E. coli strains and their sources are listed in Table 1. Strains RM9088, RM9513, RM10410, and RM14516 were isolated by cloacal swab from captured birds (Corvus brachyrhynchos and Agelaius phoeniceus) in a major agricultural region in California as described previously (Cooley et al., 2013). Wildlife sampling at all locations was approved under a set of California Department of Fish and Game (CDFG) Scientific Collection Permits issued to USDA Wildlife Services and CDFG personnel contracted to collect the samples and ship to USDA in Albany, California. Additionally, a federal permit with the U.S. Fish and Wildlife Services was obtained for sampling of geese, crows, and blackbirds. The wildlife sampling was conducted through a contract with state and federal wildlife agencies using their standard protocols. The strains were grown routinely in Luria-Bertani (LB) broth (10 g tryptone, 5 g yeast extract, and 5 g NaCl per liter) (BD Difco).

Bacterial DNA was extracted from the mid-exponential phase cultures grown in LB broth as described previously (Miller et al., 2005). Briefly, cells were lysed with SDS followed by sequential treatment with RNase A and proteinase K. The DNA was first precipitated in a sodium acetate/ethanol solution, and then purified by phenol/chloroform extraction, followed by the final ethanol precipitation. The purified DNA was re-suspended in Qiagen Elution Buffer (QIAGEN) for genome sequencing. Genomic libraries were prepared according to the PacBio (Pacific Biosciences) 20 kb library standard protocol using SMRT-bell DNA template prep kit 3.0, followed by size-selection with BluePippin size selection systems (Sage Science, Inc.), and then with P6v2 template binding step. DNA sequencing was performed on an RS II instrument (Pacific Biosciences, Menlo Park, CA, United States) with P6/C4 sequencing chemistry, and 360-min data collection protocol. The sequence reads were filtered with PreAssembler Filter prior to de-novo assembly with RS_HGAP_Assembly v.3. The complete genome sequences were submitted to GenBank for annotation using Prokaryotic Genome Annotation Pipelines. The GenBank accession numbers are listed in Table 1. The phylogroups were determined using the Clermont method (Clermont et al., 2013). Multi-locus sequence typing (MLST) was conducted using MLST 2.0 service at the Center for Genomic Epidemiology with the Escherichia coli #1 configuration (Achtman 7 gene scheme, adk, fumC, gyrB, icd, mdh, purA, and recA) (Wirth et al., 2006). The serotypes were determined by BLASTn searches of E. coli O-antigen and H-antigen gene databases described previously (Wang et al., 2003; Iguchi et al., 2015).

E. coli virulence genes listed in The Virulence Factor DataBase (VFDB) as of December 2022 were used as queries of BLASTn to search a custom database containing all genomes examined in this study. The listed genes are present in E. coli pathotypes belonging to adherent invasive E. coli (AIEC), avian pathogenic E. coli (APEC), EAEC, Shiga toxin-producing enteroaggregative E. coli (StxEAEC), EHEC, enteropathogenic E. coli (EPEC), ETEC, neonatal meningitis-associated E. coli (NMEC), and uropathogenic E. coli (UPEC) (Supplementary Table 1). The BLASTn was performed in Geneious Prime^® software (Dotmatics) with a threshold of 65% for gene coverage and 70 or 25% for sequence identity at nucleotides or amino acids level, respectively.

Three GIs and seven PAIs known to contribute to stress resistance and/or pathogenicity in enteric pathogens as described previously (Carter et al., 2022a,b) were examined by performing BLASTn searches against a custom database containing all genomes described in this study. The three GIs included AFI, LHI, and TRI; the seven PAIs included LEE, OI-57, OI-122, LAA, HPI, locus of proteolysis activity (LPA), and subtilase encoding pathogenicity island (SE-PAI). When a complete GI or PAI was not detected, each coding sequence (CDS) encoded by the query GI or PAI was used to search the genome of the testing strain by BLASTP to reveal if any homologs were present in the genome of the testing strain. Presence of key virulence genes on each PAI were revealed by BLASTn of each virulence gene followed by mapping their chromosomal locations with the Geneious Prime^® software. Strain EDL933 harbors two TRIs, OI-43 and OI-48, which exhibit over 99% identity. The OI-43 was used as the query in this study.

The complete genome sequence of each avian STEC strain was submitted to PHASTER (Arndt et al., 2016) for identification of prophage and prophage-like elements. The putative integration sites were initially identified by PHASTER and confirmed in Geneious Prime^® software using “Find Repeats” in the defined chromosomal regions. The Stx-prophage genomes were extracted from the corresponding bacterial genomes. Comparative analysis was performed in Geneious Prime^® software. Briefly, Stx-prophages genomes were aligned using Geneious Alignment with the following parameters: Alignment type: Global alignment with free end gaps; Cost Matrix: 65% similarity; Gap open penalty: 12; Gap extension penalty: 3; and Refinement iterations: 2. A neighbor-joining consensus tree was constructed with the following parameters: Genetic Distance Model, Jukes-Cantor; Resampling Method, bootstrap; and number of replicates, 10,000.

Curli fimbriae were examined by growing each strain at 26°C for 1 or 2 days on the Congo Red indicator (CRI) plates, consisting of LB agar plates without sodium chloride and supplemented with 40 μg/ml of Congo Red dye and 10 μg/ml of Coomassie Brilliant Blue, as described previously (Carter et al., 2011). Curli-producing strains were indicated by red colonies whereas curli-deficient strains were indicated by white colonies on CRI plates. Biofilm assays were carried out as described previously (Carter et al., 2016, 2018). Briefly, one ml of LB broth inoculated with 1 × 10⁶ cells/ml was aliquoted into a borosilicate glass tube and then incubated statically at 28°C for 24, 48, and 120 h. At the end of each incubation, the planktonic cells were removed carefully, and the tubes were rinsed twice with one ml sterile distilled water and then stained with one ml 0.1% crystal violet at room temperature for 30 min. The dye was then removed gently, and the tubes were washed twice with sterile distilled water. The crystal violet bound to the glass tube was solubilized in 0.5 ml of 33% acetic acid and the absorbance was determined at 570 nm using a microplate reader (SpectraMax 340; Molecular Devices, Sunnyvale, CA). Tubes with uninoculated media served as negative controls. Each data set was the average of results from at least three biological replicates. All data were first evaluated for normal distribution by the Shapiro-Wilk test using Graph Pad Prism 10 software (Dotmatics). The difference in biofilm formation, represented by the absorbance at 570 nm, among the avian strains was assessed by the adjusted P-value of the Tukey’s multiple comparisons test after a One-way ANOVA test (*P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; ****P ≤ 0.0001). Similarly, the difference in biofilm formation of each strain at various incubation times was assessed by the adjusted P-value of the Tukey’s multiple comparisons test after a One-way ANOVA test.

The Stx activity of the avian E. coli strains was measured using a mammalian Vero cell line, Vero-d2EGFP, that harbored a destabilized variant (t_1/2 = 2 hr) of the enhanced green fluorescent protein (EGFP) as described previously (Quiñones et al., 2009; Quiñones and Swimley, 2011). Strains EDL933 and K-12 sub-strain MG1655 were used as a positive and a negative control for the assay, respectively. Cell-free culture supernatants for all tested strains were prepared as previously described (Quiñones et al., 2009; Quiñones and Swimley, 2011; Silva et al., 2019). One-day prior to intoxication, Vero-d2EGFP cells were seeded at 9,500 cells per well in Greiner Bio-One CELLSTAR^® black 96-well microplates with μClear^® bottoms (VWR International) and were grown at 37°C under humidified conditions with 5% CO₂ in Ham’s F-12 Nutrient medium (Life Technologies Corp.), supplemented with 10% fetal bovine serum (American Type Culture Collection) and 1% penicillin-streptomycin (Life Technologies). The Vero-d2EGFP cells were then exposed to Ham’s F-12 medium containing tenfold dilutions of cell-free culture supernatants from each bacterial strain and were incubated at 37°C for 20 h in a 5% CO₂ humidified incubator. Following intoxication, the Vero-d2EGFP cells were briefly rinsed three times with 1X phosphate-buffered saline, and the EGFP fluorescence from the Vero-d2EGFP cells was measured using a BioTek Synergy HT Multi-Detection Microplate Reader (Agilent Technologies) with the 485/20 nm excitation filter and the 528/20 nm emission filter. All experiments were performed in triplicate and each experimental condition consisted of six replicates. The results were expressed as percentages of the GFP fluorescence values for culture supernatant-treated Vero-d2EGFP cells when compared to the fluorescence values from control Vero-d2EGFP cells incubated without bacterial supernatants. The significant differences in the cytotoxicity levels among the STEC strains were determined with a One-way ANOVA test followed by a Tukey-HSD post hoc analysis (P-value < 0.05) in R version 4.2 (R Core Team, 2022).

Unlike STEC O157:H7 strain EDL933, in silico phylo-typing placed the three E. coli crow strains, RM9088, RM9513, and RM10410, in phylogroup A, while the blackbird E. coli strain RM14516 in phylogroup D (Table 1). For the genomes of the three crow strains, different plasmid profiles were detected. RM9088 contained two large plasmids and one of them harbored enterohemolysin genes (hlyCABD), thus was designated as pEHEC; RM9513 contained one large plasmid that lacked the enterohemolysin genes; while no plasmids were identified in RM10410. These three crow strains also carried different stx genes and had different serotypes and genome types, determined by the MLST-based sequence types (STs) (Table 1). Strain RM9088 carried the stx_1a genes only, had a serotype of O109:H48, and belonged to ST339; strain RM9513 did not carry any stx genes, had a serotype of O9:H30, and belonged to ST1294; and strain RM10410 carried both stx_1a and stx_2d genes, had a serotype of O113:H4, and belonged to ST10, the same sequence type as the E. coli K-12 sub-strain MG1655. Like strain EDL933, the genome of the blackbird strain RM14516 was composed of a chromosome and a pEHEC. This blackbird strain carried only stx_2a genes, had a serotype of O17:H18, and belonged to ST6507.

A total of 313 genes related to bacterial adherence, biofilm formation, bacterial invasion, toxins/effectors production and delivery, and iron acquisition were examined systematically (Supplementary Table 1). Among the three stx-positive avian strains, 102, 55, and 87 virulence genes were detected in RM9088, RM10410, and RM14516, respectively, while 58 virulence genes were detected in stx-negative strain RM9513 (Figure 1). Strain RM9088 had the greatest number of genes related to fimbriae, Type VI Secretion System (T6SS), toxins and toxin delivery systems, and iron uptake and storage (Figure 1).

Like the EHEC prototype strain EDL933 and the nonpathogenic K-12 sub-strain MG1655, all four avian E. coli strains carried the genes encoding curli fimbriae, hemorrhagic E. coli pilus, and type I fimbriae. The E. coli common pilus genes were also identified in these avian strains although with a point deletion and an amber mutation in the ecpC of strain RM9088 and the ecpD of strain RM10410, respectively. In contrast, none of the avian strains carried the genes related to biosynthesis of the afimbrial adhesin AFA-I, the S fimbriae, or the aggregative adherence fimbriae that are commonly present on the large aggregative plasmid pAA in EAEC strains (Supplementary Table 1). Difference in fimbriae genes content was also revealed. For example, the E. coli laminin-binding fimbriae genes were detected in all crow strains, whereas the F1C fimbriae genes were present only in the blackbird strain RM14516 (Table 2). The crow strain RM9088 appeared to carry more fimbriae genes than any other avian strains examined. For example, homologs of adhesive K88 fimbriae genes were present on the pEHEC of RM9088; Homologs of colonization factor antigen I (CFA/I) genes were detected in the chromosome of RM9088 (Table 2). The crow strain RM10410 was the only one carrying homologs of P fimbriae genes although a single base deletion and an IS66 insertion were detected in papH and papA, respectively (Table 2).

Similarly, differences in adhesin genes profile were detected among the avian strains. The genes encoding intimin-like adhesin EaeH and an AIDA-1 type autotransporter EhaB were present in all avian strains; the upaG gene was detected in all strains except RM10410; and the genes agn43 and upaH were detected in all strains except RM9513 (Table 2). Moreover, cah, encoding the autoaggregative protein Cah was detected in strains RM9513 and RM10410; paa, encoding a porcine attaching-effacing associated protein, was only detected in the crow strain RM9088; and eaeX, encoding an inverse autotransporter adhesin IatC, was only detected in the blackbird strain RM14516 (Table 2).

Subsequent examination of the genes related to the production and delivery of toxins and effectors revealed that, unlike the strain EDL933, none of the avian E. coli strains carried homologs of genes involved in biosynthesis of the Type III Secretion System (T3SS) apparatus (Supplementary Table 1). Consistently, no LEE island was identified in any of the avian strains. Among the 60 genes encoding T3SS effector proteins examined, homologs of 13 effector genes were identified in strain RM14516, while five effector genes were identified in other avian strains (Supplementary Table 1).

The E. coli T6SS gene clusters are classified in three distinct phylogroups (T6SS-1 to T6SS-3) based on the gene organization and sequence similarities (Journet and Cascales, 2016), and among them, T6SS-1 and T6SS-2 are the most prevalent. Examining the three T6SS gene clusters revealed that, although homologs of T6SS-1 genes were not identified in any of the avian strains, homologs of T6SS-2 genes were identified in all strains (Figure 2A). A total of 18 genes in strain RM9088 showed homology to T6SS-2 genes and were clustered on a 23-Kb GI located upstream of the gene encoding a RHS repeating protein (locus tag: FRV13_02745). All genes were predicted to be functional except tssM (locus tag: FRV13_02665), which carried an amber mutation. A T6SS-2 gene cluster was also identified in strain RM9513, located upstream of the gene encoding a RHS repeating protein (locus tag: FS836_01975) (Figure 2A). However, unlike the T6SS-2 in strain RM9088, a large deletion (∼2.4 Kb) spanning from the coding sequence of tagO to the coding sequence of tssH was revealed in strain RM9513. This deletion appeared to be mediated by the recombination between the two direct repeats located within the coding sequence of tagO and tssH, respectively, based on sequence analyses (Figure 2A). All 17 T6SS-2 genes in strain RM14516 were predicted to be functional. In strain RM10410, a short DNA segment containing the T6SS genes hcp, impA, and a truncated tssM was identified, located at the same chromosomal locations as the T6SS-2 in other avian strains. Sequence alignment revealed that the T6SS-2 genes in strain RM14516 was highly similar to the T6SS-2 in APEC strain O1, while the T6SS-2 genes in both RM9088 and RM9513 were more closely related to T6SS-2 genes in the STEC O104:H4 outbreak strain 2011C-3493 (Figure 2B). Homologs of T6SS-3 genes were only revealed in strain RM9088, which were located on the plasmid pEHEC. High sequence similarity was detected for a total of 15 genes when compared to the T6SS-3 genes identified in the STEC O104:H4 outbreak strain 2011C-3419, regardless of an IS110 insertion sequence between the genes aaiL and aaiN in strain RM9088 (Figure 2C).

In addition to T3SS and T6SS, other secretion systems and autotransporters to deliver toxins or effector proteins were examined systematically. Among the 13 genes related to delivery of toxins or effector proteins, only a few were detected, including cdiAB in all four strains, pic in strain RM9088, espI in strain RM10410, and espP in strain RM14516 (Supplementary Table 1). Besides the stx genes and enterohemolysin genes described previously, the chromosome-borne hemolysin gene hlyE was detected in all avian strains, although a large deletion and an IS4-insertion in the coding sequencing of hlyE was detected in RM9088 and RM10410, respectively. Interestingly, the heat-labile toxin genes eltAB were detected in strain RM9088, implying that RM9088 carried a hybrid pathotype of STEC and ETEC.

Diverse integration sites for Stx-prophages were revealed in the avian STEC strains. The Stx1a-prophage in RM9088 was inserted within the gene ompW, the same insertion site utilized by the Stx2a-prophages in STEC O121:H19 strains (Carter et al., 2022b; Table 3). This integration site in strain EDL933 is occupied by the O-island #102. In RM10410, the Stx1a-prophage was inserted within the ynfG gene (also known as dmsB), while the Stx2d-prophage was inserted at the same site as the Stx1a-prophage in strain EDL933 (Table 3). The Stx2a-prophage in strain RM14516 is inserted within gene encoding the tRNA dihydrouridine synthase DusA (Figure 3A). Sequence comparison of the Stx-prophages in the avian strains with those in EDL933 revealed two clusters (Figure 3A). Both Stx1a-prophages in the crow strains RM9088 and RM10410 were grouped together with the Stx1a- and Stx2a-prophages in EDL933, while the Stx2d in the crow strain RM10410 displayed the highest sequence similarity with the Stx2a-prophage in the blackbird strain RM14516.

The regulatory elements of the stx_1a in the crow strains RM9088 and RM10410, including antitermination protein Q, the late promoter P_R’, and the qut site, were highly similar to those carried by the EDL933 Stx1a-prophage (Figure 3B). The regulatory elements of the stx_2d in the crow strain RM10410 was highly similar to those carried by the EDL933 Stx2a-prophage (Figure 3C). However, this region exhibited a large sequence divergence in the blackbird strain RM14516. Specifically, the Stx2a-prophage in strain RM14516 encoded an antitermination protein exhibiting 47.6% sequence identity with the Q protein encoded by EDL933 Stx2a-prophage, carried an altered qut site, and lacked the late promoter P_R’. Moreover, there were two CDSs located between the coding sequences of the antitermination protein gene and the stx_2a.

A 13-Kb AFI was present in strain RM9088, carrying the genes related to acid resistance and the genes encoding the efflux pump. This AFI was nearly identical to the AFI in E. coli strain MG1655 (% Identity, 99.1) (Table 4). The complete LAA is about 81 Kb and carries 89 CDSs as previously reported in STEC O91:H21 strain B2F1 (Montero et al., 2017). The LAA virulence genes are distributed in the four modules, including sisA and hes on module I, iha and lesP on module II, pagC, tpsA and tpsB on module III, and agn43 on module IV. A complete LAA was not detected in RM9088, but homologs of module I and module IV were identified (Table 4). The module I carried the genes atoSC and atoDAB, related to acetoacetate metabolism, and atoE, a short chain fatty acid transporter gene. Neither of the two virulence genes (sisA and hes) were present in module I and no homologs of sisA or hes were identified after searching the entire RM9088 genome. The LAA module IV was similar in length when compared to the reference strain B2F1, exhibiting over 74% sequence identity, and carrying the virulence gene agn43. On the identification of additional PAIs in strain RM9088, no homologs of OI-57, OI-122, and TRI were revealed, however, a complete HPI was detected. In Y. pestis, the 36-Kb HPI is located on a 102-Kb mobile GI adjacent to the tRNA gene asnT, and harbors genes involved in iron uptake (Buchrieser et al., 1998; Carniel, 1999). The 32-Kb HPI in strain RM9088 was found also integrated downstream of the tRNA gene asnT (Figure 4A). Furthermore, this HPI exhibited 99% sequence identity with the HPI in Y. pestis and carried all genes required for the biosynthesis, transport, and regulation of the siderophore yersiniabactin (Figure 4B). The putative integration site for HPI is conserved in the crow strains RM9513 and RM10410, and in several STEC strains belonging to the clinically relevant serotypes including STEC O104:H4 and STEC O26:H11 (Data not shown).

As observed for strain RM9088, strain RM9513 carried an AFI highly similar to the AFI in strain MG1655, and an incomplete LAA with gene homologs in the module I and IV fragments, located apart on the chromosome (Table 4). The LAA module I fragment carried the genes related to acetoacetate metabolism but lacked the virulence genes sisA and hes. The module IV carried the autotransporter gene cah, a homologue of agn43 (Supplementary Table 1). No homologs of LAA module II or III were identified in RM9513. Unlike strain RM9088, strain RM9513 carried an incomplete OI-57, exhibiting sequence similarity with a 20-Kb DNA segment in the OI-57 in strain EDL933. The complete OI-57 is about 80 Kb, carrying virulence genes adfO (paa) and ckf, and is associated with the highly pathogenic seropathotypes A and B (Imamovic et al., 2010). The virulence gene ckf was present in the OI-57 in strain RM9513.

The AFI in strain RM10410 was highly similar to the AFIs in strains MG1655, RM9088, and RM9513 (Table 4). However, unlike the crow strains RM9088 or RM9513, a 47-Kb TRI was detected in RM10410, which was located adjacent to the tRNA gene ileX and contained the genes conferring resistance to tellurite and the urease gene cluster (Figures 4C, D). Similar to the TRIs in EHEC strain EDL933, two 12-bp DRs were identified flanking the TRI in strain RM10410 (Figure 4C). Homologs of LAA module I, II/III, and IV were detected at three separate locations on the RM10410 chromosome (Table 4). The module I was highly similar to the LAA module I in strains MG1655, RM9088, and RM9513. However, unlike the above strains, both virulence genes (sisA and hes) were detected in the RM10410 genome but located outside of module I (sisA: 4,940,577–4,939,534; hes: 3,443,911–3,443,165). Homologs of LAA module II and III were located adjacently, containing all virulence genes on the LAA module II and III in the reference strain B2F1, including iha (locus tag: FS611_15505) and lesP (locus tag: FS611_15475) on the module II, and pagC (locus tag: FS611_15370), tpsA (locus tag: FS611_15350), and tpsB (locus tag: FS611_15330) on the module III. Additionally, two homologs of LAA module IV were detected in strain RM10410 (Table 4), one contained virulence gene agn43 (locus tag: FS611_21665) while the other one contained the gene cah (locus tag: FS611_24255).

Among the avian strains examined, only the blackbird strain RM14516 harbored a 22-Kb AFI highly similar to the one in strain EDL933 (Table 4), in which, the iron acquisition genes chuSA and chuTWXYU were inserted between the genes encoding a transcription regulator DctR (locus tag: FS841_18470) and a transport ATPase YhiD (locus tag: FS841_18520). Homologs of LAA module I, III, and IV were also detected in strain RM14516 but were located in different regions on the chromosome. The LAA module I carried the genes related to acetoacetate metabolism but lacked the virulence genes sisA and hes, similarly to the LAA module I in strains RM9088, RM9513, and MG1655 (Table 4). Although no homolog of LAA module II was identified in strain RM14516, the virulence genes iha and lesP were present on the plasmid pEHEC. The module III was highly similar to the LAA module III in the reference strain B2F1 (% identity, >90), containing virulence gene tpsA and tpsB. No homolog of pagC was revealed in the genome of RM14516. The LAA module IV in strain RM14516 contained virulence gene agn43 and exhibited 76% sequence identity with the LAA module IV in the reference strain B2F1. Although a complete SE-PAI was not detected in strain RM14516, the virulence genes subAB (locus tags: FS841_23955 and FS841_23960), encoding AB₅ subtilase cytotoxin, were present on the plasmid pEHEC.

Biofilm formation by E. coli avian strains was evaluated quantitatively. Following 24 h incubation, a great amount of surface-associated biomass, exemplified as a strong ring on the glass surface, was observed on all tubes inoculated with strain RM9088 (Figure 5A). Consistently, the biofilm of RM9088 was significantly greater than that of any other strains examined (One-way ANOVA test followed by Tukey’s test, adjust P < 0.0001) (Figure 5B). Although a visible dye ring was observed on glass tubes inoculated with strain RM14516, the biofilm was not significantly greater than that of strains RM9513 or RM10410. No visible rings were observed for tubes inoculated with either RM9513 or RM10410. Following 48 h of incubation, biofilms of strain RM9088 increased significantly compared with the biofilms at 24 h (One-way ANOVA test followed by Tukey’s test, adjust P = 0.0064) and was significantly greater than that of any other strains examined (One-way ANOVA test followed by Tukey’s multiple comparisons test, adjust P < 0.0001) (Figure 5D). Biofilms of strain RM14516 at 48 h were significantly greater than that of strains RM9513 or RM10410 when the ANOVA test was performed for strains RM14516, RM9513, and RM10410 (Tukey’s test, adjust P < 0.05). No visible rings were observed for tubes inoculated with strain RM9513 or RM10410 following either 24 h or 48 h incubation (Figures 5A, C). The levels of biofilm by both strains RM9088 and RM14516 continued to increase when the incubation time was extended to 120 h (Figure 5E). At 120 h, the biofilms of strain RM9088 were 17.7- and 2.2-fold larger in size when compared to the biofilms detected at 24 and 48 h, respectively. Moreover, the levels of biofilm by strain RM14516 were 12.4- and 4.6-fold higher than those detected at 24 and 48 h, respectively (Figure 5F). In contrast, no biofilms were detected for strains RM9513 or RM10410 even when incubation time was increased to 120 h.

Because all avian strains harbored intact curli genes, production of curli fimbriae were examined. Strains RM9088 and RM14516 were found to be a strong and moderate curli producer as colonies exhibited a dark red and red color on CRI, respectively. By contrast, RM9513 was a weak curli producer and RM10410 was a curli-deficient strain as colonies grown on CRI plates appeared pink and white, respectively (Figure 5G).

The Stx-mediated cytotoxicity of the avian strains was assessed by using the mammalian host Vero-d2EGFP cells and bacterial cell-free culture supernatants. Incubation with high concentration of active Stx resulted in low levels of fluorescence in the host Vero-d2EGFP cells. The detected levels of cytotoxicity in the examined avian strains were compared to those levels detected when testing Stx-negative E. coli strain MG1655 and the Stx-positive EHEC prototype strain EDL933. When testing a 10-fold dilution of culture supernatant from avian strain RM10410 (stx_1a + stx_2d), the cytotoxicity level was comparable to the EHEC strain EDL933 (stx_1a + stx_2a) and was significantly higher in Stx activity than the negative control strain MG1655 and the stx-negative crow strain RM9513 (Figure 6A). Interestingly, the greatest cytotoxicity was detected when testing culture supernatants from strains RM9088 (stx_1a) and RM14516 (stx_2a), which both were higher than that of the EHEC strain EDL933. A similar result was observed when the Vero-d2EGFP cells were intoxicated with the 100-fold dilution of the Stx-containing supernatants (Figure 6B). When the Vero-d2EGFP cells were intoxicated with the 1000-fold dilution of the Stx-containing supernatants, the avian strain RM9088 exhibited greater cytotoxicity than strain RM14516, which both were significantly greater than any of other strains examined. The avian strain RM10410 displayed similar cytotoxicity levels as the strain EDL933, both of which were significantly higher than the negative control strain MG1655 and the stx-negative avian strain RM9513 (Figure 6C).