring 2008 with the first registered cases in February, May, and April, respectively (Fig. 1). Cluster 4 cases were detected in low numbers throughout the 2-year period with a slightly higher incidence in January and February 2009. Cluster 1 was by far the largest and peaked during the summer 2008. During 2008 the number of cases in clusters 2 and 3 followed that of cluster 1 at an ∼10-fold smaller scale. Throughout 2009, cases were continuously reported in all clusters, but at a lower level. A few cases with isolates belonging to clusters 1, 2, and 3 have been detected during the first months of 2010; however, these isolates are not included in this study.
Number of confirmed cases in clusters 1, 2, and 3 during 2008 and 2009. Numbers of isolates are confirmed cases.
Patients in all clusters were geographically confined to Denmark, but were spread in the whole country. The age distribution was skewed toward younger age groups than usually seen for Typhimurium infections in Denmark. Patient interviews showed that the overall food consumption habits of cases in all clusters were similar (e.g., preference for traditional Danish food, including pork). The epidemiological investigations were partly presented previously (Ethelberg et al., 2008). The clusters were investigated as three separate outbreaks based on the case-definition described below. Cluster 4 was identified as a separate cluster late in 2008, and these cases were then excluded from the epidemiological analyses of cluster 1.
Cluster detection and definition using molecular typing
Clusters were detected and defined by MLVA during routine surveillance of Salmonella Typhimurium (Table 1). All isolates with the MLVA cluster type or differing in one locus were initially included in the cluster. However, one-locus variant isolates were PFGE typed and isolates were included only if they expressed the cluster PFGE type. The MLVA411 profile differed from MLVA822 (cluster 1) in only one locus, but PFGE typing resulted in the inclusion of only two MLVA411 isolates in cluster 1. Twenty-one isolates of MLVA411 expressed a distinct PFGE profile (PFGE333) and were defined as cluster 4 (Table 1). The MLVA cluster types of clusters 1, 2, and 3 were new types that had not been registered in the previous 4 years of routine MLVA typing of all human isolates.
One-locus MLVA variants.
Other phage types included in a cluster were justified by MLVA and PFGE.
MLVA, multilocus variable number of tandem repeat analysis; NT, nontypeable; PFGE, pulsed-field gel electrophoresis; RDNC, reaction-does-not-conform.
A minimum spanning tree (MST) was created (Fig. 2) based on MLVA profiles of 5525 Salmonella Typhimurium isolates from humans, foods, and animals in the Danish surveillance database. In general, branching of the tree reflected the size variation of the two most stable loci, STTR9 and STTR3. Clusters 1, 2, and 4 were grouped in the same branch, as they had the same STTR9 and STTR3 alleles. Cluster 3 was located on a different branch as this cluster varies from the other clusters at three or four loci.
Minimum spanning tree calculated from multilocus variable number of tandem repeat analysis (MLVA) profiles. Each circle represents a distinct MLVA profile. The size of the circles is correlated with the number of isolates contained. The tree is based on 1052 MLVA profiles of Salmonella Typhimurium isolates in the Danish surveillance database (5525 isolates from humans, animals, and food). Clusters 1–4 are colored gray and indicated by arrows.
In most cases the alternative MLVA profiles varied by the addition or deletion of one repeat unit in loci STTR5, STTR6, or STTR10. Variations of two or more repeat units were rarely detected.
Pulsed-field gel electrophoresis
A random selection of isolates with the dominant MLVA profiles and all isolates with a one-locus difference were typed by PFGE (209 isolates). The majority of isolates within a cluster were assigned the cluster PFGE types (Table 1). The cluster profiles were unique for the particular clusters (Fig. 3). PFGE profiles of Salmonella Typhimurium isolates generally have a high degree of similarity, but the four cluster profiles differed by only 2, 3, or 4 bands. Compared to all Typhimurium PFGE profiles in our database, the four cluster profiles were situated in the same major branch of the dendrogram (this branch represented approximately one-third of all profiles). A few isolates with the main cluster MLVA profiles showed variations in the PFGE profile (3–6 variant PFGE profiles in each cluster). In general, the alternative types were single-band differences and were represented by a single or a few isolates only. Interestingly, we observed isolates with the main MLVA profile of a cluster expressing the main PFGE profile of another cluster, for example, a cluster 3 isolate (MLVA767) expressing the main cluster 1 PFGE type (PFGE061).
A UPGMA comparison (band based) of pulsed-field gel electrophoresis (PFGE) XbaI profiles of clusters 1–4 showing number of different bands: Cluster 1 (93% of typed isolates in cluster 1 express PFGE 061), cluster 2 (83% of typed isolates express PFGE 325), cluster 3 (75% of typed isolates express PFGE 328), and cluster 4 (100% of typed isolates express PFGE 333).
Phage typing and antimicrobial resistance
Most isolates were found to have the typical cluster phage type although we observed an overlap of phage types between clusters (Table 1). Cluster 4 was described by the primary phage type DT135; however, 25% of isolates in cluster 4 were DT3. Each of the four clusters contained isolates designated RDNC (reaction-does-not-conform) or NT (nontypeable). When repeating the phage typing of isolates, we observed an occasional change between phage types DT135 and DT3. The separation of phage types has previously been described as problematic due to only minor differences in lysis pattern and strength (Baggesen et al., 1997). Further, the interpretation of the lysis pattern of cluster 1 isolates (phage type U292) was found to be difficult to standardize between laboratories as discussed previously (Baggesen et al., 2010). All isolates showed full sensitivity to antimicrobial agents.
Optical Mapping
Six isolates representing the four clusters were selected for Optical Mapping. Two isolates from cluster 1 were analyzed: an isolate with the primary MLVA and PFGE profile and an isolate with variation in one MLVA locus (MLVA411) and the primary PFGE profile (PFGE061). In addition, an isolate was selected from a sporadic case in year 2000 with the cluster 1 MLVA and PFGE profile. From each of the clusters 2, 3, and 4, one isolate representing the primary type was selected for Optical Mapping.
The optical maps of the cluster isolates were compared to five in silico maps produced from publicly available genome sequences of Salmonella Typhimurium (LT2, SL1344, 14028s, D23580, and a DT104). Based on the optical maps of the sequenced genomes, the gene content was inferred to the optical maps of the unsequenced strains.
All cluster isolates had the Salmonella pathogenicity islands one to six and the prophages Gifsy-1 and Gifsy-2. All cluster isolates lacked the prophage Fels-1 and Fels-2 that are present in LT2, but the cluster 2 and 4 strains had a smaller insert at the Fels-2 site. The prophage ST64B was observed in all strains but LT2 (data not shown). The outbreak strains seem quite similar to the sequenced genomes in terms of content of virulence markers, corresponding to the fact that the outbreak did not cause especially severe symptoms of infection (Ethelberg et al., 2008).
We observed differences between the optical maps that were caused by insertions and deletions. Compared to LT2, an extra restriction site was detected in the cluster 1, cluster 3, and D23580 strains (data not shown). The extra restriction site was seen in the fragment that harbors the gene ratB; correspondingly, ratB is a pseudogene in D23580 and presumably also in our cluster 1 and cluster 3 strains. An ∼31-kb insert was detected in both D23580 and the cluster 4 strain. The sequence from the insert in D23580 showed similarity to enterobacteria phage PsP3 (∼97%) and to bacteriophage 186 (∼95%).
Prophage elements added to the diversity observed between optical maps. The cluster 3 strain had a small insertion of ∼12 kb at the ST104 phage site; D23580 also had an insert in this region, with several NcoI restriction sites not detected in ST104 (Fig. 4). The cluster 1 strain was shown to have an insertion at the ST104 phage insertion site, although this insertion was 18 kb larger than the insert in D23580 and with a different restriction pattern (Fig. 4).
Optical maps of the six selected cluster strains and LT2 and D23580. The lateral lines represent NcoI restriction sites, light gray boxes are SPI-6, medium gray box is prophage BTP1, and dark gray boxes represent the unknown inserts that probably contain prophages.
Discussion
Denmark experienced three large clusters of Salmonella Typhimurium infections in 2008 and 2009 with a total of 1732 registered cases. To investigate the possible genetic relationship between the cluster strains, these were further characterized by molecular and phenotypic methods. The cluster strains shared some characteristics such as related PFGE types and full sensitivity to antimicrobial agents. Further, the cluster strains expressed rare phage types; the three phage types combined were responsible for <3% of all human cases of Typhimurium infections in the previous 10 years in Denmark. Interestingly, a slight increase was detected in all three phage types around 2001.
In Denmark, MLVA is used as a routine method for surveillance and cluster detection of Salmonella Typhimurium as this is an inexpensive, high-throughput method that allows fast typing as well as easy analysis of data in a standardized manner (Larsson et al., 2009). MLVA is measuring changes in DNA sequences that evolve at a relatively fast rate. These genomic changes do not reflect the true evolutionary relationship between strains; however, they indicate a clonal affiliation, which makes the method suitable for cluster detection as previously shown (Torpdahl et al., 2007; Bruun et al., 2009). The Salmonella Typhimurium outbreaks in Denmark in 2008–2009 of different sizes provided us with valuable information on the MLVA method for epidemiological surveillance. The MLVA markers were stable during the long lasting outbreaks of almost 2 years, emphasizing that the MLVA method reliably serves the purpose of defining clusters during outbreak investigation. On the other hand, our data indicate that MLVA should not stand alone as the only cluster defining method in complex outbreak situations. In Denmark, PFGE is currently used as the secondary molecular typing method complementing MLVA by being less discriminative for Salmonella Typhimurium (Torpdahl et al., 2007).
This study also provided information about strain stability and the evolutionary processes of the variable number of tandem repeats (VNTR) loci involved. The two VNTR loci, STTR9 and STTR3, were stable during the cluster period. Changes observed in MLVA profiles within epidemiologically related clusters were in loci STTR5, 6, or 10. The gain or loss of a single repeat unit was primarily observed, and only occasionally changes involving more repeat units were observed. Similar observations have been reported previously for Salmonella Typhimurium (Hopkins et al., 2007) and E. coli O157 (Noller et al., 2003), whereas Malorny et al. (2008) did not detect single locus variants within epidemiologically related Salmonella Enteritidis strains. According to our cluster criteria, isolates with single-locus MLVA variation must have the cluster PFGE profile to be included. The data based on a combination of several typing methods confirm the hypothesis that single-locus variations may occur during an outbreak of Salmonella Typhimurium. Although PFGE is the gold standard for cluster detection, our data indicate that PFGE patterns can vary within an outbreak, for example, due to insertion of prophages as shown by the optical maps.
In general, the molecular data that we have obtained clearly suggest that the cluster strains are distinct strains. In our study, the combination of MLVA and PFGE assigned isolates to four different clusters. Although these clusters do not reflect an evolutionary relationship, the MLVA MST analysis suggests linkage of clusters 1, 2, and 4 and a more distant relationship to cluster 3 (Fig. 2). The three phage types (U292, DT135, and DT3) assigned for each cluster were also observed in single isolates in other clusters. Retyping a number of outbreak isolates generally confirmed this, but also showed an occasional change between phage types DT3 and DT135, which is most likely due to the high resemblance in the reaction pattern of these two phage types.
PFGE typing supported discrimination between the three clusters defined by MLVA and identified a separate cluster (cluster 4) that was not detected initially by MLVA (Fig. 3). Cluster 4 isolates were excluded from cluster 1 due to the distinct PFGE profile. Comparison of PFGE profiles showed that the profiles of clusters 1 and 3 differed by two bands and clusters 2 and 4 differed by three bands (Fig. 3), which is usually considered closely related (van Belkum et al., 2007).
To get more detailed genetic information than given by the routine typing methods, Optical Mapping was applied to the cluster strains. Optical Mapping emphasized that cluster 1–4 are distinct strains (Fig. 4). The optical maps of the cluster 1 isolates were indistinguishable, regardless of MLVA profile and time of isolation. This suggests that the cluster 1 outbreak profile has persisted in the environment from 2000 until 2008. The cluster 1 and cluster 3 strains had optical maps similar to the suggested highly virulent sub-Saharan clone D23580 (Kingsley et al., 2009). The strains had possible prophage inserts at identical sites in the genome and additionally an extra restriction site in the ratB gene. However, many genes in the D23580 clone are pseudogenes (Kingsley et al., 2009), and even though our strains have similar optical maps, we do not have the sequence of our strains and are therefore unable to detect differences at the nucleotide level.
Based on the epidemiological investigations, we presume that the cluster 1 strain is not especially virulent compared to other Salmonella Typhimurium strains and this is in concurrence with our observation of the genomic content of this strain. It could be hypothesized that the large insert of ∼56 kb contains a functional prophage. It was previously described how prophages may excise from a strain and kill the competitors in the current growth environment, thereby increasing the competitive fitness of the host strain (Bossi et al., 2003).
We observed a molecular resemblance between the four clusters. This resemblance can reflect a common ancestor, or, alternatively, it may reflect convergent evolution of the four cluster strains, where isolates of different origins evolve independently along parallel paths, resulting in similar traits. This last possibility could, for example, be isolates sharing a common niche to which they have adapted and thus acquired similar molecular content. Without proper sequencing data we are unable to understand the true evolution between the four cluster strains.
Conclusions
The molecular and genetic analysis suggests that the four cluster strains are distinct strains; however, they share genotypic characteristics and seem to be epidemiologically linked and possibly originate from a common reservoir. The molecular resemblance of the cluster strains may account for a common progenitor at some point, or it may be the result of genetic recombination. Alternatively, the strains may have adapted to a common niche. In any case, our data are not sufficient to determine which scenario is more reliable.
Footnotes
Acknowledgments
We thank the regional clinical microbiological laboratories for submitting clinical Salmonella isolates to Statens Serum Institut for national surveillance.
Disclosure Statement
No competing financial interests exist.