The Animal Ethics Committee of Vietnam National University of Agriculture, Hanoi, Vietnam, reviewed and approved the study protocol (Approval No. VNUA-2023/05; approved on August 10, 2023). Spleen samples were collected from naturally diseased, dead, or clinically healthy chickens during routine veterinary diagnostic investigations and with permission from the respective farm owners. The study did not involve experimental infection, deliberate disease induction, or any additional invasive procedures performed specifically for research purposes. Sample collection, animal handling, and subsequent laboratory investigations were conducted in accordance with the conditions stipulated in the approved study protocol and applicable institutional animal welfare requirements. Only samples obtained during routine diagnostic or veterinary examination procedures were included in the study, thereby minimizing additional handling and potential distress to the animals.

The study was conducted from August 2023 to August 2025 on 35 commercial poultry farms located in five major poultry-producing provinces of northern Vietnam: Hanoi, Hai Phong, Phu Tho, Bac Ninh, and Thai Nguyen. The selected farms represented both intensive and backyard production systems and included broiler, layer, and breeder flocks of different sizes and age groups.

A cross-sectional study was conducted to characterize the distribution, profiles, and co-occurrence patterns of VAGs among APEC and NPEC isolates recovered from chickens. A total of 515 chickens from 35 poultry farms were included. Spleen samples were collected from 365 sick or dead chickens showing clinical signs or postmortem lesions consistent with colibacillosis and 150 clinically healthy chickens without apparent clinical signs or gross pathological lesions at sampling. Isolated E. coli strains were screened for 19 VAGs using multiplex polymerase chain reaction (mPCR), and isolates were subsequently classified as APEC or NPEC according to their VAG profiles. The study focused specifically on VAG distribution and co-occurrence; therefore, serotyping, phylogrouping, and multilocus sequence typing analyses were not included in the study design.

The minimum sample size was estimated using the standard formula for cross-sectional prevalence studies described by Naing et al. [31]. At the farm-level, an expected APEC prevalence of 90% [32], a 95% confidence level, and 10% precision yielded a minimum required sample size of 35 farms. Accordingly, 35 commercial poultry farms were included. At the individual level, based on the expected APEC prevalence of 20%–30% reported by Kathayat et al. [1] and a 95% confidence level, the recommended sample size was 9–14 chickens per farm. Approximately 14–15 chickens were therefore randomly sampled from each farm, exceeding the minimum requirement while maintaining an appropriate balance between statistical power and field feasibility. This resulted in a total sample size of 515 chickens.

A total of 515 spleen samples were randomly collected from broiler, layer, and breeder flocks on the 35 study farms. The spleen was selected because, as a major component of the reticuloendothelial system, it reflects systemic bacterial dissemination during colisepticemia. Splenomegaly and congestion are recognized postmortem findings associated with acute E. coli septicemia in poultry [33], and spleen sampling together with other internal organs is widely used in APEC isolation protocols [1]. Uniform collection of spleen samples from both non-healthy and clinically healthy chickens also reduced the potential for environmental or fecal cross-contamination associated with organs adjacent to the gastrointestinal tract and provided a consistent basis for comparison between APEC and NPEC isolates.

Of the 515 samples, 365 were collected from sick or dead chickens exhibiting clinical signs or characteristic postmortem lesions consistent with colibacillosis, including airsacculitis, fibrinous pericarditis, perihepatitis, septicemia, and splenomegaly. These birds were categorized as non-healthy chickens. The remaining 150 samples were obtained from clinically healthy chickens that showed no apparent clinical signs or gross pathological lesions at the time of sampling.

Spleen samples were cultured on MacConkey agar (Merck KGaA, Darmstadt, Germany) and incubated at 37°C for 24 h. Bright pink colonies surrounded by precipitate were aseptically subcultured onto eosin methylene blue agar (Merck KGaA) and incubated at 37°C for 24 h. Single colonies with a greenish-black appearance and metallic sheen were selected for biochemical characterization using the API 20E kit (bioMérieux, Marcy-l’étoile, France) according to the manufacturer’s instructions.

Presumptive isolates were subsequently confirmed as E. coli by PCR using primer pairs described by Bej et al. [34]. All confirmed isolates were stored in brain heart infusion broth (Merck KGaA) supplemented with 20% glycerol at −80°C until further analysis.

DNA was extracted from E. coli isolates using the boiling–freeze-thaw method [10, 35, 36]. Briefly, isolates were cultured on nutrient agar (Merck KGaA) at 37°C for 24 h. Smooth, circular, white colonies were suspended in 200 µL of deionized water, thoroughly vortexed, and boiled at 100°C for 20 min. The bacterial suspensions were subsequently cooled on ice for 20 min and centrifuged at 14,000 x g for 10 min at 4°C. The supernatant containing bacterial DNA was transferred to a new sterile microcentrifuge tube, adjusted to 100 ng/µL, and stored at −20°C until mPCR analysis.

The mPCR assay was used to detect 19 VAGs in the E. coli isolates. Each reaction contained DNA at 100 ng/µL, 10 pmol of each primer pair (Integrated DNA Technologies, Coralville, IA, USA), 10× PCR buffer (Sigma-Aldrich, St. Louis, MO, USA), magnesium chloride (MgCl₂) (50 mM), 2.5 U of Taq DNA polymerase (Quantabio, Beverly, USA), deoxyribonucleotide triphosphates (dNTPs) (Invitrogen, Carlsbad, CA, USA), and PCR-grade water. The total reaction volume was 25 µL and consisted of 5 µL of DNA, 1 µL of each primer, 2.5 µL of 10× PCR buffer, 1.25 µL of MgCl₂, 1 µL of dNTPs, 0.5 µL of Taq DNA polymerase, and PCR-grade water to a final volume of 25 µL, according to the corresponding multiplex group presented in Table 1.

Thermal cycling consisted of an initial denaturation at 94°C for 3 min 30 s, followed by 35 cycles of denaturation at 94°C for 30 s, annealing for 1 min at the temperature specified for each multiplex group (Tm1 = 56°C, Tm2 = 57°C, Tm3 = 54°C, Tm4 = 58°C, and Tm5 = 56°C; Table 1), and extension at 72°C for 1 min 30 s. A final extension was performed at 72°C for 6 min, after which the products were maintained at 4°C.

The mPCR products were separated by electrophoresis on 1% agarose gels containing ethidium bromide together with a 100-bp DNA ladder (Cleaver Scientific, Rugby, UK) for 30 min and visualized under ultraviolet illumination using a gel documentation system (Bio-Rad Laboratories, Inc., Hercules, CA, USA). A reaction was considered positive when a distinct band corresponding to the expected amplicon size was observed relative to the 100-bp DNA ladder. Because amplification results were interpreted based on discrete DNA fragments of predetermined sizes, dedicated gel image-analysis software was not required.

All primers were obtained from previously published and validated studies and were commercially synthesized. Primer specificity and expected amplicon sizes were verified according to the original publications and further confirmed in the present study using a previously characterized APEC-positive-control strain. Each multiplex group included an APEC reference strain carrying the corresponding target VAGs as a positive-control and nuclease-free water as a negative control. The positive-control strain was isolated from chickens and maintained at the Key Laboratory of Veterinary Biotechnology, Faculty of Veterinary Medicine, Vietnam National University of Agriculture. Primer sequences, expected amplicon sizes, multiplex groups, and corresponding annealing temperatures are presented in Table 1[9, 10, 12, 14, 37–39].

Classification of E. coli isolates as APEC or NPEC was based on the VAG profiles determined by mPCR. Isolates harboring five or more VAGs (≥5 VAGs) were classified as APEC, whereas isolates carrying fewer than five VAGs were classified as NPEC, following previously reported criteria [20, 23, 24]. This threshold has been applied because isolates harboring multiple VAGs exhibit greater pathogenic potential and are more frequently associated with systemic colibacillosis.

Johnson et al. [9] identified five minimal predictor genes (iutA, hlyF, iss, iroN, and ompT) strongly associated with APEC pathogenicity. Challenge experiments in 1-day-old chicks further demonstrated that isolates harboring fewer than two of these five marker genes rarely exhibited pathogenicity, whereas most virulent APEC strains possessed two or more [40]. Furthermore, Johar et al. [23] classified chicken E. coli isolates harboring at least five of the investigated VAGs (ompT, hlyF, iroN, iucD, iss, tsh, vat, and cva/cvi) as APEC, whereas isolates carrying fewer than five genes were classified as NPEC.

Based on these previously validated criteria, nine selected APEC-associated VAGs (ompT, hlyF, iroN, iutA, tsh, vat, iss, cva/cvi, and iucD) were used for APEC classification in the present study. An E. coli isolate was classified as APEC when it carried ≥5 of these nine selected VAGs, whereas isolates carrying <5 of these nine VAGs were classified as NPEC. A chicken was considered APEC-positive when its E. coli isolate carried ≥5 selected APEC-associated VAGs. A farm was considered APEC-positive when at least one sampled chicken was APEC-positive [32].

Statistical analyses were performed using R software (version 4.3.0; R Foundation for Statistical Computing, Vienna, Austria). The prevalence of each VAG was expressed as a percentage with a 95% confidence interval (CI), calculated using the Clopper–Pearson exact binomial method. Differences in the prevalence of individual VAGs between groups were assessed using Pearson’s chi-square test or Fisher’s exact test, as appropriate. All tests were two-sided, and p < 0.05 was considered statistically significant.

Pairwise co-occurrence of VAGs was quantified using the Jaccard similarity coefficient based on gene presence/absence data. The Jaccard similarity coefficient was selected because the analysis focused specifically on the co-occurrence of gene presence, and several VAGs had very high prevalence, making correlation-based measures such as the phi coefficient less informative. Co-occurrence patterns were visualized using hierarchical clustering heatmaps and network analysis, with edge thickness proportional to the Jaccard similarity coefficient. UpSet plots were generated to visualize the frequency and complexity of VAG combination patterns in APEC and NPEC isolates.

A total of 160 E. coli isolates were recovered from 365 spleen samples collected from non-healthy chickens, corresponding to an isolation rate of 43.84%. Among the 150 spleen samples collected from healthy chickens, 40 (26.67%) were positive for E. coli. Overall, E. coli was isolated from 200 of 515 samples (38.83%), and all 35 surveyed farms (100%) had at least one E. coli-positive sample. The isolation rate was significantly higher in non-healthy chickens than in healthy chickens (p < 0.001). All 200 isolates were subsequently subjected to mPCR for VAG detection.

All 19 investigated VAGs were detected among the E. coli isolates recovered from non-healthy chickens: fyuA, eaeA, hlyF, iroN, tsh, iucA, iss, irp2, cva/cvi, hlyA, fimC, ompT, vat, colV, papC, iucD, astA, iutA, and sfa_foc. In contrast, 15 of the 19 VAGs were detected among isolates from healthy chickens; papC, eaeA, sfa_foc, and hlyA were not detected in this group (Tables 2, 4, and 5 and Figures 1 and 2).

Among the 160 E. coli isolates recovered from non-healthy chickens, fimC was the most prevalent VAG, detected in 97.5% of isolates (95% CI: 93.7–99.3), followed by ompT (91.3%; 95% CI: 85.8–95.1), iutA (90.6%; 95% CI: 85.0–94.7), hlyF (78.1%; 95% CI: 70.9–84.3), iucD (75.0%; 95% CI: 67.6–81.5), iss (70.0%; 95% CI: 62.3–77.0), iroN (62.5%; 95% CI: 54.5–70.0), and irp2 (61.3%; 95% CI: 53.2–68.8). The tsh, colV, papC, astA, iucA, cva/cvi, vat, and fyuA genes were detected in 22.5%–41.9% of isolates. In contrast, eaeA (16.3%; 95% CI: 10.9–22.9), sfa_foc (11.9%; 95% CI: 7.3–17.9), and hlyA (3.1%; 95% CI: 1.0–7.1) were detected at relatively low frequencies (Table 2 and Figure 2).

Among the 40 E. coli isolates recovered from healthy chickens, fimC was the most prevalent VAG, detected in 75.0% of isolates (95% CI: 58.8–87.3), followed by ompT (65.0%; 95% CI: 48.3–79.4), iutA (62.5%; 95% CI: 45.8–77.3), hlyF (32.5%; 95% CI: 18.6–49.1), iucA (27.5%; 95% CI: 14.6–43.9), and iucD, iss, iroN, and cva/cvi (25.0% each; 95% CI: 12.7–41.2). The remaining VAGs were detected at lower frequencies: fyuA, 5.0% (95% CI: 0.6–16.9); vat, 10.0% (95% CI: 2.8–23.7); astA, 12.5% (95% CI: 4.2–26.8); tsh, 12.5% (95% CI: 4.2–26.8); colV, 15.0% (95% CI: 5.7–29.8); and irp2, 17.5% (95% CI: 7.3–32.8). Notably, papC, eaeA, sfa_foc, and hlyA were not detected in isolates from healthy chickens (Table 2 and Figure 2).

Comparison between the two groups showed that several VAGs were significantly more prevalent among isolates from non-healthy chickens. Significant differences were observed for fimC, ompT, iutA, hlyF, iucD, iss, iroN, irp2, and fyuA (p < 0.001), as well as for vat, astA, papC, eaeA, and sfa_foc (p < 0.05). In contrast, the prevalence of cva/cvi, iucA, colV, tsh, and hlyA did not differ significantly between isolates from non-healthy and healthy chickens (p > 0.05) (Table 2 and Figure 2).

Based on the presence of the selected VAGs ompT, hlyF, iroN, iutA, tsh, vat, iss, cva/cvi, and iucD, 121 of the 200 E. coli isolates (60.5%) carrying five or more VAGs were classified as APEC. Among these, 115/121 (95.0%) were recovered from non-healthy chickens, whereas 6/121 (5.0%) were recovered from healthy chickens. The remaining 79/200 isolates (39.5%), which harbored fewer than five VAGs, were classified as NPEC; 45/79 (57.0%) were recovered from non-healthy chickens and 34/79 (43.0%) from healthy chickens (Tables 4 and 5; Figure 3). In addition, 34 of the 35 surveyed farms (97.1%) had at least one APEC-positive sample.

The respective prevalence of VAGs in APEC and NPEC isolates was as follows: hlyF, 100.0% (95% CI: 97.0–100.0) and 21.5% (95% CI: 13.1–32.2); iutA, 96.7% (95% CI: 91.8–99.1) and 67.1% (95% CI: 55.6–77.3); ompT, 95.9% (95% CI: 90.6–98.6) and 70.9% (95% CI: 59.6–80.6); iucD, 95.9% (95% CI: 90.6–98.6) and 17.7% (95% CI: 10.0–27.9); iss, 90.1% (95% CI: 83.3–94.8) and 16.5% (95% CI: 9.1–26.5); irp2, 79.3% (95% CI: 71.0–86.2) and 11.4% (95% CI: 5.3–20.5); fyuA, 57.0% (95% CI: 47.7–66.0) and 0% (95% CI: 0–4.6); vat, 48.8% (95% CI: 39.6–58.0) and 5.1% (95% CI: 1.4–12.5); papC, 38.0% (95% CI: 29.3–47.3) and 0% (95% CI: 0–4.6); astA, 36.4% (95% CI: 27.8–45.6) and 10.1% (95% CI: 4.5–19.0); tsh, 30.6% (95% CI: 22.5–39.6) and 5.1% (95% CI: 1.4–12.5); eaeA, 21.5% (95% CI: 14.5–29.9) and 0% (95% CI: 0–4.6); and sfa_foc, 15.7% (95% CI: 9.7–23.4) and 0% (95% CI: 0–4.6). The prevalence of these genes was significantly higher in APEC than in NPEC isolates (p < 0.0001).

The prevalence of iroN [65.3% (95% CI: 56.1–73.7) vs. 39.2% (95% CI: 28.4–50.9)], cva/cvi [40.5% (95% CI: 31.7–49.8) vs. 15.2% (95% CI: 8.1–25.0)], and colV [31.4% (95% CI: 23.3–40.5) vs. 13.9% (95% CI: 7.2–23.5)] was also significantly higher in APEC than in NPEC isolates (p < 0.01). Similarly, fimC was significantly more prevalent in APEC isolates [96.7% (95% CI: 91.8–99.1)] than in NPEC isolates [87.3% (95% CI: 78.0–93.8)] (p < 0.05). In contrast, the prevalence of iucA [34.7% (95% CI: 26.3–43.9) vs. 24.1% (95% CI: 15.1–35.0)] and hlyA [4.1% (95% CI: 1.4–9.4) vs. 0% (95% CI: 0–4.6)] did not differ significantly between the two groups (p > 0.05) (Table 3 and Figure 3).

The five-marker predictor set described by Johnson et al. [9], comprising hlyF, ompT, iroN, iss, and iutA, also differed markedly between APEC and NPEC isolates. Among APEC isolates, hlyF was detected in all isolates (100.0%), followed by iutA (96.7%), ompT (95.9%), iss (90.1%), and iroN (65.3%). In NPEC isolates, the corresponding prevalence was substantially lower for ompT (70.9%), iutA (67.1%), iroN (39.2%), hlyF (21.5%), and iss (16.5%) (p < 0.0001). Simultaneous carriage of all five predictor genes was detected in 70/121 (57.8%) APEC isolates but in none of the 79 NPEC isolates.

Among the 160 E. coli isolates recovered from non-healthy chickens, 33 distinct VAG combination patterns containing 3–15 genes were identified (Table 4). Isolates harboring more than eight VAGs accounted for 103/160 (64.37%) and were distributed among 18 combination patterns. Notably, 8/160 isolates (5.0%) simultaneously harbored 15 VAGs: ompT-iutA-fimC-iucD-hlyF-iss-iroN-irp2-vat-papC-cva/cvi-iucA-colV-eaeA-sfa_foc.

A further 15/160 isolates (9.4%) simultaneously harbored 14 VAGs and were distributed among three combination patterns: ompT-iutA-fimC-iucD-hlyF-iss-iroN-irp2-fyuA-vat-papC-iucA-colV-tsh; iutA-fimC-iucD-hlyF-iss-iroN-irp2-fyuA-papC-cva/cvi-astA-iucA-colV-hlyA; and ompT-iutA-fimC-iucD-hlyF-iss-iroN-irp2-vat-papC-astA-iucA-colV-eaeA. The remaining 57/160 isolates (35.63%), which harbored 3–7 VAGs, were distributed among 15 combination patterns. The 40 E. coli isolates recovered from healthy chickens were classified into nine different combination patterns containing 3–7 VAGs (Table 5). None of the isolates harbored all 19 VAGs simultaneously.

Among isolates from non-healthy chickens, those with combination patterns containing 6–15 VAGs (molecular pathotypes VAG1–VAG22) were classified as APEC, whereas those carrying 3–6 VAGs (VAG23–VAG33) were classified as NPEC (Table 4). Among isolates from healthy chickens, isolates with a five-gene combination (ompT-iucD-hlyF-vat-cva/cvi; VAG34) and a seven-gene combination (ompT-iutA-fimC-iucD-hlyF-fyuA-tsh; VAG35) were classified as APEC, whereas those harboring 3–6 VAGs (VAG36–VAG42) were classified as NPEC (Table 5). NPEC isolates harbored 3–6 VAGs in total across the 19 VAGs investigated; however, none carried ≥5 of the nine predefined APEC-associated VAGs used for classification. Overall, 23 distinct VAG combination patterns were identified among APEC isolates, compared with 13 among NPEC isolates (Figures 4 and 5).

The co-occurrence heatmap showed a denser and more structured pattern of VAG associations among APEC isolates than among NPEC isolates, particularly for genes associated with iron acquisition, serum resistance, and ColV plasmid-associated determinants (Figure 4). In contrast, NPEC isolates exhibited fewer and less extensive co-occurrence associations. Strong associations were observed among irp2, iss, hlyF, iutA, fimC, ompT, and iucD in APEC isolates. A distinct cluster was also observed between iucA and colV (Figure 4).

The genes eaeA, sfa_foc, and hlyA were detected in a subset of APEC isolates and co-occurred with the plasmid-associated determinants iroN, iss, and hlyF. Overall, APEC isolates displayed more numerous and complex VAG co-occurrence patterns than NPEC isolates (Figures 4 and 5).

A total of 23 distinct VAG profiles were identified among the APEC isolates. Most isolates (103/121, 85.1%) were distributed among 18 profiles containing 8–15 VAGs (Tables 4 and 5; Figure 5). The most prevalent profile, P1, comprised 15 VAGs and was detected in eight isolates. This profile included multiple Extraintestinal pathogenic E. coli (ExPEC)-associated virulence determinants, namely ompT, iutA, fimC, iucD, hlyF, iss, iroN, irp2, fyuA, papC, cva/cvi, iucA, colV, eaeA, and sfa_foc. Isolates carrying this profile were recovered from diseased chickens exhibiting characteristic gross pathological lesions, including perihepatitis, pericarditis, airsacculitis, swollen-head syndrome, salpingitis, osteomyelitis/synovitis, and markedly enlarged and hemorrhagic spleens. The remaining 18 APEC isolates (14.9%) were distributed among profiles containing 5–7 VAGs.

In contrast, 13 distinct VAG profiles were identified among NPEC isolates, with most profiles comprising 3–6 VAGs (Tables 4 and 5; Figure 5). The predominant NPEC profiles mainly included ompT, fimC, iutA, hlyF, and iroN, whereas papC, fyuA, eaeA, sfa_foc, and hlyA were detected infrequently. Overall, APEC isolates exhibited greater diversity and complexity of VAG combinations than NPEC isolates.

The predominance of APEC profiles containing numerous ExPEC-associated virulence determinants indicates a broader virulence gene repertoire than that observed among NPEC isolates. However, because pathogenicity was not confirmed by experimental challenge in this study, these molecular profiles should be interpreted as indicators of virulence-associated potential rather than definitive evidence of pathogenicity.

Among the APEC isolates, the most prevalent VAG profile was ompT-iutA-fimC-iucD-hlyF-iss-iroN-irp2-vat-papC-cva/cvi-iucA-colV-eaeA-sfa_foc, which was detected in 8/121 isolates (6.6%). In contrast, the predominant VAG profile among NPEC isolates was ompT-fimC-hlyF, which was detected in 12/79 isolates (15.2%).

Colibacillosis is a major infectious disease of poultry caused by APEC and is characterized by localized and systemic infections [1, 30, 41, 42]. The definition and classification of APEC remain subjects of debate; however, a multi-criteria approach incorporating the site of isolation and VAG profiles is increasingly accepted [1, 43]. Isolation of E. coli from extraintestinal organs, particularly the spleen, reflects the ability of the bacterium to invade host tissues, evade host defenses, and disseminate systemically. Detection in the spleen, an important peripheral lymphoid organ, is therefore considered a relevant biological indicator of APEC pathogenicity [33]. In addition, APEC strains typically harbor multiple VAGs and PAIs that contribute to their pathogenic potential [10, 13, 15, 16, 44, 45]. The mPCR assay has consequently become a widely used rapid and reliable molecular approach for detecting VAGs and characterizing APEC strains [46]. In the present study, 19 VAGs were investigated by mPCR: ompT, iutA, fimC, iucD, hlyF, iss, iroN, irp2, fyuA, vat, papC, cva/cvi, astA, iucA, colV, tsh, eaeA, sfa_foc, and hlyA.

Challenge experiments in 1-day-old chicks demonstrated that APEC strains harboring fewer than two of the five marker genes hlyF, ompT, iroN, iss, and iutA rarely exhibited pathogenicity, whereas most virulent APEC isolates carried two or more of these genes [40]. Furthermore, in a previous survey evaluating ompT, hlyF, iroN, tsh, vat, iss, cvi/cva, and iucD in chicken-derived E. coli, isolates carrying at least five of the investigated VAGs were classified as APEC, whereas those carrying fewer than five were considered NPEC [23]. Based on the classification criteria applied in the present study, 121/200 (60.5%) isolates were classified as APEC, whereas the remaining 79/200 (39.5%) were classified as NPEC.

Among the 121 APEC isolates, 115 (95.0%) originated from non-healthy chickens, whereas only 6/121 (5.0%) were recovered from clinically healthy birds. This distribution supports an association between the accumulation of VAGs and clinical colibacillosis. Nevertheless, recovery of APEC-classified isolates from healthy chickens indicates that clinically normal birds may carry E. coli strains with substantial virulence-associated repertoires. The presence of multiple VAGs alone may therefore be insufficient to determine whether clinical disease develops, because host susceptibility and environmental factors can also influence pathogenesis [1, 30, 43]. Clinically healthy chickens may consequently represent reservoirs of potentially pathogenic APEC within poultry populations [1, 30, 43].

All 19 investigated VAGs were detected among E. coli isolates from non-healthy chickens, whereas 15 were detected among isolates from healthy chickens; papC, eaeA, sfa_foc, and hlyA were not detected in the latter group. Similarly, all 19 VAGs were represented among APEC isolates, whereas 14 were detected among NPEC isolates, with papC, eaeA, sfa_foc, fyuA, and hlyA absent from NPEC. These findings indicate that APEC isolates possessed a broader VAG repertoire than NPEC isolates [10, 23]. The accumulation of multiple VAGs appears to be an important feature of APEC pathogenic potential, consistent with previous studies reporting that highly virulent APEC strains generally harbor larger and more diverse virulence repertoires [3, 9, 10, 47]. This genetic complexity supports the value of VAG profiling as a molecular approach for differentiating APEC from less virulent or commensal E. coli populations in poultry [10, 20].

The exclusive detection of papC, eaeA, sfa_foc, fyuA, and hlyA among APEC isolates further suggests that these genes may contribute to the broader virulence repertoire observed in this group [2, 43]. Of particular One Health relevance, papC, eaeA, and fyuA have also been reported among human ExPEC, including uropathogenic E. coli (UPEC) and neonatal meningitis-associated E. coli (NMEC), where they are involved in adhesion, iron acquisition, and extraintestinal colonization [1, 30]. Their detection in poultry-derived APEC indicates that chickens may harbor E. coli strains possessing virulence determinants shared with human ExPEC. Although zoonotic transmission was not investigated in the present study, these findings support the importance of integrated surveillance across the animal–human–environment interface.

Among APEC isolates, the most frequently detected VAGs were hlyF (100.0%), fimC (96.7%), iutA (96.7%), ompT (95.9%), iucD (95.9%), iss (90.1%), irp2 (79.3%), and iroN (65.3%) (Table 3). Previous studies have reported considerable variation in the prevalence of these genes, including hlyF (59.3%–100%), fimC (70.0%–94.9%), ompT (67.2%–100%), iutA (79.5%–93.1%), iucD (70.8%–100%), iss (55.6%–100%), iroN (67.88%–100%), and irp2 (45.26%–100%) [9, 10, 14, 16, 18–20, 38, 44, 48–51]. In the present study, the prevalence of ompT, iutA, hlyF, iss, iroN, and irp2 was generally comparable to previously reported ranges, whereas fimC occurred at a higher frequency.

The prevalence of the major APEC-associated VAGs in this study was generally comparable to or higher than that reported in other Asian countries [19, 22, 27, 52–56]. In particular, fimC (96.7%), iutA (96.7%), and hlyF (100.0%) were highly prevalent among APEC isolates and exceeded several previously reported estimates from China, South Korea, and Thailand. Similarly, ompT and iss occurred at frequencies within the upper ranges reported in Asian studies. In contrast, the prevalence of iroN (65.3%) was lower than that reported in some studies from China and South Korea but remained comparable to findings from India and Thailand (Supplementary Table S4).

The five-marker predictor set proposed by Johnson et al. [9], comprising hlyF, ompT, iroN, iss, and iutA, also showed substantial discriminatory value in the present study. Although individual markers occurred in some NPEC isolates, simultaneous carriage of all five markers was detected only among APEC isolates. These findings are consistent with those of Johnson et al. [9] and support the usefulness of this predictor set for molecular characterization and epidemiological surveillance of APEC in poultry [1, 43].

Previous studies from Vietnam reported lower frequencies of several individual VAGs, including ompT (28.7%), iss (22.5%), hlyF (18.7%), iucD (17.5%), and iroN (12.5%), than those observed in the present study [27]. Differences among studies may be related to sampling strategy, isolate source, study population, and diagnostic or classification criteria. For example, isolates obtained from fecal samples or healthy chickens may harbor fewer VAGs than isolates recovered from systemic infection sites such as the liver or spleen [2, 43].

The remaining VAGs among APEC isolates occurred at frequencies ranging from 4.1% to 57.0% (Table 3). The prevalence of fyuA (57.0%) and tsh (30.6%) was lower than that reported in China (77.8% and 61.8%, respectively) [10]. In contrast, vat (48.8%) and papC (38.0%) occurred more frequently than previously reported in China (27.01% and 18.98%, respectively) [19] but less frequently than in Australia (89.7% and 44.8%, respectively) [48].

The frequencies of cva/cvi (40.5%), iucA (34.7%), and colV (31.4%) were lower than several previously reported estimates, whereas the prevalence of astA (36.4%) fell within the broad range reported in earlier studies [10, 48–50]. Conversely, eaeA (21.5%), sfa_foc (15.7%), and hlyA (4.1%) were detected more frequently than in several previous reports, in which prevalence estimates of approximately 1.28%, 0%–4.4%, and 0%–1.28%, respectively, were reported [10, 48–50]. The prevalence of fyuA, tsh, papC, vat, cva/cvi, and iucA has varied considerably among studies, whereas astA and hlyA have generally occurred at relatively low frequencies in reports from China, South Korea, India, Thailand, and Vietnam (Supplementary Table S4).

Notably, eaeA and sfa_foc, which have rarely been reported among APEC isolates in Asian studies, were detected in a subset of the Vietnamese isolates [10, 19, 22, 27, 44, 52–56]. Their occurrence contributes to the broader VAG diversity observed in the study population, although molecular detection alone cannot establish their direct contribution to pathogenicity. Differences in VAG prevalence among APEC populations from different countries may reflect geographical variation, poultry production systems, host populations, chicken breeds, circulating bacterial lineages, sampling strategies, and differences in the VAG panels investigated [30, 48]. Such genetic variability may contribute to differences in pathogenic potential among APEC populations [57] and is epidemiologically relevant for understanding disease severity and strain circulation within poultry production systems.

APEC strains are commonly characterized by multiple marker genes associated with adhesion, iron acquisition, serum resistance, proteolysis, and other virulence-related functions. These include ompT (episomal outer membrane protease), iutA (ferric aerobactin receptor), fimC (Type 1 fimbrial adhesin), iucD (aerobactin synthesis), hlyF (putative avian hemolysin), iss (increased serum survival), and iroN (salmochelin siderophore receptor) [9, 14]. APEC isolates recovered from chickens with clinical signs and lesions of colibacillosis frequently harbor multiple combinations of these genes [9, 37, 44]. Ewers et al. [14] identified eight VAGs, papC, iucD, irp2, tsh, vat, astA, iss, and cva/cvi, as useful markers for differentiating APEC from NPEC. In the present study, APEC isolates simultaneously harbored between one and eight of these markers.

Previous studies have also demonstrated that the number and combination of VAGs are associated with differences in pathogenic potential. Highly pathogenic E. coli strains have been reported to harbor approximately 8–13 VAGs, including irp2, fyuA, iutA, papC, iss, colV, iucD, iucA, fimA, fimC, hlyA, tsh, and colBM, whereas strains with intermediate or low-pathogenicity generally harbor fewer virulence determinants [10]. Jeong et al. [18] reported iroN as the most prevalent gene among Korean APEC isolates (100%), followed by fimC (94.1%), ompT (90.1%), hlyF (87.1%), iss (78.2%), iucD (73.3%), tsh (61.4%), fyuA (44.6%), irp2 (43.6%), and vat (10.9%). These isolates were classified into 27 molecular pathotypes, with iroN-fimC-ompT-hlyF-iucD-tsh-iss-irp2-fyuA, iroN-fimC-ompT-hlyF-iucD-tsh-iss, and iroN-fimC-ompT-hlyF-iss predominating.

In the present study, 103/121 (85.1%) APEC isolates harbored more than eight VAGs and were distributed among 18 combination patterns (Tables 4 and 5; Figure 5), whereas the remaining 18/121 isolates (14.9%) contained 5–7 VAGs and were distributed among five patterns. Necropsy examination of diseased chickens revealed characteristic lesions of colibacillosis, including perihepatitis, pericarditis, airsacculitis, swollen-head syndrome, salpingitis, and osteomyelitis/synovitis, and APEC isolates recovered from spleen samples showed substantial diversity in their VAG profiles [58]. Notably, 23/121 APEC isolates (19.0%) harbored 14–15 VAGs and were recovered from chickens exhibiting typical and severe colibacillosis-associated lesions [1]. Of these, eight isolates (6.6%) simultaneously carried 15 VAGs, whereas 15 isolates (12.4%) carried 14 VAGs.

These extensive virulence repertoires are consistent with the concept that APEC pathogenic potential reflects the combined action of multiple virulence determinants rather than the presence of a single gene. Similar multi-gene combinations have been reported predominantly among highly virulent APEC strains [2, 10, 18, 23, 46]. The coexistence of genes involved in adhesion, iron acquisition, toxin production, serum resistance, and plasmid-associated functions may collectively facilitate colonization, survival in extraintestinal tissues, evasion of host defenses, and systemic dissemination [1, 3, 26, 30]. Comparable VAG-rich profiles have previously been associated with severe colibacillosis and increased pathogenicity in experimental models [1, 3, 10, 30]. Nevertheless, the present study did not experimentally determine virulence; therefore, isolates with extensive VAG repertoires should be regarded as having putatively greater pathogenic potential. Confirmation would require in vivo pathogenicity testing, such as chicken challenge experiments or embryo lethality assays [9, 10].

The present findings are consistent with previous reports from Vietnam. Vounba et al. [27] detected several APEC/ExPEC-associated VAGs, including iroN, iss, hlyF, ompT, and iucD, among E. coli isolates from healthy chickens, together with β-lactam resistance determinants, demonstrating the coexistence of virulence and antimicrobial resistance traits in poultry-associated E. coli. Similarly, Sary et al. [28] identified multidrug-resistant ExPEC-like E. coli harboring VAGs such as papC and tsh in retail chicken carcasses, indicating the potential for dissemination through the poultry food chain. More recently, Hoang et al. [29] reported that the Vietnamese APEC strain O78:H9 (HPVN24) possessed a broad repertoire of virulence and antimicrobial resistance genes.

Consistent with these studies, the APEC isolates investigated here showed a high prevalence of core VAGs, including iroN, iss, ompT, and hlyF. However, unlike previous Vietnamese investigations that primarily focused on antimicrobial resistance, selected virulence determinants, or individual strains, the present study characterized 19 VAGs across a larger isolate collection and identified 23 distinct VAG combination patterns, including profiles containing up to 15 genes. These findings extend current knowledge of the molecular diversity of poultry-associated APEC in Vietnam and provide a broader basis for epidemiological surveillance.

The co-occurrence heatmap revealed strong associations among several VAGs in APEC isolates, including hlyF, ompT, iss, iroN, iutA, cva/cvi, and colV. These genes include determinants associated with iron acquisition, serum resistance, bacterial adaptation, and systemic dissemination and are frequently linked to ColV-associated virulence plasmids [1, 3, 30]. In contrast, NPEC isolates showed a less extensive and more fragmented co-occurrence network, consistent with their more limited VAG repertoires. The UpSet plot similarly demonstrated that VAGs occurred in more numerous and complex combinations among APEC than among NPEC isolates.

The frequent co-occurrence of these determinants supports the importance of considering combinations of VAGs rather than individual genes when characterizing APEC. However, because plasmid analysis was not performed, the present data cannot confirm that the observed gene associations were physically located on the same ColV-like plasmids. Further plasmid characterization and whole-genome sequencing would be required to establish their genomic organization and patterns of co-inheritance.

Several VAGs detected in the present study, including papC, fyuA, iutA, iroN, ompT, and iss, overlap with virulence signatures reported in human ExPEC lineages, particularly UPEC and NMEC. Previous studies have demonstrated that APEC and human ExPEC may share virulence determinants, plasmid-associated pathogenicity genes, and, in some cases, phylogenetic backgrounds such as ST95 and ST131 [1, 30]. The detection of complex multi-gene VAG profiles among Vietnamese poultry APEC isolates, including strains carrying 12–15 VAGs, therefore indicates the presence of isolates with ExPEC-associated molecular characteristics.

These findings have potential One Health relevance, particularly in poultry-producing regions where humans, poultry, poultry products, and farm environments are closely interconnected. However, this study did not directly investigate zoonotic transmission, and detecting shared VAGs alone does not establish transmission between poultry and humans. Genomic characterization and comparative phylogenetic analyses of poultry, human, and environmental isolates would be required to determine whether epidemiologically related lineages circulate across these compartments [1, 30, 59, 60]. Nevertheless, the findings support continued molecular surveillance of APEC virulence profiles together with antimicrobial resistance monitoring within an integrated One Health framework.

Several limitations should be considered when interpreting these findings. First, pathogenic potential was inferred from VAG profiles and clinical or pathological associations because in vivo challenge experiments were not performed. Consequently, the presence of multiple VAGs should not be interpreted as definitive evidence of virulence. Second, serotyping, phylogrouping, multilocus sequence typing (MLST), whole-genome sequencing, and plasmid characterization were not undertaken, limiting assessment of clonal relationships, genomic backgrounds, and the physical organization of co-occurring VAGs. Third, exclusive use of spleen samples may have introduced sampling bias by preferentially representing systemic infections rather than the broader diversity of E. coli colonizing poultry. Finally, the cross-sectional design did not permit evaluation of temporal changes, persistence, or transmission dynamics.

Despite these limitations, this study provides baseline information on the prevalence, diversity, combination profiles, and co-occurrence patterns of APEC-associated VAGs in northern Vietnam. Future studies should integrate experimental pathogenicity assessment with serotyping, phylogrouping, MLST, whole-genome sequencing, plasmid analysis, and antimicrobial resistance profiling. Comparative genomic studies incorporating poultry, human, food, and environmental E. coli isolates would also help clarify transmission pathways and the One Health relevance of the VAG-rich APEC lineages identified in this study.