Whole genome sequencing as a tool to investigate a cluster of seven cases of listeriosis in Austria and Germany, 2011–2013

A cluster of seven human cases of listeriosis occurred in Austria and in Germany between April 2011 and July 2013. The Listeria monocytogenes serovar (SV) 1/2b isolates shared pulsed-field gel electrophoresis (PFGE) and fluorescent amplified fragment length polymorphism (fAFLP) patterns indistinguishable from those from five food producers. The seven human isolates, a control strain with a different PFGE/fAFLP profile and ten food isolates were subjected to whole genome sequencing (WGS) in a blinded fashion. A gene-by-gene comparison (multilocus sequence typing (MLST)+) was performed, and the resulting whole genome allelic profiles were compared using SeqSphere+ software version 1.0. On analysis of 2298 genes, the four human outbreak isolates from 2012 to 2013 had different alleles at ≤6 genes, i.e. differed by ≤6 genes from each other; the dendrogram placed these isolates in between five Austrian unaged soft cheese isolates from producer A (≤19-gene difference from the human cluster) and two Austrian ready-to-eat meat isolates from producer B (≤8-gene difference from the human cluster). Both food products appeared on grocery bills prospectively collected by these outbreak cases after hospital discharge. Epidemiological results on food consumption and MLST+ clearly separated the three cases in 2011 from the four 2012–2013 outbreak cases (≥48 different genes). We showed that WGS is capable of discriminating L. monocytogenes SV1/2b clones not distinguishable by PFGE and fAFLP. The listeriosis outbreak described clearly underlines the potential of sequence-based typing methods to offer enhanced resolution and comparability of typing systems for public health applications.


Introduction
The large majority of listeriosis cases (sporadic and outbreakrelated) are caused by foodborne transmission, which accounts for 99% of human cases [1,2]. In an era when food production and food storage rely heavily on refrigeration, the ability of Listeria monocytogenes to grow (albeit slowly) at low temperatures has opened a new ecological niche to a pathogen that previously had only mediocre relevance. Industrialized food manufacturing also constitutes an ecological niche, owing to the ability of L. monocytogenes to form biofilms for colonization of surfaces [3]. Seeliger even dubbed this zoonosis a 'man-made disease' [4].
In Austria, laboratories have a legal obligation to forward human and food L. monocytogenes isolates to the national reference centre; however, the source of the food isolates does not have to be notified to the reference laboratory. On 21 January 2013, the Austrian Food Authority mandated the Austrian Agency for Health and Food Safety to investigate the source and and public health significance of a certain pulsed-field gel electrophoresis (PFGE) clone of L. monocytogenes serovar (SV) 1/2b, which, in 2012, had become the dominant Listeria strain among food isolates; the cluster strain accounted for 29 of 50 L. monocytogenes SV1/2b food isolates received at the National Reference Laboratory during the second half of 2012. At the beginning of the cluster investigation, only two invasive cases of listeriosis with this strain were known. Active casefinding revealed seven human cases of invasive listeriosis caused by L. monocytogenes SV1/2b sharing PFGE (AscI, ApaI, and SmaI) and fluorescent amplified fragment length polymorphism (fAFLP) (HhaI and HindIII) patterns (defined as '1/2b cluster strain') ( Fig. 1). All cluster cases were female, the median age was 75 years (range: 71-83 years), and two cases had fatal outcomes; between April 2011 and July 2013, five cases (cases 1 and 4-7) occurred in Austria and two cases occurred in Germany (cases 2 and 3). We found isolates from food products of five producers (two in Austria, two in Germany, and one in Romania) that were indistinguishable from the human cluster isolates by PFGE (AscI, ApaI, and SmaI) and fAFLP (HhaI and HindIII). Epidemiological investigation suggested that cluster cases 4-7 (Austria, 2012 and 2013) were linked to either one of the two Austrian food producers (A and B). For cluster case 1 (Austria, 2011) and cluster cases 2 and 3 (Germany, 2011), the incriminated ready-to eat (RTE) meat products of producer B had been excluded as possible sources by the findings of the epidemiological investigation. To further investigate this listeriosis cluster, we applied whole genome sequencing (WGS) and subsequent gene-by-gene comparison (multilocus sequence typing (MLST)+).

WGS and data analysis
The 18 Listeria isolates were subjected to WGS in a blinded fashion. A single colony of a fresh culture was incubated overnight in brain-heart infusion broth prior to DNA extraction. One microlitre of this culture was subsequently subjected to DNA extraction with the Qiagen MagAttract HMW DNA kit (Qiagen, Hilden, Germany), according to the manufacturer's recommendations. Sequencing libraries containing 1 ng of DNA were prepared with Nextera XT chemistry (Illumina, San Diego, CA, USA) for a 75-bp, 150bp or 250-bp paired-end sequencing run on an Illumina MiSeq sequencer, according to the manufacturer's recommendations (Illumina), aiming for a minimum coverage of 75fold [9]. After quality trimming with the default parameters of the CLC Genomic Workbench software version 6 (CLC bio, Arhus, Denmark), the sequencing reads were assembled with the CLC Genomic Workbench de novo assembler with default parameters, with the exception of 'length fraction value = 0.8' (CLC bio). Gene sequences for subsequent analyses were extracted from contigs with Ridom Seq-Sphere + software version 1.0 (Ridom, Muenster, Germany). For the gene-by-gene analysis, i.e. MLST+, as described earlier, we initially compiled a list of genes that were present in all strains analysed [10]. This list, comprising 2,298 genes used for L. monocytogenes MLST+ typing, is accessible under http://www.ages.at/ages/gesundheit/mensch/ listeriose/gene-by-gene-typing-of-l-monocytogenes-isolates/. The designation of the analysed genes was based on L. monocytogenes strain EGD-e (GenBank accession no. NC_003210). For phylogenetic analysis, the defined cgMLST sequences, i.e. the 2298 gene sequences, were extracted from each assembly and assessed for quality, i.e. the absence of premature stop codons and ambiguous nucleotides, and a minimum sequence coverage of ten-fold or more over the whole gene with a minimum substitution frequency in reads of 75%. If genes fulfilled all of these quality criteria, their complete sequence was analysed in comparison with L. monocytogenes strain EGD-e, and a numerical allele type was assigned by SeqSphere + . For illustration of the phylogenetic relationship of the 18 L. monocytogenes isolates, the combination of all alleles in each strain formed an allelic profile, i.e. the typing result, which was used for subsequent construction of a minimum-spanning tree with SeqSphere + software (Ridom). The whole genome sequence reads have been deposited at the European Nucleotide Archive (study accession no. PRJEB5481), and detailed WGS characteristics are shown in Table 2.

Results
After quality control of the WGS data, 2298 genes were finally analysed: isolates AT02, AT03, AT16 and AT18 from Austrian cluster cases 4-7 had different alleles at ≤6 genes, i.e. differed by ≤6 genes from each other (defined as outbreak cases); the dendrogram placed these isolates in between the Austrian unaged soft cheese isolates (AT05-AT08 and AT17) from producer A (≤19-gene difference from the human cluster) and  the RTE meat isolates (AT09 and AT10) from producer B (≤8gene difference from the human cluster) (Fig. 2). In contrast to the German and Romanian food products, these two Austrian food products, which harboured L. monocytogenes at <30 CFU/ g, appeared on grocery bills collected from the outbreak cases (cases 4-7) after hospital discharge. MLST+ clearly separated the isolates from Austrian cluster case 1 (AT01) and German cluster cases 2 and 3 (AT14 and AT15) from the outbreak cases' isolates (≥48 different genes). The two German meat isolates (AT12 and AT13) and the Romanian meat isolate (AT11) showed a ≥56-gene difference from the outbreak cases' isolates. The control SV1/2b strain (AT04) differed from the isolates with indistinguishable PFGE/fAFLP patterns by 1574 genes.

Discussion
L. monocytogenes is a particularly important cause of illness, and is mainly found in foods that are packaged and prepared commercially, rather than those prepared in the home [11]. During recent decades, consumer lifestyles have changed, with less time available for food preparation, and higher consumption of RTE and takeaway foods. Changes in food production and technology have led to the production of foods with a long shelf-life that are typical 'Listeria risk foods', because the psychrotrophic bacteria have time to multiply, and the food does not undergo a listericidal process such as cooking before consumption. Swaminathan named the high degree of centralization and consolidation of food production and processing, the increased use of refrigerators as the primary means of preserving food and the above-mentioned changes in food consumption habits (increased consumer demand for convenient food) as the main factors increasing the incidence of listeriosis [12]. Epidemiological investigation suggested that four of the five Austrian cluster cases of L. monocytogenes SV1/2b (cases 4-7; 2012 and 2013) were linked with either one of the two microbiologically related Austrian food producers; food products from both producers appeared on grocery bills collected from these cases after hospital discharge. Already in a previous outbreak, all surviving Austrian outbreak cases had been asked to collect grocery receipts for the 3 weeks after they were discharged from hospital, to enable the collection of information on routine food consumption behaviour [13,14]. This approach successfully revealed 'Quargel', a type of acid curd cheese available in different flavours, as the source of the 2009-2010 outbreak, which involved 34 cases of invasive listeriosis (including eight fatalities) [15].
For Austrian cluster case 1 and the German cluster cases (cases 2 and 3), all of which occurred in 2011, the incriminated RTE meat products of producer B had been excluded as possible sources by epidemiological investigation (data not shown). MLST+ was able to discriminate L. monocytogenes isolates from these three cluster cases, and these isolates were related according to PFGE and fAFLP. On the basis of these MLST+ findings, cluster cases 1-3 could be excluded from the outbreak, and the meat products from Romania and Germany were ruled out as possible sources of infection. Without MLST+, the three 2011 cases (including two fatalities) would have been incorrectly considered to be outbreakrelated, resulting in the false exclusion of RTE meat products of producer B, which were sold neither in Germany nor in the sole grocery store frequented by case 1, as a possible source.
Our investigation suggests that the outbreak cases (cases 4-7; 2012 and 2013) were linked with either one of the two Austrian food producers. However, none of the 34 food samples positive for the SV1/2b cluster strain with quantitative data available yielded L. monocytogenes in numbers above 100 CFU/g (data not shown), the legal threshold in the EU.
We were unable to determine the causative vehicle among two biologically plausible suspects left: unaged soft cheese and shrink-wrapped deli meat. However, raised awareness caused by the outbreak investigation and the resulting intensification of in-factory sanitation stopped the detection of the cluster strain in food samples (Fig. 3). The producer of the fresh soft cheese, producer A, was first contacted on 26 February 2013, and the producer of the RTE meat, producer B, on 4 March 2013. At the fresh soft cheese production facility of producer A, routine testing for L. monocytogenes of every batch produced (with negative results being a prerequisite for any delivery) was implemented. At the RTE meat production facility of producer B, a meat-slicer that, by intensified environmental surveillance, was identified as a possible source of L. monocytogenes contamination was sanitized. To date (February 2014), no further cases caused by the L. monocytogenes SV1/2b cluster strain have been reported since 15 July 2013 (hospitalization date of case 7). This shows that outbreak investigation and general public health action can control an outbreak even in the absence of definitive proof of a causative vehicle.
In contrast to the situation in forensic human genetics, where-with the exception of monozygotic twins-indistinguishable DNA fingerprinting patterns prove epidemiological relatedness, in bacteriology even epidemiologically unrelated isolates can yield DNA fingerprinting patterns that are indistinguishable from each other. Explanations for this might be a monomorphic population structure, which is known from other pathogens such as Escherichia coli O157:H7, or the limitation of the typing method used [16]. As our data show for L. monocytogenes, pure chance can yield pairs of epidemiologically unrelated isolates with PFGE and fAFLP patterns that are indistinguishable from each other, and pairs of contemporary isolates of human and of food origin without any causal relationship. On the other hand, the maximum resolution of WGS poses a major challenge in determining meaningful similarity thresholds for grouping of related isolates to provide an appropriate level of discrimination for source attribution. The definition of such thresholds will be a major task for the scientific community and the public health authorities, using well-defined outbreak scenarios of every pathogen species of interest. The applied method has to allow for some genetic diversity between isolates from human and food sources, but only to the degree that it can still be assumed that they originate from the same source [17]. WGS and genome analyses have already proven their potential to identify genes that contribute to persistence of L. monocytogenes in food production plants and to virulence [18,19]. At present, we lack knowledge on the expected rate of change of L. monocytogenes producer of the unaged soft cheese, producer A, was first contacted on 26 February, and the producer of the ready-to-eat meat, producer B, on 4 March. Raised awareness and resulting intensification of in-factory sanitation caused the cluster strain to disappear. in our MLST+ scheme over time, on the expected diversity within outbreak strains, and on the expected diversity between outbreak strains and source isolates. We see a clear need for this diversity information before this MLST+ scheme can be used for source attribution rather than exclusion. However, we have shown that comparative WGS with geneby-gene-based data analysis (MLST+) is capable of discriminating L. monocytogenes SV1/2b clones grouped together by the genotyping methods PFGE (AscI, ApaI, and SmaI) and fAFLP (HhaI and HindIII). The meeting participants of a recent ECDC technical consultation on harnessing genomics for epidemiological surveillance (Paris, 1-2 October 2013) unanimously agreed that WGS will replace current typing techniques for surveillance and outbreak investigation [20]. The listeriosis outbreak described clearly underlines the potential of sequence-based typing methods to offer enhanced resolution and comparability of typing systems for public health applications.

Transparency Declaration
The authors declare no conflicts of interest.