A rapid cross-jurisdiction foodborne outbreak response was jointly coordinated by the Shenzhen, Dongguan and Guangdong Centers for Disease Control and Prevention, through a multicenter effort to conduct epidemiological investigation, laboratory diagnoses, near real-time whole-genome and metagenomics analyses, genomic source tracing, and stipulate intervention strategies, as summarized in Figure 1. Epidemiological investigations and analyses began on September 22 and were conducted by the Dongguan and Shenzhen CDCs. Dongguan CDC was also responsible for the on-site field investigation at the nursery, as well as the collection of samples from food (n = 19), the environment (n = 7), food handlers, and kitchen staff (n = 11). Samples from case-patients were collected by hospital physicians in Shenzhen (n = 46) and Dongguan (n = 30). Rapid diagnostic testing using a FilmArray GI panel multiplex PCR was performed at Shenzhen CDC. Bacterial culture, isolation, and identification were performed by Dongguan and Shenzhen CDCs during September 22–26. On September 26 and the ensuing 34 h, bacterial isolates were sent to Beijing Institute of Microbiology and Epidemiology (n = 14) and the Guangdong CDC (n = 52) for whole-genome sequencing, whereas metagenomic samples were sequenced by Shenzhen CDC (n = 30). Following the identification of the implicated food source, a traceback investigation was conducted by Shenzhen CDC and Shenzhen Agricultural Product Quality and Safety Inspection and Testing Center. All results generated were collated at Shenzhen CDC for data and bioinformatics analyses.

On-site investigation included interviews with school and kitchen staff, inspections of kitchen and dining areas of the canteen, reviewing surveillance video recordings, procedures of food preparation, and meal serving records. Demographic data, onset of illness, clinical symptoms, and food consumption histories were obtained through structured questionnaires collected at hospitals. During and immediately following the investigation period, the nursery was suspended from operation; pending detailed review of food preparation procedures at the nursery canteen. Standard epidemiological investigation procedures were followed, for which a summary workflow is provided as Supplemental Material (Supplementary Figure S1). A case was defined as a person with acute onset of three or more episodes of diarrhea and/or vomiting in a 24-h period, accompanied by fever >37°C or abdominal pain, who has consumed foods prepared by the nursery canteen since September 18. Active case finding was conducted by raising public health alerts to healthcare providers in an effort to identify additional persons throughout both cities whom met the case definition.

Samples collected (n = 113) included rectal swabs or stool samples from case-patients (n = 76) and food handlers (n = 11), food samples (n = 19), and kitchenware (cutting boards, serving plates, refrigerator, and mayonnaise mixing bowl; n = 7) from the nursery canteen. Stool samples from 10 cases with severe symptoms were processed using the FilmArray GI panel (BioFire Diagnostics, Salt Lake City) multiplex PCR as previously described (Hu et al., 2020). All samples were enriched and screened for common foodborne pathogens including Salmonella, Shigella, Vibrio parahaemolyticus, Vibrio cholerae, Staphylococcus aureus, Escherichia coli O157:H7, enterotoxigenic Escherichia coli, enteropathogenic Escherichia coli, enteroinvasive Escherichia coli, enterohemorrhagic Escherichia coli, Bacillus cereus, group A Streptococcus, and Listeria monocytogenes, as previously described (Hu et al., 2008). Food samples were sliced into small pieces aseptically, and 25 g was diluted in 225 ml enrichment broth and homogenized. Samples were enriched in selenite cystine broth (SC) at 35°C for 18–24 h, and a loopful then streaked into Hektoen Enteric (HE) agar plates, which were incubated at 35°C for 18–24 h. A minimum of five suspected colonies were picked and subjected to biochemical (Vitek 2 and bioMérieux) and serological tests for identification.

DNA extraction was performed using the Qiagen QIAamp DNA Mini Kit (QIAGEN, Hilden, Germany) according to manufacturer’s instructions. Whole-genome and metagenomics sequencing were conducted in near real-time at sequencing facilities from multiple cities in China. WGS of outbreak isolates were performed at Beijing Institute of Microbiology and Epidemiology using Ion S5 or at Guangdong CDC using Illumina MiSeq, while historic strains from the Guangdong province were sequenced using Illumina MiSeq or BGISEQ-500. Metagenomics sequencing was conducted by BGI in Shenzhen using MGISEQ-2000. All sequencing data with uniform output formats were quality controlled and analyzed using same bioinformatics pipelines at Shenzhen CDC. Briefly, over 1GB of clean data were generated for each whole-genome sequenced isolate (>100x coverage, WGS) or metagenomics sample on average. Short-read sequence data were deposited in the NCBI Sequence Read Archive (BioProject PRJNA565566, accession numbers provided in Supplementary Table S1). Core-genome single-nucleotide polymorphisms (core-SNPs) were identified by the Snippy pipeline v4.3.8¹ using S. Enteritidis reference genome P125109 (accession number: NC_011294) as previously described (Jiang et al., 2020). SNPs located in repetitive and recombinogenic regions were removed prior to phylogenetic analysis. Repetitive regions in the reference genome were identified using TRF v4² and self-alignment by BLASTn, and recombinogenic regions were identified using Gubbins v2.3.4 (Croucher et al., 2015). The maximum likelihood trees were constructed based on non-repetitive and non-recombinogenic core-SNPs using IQ-TREE (Nguyen et al., 2015). Genomic source tracing was conducted by interrogating S. Enteritidis genomes (n = 241,991) from the Enterobase (Assessed September 2019; Zhou et al., 2020)³ for closely related strains (<10 core-genome multilocus sequence typing alleles differences) for inclusion into phylogenetic analysis. Pair-end libraries with short DNA fragments were constructed for whole-genome shotgun metagenomic sequencing by BGI in Shenzhen using MGISEQ-2000. After quality control, metagenomic sequencing reads were analyzed using Kraken v2.0.8 (Wood et al., 2019) for taxonomic classification and abundance estimates of S. enterica present in samples.

Between September 20 and 22, a total of 254 cases of gastroenteritis were reported in Shenzhen and Dongguan, Guangdong province, China. Preliminary investigation indicated that only those who were either children or staff from a local nursery school in Dongguan located near the city border with Shenzhen had become sick. Of 157 persons who met the outbreak case definition during initial epidemiological investigation and active case finding, demographic data and exposure questionnaires were obtained for 121 case-patients and were included in this study. The age range of cases was 2–61 years, which included young children (age 2–6; n = 116), teaching staff (n = 3), and relatives of teaching staff who consumed leftovers brought home (n = 2), with a male to female ratio of 61–39%. Predominant clinical manifestations included fever (98.4%), diarrhea (89.3%), abdominal pain (87.6%), and vomiting (79.3%). One case presented with severe diarrhea, vomiting, and a fever of 39°C was admitted to intensive care unit, none died. The distribution of cases over 3 days (September 20–22) from the epidemic curve suggested a point-source outbreak with a single incubation period, which peaked on September 21 between 4 am and 8 am with 54 cases (Figure 2).

On-site investigation conducted on September 22 found that the disease prevalence among children attending full-day classes and half-day classes at the nursery were 25.84% (116/449) were 0 (0/18), respectively, which was significantly different (χ² = 4.88, p < 0.05). Children from half-day classes were picked up at noon and subsequently were not given lunch and afternoon snacks. Combined with the time distribution of the cases, it is considered that the lunch or afternoon snack in the nursery canteen on September 20 may have led to the outbreak. Different foods were provided for teaching staff and children at the nursery, and no sandwiches were served for teaching staff on September 20. However, three sick teaching staff and two of their relatives who consumed leftovers sandwiches brought home developed symptoms of gastroenteritis and were eventually identified as cases. Interviews conducted found that mayonnaise containing raw eggs was spread directly on to bread, made into sandwiches, and served without heating. Based on these epidemiological findings, egg sandwiches served to children on September 20 as afternoon snacks were implicated as the suspected food source of the outbreak.

From a total of 113 samples, 66 were positive for S. Enteritidis, which were isolated from case-patients (n = 58), food samples (n = 4), food handlers (n = 2), and the mayonnaise mixing bowl (n = 2). Of the 10 samples tested by the FilmArray GI panel multiplex PCR, nine were found to be positive for Salmonella, while only seven samples had suspected colonies on HE plates. Using the Vitek 2 rapid diagnostic system, suspected colonies were identified to be Salmonella species, while serological testing has been determined as serotype Enteritidis.

The phylogenetic relationships inferred by WGS SNP-based cluster analysis between all sampled isolates were concordant with epidemiological findings. A maximum likelihood tree was constructed to visualize that all outbreak-associated isolates were genetically closely related to each other (Figure 3A). From a total of 860 core-SNPs identified, all outbreak-associated S. Enteritidis isolates differed by pairwise distances of one SNP or less and clustered together, including isolates from sandwiches, the mayonnaise mixer, and case-patients (SNP distance matrix provided in Supplementary Table S2). In conjunction with epidemiological analyses, the kitchen-made mayonnaise used as an ingredient of egg sandwiches was determined to be the source of the outbreak. In comparison, the minimum distance between historic isolates from the Guangdong province (n = 81) and any of the outbreak-associated isolates was 14 SNPs, larger than the common threshold (≤3 SNPs) used for delineating outbreak clusters (Taylor et al., 2015), suggesting a non-local outbreak source. From the preliminary metagenomics analysis of patient fecal samples conducted, 300–197,790 reads (median: 3,967) were assigned to S. enterica, with species-specific abundances for S. enterica ranked within top 10 of all abundances for 50% of samples (n = 15), ranging from 0.01 to 5.69% (median 0.19%) per sample (Figure 3B).

To further elucidate the possible origins of a non-local outbreak source, a genomic source tracing analysis conducted using the Enterobase (Zhou et al., 2020). A S. Enteritidis clone that shared the same SNP genotype (pairwise distance ≤ 2 SNP between genomes) with isolates from the current outbreak (Figure 4A) was found throughout a multi-provincial egg distribution network in China (n = 8), thereby revealing the possible transmission dynamics (Figure 4B). This clone appeared to have originated from an egg producer in the Hebei province, with a possible transmission chain that included intermediary distributors in the Liaoning province, before being sold at wholesale markets in Dongguan, Guangdong province, which in turn gave rise to the current disease outbreak in Dongguan. With the assistance from the State Administration for Market Regulation (SAMR), epidemiological traceback investigation revealed and confirmed that eggs used for mayonnaise production in this outbreak were purchased from the Dalingshan market, Dongguan, which was sourced from an egg distributor in Anshan, Liaoning Province, for a Hebei chicken farm.