The aEPEC strain BA92 used in this study belongs to the serotype O2:H16 and was isolated during an epidemiological study carried out in the city Salvador, State of Bahia, Brazil (Bueris et al., 2007; Abe et al., 2009). All other bacterial strains and plasmids employed are listed in Table 1 . Routinely, these strains were grown in lysogeny broth (LB) at 37°C, supplemented as appropriate with kanamycin (50 µg/mL) or ampicillin (100 µg/mL), and stored at -80°C in LB supplemented with 30% glycerol.

In the present study, the genome of aEPEC BA92 was sequenced using the Oxford Nanopore Technologies (ONT) MinION long-read sequencer. Bacterial genomic DNA was extracted using the QIAamp DNA Micro Kit (Qiagen, NW, Germany) as recommended by the manufacturer. The library was prepared using Rapid sequencing gDNA - barcoding (SQK-RBK004, ONT) and sequenced with a R9.4.1 MinION flow cell using MinKNOW (v22.03) with default settings. Basecalling was performed with Guppy (v5.1, ONT) to obtain long-read FASTQ files.

Further, this data was combined with the Illumina short reads (NCBI accession number: PIJZ00000000), that was obtained in a previous study (Hernandes et al., 2020), for hybrid genome assembly using Unicycler (v0.5.0) (Wick et al., 2017). The genome was annotated using Prokka (v1.14.5) (Seemann, 2014) and Easyfig (v3.0.0) (Sullivan et al., 2011) was employed to illustrate the insertion point of the eaa-containing region in the genome of the aEPEC BA92 strain. For comparative purposes, the draft genome sequence of the aEPEC serotype O2:H16 IAL5132 (NCBI accession number: PIKD01000024.1, contig No. 24), that lacks eaa, was used (Hernandes et al., 2020). Furthermore, the presence of prophages in the genome of the aEPEC BA92 strain was evaluated using PHASTEST (v3.0) (Wishart et al., 2023).

The amino acid sequence of Eaa was deduced from the nucleotide sequence using CLC Main Workbench 7 software (Qiagen, NW, Germany), and its molecular mass calculated with the Compute pI/MW tool available at the ExPASy website (https://web.expasy.org/compute_pi/) (Gasteiger et al., 2003). The signal peptide cleavage site was predicted using SignalP 6.0 (https://services.healthtech.dtu.dk/services/SignalP-6.0/) (Teufel et al., 2022), and the domains present in Eaa identified using Pfam and InterPro databases (https://www.ebi.ac.uk/interpro/) (Mistry et al., 2021; Paysan-Lafosse et al., 2023).

ClustalW algorithm (Larkin et al., 2007) was used in the MEGA11 software (Tamura et al., 2021) to align Eaa with other characterized autotransporter proteins ( Supplementary Table S1 ). The best model to build a maximum likelihood tree was investigated in the IQ-TREE 1.6.12 tool (Nguyen et al., 2015) in combination with the ModelFinder algorithm (Kalyaanamoorthy et al., 2017), which suggested the PMB (Probability Matrix from Blocks model (Veerassamy et al., 2003), with parameters +F+G4 and Gamma shape alpha = 7.1547, as the most appropriate model and parameters to be employed in this analysis. The maximum likelihood was calculated using the UFBoot2 algorithm (Hoang et al., 2018) with 1,000 bootstraps. The resulting Newick file, generated by the IQ-TREE tool, was used as input in the online iTOL tool (https://itol.embl.de/) (Letunic and Bork, 2021) for visualization of the maximum likelihood tree.

The in silico Eaa tertiary structure was predicted using ColabFold (https://colab.research.google.com/github/sokrypton/ColabFold/blob/main/AlphaFold2.ipynb) (Mirdita et al., 2022).

In the present study two recombinant plasmids were constructed: pHO, which corresponds to the pET-28a vector harboring the portion of the eaa gene encoding the predicted passenger domain of the autotransporter protein (from nucleotide 62 to 1,415); and pIC, which corresponds to the pBAD/Myc-His A vector harboring the entire predicted eaa gene ( Table 1 ). Both plasmid vectors were extracted from the host bacteria using the QIAprep spin miniprep kit (Qiagen, NW, Germany) and the inserts were prepared by polymerase chain reaction (PCR) amplification performed with the Platinum^® Taq DNA Polymerase High Fidelity enzyme (Invitrogen, Vilnius, Lithuania) and using the DNA from the aEPEC BA92 as template. The primers used to amplify the portion of the gene encoding the passenger domain (pET28a-BamHI-F and pET28a-XhoI-R) or the entire eaa gene (Forward-XhoI-F and Reverse-KpnI-R) are described in Table 2 .

The part of the gene that encodes the Eaa passenger domain was cloned in BamHI and XhoI sites of pET-28a, generating the pHO recombinant plasmid. This allowed the production of the passenger domain carrying a 6xHis-tag on the C-terminal region. The entire eaa gene was cloned in XhoI and KpnI sites of pBAD/Myc-His, generating the pIC recombinant plasmid ( Table 1 ). The vectors and the respective inserts were digested with the appropriate enzymes (Invitrogen, Vilnius, Lithuania) and ligated using T4 DNA ligase (Invitrogen, Vilnius, Lithuania) following the manufacturer’s recommendations. The recombinant plasmid pHO was transformed into the BL21(DE3) strain, and the pIC was transformed into the MS427 strain by heat shock (Sambrook et al., 1989).

After the bacterial transformation, the presence of the pHO and pIC recombinant plasmids in the host bacteria cells were verified by PCR performed with the GoTaq Master Mix enzyme (Promega, WI, USA) and using the eaa-F and eaa-R primers. Subsequently, the DNA sequences cloned in the pHO and pIC recombinant plasmids were confirmed through Sanger sequencing (Sanger et al., 1977) using the primers T7 promoter/T7 terminator and pBAD-F/pBAD-R, respectively. All primers were used at a concentration of 0.34 μM each and are described in Table 2 .

The production and purification of the Eaa passenger domain was performed as previously described (Chura-Chambi et al., 2022), with some modifications. The BL21(DE3)(pHO) strain was inoculated in 1 L of the rich medium 2-fold HKSII (Jensen and Carlsen, 1990) containing 50 µg/mL of kanamycin and incubated at 37°C. Induction of Eaa passenger domain carrying a His-tag expression was performed by the addition of 0.5 mM of IPTG, when culture reached the optical density of 600 nm (OD600) between 2.0 and 3.0, and cultivation continued at 30°C for 16 h.

The cultures were centrifuged at 4000 x g for 10 min at 4°C, and the resulting pellet was resuspended on lysis buffer (0.1 M Tris HCl- pH 7.0, 5 mM EDTA and 0.1% sodium deoxycholate) and lysed using the Bandelin Sonopuls HD 2070 (Bandelin, BE, Germany). Then, the pellets were washed twice with a buffer containing 0.1 M Tris HCl, 5 mM EDTA and 0.1% of sodium deoxycholate. After the washing step, the pellets were resuspended in a solution containing 50 mM of Tris and 1 mM EDTA (pH 7.0), then centrifuged and suspended in the same buffer. The suspension was transferred to a plastic bag and vacuum sealed (R4-6-40, High Pressure Equipment, USA). After that, it was pressurized at 2,4 kbar for 90 min with a pump (PS-50, High Pressure Equipment, USA) and an air compressor, using an oil-in-water mixture as transmission fluid. After decompression, the samples were centrifuged at 12,000 x g for 15 min, dialyzed in membrane against 25 mM Tris pH 7.0 for 18 h at 4°C under agitation, centrifuged and stocked at -20°C until further use.

The protein was purified using immobilized metal affinity chromatography (IMAC) with a HisTrap FF (GE Healthcare, MA, USA) affinity nickel chromatographic column. This column coupled with the Äkta (GE Healthcare, MA, USA) was balanced using a buffer containing 25 mM Tris pH 7.0 and 150 mM NaCl. Elution was performed using 25 mM Tris pH 7.0, 150 mM NaCl and 1 M imidazol, applying a gradient of 0 to 1 M of imidazol for 30 min (1 mL/min). The eluted fractions were dialyzed against 25 mM sodium phosphate pH 7.0 and later submitted to immunolabeling using anti-His antibodies.

Anti-Eaa polyclonal serum was produced as described by Evans et al. (1975), with some modifications. Of note, pre-immune serum was collected from the rabbit before the immunization protocol. Briefly, a New Zealand White female rabbit was intravenously inoculated with 100 µg of the purified passenger domain of Eaa and 200 µL of Montanide ISA 50 V2 (Butantan Institute, SP, Brazil) as adjuvant. Three immunizations were performed with 15-day intervals. 15 days after the final immunization, approximately 50 mL of blood was collected. The serum fraction was separated and depleted of the complement by heating at 56°C for 30 min. For serum adsorption, 10 glass tubes with 3 mL of LB inoculated with BL21(DE3)(pET-28a) were grown for approximately 18 h, centrifuged and the resulting pellets were stored at -20°C until the time of use. Then, an aliquot of the anti-Eaa serum was mixed with a pellet of the BL21(DE3)(pET-28a) strain and left at gentle agitation for 12 h at room temperature. At the end of this period, the preparations were centrifuged and the anti-Eaa serum was mixed with another frozen pellet of the BL21(DE3)(pET-28a) strain for 12 h. These cycles were repeated until the 10 frozen pellets were used. Finally, the anti-Eaa serum was stored at -20°C. The anti- Eaa serum production protocol was approved by the Ethics Committee on Animal Use of the Butantan Institute (CEUAIB protocol: 9452010316).

The purified recombinant protein was submitted to SDS-PAGE 12% (Laemmli, 1970) under denaturing conditions, transferred to a nitrocellulose membrane (Bio-Rad, CA, USA) and examined with the anti-Eaa serum. Then, the membrane was incubated in 10% skim milk (diluted in PBS: phosphate buffered saline pH 7.4) at 4°C for 16 h, washed three times with PBS supplemented with 0.05% Tween-20 (PBS-T), and, further, incubated with anti-Eaa serum (1:200) at room temperature for 1 h. Then, the membrane was washed with PBS-T and incubated with goat anti-rabbit IgG secondary antibodies conjugated with peroxidase (Sigma-Aldrich, HE, Germany) (1:10,000) for 1 h at room temperature. After washing with PBS-T, the membrane was developed using SuperSignal West Pico PLUS Chemiluminescent Substrate (Thermo Fisher Scientific, MA, EUA) and visualized using the transilluminator Amersham Imager 600 (GE Healthcare, MA, USA).

Bacterial strains aEPEC BA92, MS427(pBAD) and MS427(pIC) were cultivated overnight at 37°C in LB broth supplemented with 0.2% L-arabinose (Sigma-Aldrich, HE, Germany) and 100 μg/mL ampicillin, when appropriate. Bacterial cultures were then centrifuged (2,348 x g for 5 min) and the supernatant was discarded. After four washings (2,348 x g for 5 min centrifugation) with PBS, preparations were fixed with 4% paraformaldehyde in PBS (Sigma-Aldrich, HE, Germany) for 30 min at room temperature. Preparations were then washed four times with PBS containing 0.2% bovine serum albumin (BSA) (Thermo Fisher Scientific, MA, USA) (PBS-BSA), and blocked in the same solution for 60 min at room temperature. After centrifugation to remove PBS- BSA, preparations were incubated with rabbit pre-immune serum or anti-Eaa serum, diluted (1:25) in PBS, for 18 h at 4°C. After this period, the preparations were washed five times with PBS-BSA and incubated with a goat anti-rabbit serum conjugated with 10 nm colloidal gold particles (Sigma-Aldrich, HE, Germany) diluted (1:25) in PBS at room temperature for 5 h. After five washings with PBS, the preparations were applied onto 200 mesh copper grids (Electron Microscopy Sciences, PA, USA) previously prepared with 0.5% Formvar (Sigma-Aldrich, HE, Germany) in chloroform (Merck, HE, Germany), and left for 2 min. After removing the excess of liquid and drying at room temperature, the preparations were analyzed using a LEO 906E transmission electron microscope (Zeiss, BW, Germany) operating at 80 kV in the Structural and Functional Biology Laboratory of the Butantan Institute.

Autoaggregation assays were carried out as previously described (Kjærgaard et al., 2000b), with modifications. Briefly, the bacterial strains were incubated at 37°C with agitation for 18 h in 40 mL of Brain Heart Infusion (BHI) broth (Oxoid, Basingstoke, England) supplemented with 0.2% L-arabinose and 100 µg/mL ampicillin. The cultures were then centrifuged at 5,000 x g for 10 min and the supernatant discarded. The resulting pellet was resuspended in sterile BHI, containing 0.2% L-arabinose and 100 µg/mL ampicillin, and the OD600 adjusted to 1.0. Then, 10 mL of each preparation were divided into two glass tubes with 5 mL each, one of which was kept static during all the assay, while the other was homogenized before the OD600 readings. The autoaggregation assay was carried out for a total of 2 h. After 20 min, 200 µL of each preparation was collected from the very top and transferred to a 96-well plate. The OD600 was measured using the BioTek Epoch 2 with the Gen5 software (Agilent, CA, USA).

At the end of the autoaggregation assay, 10 μL of the bottom of each culture was collected and transferred to a glass microscope slide, which was heat-fixed and subsequently stained with 4’,6-Diamidino-2-phenylindole (DAPI; Thermo Fisher Scientific, MA, USA) (diluted 1:500 in PBS) for 20 min and kept protected from light during this period. Then, the slides were washed by immersion in distilled water, dried, and analyzed under a BX60 fluorescence microscope (Olympus, Kanto, Japan).

The ability of the bacterial strains to produce biofilm on polystyrene plates was performed as previously described (Oliveira-Garcia et al., 2003), with modifications. In each well of a 96-well microplate (Tissue Culture Plates 96 well – K12-096, Kasvi, PR, Brazil), 200 μl of LB were added containing 2% D-mannopyranoside (Sigma-Aldrich, HE, Germany), 0.2% L-arabinose and 100 μg/mL ampicillin, as well as 10 μL of overnight bacterial cultures. After 24, 48 or 72 h of incubation at 37 °C, each well was washed with 200 μL of PBS and then fixed with 200 μL of 3% formaldehyde for 1 h. After, the wells were washed with distilled water and the preparations stained with 1% crystal violet for 20 min. After the staining step, the wells were washed with distilled water and the plate was dried at room temperature for approximately 2 h. The crystal violet was solubilized with 200 μL of methanol for 10 min and the OD570 measured in an ELISA BioTek Epoch 2 reader with the Gen5 software.

Additionally, we also investigated the ability of the bacterial strains to produce biofilm on glass surfaces. Sterile glass coverslips were introduced into 24-well polystyrene plates (Tissue Culture Testplate 24 – 92024, TPP, SH, Switzerland). Subsequently, 950 µL of LB, containing 2% D-mannopyranoside, 0.2% L-arabinose and 100 µg/mL ampicillin, were added as well as 50 µL of the bacterial cultures. The plates were then incubated at 37°C for 24, 48 or 72 h. After the distinct periods of incubation, the wells were washed with 1 mL of PBS, followed by fixation with 1 mL of 3% formaldehyde for 1 h. Then, the wells were washed with distilled water and the coverslips were transferred to a novel 24-well plate. The glass coverslips were stained with 1 mL of 0.1% crystal violet for 20 min, washed with distilled water and dried at room temperature. The crystal violet was then resuspended with 33% glacial acetic acid for 10 min and the OD570 measured in a BioTek Epoch 2 ELISA reader with the aid of the Gen5 software.

Biofilm production on polystyrene was performed in three biological replicates, carried out on different days, with five technical replicates being performed in each assay. For the glass surfaces tests, three technical replicates were performed instead.

Eaa binding to ECM components was carried out as previously described (Salazar et al., 2014), with modifications. Collagens type I, III, (Corning, NI, USA), IV and V, cellular and plasma fibronectin, laminin, fibrinogen, vitronectin, and the negative controls BSA and fetuin (Sigma-Aldrich, HE, Germany), diluted at 10 μg/mL in PBS, were immobilized in 96-well plates (96 Well EIA/RIA Assay Microplate – high binding surface – 3590, Corning, NI, USA) and incubated at 4°C for 18 h. The wells were washed three times with phosphate buffered saline – 0.05% Tween 20 (PBS-T) and non-specific binding sites were blocked using 100 µL of 1% BSA for 2 h at 37°C. Then, 100 µL of recombinant Eaa (0.1 μM) diluted in PBS were added, and incubation proceeded for 1.5 h at 37°C. After three washes with PBS-T, anti-Eaa serum produced in rabbit (diluted 1:1,000 in PBS) was added and incubated for 1 h at 37°C. Subsequently, peroxidase conjugated goat anti-rabbit IgG (diluted 1:5,000 in PBS) (Sigma-Aldrich, HE, Germany), was added and incubated for an additional hour at 37°C. The wells were washed three times and o-phenylenediamine (0.04%) in citrate phosphate buffer (pH 5.0) plus 0.01% H₂O₂ was added. The reaction was carried out for 15 min and stopped with 50 µL of H₂SO₄ 8N. The absorbance was read at 492 nm using the Multiskan EX device (Thermo Fisher Scientific, MA, USA).

To further explore the interaction of Eaa with fibronectin, a dose-dependent binding using up to 2 μM of recombinant protein was carried out. Furthermore, mapping of the fibronectin interacting domain with the Eaa passenger domain was also evaluated. In this case, F30 (heparin binding domain – HBD) (Sigma-Aldrich, MO, USA) and F45 (gelatin binding domain – GBD) (Sigma-Aldrich, MO, USA) fibronectin domains were used, and the binding assay was performed essentially as described above. All assays were carried out with two technical and three biological replicates.

To identify statistical differences among the bacterial strains used in the autoaggregation assay, second-degree polynomial regressions adjustments were performed for all replications and subsequent analysis of variance were performed for each parameter. Biofilm data were analyzed through ANOVA followed by Tukey and ECMs binding was analyzed by ANOVA followed by Dunnett’s test. The dose-dependent binding assays were analyzed through multiple t tests. P < 0.05 was adopted to consider the differences observed between the groups as significant.

Initially, we used a dataset generated in a previous study (Hernandes et al., 2020) to investigate the possible occurrence of a set of genes unique to aEPEC of serotype O2:H16 when compared to a global collection of EPEC isolates. This analysis demonstrated that 31 genes were present only in aEPEC strains of this serotype ( Supplementary Table S2 ). Among them we highlighted one predicted to encode an uncharacterized autotransporter protein. This gene was found in 6 (85.7%) of the 7 aEPEC strains of serotype O2:H16 analyzed ( Supplementary Table S2 ) with 100% coverage and identity. Importantly, based on its predicted adhesive properties, this novel autotransporter protein was named in the present study as EPEC Autotransporter Adhesin (Eaa).

Subsequently, to better understand the genetic context in which the gene encoding the autotransporter protein Eaa is inserted in the genome of aEPEC of serotype O2:H16, the strain BA92 was selected. The ONT MinION long read sequence generated a total of 450,101 reads with a coverage of 429.56 times. The hybrid genome assembly (using the MinION long reads combined with the Illumina short reads) showed that the predicted genome size of the aEPEC BA92 was 5,134,725 bp, with 50.22% of GC content and composed of the chromosome and 5 distinct plasmids ( Supplementary Table S3 ).The eaa gene is located in the chromosome of this strain in a genomic region identified as a prophage ( Figure 1 ; Supplementary Table S4 ). Using the strain IAL5132 (an aEPEC strain of serotype O2:H16 that lacks the eaa gene) as reference, we could determine that the eaa-chromosomal containing region present in the aEPEC BA92 strain is 17,014 base pairs in length, organized in 20 coding sequences, including eaa, and inserted upstream to the DNA sequence encoding for the transfer RNA of the amino acid threonine ( Figure 1 ). All genes present in the chromosomal region containing the prophage harboring the eaa gene are listed in Supplementary Table S5 .

The deduced amino acid sequence of the complete Eaa products has 733 amino acids in length with a predicted molecular mass of 79 kDa. The signal peptide of the Eaa protein was identified as the first 21 amino acids ( Figure 2A ). Moreover, four AIDA- repeats, one pertactin domain and one β-barrel translocator domain were also identified ( Figure 2A ). As expected, Eaa is predicted to be a monomeric protein with its passenger and the β-barrel translocator domains located in the N-terminal and C-terminal regions, respectively ( Figures 2B, C ).

Furthermore, a maximum likelihood analysis, comparing the amino acid sequence of some representative autotransporter proteins ( Supplementary Table S1 ) from the four autotransporter families, indicated that Eaa belongs to the AIDA-I family and is closely related to the adhesins TibA and EhaD ( Figure 3 ). Furthermore, an amino acid alignment between Eaa and TibA demonstrated that these two proteins present 41.1% similarity and 26.2% identity ( Supplementary Figure S1 ).

First, the production of the autotransporter protein Eaa was investigated by immunoblotting using the anti-Eaa serum. The SDS-PAGE results showed the production of a 79 kDa protein in the cultures of aEPEC BA92 and MS427(pIC) strains ( Supplementary Figure S2 ), thus indicating the Eaa production by these bacteria.

Subsequently, an immunogold-labelling transmission electron microscopy was performed to verify whether the autotransporter Eaa protein would be able to reach the extracellular environment and remain anchored in the outer membrane. This analysis demonstrated that Eaa was detected on the surface of the aEPEC BA92 and MS427(pIC) strains, but not on the surface of the MS427(pBAD) strain, used as a negative control in this assay ( Figure 4 ).

To investigate the participation of Eaa in the autoaggregation and biofilm formation phenotypes, the MS427 strain (that correspond to the K-12 E. coli MG1655 deleted in the agn43 gene), was used as a host bacteria for three different plasmids, as follows: pBAD/Myc-His A, pIC (that corresponds to the pBAD/Myc-His A vector harboring the eaa gene) and pCO4 (that corresponds to the pBAD/Myc-His A vector harboring the agn43 gene), as described in Table 1 . Of note, the agn43 gene encodes an autotransporter adhesin, termed Antigen 43, whose role in bacterial-bacterial autoaggregation and biofilm formation was already documented in the literature (Kjærgaard et al., 2000a, 2000b). Considering that, MS427(pBAD) and MS427(pCO4) were employed as negative and positive controls, respectively.

In the autoaggregation assay, we observed a progressive decrease of the OD600 at each reading for two of the strains tested: MS427(pIC) and MS427(pCO4); while for the MS427(pBAD) strain, used as a negative control, the OD600 remained constant throughout the assay ( Figure 5 ). In all periods in which the DO600 was measured, the strains MS427(pIC) and MS427(pCO4) differed statistically from the negative control, as well as from each other (P<0.001).

To confirm the hypothesis that OD600 decreased due to bacterial autoaggregation, an aliquot of the bacterial pellet, formed at the bottom of the test tube at the end of the assay, was collected and subjected to immunofluorescence microscopy. While for the MS427(pBAD) strain only isolated bacteria were observed, the MS427(pIC) and MS427(pCO4) strains formed large bacterial aggregates ( Figure 6 ). Taken together, these data demonstrated the ability of Eaa in mediating bacterial autoaggregation.

Biofilm formation assays employing the Eaa-producing strain (MS427(pIC)), in comparison to the negative (MS427(pBAD)) and positive (MS427(pCO4)) controls, were preformed to investigate the involvement of Eaa in biofilm formation. The MS427(pIC) produced significantly more biofilm on polystyrene than the negative control MS427(pBAD) in all periods of times tested (24, 48 and 72 h), similarly to the observed with the positive control MS427(pCO4) (P<0.0001). On the other hand, no statistical difference (P>0.05) was observed between the MS427(pIC) and MS427(pCO4) strains that produce the autotransporter proteins Eaa and Antigen 43, respectively ( Figure 7 ). No increase in biofilm production was evidenced in the MS427(pIC) and MS427(pCO4) when compared to the negative control MS427(pBAD) in assays performed on glass coverslips surface (data not shown).

The ability of the passenger domain of the autotransporter protein Eaa to bind to several ECM and plasma components was investigated. Purified recombinant Eaa (aminoacids 22 - 470) bound to most of the macromolecules tested, i.e. fibrinogen, plasma and cellular fibronectin, laminin and type I, III and V collagens, with especially pronounced binding to collagen V and fibronectin ( Figure 8 ). Adhesion to plasma and cellular fibronectin was further explored on a quantitative basis. A dose-dependent binding to both ECM molecules was observed at increasing concentrations of Eaa ( Figure 9A ). Furthermore, the fibronectin domains responsible for binding were mapped using purified F30 (heparin binding domain, HBD) and F45 (gelatin binding domain, GBD). Eaa dose-dependently interacted with the HBD of fibronectin as shown in Figure 9B .