Nationwide genomic surveillance of Shiga toxin-producing Escherichia coli infections conducted across Belgium from 2019 through 2024 revealed that carriage of the stx2a and stx2d toxin subtypes drastically elevates a patient’s likelihood of progressing to hemolytic uremic syndrome, a potentially fatal renal complication. Retrospective examination of clinical submissions by the Belgian National Reference Centre confirmed that individuals infected with stx2a-harboring isolates exhibited more than a sevenfold increase in the odds of developing severe kidney disease compared to baseline profiles.
In Belgium, surveillance of human infections operates through voluntary referrals submitted by hospital laboratories to the National Reference Centre, alongside mandatory reporting to regional public health authorities and routine monitoring through Sciensano’s sentinel network. However, public health epidemiologists established in 2018 an empirical risk assessment scheme intended to guide clinical interventions by sorting bacterial isolates into risk tiers according to primary toxin genes. Because that earlier framework rested upon legacy datasets comprising only 411 bacterial isolates analyzed between 2011 and 2016, reference center microbiologists sought to reassess the criteria using high-resolution genomic data collected across six consecutive years of universal sequencing.
Between January 1, 2019, and December 31, 2024, the National Reference Centre received and evaluated 3,212 unduplicated human fecal samples voluntarily referred by practicing diagnostic laboratories across Belgium. The annual intake of patient samples escalated substantially during the six-year observation window, climbing from 427 specimens processed in 2019 to 852 specimens registered during 2024. Over this identical monitoring duration, the overall proportion of fecal specimens testing positive for Shiga toxin genes by culture-based polymerase chain reaction surged from 30.7 percent up to 63.0 percent.
Across the entire investigation, reference center staff recovered a total of 1,234 confirmed bacterial isolates from 1,586 polymerase chain reaction-positive specimens, establishing an overall culture recovery efficiency of 76.9 percent. Demographic and clinical metadata revealed that fourteen individuals carried two distinct bacterial strains, while three patients provided isolates originating from discrete, clinically separated infection episodes where microbiological clearance had been documented between sampling events. Genomic profiling revealed marked diversity within the national microbial population, identifying serogroup O157 as the most prevalent single lineage, comprising 31.3 percent of all recovered isolates. However, non-O157 lineages accounted for the vast majority of remaining samples, led prominently by serogroups O26 at 11.4 percent, O63 at 6.16 percent, O145 at 5.75 percent, O103 at 5.6 percent, and O146 at 5.2 percent. Collectively, these six dominant serogroups represented 65.3 percent of all characterized isolates, while advanced in silico genotyping uncovered rare and novel lineages including RKI3, RKI4, and RKI11.
To characterize isolates at nucleotide resolution, microbiologists conducted short-read paired-end whole-genome sequencing utilizing Illumina HiSeq 1500 and NovaSeq 6000 instrumentation generating read lengths between 150 and 250 base pairs. Raw sequencing data underwent automated de novo assembly using SPAdes within the BioNumerics platform, adhering to stringent quality thresholds including minimum read coverages exceeding thirtyfold, average Phred quality scores above thirty, and assembly spans between 3.9 and 6.5 megabases. Sequence data were simultaneously integrated into the EnteroBase public repository to facilitate core genome multilocus sequence typing and hierarchical clustering at the HC5 threshold. Genomic antigen predictions demonstrated remarkable stability between somatic and flagellar markers across several key lineages: every identified O26 isolate expressed the H11 flagellar antigen, every O63 isolate bore H6, all O80 isolates possessed H2, all O91 isolates carried H14, and 385 out of 386 O157 isolates displayed the classical H7 flagellar type, whereas lineages like O113, O145, and O146 displayed multiple flagellar combinations.
Beyond classical serotyping, the research team analyzed the distribution of principal virulence genes, determining that 20.9 percent of all isolates carried stx1 alone, 47.2 percent carried stx2 alone, and 31.8 percent harbored combinations of both toxin families. Among isolates possessing stx1 alone, the stx1a subtype represented the overwhelming majority at 83.3 percent, whereas stx2a was detected in 40.0 percent of isolates harboring stx2 exclusively. The co-occurring profile stx1a together with stx2c represented the single most common multi-toxin arrangement, appearing in 44.8 percent of dual-positive isolates. Notably, longitudinal analysis revealed a sharp expansion in isolates carrying the stx2f gene subtype, which rose from 2.4 percent of national isolates in 2019 to 17.6 percent in 2023 and 11.6 percent in 2024. In terms of secondary virulence factors, the intimin adherence gene eae was identified in 76.4 percent of isolates, the enterohemolysin gene ehxA was detected in 78.8 percent, and 64.9 percent of the collection harbored all three primary pathogenic determinants.
To rigorously delineate the clinical determinants of severe disease, statisticians applied a binomial generalized additive model with a logit link function to data from 1,161 patients with verified clinical outcomes, among whom 129 individuals developed hemolytic uremic syndrome. This flexible statistical approach allowed investigators to control for complex, non-monotonic relationships between patient age and complications while evaluating specific toxin subtypes through Akaike Information Criterion forward stepwise selection. The statistical model demonstrated that the presence of the stx2a subtype was overwhelmingly associated with severe renal injury, imparting an adjusted odds ratio of 7.45 with a 95 percent confidence interval spanning from 3.84 to 14.43. Carriage of stx2d also significantly heightened complication likelihood, yielding an odds ratio of 3.05 with a confidence interval between 1.02 and 9.18. Conversely, isolates carrying stx1a alone exhibited an adjusted odds ratio of 0.32, indicating a statistically significant protective association relative to severe kidney complications.
The generalized additive model uncovered an unmistakable non-linear, U-shaped relationship between patient age and the mathematical probability of developing hemolytic uremic syndrome. Peak vulnerability occurred during early childhood, showing exceptionally high predicted probabilities among infants and toddlers followed by a precipitous decline as age advanced toward adolescence, with HUS cases proving exceptionally uncommon beyond age eight. A subtle uptick in estimated probability emerged among elderly demographics, though researchers cautioned that older cohorts contained sparse sample counts. Critically, zero cases of hemolytic uremic syndrome were detected among the 122 patients carrying the stx2f subtype, nor among the 41 individuals carrying stx1c alone. Among 106 stx2f-infected individuals with detailed symptom records, 71.7 percent experienced diarrhea or bloody stool, and half were under five years old. Neither stx2b nor stx2c displayed statistically significant associations with kidney failure, showing odds ratios of 0.88 and 1.32, respectively, while rare variants like stx1d, stx2e, stx2g, and stx2i occurred too infrequently for formal statistical inference.
Genomic surveillance facilitated by the EnteroBase core genome multilocus sequence typing scheme grouped 404 isolates, representing 32.7 percent of the national collection, into 111 distinct HC5 clusters containing between two and 27 bacterial genomes each. Most identified clusters remained modest in scale, with 81.1 percent involving only two to four closely related isolates. Epidemiological field investigations conducted by regional health authorities rarely pinpointed widespread commercial food contamination or environmental sources; instead, when investigators established definitive transmission pathways, clustered isolates almost exclusively involved person-to-person spread among individuals cohabiting within identical domestic households. Serotype O157:H7 was implicated in 44.1 percent of all genomic clusters, while serotype O26:H11 accounted for 17.1 percent, reflecting their pervasive community presence. Thirteen clusters exclusively comprised stx1-bearing organisms, 58 involved stx2-bearing bacteria, 37 encompassed dual-toxin strains, and three clusters exhibited divergent toxin profiles, while thirteen of the stx2 clusters consisted entirely of emerging stx2f organisms.
Synthesizing six years of genomic and clinical evidence, the Belgian reference center revised its national risk classification scheme, officially reclassifying stx2f from medium-risk down to low-risk for hemolytic uremic syndrome while maintaining stx2a and stx2d in the highest alert category. To prevent laboratory overburdening caused by escalating clinical sample referrals, authorities established targeted sequencing criteria prioritizing isolates from diagnosed HUS cases, all stx2-positive isolates excluding stx2f, all O26 and O157 strains, and any isolates connected to active outbreak investigations. The authors acknowledged notable methodological constraints, including potential genetic overrepresentation arising from household clusters and referral bias inherent in voluntary laboratory submissions, which matched mandatory notification records closely through 2021 before tracking slightly lower from 2022 to 2024. Additionally, clinical records lacked information regarding patient comorbidities and concurrent infections, underscoring that while genomic profiling provides an indispensable early triage mechanism, clinical risk assessments must always integrate microbiological markers with patient age and physiological status.
Subject of Research: Genomic surveillance and clinical risk factors of Shiga toxin-producing Escherichia coli infections in Belgium
Article Title: A six-year genomic surveillance of Shiga toxin-producing Escherichia coli infections in Belgium, 2019–2024
Article References: Crombé, F., Piérard, D., Cools, W., Hammami, N., Audistère, R., Rolin, F., Delbrassinne, L., Laisnez, V., Lang, C., & Vanmechelen, B. (2026). A six-year genomic surveillance of Shiga toxin-producing Escherichia coli infections in Belgium, 2019–2024. Frontiers in Public Health, 14, Article 2006847. https://doi.org/10.3389/fpubh.2026.2006847
Image Credits: AI Generated
DOI: 10.3389/fpubh.2026.2006847
Keywords: Shiga toxin-producing Escherichia coli, hemolytic uremic syndrome, whole-genome sequencing, Belgian National Reference Centre, genomic surveillance, serogroup O157, serogroup O26, stx2a subtype, stx2d subtype, stx2f subtype, stx1a subtype, bacterial epidemiology, culture-independent diagnostic testing, pediatric kidney complications, cgMLST cluster detection, virulence gene profiling, public health microbiology
News Source: Juliet Wilcox. (November 13, 2026). Belgian Study Identifies Shiga Toxin Genes Driving Severe Kidney Complications. Scienmag.



