Investigation of a Staphylococcus aureus sequence type 72 food poisoning outbreak associated with food‐handler contamination in Italy

On August 2019 a staphylococcal food poisoning outbreak occurred in an elderly home in Piedmont, Italy. The epidemiological investigation performed among the persons that consumed the meal identified chicken salad as the most likely source of the outbreak. Staphylococcus aureus was isolated from a total of seven samples, namely one vomit sample from a guest of the nursing home, two food samples (chicken salad with and without mayonnaise) and nasal swabs collected from a total of four persons working in the kitchen of the nursing home. The maximum likelihood tree obtained using single nucleotide polymorphisms analysis revealed that the isolates from the aforementioned samples clustered together. Multilocus sequence typing revealed that they belonged to Sequence Type 72. Fourier transform infrared spectroscopy (FTIR) was used in parallel to single nucleotide polymorphisms and whole genome sequencing for the determination of the degree of relatedness of the isolates. The results of the FTIR showed the same clustering obtained with single nucleotide polymorphisms and whole genome sequencing and revealed the source of infection. This study underlines the importance of both laboratory evidence and epidemiological data for outbreak investigation and further confirms that FTIR is a suitable support for the short‐term epidemiological investigation on source attribution in case of a S. aureus infection.

diarrhoea and abdominal pain or nausea following a short incubation period.
Human food intoxication by S. aureus is mainly associated with inadequate handling of cooked or processed foods (Argudín et al., 2010) followed by favourable environmental conditions for its growth and enterotoxin(s) production during food storage and preparation (i.e. time and temperature). Food poisoning outbreaks associated with post-process contamination of foods with S. aureus are in part the responsibility of food handlers who carry enterotoxigenic staphylococci in their nares or on their skin (Angelillo et al., 2000;Portocarrero et al., 2002). Indeed, S. aureus represents a ubiquitous commensal that colonizes the anterior nares of healthy adults with percentages of the global population with 20% to 30% of intermittently and persistently infected respectively (Kluytmans & Wertheim, 2005;van Belkum et al., 2009). However, S. aureus can also cause serious infections, toxinoses and life-threatening diseases, including skin and soft tissue infections, toxic shock syndrome and septicaemia.
Subtyping of S. aureus is crucial to epidemiological investigations and phylogenetic studies (Johler et al., 2013;van Belkum et al., 2007) and common techniques used for subtyping of S. aureus are pulsedfield gel electrophoresis (PFGE), spa typing and multilocus sequence typing (MLST) (Aires- de-Sousa et al., 2006;Cookson et al., 2007;Strommenger et al., 2006). Ideally, a typing method needs to provide a reliable and accurate bacterial type, at the highest speed and lower cost possible (MacCannell, 2013).
The investigation of foodborne outbreaks studies is supported by many different culture-dependent or DNA-based, expensive, and/or time-consuming typing techniques such as next generation sequencing and pulsed-field gel electrophoresis, requiring at least several days until a complete characterization is available. Usually, investigations are limited to retrospective studies, nevertheless, to rapidly confine foodborne outbreaks, it is important to be able to identify and characterize the patient's pathogen and to link it with the food source, in order to take appropriate measures to prevent further spreading as quickly as possible. The development of easy, rapid and sensitive methods is, thus, still needed for implementing control measures in the case of outbreaks.
There is an increased interest in tracking, identifying and understanding the diversity of S. aureus in various settings. Fourier transform infrared (FTIR) spectroscopy is a phenotypic, rapid, nondestructive, simple, inexpensive, and high-throughput analytical tool, based on the differential vibrational modes of distinct chemical bonds when exposed to an infrared beam. Each bacterial cell exhibits a unique FTIR spectrum, corresponding to its specific fingerprint signature and correlating with genetic information Naumann et al., 1991).

| MATERIAL S AND ME THODS
On the evening of 7 August 2019, 11 persons manifested gastrointestinal symptoms, nausea and headache in a nursing home for the elderly. At 6.30 pm the doctor on duty of the Food Hygiene and Nutrition Service, Local Health Authority, Piedmont Region, Italy, received notification of a suspected food poisoning from the doctor of the nursing home for the elderly. An epidemiological investigation was carried out to determine the full extent of the outbreak and its probable source. Biological and food residual samples were collected, and microbiological analyses were performed.
Ethical approval was not required as this was a secondary data analysis and we report non-identifiable data.

| Epidemiological investigation
All the people that at lunch had consumed the meal were contacted, and structured interviews were performed by the local health authority to collect information on food exposure and illness symptoms. The guests self-reported what food exposures they had and whether or not they became ill following the exposure. Information about nature of symptoms and duration of illness was collected; the main symptoms and the time of onset were analysed to determine the possible causes of the outbreak and to draw the epidemic curve.
The case definition included persons who developed specific symptoms (abdominal pain, nausea, vomit, and diarrhoea) with onset from the consumption of the common lunch until 12 h.

| Laboratory investigations
Only one emesis sample was collected and analysed for pathogenic bacteria and toxins potentially responsible for the reported symptoms. Analyses for the presence of S. aureus, Bacillus cereus, and Norovirus were performed. In addition, nasal swabs were collected from the personnel involved in food handling that were on duty on the day and the day before the outbreak (i.e. the cook and the Impacts • In this study we report a S. aureus outbreak caused by the contamination of food from an asymptomatic food handler.
• S. aureus strain ST-72 identified as the cause of the foodborne outbreak is a common community acquired pathogen in South Korea.
• Fourier transform infrared spectroscopy allowed source tracking in a more rapid and less expensive way than WGS, thus it can be used for strain differentiation, identification and comparison for S. aureus. In addition, bacterial genomic DNA was extracted using the EXTRACTME Genomic DNA isolation kit (Blirt) and Whole Genome Sequencing (WGS) was performed on the MiSeq platform (Illumina) using paired-end libraries which were prepared following the Illumina™ DNA Library Prep Kit (Illumina), with 150-bp read length. The reads were first subjected to the Galaxy tool 'FastQC Read Quality reports', accessed via the Galaxy public server at https://usega laxy.org, (Afgan et al., 2016) to provide the quality control checks on raw sequence data. Raw reads were trimmed using the Galaxy tool Trimmomatic 0.38 (Bolger et al., 2014) by removing Nextera adaptors and other Illumina-specific sequences ("Illuminaclip" set to value "Nextera (paired-ended)"), removing low-quality residues at the start and end of the reads ("leading:10" and "trailing:10"), clipping reads when average Qscores dropped below 20 over a sliding window of four residues ("slidingwin-dow:4:20"), and dropping reads shorter than 40 bases after processing ("minlen:40").and finally the reads were assembled to genomes by means of Unicycler (ver. 0.4.1.1) via Galaxy (Wick et al., 2017) using for the bridging mode moderate contig size and misassembly rate ("Bridging mode" set to value "Normal") and contigs below 200 bp in length were excluded ("Exclude contigs from the FASTA file which are shorter than this length (bp)" set to value "200"). The assembled genomes were processed to determine the multilocus sequence typing (MLST) in silico with MLST 1.8 (accessed via https://cge.food.dtu.dk/servi ces/MLST//) (Larsen et al., 2012) selecting "5x" for minimum depth for an allele. The antimicrobial resistance genes were identified using ResFinder 4.1 (accessed via https://cge.cbs.dtu.dk/servi ces/ResFi nder/; Cosentino et al., 2013). Finally, genomes were analysed for virulence gene detection with VirulenceFinder 2.0 (accessed via https://cge.food. dtu.dk/servi ces/Virul enceF inder/) selecting 90% as threshold for identification and 60% for minimum length (Joensen et al., 2014).
The fastq files of paired reads were processed with CSI Phylogeny 1.4 (accessed via https://cge.cbs.dtu.dk/servi ces/CSIPh yloge ny/) to call and filter single nucleotide polymorphisms (SNPs) and infer phylogeny based on the concatenated alignment of the high-quality SNPs . SNP analysis was performed with the following parameters: 10 × minimum depth at SNP position, 10% minimum relative depth at SNP position, 100 bp minimum distance between SNPs, 30 for minimum SNP quality, 25 for minimum read mapping quality, 1.96 minimum Z-score for each SNP, and including S. aureus NCTC 8325 (GenBank accession number: NC_007795.1) as reference.
The evolutionary history was inferred by using the maximum likelihood method and Tamura-Nei model (Tamura & Nei, 1993). Spectra processing and visualization was performed with the ir biotyper Client software V3.1 (Bruker Daltonics), using default settings as recommended by the manufacturer. After spectra smoothing using the Savitzky-Golay algorithm, the second derivative of the spectra was calculated by the software. After vector normalization, spectra relation within a wavenumber range from 1300 to 800 cm −1 was analysed. IRTS 1 and IRTS 2 (Infrared Test Standard) were measured as quality control prior to sample spectra acquisition, in each run. All spectra were acquired intercalating a background spectrum between each sample/control measurement. Each sample spectra was evaluated by the software considering four quality parameters: absorption (range: 0.4-2 cm −1 ), noise (value: <300 cm −1 ), water vapour (value: <300 cm −1 ) and fringes (value: <100 cm −1 ). For each strain at least three good quality spectra were considered for further analysis. Exploratory analysis was performed applying principal components analysis (PCA) and linear discriminant analysis (LDA). Hierarchical cluster analysis (HCA) was done using Euclidean metric and average linkage algorithm. Results were displayed as scatter plot (PCA/LDA) and dendrogram (HCA).

| Epidemiological and clinical characteristics of cases
From the performed interviews resulted that an overall of 69 people (60 guests, 5 beneficiaries of home assistance, 2 healthcare professionals, 1 canteen cook and 1 cook helper) consumed the meal. The menu provided pasta with pesto and fresh tomatoes, pasta with oil, chicken salad with or without mayonnaise and cooked vegetables. The meal was prepared by the internal canteen. The interviewed personnel referred that out of the 33 people who consumed the chicken salad, 11 (9 guests and 2 healthcare professionals) were symptomatic, while 22 were asymptomatic (15 guests, 2 food handlers and 5 beneficiaries of home assistance; Table 1). Since the remaining 36 people did not consume the chicken salad, it resulted as the most likely source of the outbreak.

| Laboratory investigations
Emesis sample resulted positive for S. aureus but negative for B. cereus, and for the presence of Norovirus. Nasal swab culture revealed the presence of S. aureus in all of the swabs collected from the nares of the four food handlers as reported in Table 2. Among the food tested, all samples tested negative for Norovirus, B. cereus and its emetic toxin and, whereas high levels of CPS (>100,000 CFU g −1 ) were isolated from the chicken salad (served both with and without mayonnaise), no staphylococcal enterotoxins were detected.
Overall S. aureus was isolated from a total of seven samples, namely one vomit, two food (chicken salad with and without mayonnaise) samples and nasal swabs collected from a total of four persons working in the nursing home for the elderly.
The WGS run had the following overall statistics: cluster passing filter of 98.01%, quality score ≥30 of 97.54% with a total yield of 10.30 Gbp achieved. The analysed S. aureus isolates obtained from the same matrix shared the same anti-microbial resistance genes, enterotoxin genes and ST, hence only one strain has been reported in Table 2. Only for operator 3 (O3), the two detected isolates belonged to two strains differing from one another for the presence/absence of the enterotoxin genes. No staphylococcal enterotoxins were identified in food, but WGS analysis revealed the presence of the sei and seg enterotoxin genes in every sample, while sec was identified in TA B L E 1 Summary of the subjects involved in the outbreak.

Role
Chicken salad consumer Symptomatic Asymptomatic the strains of operators 3 and 4 (O2 and O4) and seh was additionally identified in the strain identified from O2 ( O3 and O4) were located on another branch of the tree (Figure 1).
IRBT acquired at least four spectra of good quality for each isolate. Regarding hierarchical cluster analysis, the clustering cut-off value, automatically calculated by the software, revealed overall five clusters, one including samples F1, F2, O1 and P1, the other four clusters including one sample each (Figure 2).
Based on these findings we hypothesize that this outbreak has been provoked by O1 that contaminated with S. aureus the chicken salad by failure to follow good hygienic practices. The results of the FTIR confirmed the clusterization previously described (Figures 2   and 3).

| DISCUSS ION
Here, we describe a confirmed food poisoning outbreak due to CPS that can be considered an outbreak with strong microbiological and epidemiological evidence according to the European Food Safety The ST-72 strain identified to be the cause of the outbreak is the most commonly community-associated methicillin-resistant or -sensitive S. aureus in South Korea (Joo et al., 2012). Indeed, recent findings have demonstrated that these isolates have been developing resistance to desiccation and adaptation to hypotonic solutions, promoting their survival in a hospital setting (Joo et al., 2017). In addition, within food ST-72 has been identified in Korea in the meat and milk production chain (Kim et al., 2015;Lim et al., 2013), and in Uruguay in food premises (Machado et al., 2020) while interestingly, the first identification of the isolate in Europe has been recently described in Italy in buffalo milk products (Normanno et al., 2020). Here we report the first foodborne intoxication due to S. aureus ST-72 in Italy and in Europe.
For operator 3, two isolates belonging to two strains differing for the presence/absence of the seg and sei enterotoxin genes were detected. Much of the variation between S. aureus strains appears to be attributable to mobile genetic elements, such as plasmids, F I G U R E 1 Maximum likelihood tree obtained using SNP analysis with CSI Phylogeny 1.2 highlights the genomic correlation between the strains isolated from the nasal swabs of the caterer (O1), the vomitus sample of a patient (P1) and the food samples from chicken salad with (F1) and without mayonnaise (F2) and that cluster differ from the other isolates of operators (O2, O3 and O4).
bacteriophages, transposons and insertion sequences (Sumby & Waldor, 2003). The seg and sei genes are present in S. aureus in a tandem orientation belonging to the same operon (Jarraud et al., 2001) and as reported by Sumby and Waldor (2003) seg is a protophageborne gene that occasionally could lead to excision and consequently the loss of SE genes.
The analyses performed with FTIR and WGS gave comparable results in terms of their ability to link different isolates, but, while whole genome sequencing provides more information, Fourier transform infrared spectroscopy is a rapid and inexpensive method that can be applied as a real-time surveillance method.
In addition, with FTIR two ST-22 isolates that differed for the absence/presence of enterotoxin genes (seg toxin and sei toxin) were discriminated from one another. Even though there is no indication that exactly this difference is responsible for the different clustering, similar results have been described in a previous study (Meyers, 2000), where spectra have been suggested to serve as "fingerprint" usable in taxonomic discrimination. Fourier transform infrared spectroscopy has already been described as a promising tool for rapid discrimination among strains in outbreak investigations (Johler et al., 2013) as well as in source attribution studies (Harmsen et al., 2003), and recently, the perfor- self-limiting and without severe consequences, but in the case of an outbreak within a home for the elderly the rapid identification of the source of contamination can be crucial.

CO N FLI C T O F I NTE R E S T S TATE M E NT
The authors have no conflict of interest to declare.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.

F I G U R E 3
Results of LDA analysis depicted as 3D scatterplot. Each geometric form represents one spectrum, while colours and shapes are attributed to identify isolates and multilocus sequence typings respectively.