In 2017, an 8-year-old girl in Japan was hospitalized for a urinary tract E. coli infection, which was treated with ceftazidime. Two days after hospitalization, she became unconscious. Laboratory results revealed anemia (hemoglobin 10.5 g/dL) with schistocytes; low platelet count (3.8 × 10⁴/µL); and elevated creatinine (1.38 mg/dL), total bilirubin (1.7 mg/dL), and lactate dehydrogenase (1,848 U/L). Magnetic resonance imaging of her head showed hyperintensity in the basal ganglia and thalamus, suggesting edema and necrosis. From the urine sample, we isolated a LEE-negative STEC (strain JNE170426) carrying the Shiga toxin 2 gene (stx2). On the basis of these findings we diagnosed her condition as HUS with urinary tract infection. We performed intravenous high-dose methylprednisolone therapy, plasma exchange, and hemodialysis for HUS encephalopathy and renal failure, but after 12 days of intensive therapy, she died of HUS encephalopathy.

The isolated STEC did not show agglutination against commercial O1–O188 antisera (Denka Company Ltd., https://www.denka.co.jp; Statens Serum Institut, https://en.ssi.dk). However, comprehensive PCR-based O serogrouping (5) revealed that the isolate was classified into OX18, an atypical O serogroup originally identified from a nonpathogenic E. coli strain from a healthy sow (6). Using OX18-specific PCR screening of O-untypeable STEC isolates obtained during 2007–2019 by the National Institute of Infectious Diseases in Japan, we found 25 additional STEC OX18 isolates (Table). To characterize these isolates, we performed whole-genome sequencing (WGS) using MiSeq (Illumina, https://www.illumina.com). WGS results were analyzed as described elsewhere (7,8) with slight modification. We used BactSNP version 1.0.2 (http://platanus.bio.titech.ac.jp/bactsnp) (9) and Gubbins version 2.4.1 (https://sanger-pathogens.github.io) (10) for core genome SNP extraction. Public database strains used for the phylogenetic analysis are shown in Appendix Table 1. We deposited draft genome sequences and short-read sequencing data into the DDBJ/National Center for Biotechnology Information/European Nucleotide Archive database (BioProject accession no. PRJDB10421; Sequence Read Archive accession no. DRA010812).

In silico analysis revealed that none of the STEC OX18 isolates carried LEE; we classified them into 5 H-genotypes: H2 (n = 2), H8 (n = 1), H19 (n = 20), H28 (n = 1), and H34 (n = 2) (Table). Core-genome SNP phylogeny revealed that OX18 isolates with the same H-types formed closely related groups (Table; Figure). Isolates from patients belonged to OX18:H2, H19, and H34; isolates belonging to OX18:H8 and H28 were obtained from asymptomatic carriers. The OX18 isolate from the case-patient who died of HUS (strain JNE170426) belonged to H2 and carried stx2a and several virulence genes, including STEC autoagglutination adhesin (saa), subtilase toxin (sub), enterohemolysin (ehx), and serine protease (espP) (Appendix Table 2). These regions showed high similarity (>99%) to a large plasmid of STEC O104:H21 strain CFSAN002236 (11). Therefore, these virulence factors are likely to be encoded in similar plasmids.

On the other hand, the other OX18:H2 isolate (strain JNE133347) from an asymptomatic carrier did not carry the virulence genes described above but carried genes for Shiga toxin 2e (stx2e), heat-stable enterotoxin (st), and Pap fimbriae (pap). Of note, the other isolates obtained from HUS belonged to OX18:H19 and were phylogenetically close to OX18:H2 (Figure). The OX18:H19 lineage showed a similar virulence profile to the OX18:H2 isolate from the fatal HUS case, and carried saa, sub, ehx, and espP virulence genes on plasmid-like elements. All bovine isolates in our study were grouped into this serotype. OX18:H19 isolates from humans and bovines could not be distinguished by their lineages, suggesting that cattle can be a reservoir for that lineage. We identified OX18:H34 in isolates that carried pap as an adhesin from a patient with bloody diarrhea and from swine. The other isolates, from asymptomatic carriers, we classified into H8 and H28. The OX18:H8 isolate carried saa, sub, ehx, and espP on plasmid-like elements, similar to the H19 lineage. Meanwhile, the OX18:H28 isolate did not carry adherence factors known in pathogenic E. coli, including LEE genes, saa, pap, aggR, afaD, F4, F6, F17, F18, or F41.

Among LEE-negative STEC isolates, saa-positive STEC has often been reported in patients with severe symptoms (2,3). WGS analyses of saa-positive STEC O104:H21 and O113:H21 revealed that they carry a large plasmid (>100 kb) with several virulence genes, including saa and sub. Because the draft genomes of saa-positive OX18:H2 and H19 showed high similarity to the plasmid, it is plausible that they carry a similar large plasmid. The source or natural reservoir of these lineages was unclear. However, some OX18:H19 isolates have been obtained from cattle, suggesting that cattle or fecally contaminated foods can be a source of the infection. In addition to these lineages, OX18:H34 was found to cause severe symptoms in humans. We were unable to elucidate the pathogenesis and natural reservoir of the lineage because of the small sample size of our study; further studies are required.

In this study, we report a HUS case with urinary tract infection caused by a STEC belonging to the emerging O serogroup OX18. Our retrospective survey revealed that the novel pathogenic STECs OX18:H2, H19, and H34 have been continually isolated from humans and cattle. However, commercial antisera cannot identify these lineages. Elucidating the transmission routes and natural reservoirs of the bacteria is essential to control infection. DNA-based serotyping methods, including Og/Hg typing (6,12,13) and whole-genome sequencing (7,14,15), would be helpful for identification and surveillance of these potentially pathogenic lineages.

Additional information on Escherichia strains and virulence genes in study of whole-genome sequencing of Shiga toxin–producing E. coli OX18 from a fatal hemolytic uremic syndrome case, Japan.