Diagnosis of STEC Infections

The NRC for E. coli receives fecal samples collected across France, from patients suspected of HUS or experiencing hemorrhagic diarrhea at any age, as well as from at-risk populations with diarrhea. STEC diagnosis at the NRC is performed by DNA extraction using the Elite Ingenius instrument (Elitech, https://www.elitechgroup.com), followed by PCR with the RIDA GENE E. coli Stool Panel (R Biopharm, https://clinical.r-biopharm.com), which enables detection of stx1, stx2, and eae genes.

We inoculated positive samples onto selective culture media, including CHROMagar STEC (CHROMagar, https://www.chromagar.com), MacConkey CT (bioMérieux, https://www.biomerieux.com), and chromogenic medium chromID CPSO (bioMérieux). Chromogenic medium enables direct identification of E. coli when selective media are negative. We identified and confirmed colonies of STEC using the eazyplex EHEC complete kit (Amplex Diagnostics, https://www.eazyplex.com), which detects stx1, stx2, eae, and ehxA genes. We tested for 10 major STEC serogroups by in-house multiplex PCR as previously described (5). Finally, we conducted antimicrobial susceptibility testing on STEC strains by disk diffusion method, before storing them at –80°C.

Sequencing

We sequenced all strains at Institut Pasteur (Paris, France) using Illumina technology, as previously described (6). For 3 isolates, we performed hybrid assemblies with Unicycler (https://github.com/rrwick/Unicycler) to obtain circularized chromosomes and plasmids, using long-read sequencing with the Native Barcoding Kit 96 V14 (SQK-NBD114.96) on a GridION Flow Cell R10 with MinION (Oxford Nanopore, https://nanoporetech.com), combined with Illumina short-read data generated on a MiniSeq (https://illumina.com). We performed quality control short reads and long reads by fastQC (https://github.com/s-andrews/FastQC) combined with multiQC (https://github.com/MultiQC/MultiQC) for short reads and fastQC combined with nanoQC (https://github.com/wdecoster/nanoQC) for long reads.

Bioinformatics Analysis

We determined the isolate’s phylogroup using Clermontyping (http://clermontyping.iame-research.center). We obtained sequence type (ST) and hierarchical clustering (HC), notably HC5, from Enterobase (https://enterobase.warwick.ac.uk). We performed chromosome and plasmid annotations, including the search for potential virulence genes, with Prokka version 1.14.6 (https://github.com/tseemann/prokka) using a LA database (https://github.com/Kevi84LA/LA-database/blob/main/LA_database.faa). The LA database was built by compiling an updated set of reviewed bacterial entries from UniProt (https://www.uniprot.org), gene sets from the Center for Genomic Epidemiology (https://www.genomicepidemiology.org), the virulence factor database (https://www.mgc.ac.cn/VFs), and in-house sources as previously described (7), translated into amino acid format. To avoid redundancy across the multiple databases, we used CD-HIT version 4.8.1 (https://github.com/weizhongli/cdhit) to cluster identical genes into a single representative sequence with 100% identity and 100% coverage.

We performed resistance gene search using Abricate tool on Resfinder database (8,9). We used PlasmidFinder (https://cge.food.dtu.dk/services/PlasmidFinder) and pMLST (https://cge.food.dtu.dk/services/pMLST) (10) for plasmid characterization and performed plasmid alignment using BRIG version 1.0.0 (https://github.com/happykhan/BRIG).

We performed phylogenic analysis with IQ-Tree version 2.4.0 (https://iqtree.github.io); we based the analysis on core-genome alignment of isolates produced by Panaroo version 1.5.2 (https://github.com/gtonkinhill/panaroo). We identified single-nucleotide polymorphisms (SNPs) on the basis of core-genome alignment using pairsnp (https://github.com/gtonkinhill/pairsnp).

Serotyping

Because the O serogroup deduced from the wzx and wzy sequences by Enterobase could not distinguish between O17/O44/O73/O77/O106, defining the O77-group (O77 g) as previously described (11), we further investigated the O serogroup and H serogroup by phenotypic method at the NRC for E. coli in Spain. We determined O and H antigens using the method previously described (12); all available O (O1 to O181) antiserums were produced in the Laboratorio de Referencia de E. coli—University of Santiago de Compostela (Lugo, Spain). Finally, we used the K1/Neisseria meningitidis B antiserum (Bio-Rad Laboratories, https://www.bio-rad.com) to characterize the capsular serogroup.

Specific PCR

To identify patients potentially infected with the outbreak strain (i.e., patients with fecal PCR positive for stx2 and negative for eae during the outbreak period but without a cultured STEC isolate), we developed an in-house quadruplex PCR targeting specific regions. We designed 2 primer pairs to target the large virulence plasmid characteristic of the outbreak strain: the first pair amplifies a 285-bp fragment located between 2 open-reading frames encoding phytase/esterase activity and a putative thiol peroxidase (tpx gene), and the second pair amplifies a 163-bp fragment of an open-reading frame encoding a putative thermolabile enterotoxin (elt chain A). We designed 2 additional primer pairs to target chromosomal genes, the wzy gene encoding the O77-group and the neuS K92 capsule gene (Appendix).

Case Definitions

We classified patients with HUS, diarrhea, or both in whom the outbreak strain was isolated from stool sample as confirmed cases. We classified those patients without the outbreak strain but with a positive quadruplex PCR in stool sample as probable cases. We classified those patients without isolation of the outbreak strain and showing only the chromosomal (plasmid markers negative) on quadruplex PCR in their sample as possible cases.

Outbreak Cases

Figure 1

Figure 1. Unrooted phylogenetic tree based on core genome alignment of 51 Shiga toxin–producing Escherichia coli isolates from study of hemolytic uremic syndrome outbreak in adults and Shiga toxin–producing ...

Figure 4

Figure 4. Epidemic curve of confirmed, probable, and possible cases of Shiga toxin–producing Escherichia coliinfection caused by the outbreak strain (HC5 326896) from study of HUS outbreak in adults and...

At the end of December, the NRC for E. coli identified 4 adult HUS cases, caused by a LEE-negative STEC that did not belong to the 10 usual major serogroups determined by routine PCR and harbored that stx2 but not eae or ehxA genes. Of note, isolates failed to grow on selective STEC media and were recovered only on CPSO plates. Illumina sequencing confirmed that the outbreak was caused by a unique strain, O77 g:H18, belonging to ST69 of phylogroup D, harboring the stx2d subtype and lacking eae and ehxA. The strain belonged to the novel cluster at the HC5 level 326896, consistent with the core genome multilocus sequence typing–based hierarchical clustering scheme proposed by EnteroBase; outbreak-related isolates presented allelic distances <5. We further confirmed isolate similarity by a core-genome SNP phylogenetic analysis with a SNP median of 4 (range 1–29) (Figures 1–3; Appendix Figure 1).

Figure 5

Figure 5. Quadruplex PCR results from study of hemolytic uremic syndrome outbreak in adults and Shiga toxin–producing Escherichia colinegative for locus of enterocyte effacement, France, 2025. We performed quadruplex PCR...

In total, we identified 18 confirmed cases of STEC infection with the outbreak strain. The cases occurred during December 15, 2024–April 11, 2025 (2024 epidemiologic week 50 through 2025 epidemiologic week 15) (Figure 4); most cases were reported during December 15, 2024–February 3, 2025 (2024 epidemiologic week 50 through 2025 epidemiologic week 6). Cases were distributed nationwide with no clusters (Appendix Figure 2).

Figure 6

Figure 6. Representation of the 134-kb plasmid from isolate CNREC-004-31 in study of hemolytic uremic syndrome outbreak in adults and Shiga toxin–producing Escherichia colinegative for locus of enterocyte effacement, France,...

To explore the potential extent of the outbreak in patients without STEC isolation, we implemented an in-house specific multiplex PCR. During the outbreak period, extended to 1 month before and 1 month after, we found 38 patients (3 experiencing HUS symptoms and 35 with diarrhea) from whom no STEC was isolated from their fecal cultures to be positive for stx2 and negative for eae. We validated our in-house quadruplex PCR, targeting 2 chromosomal genes and 2 genes present in the 134-kb plasmid, for specificity (Appendix Figure 3) and then applied directly to fecal DNA extracts to identify additional cases. We classified 4 additional patients who tested positive for all 4 target genes as probable cases; 1 patient, whose sample was positive for 2 chromosomal targets, was classified as a possible case (Figures 4, 5).

All patients were adults; median age was 72.1 (interquartile range 66.7–81.8; range 34.3–89.4) years, and female/male ratio was 0.56. Among the 18 confirmed cases and 5 probable and possible cases, 17 (94%) confirmed and 4 (80%) probable and possible, had HUS. Among the 23 case-patients, 13 experienced diarrhea; 4 of those had bloody diarrhea. Three (13%) patients died. We expect to publish clinical details of patients and their outcomes in the future.

Investigations and Control Measures

Epidemiologic and traceback investigations, including food questionnaires and supermarket loyalty card records, indicated that by the end of January 2025 a total of 15/17 (88%) confirmed and 2/4 probable case-patients had consumed the same brand of raw cow’s-milk cheese. We excluded 2 probable and 1 possible cases, as well as 1 case confirmed late in our investigation on April 11. We identified the outbreak strain in a cheese sample collected from a patient’s refrigerator. As of January 24, 2025, authorities initiated recall and withdrawal of the suspected cheese produced starting November 12, 2024. In February 2025, we identified 2 additional cases through Enterobase; both had STEC isolates belonging to the same HC5 cluster (Figure 1). One case-patient in Belgium experienced HUS; the patient had not consumed the implicated cheese. The other case was in Scotland; the patient’s clinical condition and exposure data were unknown.

Molecular Characterization of Outbreak Strain

We submitted the 18 isolates from France to short-read whole-genome sequencing; they belonged to ST69 (phylogroup D) and HC5 326896. The rfb gene cluster sequence did not allow discrimination between O17/44/73/77/106 antigens, defining the O77-group (O77 g) described previously (11), whereas the fliC gene corresponded to H18. Agglutination with O antiserum specific to those 5 antigens was negative; therefore, the outbreak strain was designated O77 g:H18. All isolates harbored the stx2d-O73-C165-02 variant and were negative for eae and ehxA, as expected. We long-read sequenced 3 isolates (CNREC-004-07, CNREC-004-31, CNREC-004-43), revealing a chromosome of ≈4,934 kb, a plasmid of ≈96 kb, and, in the first 2 isolates, an additional plasmid of ≈134 kb (Appendix Table 2). The 134-kb plasmid was an IncF plasmid with a complete tra operon. The pMLST tool could not precisely identify the F replicon ST (F-:A17*:B24–79*, in which the asterisks indicate new alleles closely related to A17 and B24/B79). The 96-kb plasmid was an IncY plasmid, lacking a tra operon (Appendix Table 2).

Putative Virulence Factors

Figure 2

Figure 2. Origins and serotypes of 51 Shiga toxin–producing Escherichia coli isolates from study of HUS in adults and Shiga toxin–producing Escherichia colinegative for locus of enterocyte effacement,...

We identified several putative virulence factors (Figure 3), some of which were located on the 134-kb plasmid (Figure 6), whereas we detected no known virulence factors on the 96-kb plasmid (data not shown). In addition to the main virulence factor stx2d-O73-C165-02, we identified several toxin-encoding genes, including cdt-VABC (cytolethal distending toxin subtype V), a putative eltAB (heat-labile enterotoxin II), estb-STb2 (heat stable enteroxin), and hlyE (avian hemolysin). We also detected genes encoding adhesion or invasion factors, particularly fdeC, lpfA, yehABCD, aslA, air, fimH47, and nlpI. Furthermore, protectin-encoding genes such as traT and neuABCDES were present. Other virulence genes with diverse putative functions included the transcriptional regulators prsX/papX and anr, as well as eilA, a transcriptional regulator of the enteropathogenic E. coli type III secretion system 2, characteristic of ST69 strains (13), and several type III secretion system effectors. None of the 18 outbreak isolates we sequenced carry the iha and aggR genes, which have been described in LEE-negative STEC (14).

The detection of the neuC gene suggested the presence of a K1 capsule. However, agglutination with K1-specific antiserum was negative for all isolates except CNREC-004-43. Sequence analysis of the entire neu gene cluster showed high homology across all genes (>98%), except for neuS (87%). BLAST analysis of the neuS sequence revealed 99.92% identity with the K92 neuS gene (Appendix Figure 4). Therefore, we considered the outbreak strain serotype O77 g:K92:H18. In isolate CNREC-004-43, which tested K1-positive, we identified a C154A substitution in neuS that led to an H52N amino acid change (Appendix Figure 4). That mutation is known to be sufficient to convert a K92 capsule into K1 capsule, explaining the phenotype observed in this isolate (15). The outbreak strain carried no antimicrobial resistance genes, confirmed by phenotypic tests. Only a rarely reported parC.S57T mutation was identified, with no effect on quinolone susceptibility.

Similar Strains Worldwide

To determine whether other isolates closely related to our outbreak strain exist worldwide, we searched Enterobase for isolates with the same HC200 2073. We identified 30 such isolates (Figure 1), spanning 3 continents. Moreover, we found in GenBank the strain C165-02, linked to stx2d-O73-C165-02, which is the oldest strain we reported here, isolated in 2005 in Maryland from a patient with bloody diarrhea (16). Almost all isolates were of human origin, but clinical information, including patient age, was missing for most. Several cases were associated with HUS, consistent with the virulence of the outbreak strain. All those isolates are ST69 and have common features, including the stx2d-O73-C165-02 variant (except 1), cdtV (except 1 outbreak isolate), hlyE, and multiple adhesion and miscellaneous genes. We divided the population into 2 main groups, A and B, on the basis of phylogenetic analysis and gene distribution (Figure 1).

Group A (n = 32), which includes our outbreak isolates and 3 circularized strains from Australia, carried nearly all of the putative virulence factors described, with the exception of 3 isolates. The 3 exceptions were characterized by partial or complete loss of 134 kb plasmid-borne genes (putative eltAB, STb, traT, prsX/papX, and anr) (Figure 1–3). To further investigate the plasmid content of all HC200 2073 strains, we performed multi-alignments using BRIG. As previously indicated, CNREC-004-43, one of our long-read sequenced isolates, had completely lost the 134-kb plasmid, whereas CNREC-004-49 and PNUSAE003863 had deletions encompassing large portions of the plasmid (Appendix Figure 5, panels A, B). Of interest, our in-house quadruplex PCR detected plasmid target signals in stool samples containing CNREC-004-49, suggesting in vitro plasmid loss. Altogether, those findings indicate that the 134-kb plasmid is relatively unstable and can be lost either partially or entirely, even in the context of an outbreak and over a short time. In contrast, in group A, only 3 isolates outside the outbreak carried the 96-kb plasmid (Appendix Figure 6, panels A, B).

Group B isolates (n = 19) were all characterized by the loss of >1 plasmidborne gene (putative eltAB, STb, traT, prsX/papX, and anr) (Figure 1–3). Of note, although isolates were from different countries in Europe or North America, none carried the putative elt chain A gene. Only 4 group B isolates harbored an almost complete 134-kb plasmid, and all shared the same HC5 14945 (Figure 1; Appendix Figure 5, panel C). Nine isolates retained a large portion of the 134-kb plasmid but lacked the tra operon (Appendix Figure 5, panel C). Ten isolates carried the 96-kb plasmid, including the oldest strain, C165-02 (Appendix Figure 6, panel C). Six group B isolates might not produce a K92 capsule (Figure 1), because neuC and neuS K92 were absent. Sequence alignment of those 6 isolates with the capsular operon of CNREC-004-31 revealed the presence of a complete kps operon but high fragmentation of neu genes (Appendix Figure 7).