Fecal samples were collected from 15 severely diarrheic calves (aged 1–4 months), with one calf selected from each of 15 distinct farms in Gansu Province, China. The selected subjects represented the most severe clinical cases of diarrhea on each farm at the time of sampling, with all calves exhibiting hemorrhagic diarrhea, to increase the likelihood of isolating pathogenic E. coli strains (totally 72 calves with diarrhea). To capture the broad genetic diversity of E. coli within the region and avoid over-representation of a single clonal type, only one isolate was included from each farm. Any calves that had received antibiotic treatment were excluded from the study. Prior to sample collection, the perianal area was disinfected using 0.1% benzalkonium chloride. Subsequently, a fecal sample should be collected directly from the rectum using a sterile swab. Samples must be stored in individually packaged Carr-Blair transport medium (Hopebio, China) and transported at a temperature between 5 °C and 25 °C. Storage time should not exceed 24 h.

Samples were streaked onto eosin-methylene blue agar and incubated at 37 °C for 16 h. A single metallic-sheen colony per calf was selected, cultured in LB broth (Hopebio, China), and identified as Escherichia coli via 16S rRNA gene PCR. From each confirmed isolate, an aliquot was used for DNA extraction, and a glycerol stock was prepared for long-term storage at −80 °C.

The disk diffusion method was employed to test the isolated strains for susceptibility to ampicillin (10 μg), ceftriaxone (30 μg), cefepime (30 μg), cefoxitin (30 μg), meropenem (10 μg), gentamicin (10 μg), amikacin (30 μg), kanamycin (30 μg), chloramphenicol (30 μg), tetracycline (30 μg), doxycycline (30 μg), levofloxacin (10 μg), ofloxacin (5 μg), ciprofloxacin (5 μg), and Trimethoprim-sulfamethoxazole (1.25/23.75 μg). This method was selected in accordance with the antimicrobial susceptibility testing criteria established by the Clinical and Laboratory Standards Institute (CLSI). The quality control strain employed in this study was Escherichia coli ATCC 25922.

Genomic DNA was extracted from overnight cultures of each E. coli isolate using a commercial kit (Tiangen, China). DNA quality and concentration were verified by agarose gel electrophoresis and NanoDrop spectrophotometry.

Sequencing libraries were prepared from fragmented DNA, and paired-end sequencing was performed on an Illumina HiSeq X platform (Illumina Inc., USA). After quality control and adapter trimming, the high-quality reads were assembled de novo using SPAdes v3.11.1. The resulting genomes were annotated using Prodigal v2.6.2.

The sequence type (ST) of each isolate was identified using the MLST tool¹ in conjunction with the PubMLST database. In order to ascertain the distribution of O and H serotypes, analysis was performed using the ECTyper.² The phylogroup for each strain was identified in silico using the ClermonTyping.³ The virulence factor profiles of E. coli were identified using the VFDB⁴ and antibiotic resistance genes using CARD,⁵ ResFinder,⁶ MEGARes⁷ and ARG-ANNOT⁸ databases, respectively. The PointFinder⁹ was utilized for the purpose of detecting site mutations in antibiotic resistance genes within the chromosome. The minimum coverage and sequence similarity thresholds for screening were both set at 80%. Plasmid replicon types were identified using the PlasmidFinder v2.1.¹⁰ Utilizing E. coli K12 (Genebank Accession: GCA_000005845) as the reference genome, core single-nucleotide polymorphisms (SNPs) were generated with Snippy,¹¹ and a maximum likelihood tree was constructed for 15 E. coli strains using FastTree v2.1.10¹².

To contextualize the isolates from this study within the broader population of E. coli associated with bovine diarrhea in China, a complete reference genome was retrieved from the National Center for Biotechnology Information (NCBI) database via the Bacterial and Viral Bioinformatics Resource Center (BV-BRC)¹³ in September 2025. The search and screening criteria were as follows: Initially, searches were conducted with the keywords “cattle” and “diarrhea,” employing “Escherichia coli” as the host species. Secondly, only domestic isolates for which complete genome sequencing and assembly were available were selected. Thirdly, in order to ensure data quality and comparability, only genomes with a completion status of “Complete” or “WGS” and a genome quality of “Good” were included. In order to provide a comprehensive reflection of the national epidemiological landscape, genomes originating from diverse geographic regions within China were included in the study. Phylogenetic analysis was performed using the Roary v3.13.0¹⁴ workflow to generate alignment sequences of the core genomes of isolated strains. Subsequently, the maximum likelihood phylogenetic tree was constructed using the FastTree v2.1.10 software, based on the GTR + CAT evolutionary model. The Average Nucleotide Identity analysis of E. coli was conducted using the FastANI.¹⁵ Principal coordinate analysis is predicated on COG functions of bacterial strains, utilizing the Bray-Curtis Faith distance algorithm.

The assembly of each draft genome for the E. coli isolate was evaluated, and sequencing depth and number of contigs are reported in Supplementary Table 1. The Sequencing Depth for individual genomes was between 201X and 385X, and the number of contigs ranged from 51 to 94.

The analysis of the sequences enabled the identification of nine distinct serotypes, including O86:H28 (n = 2), O75:H9 (n = 2), O5:H10 (n = 2), O115:H21 (n = 2), O9:H9 (n = 1), O88:H8 (n = 1), O117:H12 (n = 1) and O110:H2 (n = 1). One strain (Eco11) was classified as a novel serogroup, designated OgN31. A total of ten sequence types were identified among the analyzed isolates: ST58 (n = 4), ST156 (n = 2), ST43 (n = 2), and one isolate each of ST101, ST187, ST446, ST744, ST2522, ST3519, and ST12547. Phylogroup assignment revealed that 11 isolates (73.3%) belonged to phylogroup B1, while the remaining 4 isolates (26.7%) were classified as phylogroup A.

In the present study, a total of 15 E. coli isolates were subjected to a detailed plasmid replicon typing analysis, which yielded noteworthy results. The analysis revealed the presence of distinct plasmid carriage patterns, with a particular emphasis on the prevalence of IncF-type plasmids. Specifically, the IncFIB replicon was identified in 12 out of 15 strains, frequently in combination with IncFII (e.g., Eco2, Eco5, Eco6, Eco7, Eco12, Eco13) or IncFIC (e.g., Eco10, Eco14, Eco15). Strains such as Eco1 and Eco9 exhibited more diverse and complex replicon profiles, including IncHI2 and IncI1-I(Alpha). Conversely, Eco3 and Eco8 exhibited no detectable plasmids (Table 1).

A comprehensive screening of virulence genes across 15 E. coli isolates Eco1–Eco15 was conducted, resulting in the observation of significant heterogeneity in virulence gene profiles (Supplementary Table 2). Isolates such as Eco1, Eco5, Eco6, and Eco7 exhibited a high abundance of virulence determinants, including genes associated with adhesion (e.g., afa cluster, fim genes), toxin production (e.g., cdt, cnf1, hlyA), and iron acquisition systems (e.g., fyuA, irp2, iuc/iut). In contrast, strains such as Eco11, Eco12, and Eco13 exhibited a more limited virulence repertoire, predominantly retaining only core colonization genes such as those involved in curli formation (csg operon) and enterobactin synthesis (ent genes). As depicted in the heat map (Figure 1A), a comparison of the virulence profiles of the strains under analysis is available.

Based on the circular phylogenetic tree constructed from SNP data, the 15 Escherichia coli isolates are clearly segregated into 3 distinct clades, revealing strong phylogenetic signals consistent with their sequence types (STs) and serogroups (SGs) (Figure 1B). Isolates of ST58 (Eco10, Eco12, Eco13, Eco14) form a closely related cluster, further subdivided by serotypes O115:H21 and O75:H9, indicating microevolution within this lineage. Similarly, ST43 (Eco5, Eco6) and ST156 (Eco3, Eco8) each constitute monophyletic groups, corroborating their genetic relatedness. Notably, strains sharing the same ST and SG (e.g., Eco5 and Eco6: ST43, O5:H10) are positioned adjacently, underscoring the concordance between genomic background and surface antigen profile. All ST58, ST156, ST101, ST187, ST446, and ST2522 strains were identified as phylogroup B1, whereas ST43 and ST744 strains were assigned to phylogroup A.

A subsequent analysis of the antimicrobial resistance profiles indicated a high prevalence of multidrug resistance among bovine E. coli isolates. However, the genetic basis of this resistance was found to be more complex than initially assumed (Figure 2). While phenotypic resistance to sulfonamides was widespread, the canonical genes sul1, sul2 were only detected in a limited number of strains (e.g., Eco1 and Eco3). Detection of mutations in the chromosomal drug resistance gene revealed mutations in gyrA and parC in the Eco1, Eco3 and Eco8 strains, as detailed in Supplementary Table 3. This finding suggests the probable existence of alternative, hitherto uncharacterized resistance mechanisms that are likely responsible for the predominant resistance phenotypes observed. Conversely, the presence of efflux pump genes (e.g., acrAB-tolC) was observed across all samples, indicating a fundamental contribution to the intrinsic resistance characteristics of this population. The ST43 lineage (Eco5, Eco6) demonstrated clonal dissemination of an identical, complex resistance profile, whereas ST58 strains exhibited marked heterogeneity, thereby underscoring the combined roles of clonal spread and horizontal gene transfer in shaping the AMR landscape.

The pangenome-based phylogenetic analysis of Escherichia coli isolates from bovine diarrhea in China revealed that the population structure was primarily shaped by serotype, with distinct bovine-associated lineages identified (Supplementary Table 4). The study identified clinically relevant serotypes, including O5:H10 (Eco5, Eco6), O75:H9 (Eco12, Eco13), and O115:H21 (Eco10, Eco14), which formed well-supported, monophyletic clusters. These findings suggest the emergence and clonal dissemination of specific pathogenic lineages within Gansu China cattle. Conversely, the polyphyletic distribution of sequence types, such as ST58, suggests that this multi-locus genotype has been acquired independently by genetically distinct strains, likely through horizontal gene transfer of housekeeping genes within the bovine environment. It is noteworthy that the phylogenetic tree did not exhibit clear clustering by geographic origin, which could be indicative of potential inter-regional transmission events or the circulation of common ancestral lineages across China. The inclusion of Chinese reference genomes from diverse sources further contextualizes these clinical isolates within the broader population of bovine E. coli in China, highlighting the circulation of both common and unique serotypes associated with bovine diarrhea (Figure 3).

The ANI and COG-based PCoA analyses consistently revealed pronounced genomic divergence among the bovine diarrheagenic E. coli isolates, with clustering patterns largely concordant with serotypes and sequence types. In the ANI heatmap, isolates such as Eco5 and Eco6 (both ST43, O5:H10) exhibited high genomic similarity (ANI > 99%), forming a tight cluster with reference genomes, indicating clonal relatedness. Similarly, strains Eco12 and Eco13 (both ST58, O75:H9) showed high ANI values, supporting their close evolutionary relationship. In contrast, strains from polyphyletic STs (e.g., ST58 members Eco10, Eco14) displayed lower ANI with other ST58 isolates, confirming genetic heterogeneity within this lineage (Figure 4A). The PCoA based on COG functional profiles further supported this structuring, with the first principal coordinate (PCo1) explaining 75.92% of the variance, clearly separating strains into distinct functional groups. Notably, the clustering in PCoA correlated with both phylogenetic clades and virulence gene profiles, suggesting that functional differences in COG categories–particularly those involved in adhesion, toxin production, and iron acquisition–may underpin the ecological adaptation and pathogenic specialization of these lineages (Figure 4B).