A total of 948 fecal samples were collected between June 2023 and March 2024 from abattoirs and farms located across seven regions, representing the majority of Chile’s cattle population (25). The sampled animals included both juveniles and adults, with most being of mixed breed. Approximately 20 g of fecal material per animal was aseptically collected directly from the rectum by trained veterinarians and transported under refrigerated conditions in sterile flasks until laboratory processing. All sampling procedures were approved by the Institutional Committee of Care and Use of Animals, Universidad de Chile (Protocol No. 23658—VET—UCH).

Samples were processed following protocols from previous studies (26, 27). Briefly, 5 g of each fecal sample were enriched in 9 mL of tryptone soy broth (Becton Dickinson and Co., United States) and incubated overnight at 42 °C. A 25 μL aliquot of the enrichment culture was then plated onto MacConkey agar (Becton Dickinson and Co., United States) and incubated at 37 °C for 18–24 h.

Bacterial growth from confluent areas was resuspended in 500 μL of sterile nuclease-free water, subjected to heat treatment at 100 °C for 15 min, and centrifuged at 26,480 × g for 5 min. DNA concentration and purity (A260/A280 ratio) were determined using a NANO-400 micro-spectrophotometer (Hangzhou Allsheng Instruments Co., China). Samples with optimal purity (1.8–2.0) were stored at −20 °C for subsequent analyses.

The presence of stx1 and stx2 genes was confirmed by multiplex PCR (28) in a LifeECO^® thermal cycler (Hangzhou Bioer Technology Co., China). The STEC97 strain [stx1-positive, eae-positive, stx2-positive (29)] was used as a positive control, while E. coli ATCC 25922 served as the negative control. Up to 30 colonies per positive sample were individually plated on MacConkey agar (Becton Dickinson and Co., United States) and CHROMagar^™ STEC (CHROMagar Microbiology, France) (30, 31). After 24 h of incubation, colonies were screened by multiplex PCR to confirm the presence of stx1 and/or stx2 genes. PCR-confirmed colonies were then tested for the uspA gene, encoding the universal stress protein A, to verify E. coli species identity (32). A single confirmed isolate per sample was stored at −80 °C for further analysis.

Genomic DNA from all STEC strains was extracted using the Wizard Genomic DNA Purification Kit (Promega, United States), following the manufacturer’s instructions. DNA concentration was measured using fluorometry with the Qubit^® dsDNA BR Assay kit (Life Technologies, United States), and DNA quality was assessed with an Epoch microplate spectrophotometer (Biotek Instruments, United States). A total of 1 ng of DNA was used for library preparation using the Nextera XT DNA Library Prep Kit (Illumina, United States), following the manufacturer’s protocol. The average fragment size of libraries was determined by capillary electrophoresis using the High Sensitivity NGS Fragment Analysis Kit (Advanced Analytical Technologies, United States). Libraries were quantified using the KAPA Library Quantification Kit (Kapa Biosystems, United States) on a Rotor-Gene Q platform (Qiagen, Germany). Whole-genome sequencing (WGS) was performed on a NovaSeq X Plus platform (Illumina, United States) with a 150-cycle paired-end reagent kit at Haplox Company, Hong Kong.

Additionally, two STEC strains isolated from clinical human stool samples, corresponding to serotypes O157:H7 and O26:H11 and provided by the Instituto de Salud Pública de Chile, were included in the analysis for comparison purposes. These strains were processed for WGS as described above. All genome sequences were deposited in GenBank under BioProject number PRJNA656305.

To enhance the comparative analysis, 568 publicly available E. coli genome sequences were retrieved from GenBank’s Sequence Read Archive (SRA)¹ on December 15, 2024. Genomes were selected based on diverse geographical origins, host species, and serotypes to broadly represent the global diversity of STEC strains.

All genomes were screened for the presence of stx subtypes by mapping reads against reference sequences using BWA (33). After filtering, a final dataset comprising 546 confirmed STEC genomes was obtained for downstream analysis. Metadata, including host origin, country of isolation, and year, were retrieved using the eSearch tool from the EMBOSS suite (34).

All FASTQ Illumina reads were assembled de novo using SPAdes (v.3.15.2) with default parameters (35). Genome assembly quality was evaluated using CheckM2 (36), which estimates genome completeness and contamination based on machine learning models trained with lineage-specific marker sets available in the DIAMOND² database.

The program was not run using a specific model; a cosine similarity calculation was performed to determine the appropriate completeness model for each isolate. The program predicts protein sequences to annotate all the genomes with DIAMOND. Finally, in all cases the Neural network contamination model was used.³ This quality control step was critical to ensure the reliability of assemblies, particularly for genomes retrieved from the SRA. Only genomes with a completeness score greater than 99.6%, as estimated by CheckM2, were included in the analysis.

Prediction of stx subtypes was performed on ABRicate (v.0.8.13).⁴ Sequence types (STs) of all STEC strains were predicted by Achtman’s multilocus sequence typing (MLST) scheme using the GitHub platform.⁵ The housekeeping genes used included adk, fumC, gyrB, icd, mdh, purA, and recA. Additionally, SerotypeFinder 2.0⁶ was used to determine serotype (37).

Adhesin gene analysis was conducted using AdhesiomeR,⁷ an R-based tool that executes BLASTn searches against a curated database of 427 adhesin genes identified across diverse E. coli strains. Adhesin sequences were classified according to sequence identity thresholds: highly similar (>95%), moderately similar (75–95%), and unrelated (<75%) (12). In order to enhance confidence in adhesin identification, a strict mode with gene-specific bit score threshold was applied. Comparative analyses were based on presence/absence matrices (1 = present; 0 = absent) and clustering profiles to detect host-specific adhesin signatures between human- and cattle-associated STEC isolates. Differences in the detection rates of adhesin genes between cattle- and human-associated isolates were analyzed using a Z-test for two proportions. The analysis was performed in Microsoft Excel (Microsoft Office 365, version 16.100.2), with statistical significance set at p < 0.05.

To further explore genetic variability, we performed genome-wide association analysis (GWAS) using two complementary approaches. First, we used the complete genome of E. coli K-12 (ASM584v2) as the reference, which contains the majority of core adhesin genes and allowed variant calling and functional annotation at the whole-genome level. Second, recognizing that several adhesin genes were absent from the K-12 genome, we conducted an additional GWAS focused on the adhesiome. For this, we extracted all 427 adhesin gene sequences identified in our adhesiome analysis—including those shared between human- and cattle-associated strains—to ensure a more comprehensive assessment. These sequences were retrieved from AdhesiomeR⁸ and used as templates for further analysis. The complete genome sequences of STEC strains isolated from stool samples (n = 158 from cattle and n = 205 from humans) were mapped to the E. coli K-12 genome and to the adhesiome sequences using BWA (38). After sorting using BamTools (39), the PCR duplicates were removed using Sambamba (40). Single nucleotide polymorphisms (SNPs) and short insertions/deletions (indels) were identified using FreeBayes and jointly using all the samples in the analysis (41).

GWAS was conducted based on SNPs and indels identified from variant call format (VCF) files generated by FreeBayes. Quality control of variants was performed using VCFtools, applying a minimum quality threshold (MinQ) of 30 and a mean depth filter of 100× (42). A logistic regression model was fitted to the data, using host species (cattle = 1, human = 0) as the binary outcome, following the approach described by Pérez-Enciso et al. (43).

We used Pyseer (44) to discover variants significantly associated with the cattle host, being humans deemed as controls in the standard association analysis. To account for potential confounding by population structure, we conducted a principal component analysis (PCA) on variant call format (VCF) files using PLINK [-pca command (45)]. The model included the first five principal components as fixed effects, as eigenvalues showed no substantial drop beyond this threshold. The pairwise distance matrix was generated using Mash (46). Following the Benjamini and Hochberg correction, the p-values were adjusted for multiple comparisons using the false discovery rate (FDR) method, implemented through the p.adjust function in R (FDR ≤0.05). Model coefficients were also presented as odds ratios (OR), with values greater than one indicating an association of the alternative allele with the cattle host.

Annotation and functional impact prediction of host-associated variants were performed using snpEff (47). For GWAS analyses based on the E. coli K-12 genome, the pre-compiled E. coli str. K-12 substr. MG1655 database provided by SnpEff was employed. For the adhesiome sequences that belonged to different bacterial strains, we constructed a custom SNPeff database. The adhesiome sequences were annotated using AUGUSTUS (48), with the E. coli K-12 database selected as the reference strain to guide the annotation (obtained as a GFF file format). This annotation file was then converted to GTF format using AGAT to develop the snpEff custom database of the adhesiome. Predicted coding sequences (CDS) and protein sequences of the identified genes were obtained using the script getAnnoFasta.pl (available at https://github.com/nextgenusfs/augustus.git). All these files were used to compile the snpEff database.

Of the 948 fecal samples collected from cattle, 71 (7.5%) were confirmed as STEC-positive, harboring stx1 and/or stx2 genes along with the uspA marker. Among these, 70.4% harbored stx2, 26.8% carried both stx1 and stx2, and only 2.8% were positive for stx1 alone. The geographical distribution of STEC varied across Chile, with regional differences in detection rates (Figure 1).

For comparative purposes, 546 publicly available STEC genomes were retrieved from NCBI, encompassing isolates from humans (50.0%), food (23.8%), and cattle (19.4%). These genomes spanned a broad temporal range (1978–2021) and were predominantly collected from Germany (43.8%), Chile (16.1%), and France (13.2%).

Among the 619 genomes analyzed (71 Chilean isolates plus 546 public genomes), we identified 29 distinct stx subtypes profiles, with stx2a (20.0%) and stx1a (18.9%) being the most prevalent. Serotype analysis indicated O157:H7 (10.8%), O26:H11 (7.1%), and O130:H11 (6.0%) as the most frequently detected. MLST classification identified 141 STs, with ST11 (13.4%) and ST297 (8.7%) being the most common. Complete metadata is provided in Supplementary Table 1. Figures 2, 3 depict the distribution of serotypes and STs according to geographical origin and host.

Among the 71 STEC strains isolated in this study, eight stx subtypes profiles were identified, with stx2d (38.0%), stx2c (26.8%), and stx1a + stx2a (12.7%) being the most prevalent. Among the 12 predicted serotypes, O130:H11 (42.3%), O185:H7 (21.1%), and O113:H21 (8.5%) were the most common. Likewise, the most frequently detected STs were ST297 (49.3%), ST2387 (20.5%), and ST223 (9.6%).

Due to the high diversity in serotypes, stx profiles, and STs, we examined the genomic structure of STEC isolates from food, cattle, and humans (n = 572) to assess whether genomic similarity clusters corresponded to host origin or geographic region. Based on STs, we generated MinHash sketches of draft whole-genome assemblies using k-mers of length 31 and a sketch size of 100,000 in Sourmash. This analysis revealed substantial genomic variability between groups, potentially enhancing STEC adaptability to diverse environments and supporting a wide range of virulence strategies (Figure 4). In parallel, genetically homogeneous clades were observed within specific STEC STs, suggesting associations with ecological factors or adaptive niches, further supported by the multidimensional scaling analysis (Supplementary Figure 1).

In this study, 71 STEC strains (7.5%) were identified among the collected stool samples. This detection rate is lower than in previous reports (27, 52, 53) and may reflect several factors, including improved farm biosecurity, differences in feeding practices, and environmental conditions. Seasonal and geographical variation in STEC prevalence has been widely documented, with higher recovery rates during spring and summer (54–56). Ambient temperature, rainfall, and vector abundance have been suggested as drivers of these seasonal trends, since warmer conditions may favor STEC persistence outside the host and increase exposure opportunities. Feeding practices can also influence bacterial shedding; animals fed forage typically shed fewer STEC than grain-fed cattle (22, 57). In our study, most samples were collected during winter and early spring, a period characterized by lower temperatures and high rainfall, which may reduce environmental persistence and transmission. Moreover, in the regions analyzed, most animals were pasture-fed. Together, these factors may at least in part explain the relatively low prevalence observed. Future longitudinal studies will be required to further elucidate these factors. Importantly, despite this lower detection rate, the strains recovered are consistent with those reported in previous studies (26, 27, 58) and are representative of the STEC populations currently circulating in Chilean cattle.

Among the isolates recovered, the majority carried stx2 subtypes, which are strongly associated with severe human disease (59). From a global perspective, stx2 was also the predominant subtype among all analyzed STEC genomes, reinforcing concerns about the threat posed by circulating STEC strains, as Stx2 exhibits both greater cytotoxic activity and higher affinity for host ribosomes (60).

In Chile, the predominant STEC serotypes identified were O130:H11 (ST297), O185:H7 (ST2387), and O113:H21 (ST223), consistent with previous reports (26, 58), suggesting the persistence of specific strains within cattle reservoirs. In contrast, global data showed a higher prevalence of O157:H7 (ST11) and O26:H11 (ST21), along with O130:H11 (ST297). Our results underscore the considerable genomic variability in STs and serotypes among circulating STEC strains.

Notably, O130:H11 (ST297) predominated among cattle isolates, whereas O157:H7 (ST11) was more common among human isolates, consistent with global patterns (26, 61–64). O157:H7, which carries the LEE pathogenicity island and is strongly associated with the stx2 gene, is the serotype most frequently linked to severe human disease (59, 65). Conversely, O130:H11, a LEE-negative serotype, has been implicated in sporadic outbreaks (66–68). Despite lacking the LEE locus, O130:H11 strains harbor virulence factors such as ehxA, saa, sab, lpfA, and iha, which may contribute to severe human disease (62, 69, 70).

Although seropathotype classification (based on ST or serotype) has historically been employed to assess the association of STEC strains with HUS and human outbreaks (71), its reliability is limited by the high genetic plasticity of STEC. The distribution of diverse serotypes across regions and hosts complicates epidemiological surveillance and strain tracking (72). Given these challenges, prioritizing molecular markers related to virulence and host adaptation over traditional serotyping could improve monitoring protocols and control strategies. This strategy would strengthen epidemiological surveillance, veterinary health measures, and public health interventions, ultimately enabling more effective and targeted control of STEC.

The high genomic variability of STEC facilitates its adaptation to diverse hosts through mechanisms such as recombination and horizontal gene transfer, impacting both virulence and transmission dynamics (49, 73, 74). However, despite this genetic plasticity, the existence of genetically homogeneous clades within specific STs suggests that certain STEC lineages have undergone niche specialization, likely driven by host- or environment-specific selective pressures. For example, at the bovine RAJ, selective pressures such as predation by bacterivorous protozoa may have favored strains encoding Stx and LEE genes, which inhibit protozoal grazing and promote persistence (50).

The evolutionary strategies of STEC may account for the greater adhesin diversity observed in human-associated strains compared to cattle-associated strains (49). Our analysis, combined with the AdhesiomeR cluster classification, confirmed that ehaA, ehaG, stgA-C, yadL-N, and iha were significantly more prevalent in cattle strains. EhaA, an autotransporter protein, facilitates rapid cell aggregation, biofilm formation, and adhesion to bovine RAJ epithelial cells when overexpressed in E. coli K-12 (19). Similarly, EhaG enhances autoaggregation, biofilm formation, and binding to collagens I–V, laminin, fibronectin, and fibrinogen (75). Although Stg adhesins have been implicated in adhesion to human and avian epithelial cells (76), their specific role in STEC gut colonization remains unclear. The Yad fimbrial autotransporter family modulates gene expression and virulence in STEC O157:H7 (77), potentially contributing to RAJ colonization (78). Additionally, Iha, a dual-function adhesin and siderophore receptor, is widely distributed among LEE-positive and LEE-negative STEC strains, and is often carried on mobile genetic elements, facilitating horizontal dissemination (11, 79, 80). Collectively, these findings suggest that cattle-adapted STEC strains rely on a relatively conserved adhesin repertoire optimized for stable colonization, biofilm formation, extracellular matrix adhesion, and iron acquisition.

In contrast, human-associated STEC strains exhibited greater diversity in adhesins, including fimbrial and non-fimbrial types, suggesting a more flexible colonization strategy adapted to heterogeneous environments. These strains showed a significantly higher representation of eae, cah, ypjA, and paa genes. Eae is a key virulence factor in LEE-positive STEC, leading to attaching and effacing lesions (81). Similarly, cah (calcium-binding antigen 43 homologous) is predominantly found in LEE-positive strains (82), promoting autoaggregation, biofilm formation, and persistence in STEC O157:H7 (81, 83). YpjA (homologue of EhaD) has been shown to enhance biofilm formation in STEC O157:H7, reinforcing its contribution to intestinal colonization (19). Meanwhile, paa (porcine attaching and effacing-associated protein) has been associated with eae-positive strains from both animals and human origins, suggesting a synergistic role with intimin-mediated adhesion (49, 84). These findings indicate that human-associated STEC strains rely on a broader adhesin repertoire than cattle strains, which may reflect adaptation to a more variable intestinal environment and stronger immune pressures.

Differences in adhesin gene abundance between cattle- and human-associated STEC strains align with distinct functional enrichments revealed by gene ontology analysis, supporting the notion of host-specific adaptation strategies. In cattle-associated strains, enriched biological processes point to strong selection for biofilm formation and attachment mechanisms, particularly involving fimbrial and non-fimbrial adhesins (85), autotransporter adhesins such as EhaA (86), chaperone-usher adhesins like Yad (87), and the adhesin-siderophore Iha (79). Conversely, human-associated strains exhibit significant enrichment in methylglyoxal detoxification, a response to a toxic glycolysis byproduct (88), potentially indicating a metabolic shift toward glyoxylate cycle activation. This finding suggests a potential metabolic shift toward glyoxylate cycle activation, favoring carbon conservation and survival under nutrient-limited conditions, typical of the human intestine (89). Moreover, methylglyoxal detoxification mechanisms also contribute to the neutralization of reactive oxygen species (ROS), enhancing bacterial survival against host immune defenses (90). Altogether, these observations indicate that adaptation to oxidative and metabolic stress represents a key selective force shaping the genomic and functional landscape of human-associated STEC strains.

GWAS represent powerful tools for identifying genetic variants associated with bacterial adaptations, including adhesion, by linking genotype to phenotype while accounting for confounding factors such as population structure (91). Applying GWAS to pathogenic bacteria improves our understanding of virulence mechanisms and host-pathogen interactions; however, relatively few studies have explored the population structure and adaptation of STEC using this approach. Some investigations have focused on associating genomic features with pathogenic properties in STEC. For instance, Matussek et al. (92) analyzed 238 STEC genomes from patients with and without HUS to identify genetic predictors of disease severity. Their integrative approach, combining serotyping, stx subtyping, virulence profiling, phylogenomics, and a pangenome-wide association study (PWAS), showed that O157:H7 clade 8 strains and stx2a or stx2a + stx2c subtypes were strongly associated with HUS, while stx1a was more frequent in non-HUS cases. Virulence genes related to adherence (eae, tir, paa), toxins (toxB, astA), and type III secretion system proteins were enriched in HUS isolates, and hundreds of accessory genes were linked to severe disease, including adhesins (yfcP, yehD, elfG, sfmA) and regulators, although many encoded hypothetical proteins. The authors concluded that severe human disease results from the interplay between canonical virulence determinants, accessory genetic elements, and host–pathogen interactions. Similarly, Peroutka-Bigus et al. (93) compared genomic and phenotypic features of three human outbreak-associated and one cattle-derived STEC O157:H7 isolates to assess host adaptation. Despite differences in virulence gene expression, adherence, and Stx production among outbreak isolates, no significant differences were detected in cattle colonization or shedding compared with the cattle-associated strain. This highlights that genomic and phenotypic variation in STEC O157:H7 does not necessarily correspond to host-specific adaptation, emphasizing that host specificity cannot be inferred solely from genetic or phenotypic traits. More recently, Espadinha et al. (94) performed a PWAS of 531 STEC isolates, identifying associations between the development of HUS and the presence of stx2a, stx1a + stx2a, or stx1a + stx2c, as well as the co-occurrence of genes such as ygiW (stress-induced protein) with group_5720 (transcriptional regulation) and pfkA (6-phosphofructokinase-1) with fieF (Zn²⁺/Fe²⁺/Cd²⁺ efflux transporter), among other epidemiological factors. In the same context, Marques Da Silva et al. (95) explored the genomic determinants of cattle colonization by comparing the genomes of STEC O22:H8 and O157:H7 strains. They identified 28 virulence-associated genes unique to O22:H8, primarily involved in adherence (e.g., cfaA, cfaB, cfaC, cfaD/cfaE, sisA, lesP, hes, pagC, tpsA, and tpsB), autotransporters (ag43), and invasion (tia). These findings highlight the complexity of adhesin gene distribution and function in STEC, reinforcing the need for further research to elucidate the genetic basis of host-specific colonization.

Beyond the host-specific differences between cattle- and human-associated STEC strains, our GWAS comparing these isolates with the non-pathogenic E. coli K-12 genome provides additional evidence for host-driven selection, highlighting potential adaptive traits that distinguish STEC from commensal E. coli strains. This analysis revealed significant enrichment of trehalose transport and protein-phosphocysteine-trehalose phosphotransferase system activities, associated with the bglF and ascF genes. It is well established that E. coli can use trehalose as an alternative carbon source and synthesize it intracellularly to counteract osmotic stress by stabilizing membrane integrity (96). The BglF and AscF proteins, key components of the phosphotransferase system, mediate the transport and phosphorylation of β-glucosides, including trehalose, thus regulating its metabolism and contributing to the bacterial stress response (97). These metabolic adaptations may enhance the resilience of STEC strains to osmotic and environmental stresses encountered in the bovine gastrointestinal environment.

Our GWAS comparing the adhesiome of cattle- and human-associated STEC with E. coli K-12 revealed distinct host-specific adhesion strategies. In cattle strains, yadK and ybgP were significantly associated. YadK, a chaperone-usher adhesin, enhances acid stress resistance, biofilm formation, and epithelial attachment, promoting persistence in the bovine gastrointestinal tract (98, 99). YbgP facilitates adhesion to epithelial and abiotic surfaces, promoting environmental persistence (87, 100). In contrast, human-associated strains exhibited significant enrichment of ybgQ, yfcS, and flu. ybgQ encodes the usher protein necessary for YbgP fimbriae assembly, suggesting a role in epithelial colonization (101). YfcS enhances biofilm formation and bacterial aggregation (87, 100), while flu (also known as agn43), an autotransporter adhesin, promotes microcolony formation, biofilm stability, and immune evasion in both LEE-positive and LEE-negative STEC strains (11, 102). These findings suggest that cattle-associated STEC strains prioritize adhesion mechanisms suited for long-term gut colonization, whereas human-associated strains exhibit a broader adhesin repertoire, likely reflecting selective pressures favoring host invasion, immune evasion, and adaptation to the intestinal niche.

Notable differences emerged when analyzing the complete adhesiome of cattle- and human-associated STEC strains compared to the analysis based solely on the E. coli K-12 genome. In cattle-associated strains, yeeJ, espP, and fimC exhibited the highest significance values and OR, suggesting a prominent role in host-specific colonization and adaptation, likely driven by selective pressures within the bovine gastrointestinal tract. yeeJ encodes an autotransporter protein with structural similarity to intimin, facilitating biofilm formation and enhancing STEC O157:H7 binding to eukaryotic cells (103, 104). EspP, a serine protease autotransporter of Enterobacteriaceae family, encoded on STEC virulence plasmids, promotes adhesion to bovine rectal epithelial cells, intestinal colonization, and supports biofilm formation and HeLa cell adherence (105, 106). Its proteolytic activity targets extracellular matrix and mucus proteins, facilitating tissue penetration and receptor exposure for other adhesins. fimC, a key component of the fim operon, plays a critical role in the assembly of type 1 fimbriae, which promote epithelial adhesion and biofilm formation in E. coli. In STEC O157:H7, type 1 fimbriae have been implicated in RAJ cell colonization, highlighting their potential contribution to cattle adaptation and long-term persistence (107, 108).

Conversely, clpV, ybgQ, and sab exhibited a highly significant OR below one, suggesting their association with human-host adaptation. ClpV, a cytosolic ATPase and essential component of the type VI secretion system (T6SS), facilitates bacterial competition by delivering effector proteins—such as peptidoglycan hydrolases, phospholipases, and DNases—into target cells (109, 110). In the STEC O157:H7 strain EDL933, ClpV mediates the translocation of catalase into macrophages, promoting immune evasion; notably, deletion of clpV reduces lethality in murine infection models (111). Furthermore, ClpV has been associated with HUS-producing STEC strains, supporting its potential role in virulence (112, 113), although its precise contribution to intestinal colonization remains unclear. Similarly, Sab, a plasmid-encoded autotransporter, enhances adherence to human epithelial cells and biofilm formation in LEE-negative STEC strains (114), potentially facilitating intestinal colonization and persistence.

In addition to the gene presence/absence patterns identified through GWAS, variant analysis offered deeper insights into host-specific selective pressures shaping adhesin functionality. Human-associated STEC exhibited greater genetic diversity, although the predicted functional impacts on adhesins were largely conserved across hosts. A key observation was the presence of SNPs with moderate-to-high effects on protein sequences. In particular, given ybgQ’s role in adhesion and outer membrane integrity, structural disruptions caused by SNPs may significantly affect bacterial adherence and survival within the human gut. Conversely, cattle-associated strains exhibited alternative mutations with moderate-to-high effects, especially in yadK, espP, and ybgP, where near-fixation frequencies suggest strong host-driven selection favoring these variants for bovine colonization. Notably, different espP alleles exhibit distinct biological activities: EspPα and EspPγ are secreted and enzymatically active, whereas EspPβ and EspPδ display reduced or absent proteolytic function (115). Among them, EspPγ specifically cleaves pepsin and coagulation factor V in humans (116), while EspPα is more frequently found in human isolates. In contrast, other EspP variants are predominantly associated with animal reservoirs and environmental sources (117). These host-driven selective pressures likely induce functional modifications that may either enhance or attenuate virulence. Further experimental studies, including in vitro and in vivo infection models, are necessary to clarify the impacts of these alternative variants on colonization efficiency and adhesin functionality.

A potential limitation of our GWAS is the uneven geographic distribution of publicly available STEC genomes, with an over-representation of isolates from countries such as Chile, Germany, and France. However, our analyses were focused on host origin rather than country of isolation, and GWAS methods corrected for population stratification were applied to mitigate this potential bias. In addition, the dataset encompassed diverse serotypes and lineages, which supports the robustness of the host-specific associations identified.

Targeting adhesins in STEC offers a promising avenue for developing effective intervention strategies. Given the essential role of adhesins in host colonization, strategies such as anti-adhesin antibodies, competitive inhibitors, or adhesin-based vaccines could significantly reduce bacterial adherence and gut colonization, ultimately lowering transmission risk. These approaches could be particularly beneficial in pre-harvest control programs designed to decrease STEC carriage in cattle. Future research should focus on identifying adhesins with high conservation across STEC strains and evaluating their potential as preventive targets, including efficacy assessments through animal model studies.