Epidemiological investigations identified an outbreak of Shiga toxin-producing Escherichia coli serotype O26:H11 associated with pre-packed sandwiches

In October 2019, public health surveillance systems in Scotland identified an increase in the number of reported infections of Shiga toxin-producing Escherichia coli (STEC) O26:H11 involving bloody diarrhoea. Ultimately, across the United Kingdom (UK) 32 cases of STEC O26:H11 stx1a were identified, with the median age of 27 years and 64% were male; six cases were hospitalised. Among food exposures there was an association with consuming pre-packed sandwiches purchased at outlets belonging to a national food chain franchise (food outlet A) [odds ratio (OR) = 183.89, P < 0.001]. The common ingredient identified as a component of the majority of the sandwiches sold at food outlet A was a mixed salad of Apollo and Iceberg lettuce and spinach leaves. Microbiological testing of food and environmental samples were negative for STEC O26:H11, although STEC O36:H19 was isolated from a mixed salad sample taken from premises owned by food outlet A. Contamination of fresh produce is often due to a transient event and detection of the aetiological agent in food that has a short-shelf life is challenging. Robust, statistically significant epidemiological analysis should be sufficient evidence to direct timely and targeted on-farm investigations. A shift in focus from testing the microbiological quality of the produce to investigating the processes and practices through the supply chain and sampling the farm environment is recommended.


Introduction
Shiga toxin-producing Escherichia coli (STEC) belong to a pathogenic group of zoonotic E. coli that cause gastrointestinal disease in humans due to their ability to produce Shiga toxin (Stx) [1]. There are two types of Stx, Stx1 and Stx2 and 10 subtypes Stx1a, Stx1c, Stx1d and Stx2a-Stx2 g [2]. Shiga toxin targets a specific subset of cells that express the receptor GB3, including podocytes, microvascular endothelial cells in the kidney, platelets, germinal centre B lymphocytes, erythrocytes and neurons [3][4][5]. Certain strains of STEC have the potential to cause haemolytic uraemic syndrome (HUS), a life-threatening condition characterised by renal failure, sometimes with cardiac and/or neurological complications [2][3][4][5]. The majority of STEC isolated from patients with severe symptoms also have a gene designated eae that encodes the protein intimin and is commonly used as a molecular marker for the locus of enterocyte effacement (LEE) pathogenicity island. Intimin is one of the large number of proteins encoded in the LEE and is required for the formation of attaching and effacing lesions on host intestinal cells [6,7]. STEC infection damages the lining of the gut reducing its capacity to reabsorb fluids resulting in diarrhoea, which may contain blood.
The outbreaks of STEC-HUS in the UK in the 1980s were caused by STEC O157:H7 and in response, laboratory methods focused on the use of cefixime-tellurite sorbitol MacConkey (CT-SMAC) agar [12]. CT-SMAC is selective for the growth of this specific serotype, thus compromising surveillance of the other STEC, including STEC O26:H11. In contrast, commercial and in-house polymerase chain reaction (PCR) assays targeting the presence of stx have the potential to detect STEC of all serotypes [13]. In England and Wales, surveillance for STEC O26:H11 has been continually improving since 2013, in parallel with the increasing number of frontline hospital laboratories implementing PCR ( [13,14] https://www.gov.uk/government/publications/ escherichia-coli-e-coli-o157-annual-totals/shiga-toxin-producingescherichia-coli-stec-data-2018). However, although all patients who present to primary healthcare and submit a faecal specimen are tested regardless of symptom severity, only the regions of the country where the local hospital laboratory has implemented PCR have the potential to detect STEC O26:H11. In Scotland, the burden of STEC O26:H11 for patients with severe clinical outcomes is well-established as specimens from cases with bloody diarrhoea and HUS have been forwarded to the Scottish E. coli O157/STEC Reference Laboratory (SERL) for STEC testing since 2002 (https:// www.foodstandards.gov.scot/publications-and-research/publications/ whole-genome-sequence-typing-and-analysis-of-non-o157-stec).
In October 2019, routine microbiological surveillance at SERL identified an unexpected increase in the number of faecal specimens that tested positive by PCR for stx1 but were negative for stx2 and rfbE O157 (a gene encoding a lipopolysaccharide synthesis enzyme unique to E. coli O157), and a multi-agency incident management team (IMT) was convened. This report describes the outbreak investigation, highlights the microbiological and epidemiological challenges encountered and makes recommendations for future practice.

Microbiology investigations
In the UK, faecal specimens from hospitalised or community cases with symptoms of gastrointestinal disease are cultured in local hospital microbiology laboratories for identification of Salmonella, Campylobacter, Shigella spp. and STEC O157:H7. In England, a sub-set of laboratories, approximately 20%, use commercial PCR assays for the detection of gastrointestinal pathogens, including STEC other than serotype O157:H7 (non-O157 STEC) ( [14], Public Health England (PHE) in-house data). Faecal specimens from patients where there is a clinical suspicion of HUS and/or those testing positive for STEC by PCR and culture-negative for STEC O157:H7 on cefixime-tellurite sorbitol MacConkey agar are submitted to the Gastrointestinal Bacteria Reference Unit at PHE for confirmation by PCR and culture [15,16]. In Scotland, all faecal specimens from patients with severe gastrointestinal symptoms, specifically bloody diarrhoea and HUS, are submitted to the SERL for PCR and culture. During the outbreak, faecal specimens testing positive for stx1 but negative for rfbE O157 at both SERL and PHE were tested using a PCR for the detection of rfbE O26 (http://old.iss.it/binary/vtec/ cont/EU_RL_VTEC_Method_02_Rev_0.pdf), while whole genome sequencing (WGS) results were pending. All strains of STEC isolated from faecal specimens were sequenced, and serotype, stx subtype profile and single nucleotide polymorphism (SNP) type were derived from the genome, as described previously [17][18][19][20].

Data availability statement
FASTQ reads from all sequences in this study can be found at the PHE Pathogens BioProject at the National Center for Biotechnology Information (accession number: PRJNA315192).

Case definitions
Confirmed: A case of STEC O26:H11 stx1/eae with a sequence that falls within five SNPs of the outbreak SNP type.
Previous analysis of the relatedness of isolates of STEC has shown that isolates from cases epidemiologically linked to the same outbreak fall within the same five SNP single linkage cluster [21,22].

Epidemiological investigations
Prospective and retrospective case ascertainment was undertaken by reviewing stx profiles from faecal specimens where STEC had been isolated but that were pending WGS, and by reviewing all WGS data held in the PHE and SERL databases.
Public Health Agencies in England, Scotland and Wales operate national enhanced surveillance systems for STEC [11]. Every laboratory confirmed case of STEC O157:H7 is asked to provide a detailed account of their food history, contact with animals and environmental exposures for the 7 days prior to onset of illness using a standardised enhanced surveillance questionnaire (ESQ). In England and Wales, public health follow-up of cases of non-O157 STEC focuses on those that are infected with STEC that are positive by PCR for stx2, because of the association between stx2a and severe clinical outcomes. Cases that are infected with STEC that are positive for stx1 only strains are not routinely administered an ESQ (https://www.gov.uk/government/ publications/shiga-toxin-producing-escherichia-coli-public-healthmanagement). In Scotland, ESQs are administered to all SERL confirmed cases of non-O157 STEC regardless of the stx profile. Following an initial review of the ESQ data, outbreak cases were re-interviewed using a trawling questionnaire to collect a more detailed food history.
The UK posted the outbreak on the European Communicable Disease Centre's Epidemic Intelligence Information System sharing the WGS accession numbers to ascertain whether related cases had been seen elsewhere.

Case−case analysis of exposures
Data were extracted from the national enhanced surveillance database maintained by Health Protection Scotland (HPS) for the confirmed outbreak cases residing in Scotland, and a bespoke dataset was created by combining standardised data collected from ESQs and the trawling questionnaires. The null hypothesis for testing was that there was no association between any exposure to food from any specific food outlet and being infected with the outbreak strain.
Controls were selected from among sporadic STEC cases reported in 2018 in the HPS national database. Controls were excluded if travel outside of the UK was reported in the 14 days before onset, if an ESQ was not completed, or if they were younger or older than the age range of the cases associated with the outbreak (between 15 and 64 years). Food exposures were re-coded as binary responses. Binary variables were created for each food outlet, coded as 1 if premise was mentioned as the source of any food items. Odds ratios (ORs) were calculated for each food outlet exposure comparing the odds of outbreak cases reporting the exposure with the odds of controls cases reporting the same exposure. P values were adjusted for multiple testing.
Multivariable logistic regressions were conducted for each exposure, adjusting for age and sex. This analysis was also repeated using all the 2018 STEC cases regardless of age and using logistic regression to adjust for age and sex.

Case−control analysis of exposures
A further analysis was undertaken in order to include exposures reported by cases in England, and to account for overmatching in the case−case analysis by using an existing database of healthy controls. Data were extracted from the national enhanced surveillance system for the confirmed outbreak cases residing in England, and exposure data were provided by HPS for the confirmed cases residing in Scotland. The null hypothesis for testing was: There is no association between any food exposures and being infected with the outbreak strain.
Controls were selected from the national control database maintained by PHE Field Services North West, which is frequency-matched by age with all confirmed STEC cases during the same time period. Controls were excluded if travel outside of the UK was reported or if an ESQ was not completed, if gastrointestinal illness in the month prior was reported, or if they were not resident in England. Food exposures were re-coded as binary responses. ORs were calculated for each food exposure reported by at least 25% of cases, comparing the odds of outbreak cases reporting the exposure with the odds of controls cases reporting the same exposure. The Fisher's exact test was conducted to produce P values, to account for the small numbers of cases (<5 exposed). A multivariable logistic regression was then created to include any food item where the results of the univariate analysis showed P < 0.2 and OR >1.

Food chain investigations
Food, water and environmental samples were collected by local environmental health officers (EHOs) from two food processing sites belonging to a national food chain franchise (food outlet A), and one site belonging to the food business operator (FBO) A, and transported in accordance with the Food Standards Agency Food Law Code of Practice (https://www.food.gov.uk/ enforcement/codes-of-practice/food-law-code-of-practice-2015) to Food, Water and Environmental Microbiology Laboratories in Edinburgh and York in cold boxes at a temperature of between 0 and 8°C and tested within 24 h of collection. Testing of the food samples followed PHE Standard Method F17 based on BS EN ISO 16654:2001 http://img.21food.cn/img/biaozhun/20100729/181/ 11294219.pdf, as previously described. STEC isolated from food samples were submitted to SERL for confirmation and typing. Growing and/or processing procedures for suppliers and wholesale distributors identified in the supply chain investigation were reviewed.

Analysis of microbiological data
During the week commencing 14th October 2019, SERL detected a higher than average number (n = 11) of faecal specimens that tested positive for stx1 compared to any other week in 2019 (median = 1.2 samples/week, range 0-6). STEC O26:H11 was subsequently cultured from 10/11 of these specimens. During the same week, PHE detected 25 faecal specimens that tested positive by PCR for stx1 which was higher than numbers in previous weeks in 2019 (median = 10 samples/week, range 3-22). STEC O26:H11 was subsequently cultured from 11/25 of these specimens. All isolates were sequenced. Sequencing confirmed the serotype of 21 isolates (SERL = 10; PHE = 11) as STEC O26: H11, stx1a/eae, and all 21 isolates fell within a five SNP single linkage cluster (Fig. 1). Ultimately, between 1st October and 30th November, 32 isolates of STEC O26:H11 fell within the same five SNP single linkage cluster were identified.

Analysis of epidemiological data
The 32 confirmed cases were distributed across the UK, with 14 cases in Scotland, 16 in England and two in Wales (Fig. 2). Onset dates ranged from 2nd October 2019 to 18th November 2019 (Fig. 3). Twenty-one cases were male (64%) and ages ranged from 3 to 77 years of age, with a median of 27 years (Fig. 4). Clinical outcome data were available for 26 cases, of which 22 (85%) cases developed bloody diarrhoea and six (23%) were hospitalised (Table 1). No cases of HUS or deaths were reported. None of the cases reported travel abroad in the incubation period.
Exposure data were available for 28 of the 32 cases, of which 26/28 (92.9%) cases reported eating out at various food outlets within 7 days of onset of symptoms. Of these, 17/26 (65.4%) cases reported purchasing food from food outlet A, 3/26 (11.5%) cases reported purchasing sandwiches from food outlet B, and 3/26 (11.5%) cases reported eating out at food outlet C. The remaining three cases ate at three different food outlets.

Case−case analysis of exposures
At the time of the case−case analysis, 14 cases had been confirmed in Scotland, and exposure information was available for 13 cases. In addition to the cases linked to the outbreak, a total of 221 records were extracted from the HPS database, of which Epidemiology and Infection 3 59 reported a history of overseas travel in the 14 days prior to onset and were excluded from the analysis leaving 162 cases not linked to this or any outbreak. In univariable analysis, among food outlet exposures there was evidence of an association between being infected with the outbreak strain and consuming food purchased at food outlet A (OR = 183.89, P < 0.001), any bakery (OR = 17.6, P = 0.019) and food outlet B (OR = 13.20, P = 0.006) ( Table 2). In multivariable logistic regressions for each food item adjusted for age and sex, association with the same exposures were maintained, with the magnitude of association increasing for food outlet A (OR = 197.05, P < 0.001), decreasing for any bakery (OR = 15.73, P = 0.030), and remaining stable for food outlet B (OR = 13.58, P = 0.007) ( Table 3).

Case−control analysis of exposures
Of 27 confirmed cases at the time of analysis (21 November 2019), questionnaires were available for 24 cases (89%), at varying levels of completion. Three hundred ninety-three controls were included in the analysis, who completed questionnaires from February to October 2019. Fifty-four per cent of controls were female, and the age range was less than 1 year to 71 years with the median age being 26 years. In univariable analysis, among food outlet exposures, there was evidence of an association between pre-packaged sandwiches and a higher OR of being a case (OR = 11.01, P < 0.001). The analysis also showed evidence for eating outside of the home (OR = 5.01, P = 0.020), handling raw beef (OR = 2.84, P = 0.019) and handling or consuming processed meat (OR = 2.48, P = 0.034) ( Table 4). In the final multivariable model for food items, adjusted for adulthood and sex, nine food items were retained, of which three were associated with increased risk of the outbreak infection. As all cases were adults, controls under 15 years of age were dropped from the analysis. The most notable exposures were pre-packaged sandwiches, processed meats and raw beef (Table 5).

Food chain investigations
The ingredients list of foods consumed by the cases at food outlet A, food outlet B and food outlet C, showed the most common ingredient was a salad mix of Iceberg lettuce, Apollo lettuce and Spinach. Five samples of mixed leaf salad were taken from two sandwich preparation sites for food outlet A (Scotland = 3; North of England = 2), and one sample tested positive for a non-O157 STEC isolate. However, microbiological typing showed that the isolate belonged to STEC O36:H19 stx2e and was unrelated to the outbreak strain. Although this does not constitute direct evidence that this salad was the source of the outbreak as it is a different strain, it does highlight a potential pathway of human exposure to STEC from the mixed leaf salad.
The salad mix was supplied by FBO A. Test results provided by FBO A for samples of this product from 1st August to 28th October 2019, showed that samples had <10 cfu/g for E. coli. EHOs from the local authority for FBO A visited the food processing and distribution site, and took seven samples of the mixed leaf product supplied to food outlet A. One sample with a production date of 30/09/2019 and a use-by date of 04/10/ 2019 had a result for E. coli of 30 cfu/g but was negative for STEC. The local authority for FBO A took nine environmental swabs including swabs of drains, multi-head packers, lines and post wash product samples on the 4th November 2019. All samples tested negative for E. coli. FBO A reported that they routinely test environmental swabs on a weekly basis to confirm that cleaning has been effective. All swabs taken between August and October 2019 tested negative for E. coli. The EHOs visited the site and were satisfied with the processes that FBO A had in place.
An investigation of the growers supplying FBO A was carried out for the period between 6th of September and 22nd of October. All businesses confirmed that they did not have any retained product or soil samples from this specific time frame and that the results of microbiological sampling during the time frame under investigation had been satisfactory. During the time frame of interest, FBO A was supplied by 11 different growers, 10 farms were located in the UK and one farm was located in France. The IMT reviewed documentation, quality standards  Saira Butt et al.
and audit reports for all growers supplying FBO A and found them to be satisfactory. Due to the large number of farms involved in the supply chain, the time that had elapsed since the production of the implicated batch of salad leaves, and the absence of an organism linking the mixed leaf salad directly to the outbreak, the IMT concluded it would be disproportionate to investigate further at the on-farm level.

Discussion
In the UK, outbreaks caused by STEC O26:H11 occur less frequently than outbreaks of STEC O157:H7 (https://www.gov. uk/government/publications/escherichia-coli-e-coli-o157-annualtotals). It is uncertain whether this is a true reflection of differences in the burden of disease, animal reservoirs and/or  In England, the STEC Operational Guidance focuses on public health follow up, with respect to administering the ESQ and requiring microbiological clearance of patients in risk groups, of those cases with STEC harbouring stx2 (https://www.gov.uk/government/publications/shiga-toxin-producing-escherichia-coli-public-health-management). Consequently, ESQs on a subset of patients linked to this outbreak and resident in England were not initially available thus hindering the epidemiological analysis in the early phase of the investigation. Evidence in the literature for the association between STEC with stx2a and the potential to cause HUS is well-established [2]. However, other studies have highlighted the role of stx1a as a marker for severe clinical outcomes, specifically as a cause of bloody diarrhoea and increasing the risk of hospitalisation [2,23]. During this outbreak, the number of patients in England reporting bloody diarrhoea (85%) was higher when compared to those expected for STEC O157:H7, although hospitalisation rates (23%) were lower (bloody diarrhoea 61% and hospitalisation 34% from Byrne et al. 2015 [11]). In Scotland, hospitalisation rates for all cases of STEC in 2019 were 38% (STEC O157:H7 45%; non-O157 STEC 28%) and bloody diarrhoea was reported by 79% (STEC O157:H7 71%; non-O157 STEC 89%) https://hpspubsrepo.blob.core.windows.net/hps-website/nss/3109/documents/2_stec-in-scotland-2019-full-report.pdf. The high rates of bloody diarrhoea in Fig. 3. Epidemic curve of confirmed cases by onset date, or sample date where onset date was not available (n = 32). Both cases positioned in mid-November were based on sample dates; one case reported being symptomatic since October and no information was available for the second case.  non-O157 cases in Scotland may reflect the selection bias inherent in the procedure for referral of cases to SERL. Globally, foodborne outbreaks of STEC O26:H11 have been previously associated with beef products [24,25], dairy products [26,27], and vegetables and salad items [28,29]. Outbreaks of STEC O26:H11 caused by person-to-person contact, specifically in nursery school settings, have also been described [30][31][32][33][34]. These reports of outbreak associated with person-to-person spread provide evidence of the transmissibility of STEC O26: H11, and the need for public health follow-up of cases and exclusion of children aged five and under from school or childcare facilities, until microbiologically clear of the aetiological agent. In England, Scotland and Wales, it is also a requirement that food handlers and clinical/social care staff working with susceptible patients are microbiologically clear of the aetiological agent before returning to work.
Although the outbreak strain was not recovered from the implicated food source, the epidemiological analysis provided robust evidence that one of the components of the mixed leaf salad was the contaminated vehicle. Outbreaks caused by contaminated leafy greens and other salad vegetables are challenging to investigate because these food items are often minor ingredients of a meal that the case may fail to recall [35]. With respect to the outbreak described here, the implicated mixed leaf salad was a component of sandwich fillings, or garnish. The link between the cases was established because the majority of infected individuals reported buying takeaway food products from the same national food chain franchise (food outlet A), and other food outlets supplied by FBO A. Contamination of produce during cultivation or processing is often transient and caused by a one-off event, such as flooding, or a temporary failure in the production process caused by poor manufacturing processes [36][37][38]. Another problematic aspect of investigating outbreaks of STEC is that contaminated food capable of causing illness may contain the pathogen at low levels, and/or be heterogeneously distributed in the food matrix, and so detection of the pathogen is challenging, even when large analytical units of food are tested, and a robust sampling plan is in place [39,40]. The fact that non-O157 STEC, including STEC O26:H11, are not part of the testing algorithm of food samples in the UK, and are difficult to differentiate from other non-pathogenic E. coli that may be present, further confounds detection.
During this outbreak investigation a strain of STEC was isolated from a sample of the mixed salad, although it was not identified as the outbreak strain. The confirmed presence of STEC in a batch of ready-to-eat (RTE) food falling is considered a serious risk to public health. Although not all strains of STEC have been found to cause bloody diarrhoea and/or STEC-HUS, a precautionary approach is appropriate given the uncertainty in the evidence and the potential for severe clinical outcomes. The FSA's current view is that the confirmed presence of STEC in RTE food is an unacceptable risk to public health and that it is Epidemiology and Infection appropriate to take action to remove contaminated food from the market (https://acmsf.food.gov.uk/sites/default/files/acm_ 1191_stec.pdf). The competent authority should be notified through incident reporting procedures and action should be taken to withdraw affected batches from the market in accordance with Article 19 of Regulation (EC) 178/2002. Information on the onward supply of the product is required to determine whether a product recall from end users/ consumers would also be appropriate. Following this incident, the product was no longer available for recall because of the time delay between the sampling and obtaining the result. However, follow up investigations were initiated by the FBO to investigate the source of STEC contamination, and the HACCP-based food safety processes were reviewed.
During the outbreak described in this study, on-farm sampling was deemed inappropriate due the lack of microbiological evidence that the mixed salad was the vehicle, the time since the contaminated product would have been on farm, the absence of evidence of ongoing contamination, and the resource commitment required. Moreover, during the time frame under investigation, the components of the mixed leaf salad had been sourced from many different farms, and that since that time there had been a seasonal shift in growers providing the supplier. However, it should be noted that STEC can survive in soil for many months [41], so even if the fields have been harvested and it has been some time since the outbreak, it may be still possible to detect the outbreak strain, and given the limitations of testing produce for STEC described above, prioritising on-farm investigations may be the most appropriate course of action. On-farm investigations during previous outbreaks linked to contaminated produce identified ruminants grazing in close proximity to the crops, and poor practice, such as the use of river or pond water for irrigation [37][38][39]42]. In North America, in order to gain a better understanding of the risks associated with fresh produce there has been an increase in the emphasis placed on root cause investigations, such as sampling the farm environment, specifically soil, water sources and animal faeces, on-farm environmental audits and assessments of the processes and practices throughout the supply chain (https://www.foodsafetynews.com/ 2021/01/fda-tests-show-cattle-lot-implicated-in-leafy-greens-ecoli-outbreak/#more-200609, https://www.fda.gov/food/foodborne-pathogens/2020-leafy-greens-stec-action-plan).
Outbreaks of severe gastrointestinal disease caused by salad vegetables and other types of raw produce contaminated with STEC continue to be a major public health concern. Salad vegetables are sold as a ready-to-eat food, as defined in Article 14 of Regulation (EC) 2073/2005 (https://www.fsai.ie/food_ businesses/micro_criteria/reg_2073_05.html).
Contaminated ready-to-eat foods are regarded as a high risk compared to food that requires cooking prior to consumption. Although the production process involves a stringent washing process that removes 90-99% of external bacteria, low numbers of residual bacteria may be sufficient to cause a significant risk of infection. Outbreaks timely and targeted on-farm investigations. Such an approach is essential to identify the root cause of outbreaks linked to salad and raw vegetables, and to establish an evidence-base for improving guidance and policy.

Conflict of interest.
There are no conflicts of interest