Sporadic deaths were reported in a newly established breeding farm in Wenchang, Hainan during the winter 2021–2022 following the introduction of new breeds. During this period, a total of six dead sheep including five adults and a lamb were collected for necropsy. According to a previous study (Rahsan et al., 2018), clinical signs including wounds, ulcers, skin lesions, excrements, and secretions from nose, eyes and mouth of each carcass were recorded before autopsy. Following dissection, bronchial exudates and lesions of the internal organs were examined visually and multiple systematic incisions.

Following collection, tissue samples from lung, liver and other organs were fixed with 10% neutral buffered formalin. Subsequently, the samples were sent to Wuhan Saville Biotechnology Co., Ltd.¹ for paraffin embedding, sectioning and staining. The stained slides were examined under a light microscope at 10× and 100× for histopathological evaluation (Rehman et al., 2021; Mekibib et al., 2019).

For mycoplasma detection, a piece of lung tissue (0.2–0.5 g) was homogenized and centrifuged at 8, 000 rpm for 10 min followed by DNA extraction from 200 μL supernatant using the TIANamp Genomic DNA Kit (TIANGEN, Beijing, China). Specific primers were designed (Supplementary Table S1) and used in a polymerase chain reaction (PCR) (Young et al., 2010). In this paper, double-distilled water (ddH₂O) is employed as the negative control in all PCR. Furthermore, for bacterial isolation samples from various internal organs, blood, pleural effusions and bronchial exudates were inoculated onto brain heart infusion (BHI) agar plates following the method described by Rehman et al. (2021). Subsequently, the colonies were then primarily identified based on the conventional biochemical methods (Jia et al., 2022). For molecular identification, DNA was extracted from the samples using DNA isolation mini kit (Novazan, Nanjing, China) and 16S rRNA gene was amplified with the universal primers listed in Supplementary Table S1 followed by sequencing (Nanshan Biotech). Finally, the species level of each isolate was identified as matching the reference strain based on the new canonical clustering threshold of >99% similarity of full-length 16S rRNA gene (Edgar, 2018).

E. coli isolates were tested by PCR with primer sets (Supplementary Table S1) of the following five virulence marker genes associated with ExPEC: papA/C (P fimbriae; counted as 1), sfa/foc (S and F1C fimbriae), afa/dra (A fimbrial and Dr-binding adhesion), kpsM II (group 2 capsule), and iutA (aerobactin system). Isolates positive for two or more of these markers were classified as ExPEC (Johnson et al., 2003).

The E. coli isolates were characterized according to phylogrouping and multilocus sequence typing (MLST) analysis. Phylogroups were determined using the Clermont PCR method (Clermont et al., 2013) (primers listed in Supplementary Table S1). In addition, MLST was performed using two schemes: the Achtman scheme targeting seven house-keeping genes (adk, fumC, gyrB, icd, mdh, recA and purA) (Manges et al., 2015), and the Pasteur scheme targetting eight housekeeping genes (dinB, icdA, pabB, polB, putP, trpA, trpB and uidA) (Primers in Supplementary Table S1) (Clermont et al., 2015).

Whole genome sequencing (WGS) was performed using Illumina HiSeq^™ platform (PE150) at Sangon Bioengineering Co., Ltd. (Shanghai, China) on representative isolate of each E. coli sequence type. Serotypes were determined for these isolates using online SerotypeFinder2.0 services (Joensen et al., 2015). PCR based serotyping with primers designed for the representative isolates (Supplementary Table S1) was subsequently used to identify serotypes of the remaining isolates.

For antimicrobial susceptibility assay, Kirby-Bauer disc diffusion method was used for 19 drugs following the CLSI guidelines (Humphries et al., 2021). However, for two drugs polymyxin B and tigecycline (Solarbio Life Sciences Co, Beijing, China), the minimal inhibitory concentration (MIC) was determined according to the 2017 EUCAST methods.² The antibiotics employed altogether fall into 11 major categories. Firstly, β-lactams include ampicillin, amoxicillin, and penicillin. Secondly, aminoglycosides consist of streptomycin, neomycin, and gentamicin. Thirdly, chloramphenicols are represented by chloramphenicol and florfenicol. Fourthly, fluoroquinolones comprise norfloxacin, ofloxacin, ciprofloxacin, and nalidixic acid. Fifthly, tetracyclines consist of tetracycline and tigecycline. Sixthly, cephalosporins are exemplified by cephalothin. Seventhly, sulfonamides include sulfisoxazole and trimethoprim. Eighthly, polypeptides are represented by polymyxin B. Ninthly, carbapenems are embodied by imipenem. Tenthly, macrolides are exemplified by erythromycin. Lastly, glycopeptides are represented by vancomycin. While E. coli ATCC 25,922 was subjected as quality control. Furthermore, the resistance gene pattern of each isolate was screened by PCR with the primer sets of 19 acquired antimicrobial resistance genes (Supplementary Table S1).

Based on the WGS of each representative E. coli isolate, the virulence factors were further obtained by VFDB tool³ following the method of Rehman et al. (2023).

Phylogenetic relationship between the isolated strains in this study and previously observed E. coli strains was determined according to a previous study with slight modifications (Rehman et al., 2023). The GrapeTree tool was used from pubMLST⁴ to analyze core genome multilocus phylogenetic analysis (cgMLPA) data for the isolated strains and other E. coli strains (Supplementary Table S2). iTOL v.4⁵ program was used for visualization and incorporation of available metadata into a phylogenetic tree.

Specific pathogen free (SPF) KM mice were employed from the Laboratory Animal Center, School of Pharmaceutical Sciences, Hainan University (SYXK0230021). The experimental protocol was approved by the Animal Management and Ethics Committee of Hainan University (HNUAUCC-2023-00020), and adhered to the recommendations of the Animal Ethics Procedures and Guidelines of the People’s Republic of China.

Mice were randomly allocated into six groups (five challenge groups and one control group) with six mice per group. The weight of each mouse was documented. Intraperitoneal injections were administered following the method of Long et al. (2022). Challenge groups were injected with 0.2 mL of E. coli solution (approximately 2 × 10⁶ CFU/mL), while control group received 0.2 mL PBS solution. Mice mortality rate and weight fluctuations were monitored every 6 h for 72 h. Pathological changes in the vital organs (heart, lung, liver, kidney and spleen) were recorded, and lungs were sent to Wuhan Saville Biotechnology Co., Ltd. (see text footnote 1) for further histopathological analysis.

Bronchopneumonia was diagnosed in six autopsied sheep collected at different time intervals. The common signs of the disease were sticky mucus in nostrils (Figure 1a) and trachea (Figure 1b), as well as irregular red consolidation and large-scale abscesses on the surfaces of extremely enlarged lungs (Figure 1c). These abscesses likely resulted from the internal accumulation of abundant pale-yellow mucoid exudate. (Figure 1d). Histopathological sectioning of lungs further revealed the features of lobular pneumonia of alveoli obliteration with the infiltration of neutrophils and capillary occlusion with intra-lumen fibrin exudates (Figure 1e,f).

Gram-negative, rod-shape bacteria were isolated from various tissues, especially lungs, liver, and blood. Further, PCR detection with specific primers for M. ovipneumoniae and M. mycoides was negative. Based on traditional 16S rRNA sequencing, all the isolates were identified as E. coli (Table 1).

A total of 29 isolates were recovered and considered for ExPEC identification. In term of pathogenic E. coli, ExPEC strains are responsible for the majority of human or animal extraintestinal infections, including pneumonia, urinary tract infections, and sepsis. However, according to the classic ExPEC virulence genes defined by Johnson et al. (2003), our results demonstrated that only three are categorized in ExPEC (Table 1). Interestingly, all 29 isolates originated from outside of the gastrointestinal tract. This finding suggests that the current method may not effectively differential between facultative ExPEC pathogens and commensal E. coli.

A total of three phylogroups were predicted and among 29 isolates, five belonged to A, 23 belonged to B1, and only one belonged to E (Figure 2). According to the two MLST schemes, the isolates were further subdivided into nine sequence types (ST) (Figure 2). Three isolates (spleen, blood, and trachea) from the second sheep exhibited novel ST according to Achtman scheme. This new ST was closely related to ST1704, but it remained as unassigned as a new identifier. Interestingly, Pasteur scheme classified this same ST as the existing ST1107 (Figure 2). Moreover, most of the STs in this study were very consistent between the both MLST schemes, except ST2521 (Achtman), which was subdivided into ST1109 and ST1110 in the Pasteur MSLT scheme, and ST1107 of Pasteur scheme into ST1704 and ST14878 based on the Achtman scheme (Figure 2).

WGS and PCR data were combined to classify the 29 isolates. According to the analysis, eight serotypes were obtained including both lipopolysaccharide O antigen and flagella H antigen. One exception case was the O-antigen serotype of three isolates (Y2XX1 from blood and Y2P1 and Y2P3 from sheep 2 spleen) was completely unmatched in the database of the Center for Genomic Epidemiology.⁶ This could be due to sequencing gaps within O-antigen and H-antigen gene clusters in their assembled genomes (Figure 2). By comparing the correlations of serotypes and ST, the serotypes O147:H7 and O153:H12 subdivided the isolates of ST2521 (Achtman scheme) into ST1109 and ST1110 (Pasteur scheme). Similarly, the O26:H11 and O8:H7 serotypes distinguished isolates ST1107 (Pasteur scheme) into ST1704 and ST14878 (Achtman scheme) (Figure 2). These finding suggest that the serotyping may be more effective in resolving the genetic distance between isolates than sequence typing. However, the two MLST schemes (Achtman and Pasteur) can be applied for co-analysis.

WGS data of 10 isolates representing unique serotypes were analyzed using the VFDB database. A total of 25 virulence factors categorized into seven classes were predicted. The most prevalent virulence factors included six adhesion-associated elements (CFA/I fimbriae, pilus, laminin-binding fimbriae, EaeH, hemorrhagic E. coli pilus, and type I fimbriae), two autotransporter associated elements (EhaA/B and UpaG adhesin), one invasion-associated element (invasion protein), two secretion system factors (T3SS translocator protein and Esp family protein), and one toxin factor (hemolysin/cytolysin A). All strains harbored at least 11 virulence factors. The highest and lowest virulence factors were predicted in isolates Y13F1 (17) and Y3X1 (11), respectively (Figure 2).

To assess the pathogenicity of isolated strains with different virulence factor combinations, a mouse infection model was established. Mortality rate and weight changes were analyzed after intraperitoneal injection of isolates. The Y13F1 isolate exhibited the most lethal effect, resulting in the death of all mice within 36 h. Conversely, Y2P1 isolate demonstrated the least lethal effect and did not cause any mortality (Figure 3A). By examining the average weight changes of mice post-intraperitoneal injection, a significant increase in weight was noticed for the control group. All mice within the challenge group experienced weight loss, with Y5YF5 resulting in the most significant weight loss and Y13F1 the least (Figure 3B). Gross pathology examination shows that the main organs of all experimental mice exhibited either localized or extensive hemorrhages except control mice. Lung hemorrhages were most severe in mice challenged with Y13F1 and Y3YF1 isolates compared to other experimental groups (Figure 3C). Additionally, the spleens of Y2P1 and Y3YF1 challenged mice were notably swollen, while the spleens of Y5YF3 and Y3YF1 challenged mice appeared more blackened (Figure 3C).

Histopathological changes in the lungs show inflammatory responses compared to the control group. The lung inflammatory response in the Y3YF1 group was particularly intense, characterized by inflammatory cell infiltration, fibrinoid exudation, and bronchial epithelial cell shedding (Figure 3D). This study demonstrates that isolates with different virulence factor combinations exhibit variation in lethality, target organ tropism, and their capacity to cause pneumonia. However, all strains displayed a degree of pathogenicity, highlighting the importance of further investigation into these isolates.

Genotypes of E. coli isolated in this study were compared to the pubMLST database. We discovered that E. coli with the same genotype as Y2F2 has only been reported in cows from the USA, UK, and Kenya, and in birds from Denmark (Figure 4A). However, E. coli with the same genotype as Y11S2 in this study exhibited a broader range of sources, including healthy individuals’ feces in China and France; clinical patients in the USA; cows from the USA, UK, Japan, and France, and chickens from Kenya and Peru (Figure 4B). E. coli with the same genotypes as Y5YF3 and Y2S3 have been found in clinical patients in the United States; feces from healthy people in France; cows in the USA, UK, and France; and pets, pigs, rabbits, and other animals in various countries (Figure 4C). E. coli with the Y2P1 genotype has only been reported in cows from the UK while the Y3X1 genotype was found only in individuals from China and Kenya (Supplementary Table S2). No reports of E. coli with the same genotypes as Y2XX1 and Y2P2 were found (Supplementary Table S2). The aforementioned E. coli genotypes were initially discovered and reported in sheep. Similarly, E. coli with the Y13F1 genotype in this study was primarily found in sheep, with a few individual cases detected in pigs and healthy people in China (Figure 4D).

Twenty-one antibiotics were used in susceptibility tests, with isolates showing resistance to at least six and up to sixteen antibiotics (Figure 5 and Table 2). The total resistance rates were 80% to tetracycline and sulfonamides (sulfamethoxazole and trimethoprim), 72% to amoxicillin, 66% to chloramphenicol, 59% to florfenicol, 45% to ciprofloxacin, 31% to nalidixic acid, 24% to streptomycin, and 17% to norfloxacin and ofloxacin. High or intermediate susceptibility was observed for four drugs (polymyxin B, imipenem, tigecycline, and gentamicin) by all isolates. On the other hand, all isolates were resistant to six agents (penicillin, ampicillin, cephalothin, neomycin, erythromycin, and vancomycin) (Figure 5).

A total of ten antimicrobial resistance genes (ARGs) or their variants (such as aadA1, aadA5, and aadA8 of aadAs) were detected by PCR among the 21 isolates of E. coli (Table 3). The number of ARGs harbored by each isolate varied, ranging from 2 (observed in 1 isolate from sheep Y2) to 8 (observed in 6 isolates: 4 from sheep Y2 and 2 from sheep Y11). However, no single ARG was consistently present in all isolates. The frequencies, from high to low, were 89.7% (26/29) for bla_TEM and tetA, 82.8% (24/29) for bla_EC, 72.4% (21/29) for dfrA, 65.5% (19/29) for sul, 55.2% (16/29) for qnrS, 48.3% (14/29) for aadA, 41.4% (12/29) for floR, 34.5% (10/29) for aph, and 17.2% (5/29) for ampC (bla_CMY). Furthermore, at least one of the three ARGs targeting β-lactamases was found in all 29 isolates.