Bacterial strains and plasmids used in this study are described in Supplementary Tables S1 , S2 . Strain 17–2 is a clinical isolate of EAEC. Cloning was performed using E. coli DH10b as the host strain. The deletion mutants, 17-2 ΔacrA and 17-2 ΔacrB, were obtained using the one-step method of gene inactivation (Datsenko and Wanner, 2000). Specifically, the kanamycin resistance gene was amplified from the pKD4 or pKD13 plasmids using the oligo pairs AFW/ARV and BFW/BRV for acrA and acrB deletion, respectively. The resulting amplicons were electroporated into the 17–2 pKD46 strain, which expresses the Red recombinase. The obtained mutants were then deprived of Km resistance by transformation with the pCP20 plasmid, which encodes a flippase that facilitates flippase/flippase recognition target (Flp/FRT) recombination (Datsenko and Wanner, 2000). The pEAacrB plasmid was constructed by cloning the acrB gene under the control of the tac promoter from the pGIP7 plasmid, using the oligo pair pEAacrBFW/pEAacrBRV. Plasmid pEAacrB served as the template for site-directed mutagenesis. As previously described (Fanelli et al., 2023), GeneArt^™ Site-Directed Mutagenesis System (GeneArt^™ Invitrogen-Thermo Fisher Scientific, Waltham, MA, USA), was employed along with the oligo pair pEAacrB _D408AFW/pEAacrB _D408ARV to introduce a codon change from GAC to GCC (D408A), replacing the A at position 1223 (relative to the ATG of acrB gene) with a C.

All bacterial strains were grown aerobically in Luria-Bertani (LB) (Sigma-Aldrich, St. Louis, MO, USA) medium at 37°C with shaking at 200 rpm, unless otherwise specified. Solid media contained 1.6% agar. Antibiotics (Sigma-Aldrich, St. Louis, MO, USA) and inhibitors were used at the following concentrations: kanamycin at 30 μg/mL, chloramphenicol at 25 μg/mL, ampicillin at 50 μg/mL and 1-(1-naphthyl-methyl)-piperazine (NMP) at 100 μg/mL. Growth kinetics of 17–2 and its derivatives in different media (LB and DMEM) were measured using a CLARIOstar plate reader (BMG LABTECH, Offenburg, Germany).

The mutations and constructs obtained in this study were confirmed using PCR reactions, routinely performed with DreamTaq DNA polymerase (Thermo Fisher Scientific, Waltham, MA, USA), and verified by DNA sequencing (Biofab, Rome, Italy). For amplifying coding sequences necessary for cloning, Pfu Taq DNA polymerase (Thermo Fisher Scientific, Waltham, MA, USA) was used. The sequences of the oligonucleotides, designed based on the 17–2 genome, are listed in Supplementary Table S3 . DNA and protein sequence comparisons were conducted using the NCBI BLAST tool (http://blast.ncbi.nlm.nih.gov/Blast.cgi).

Minimum inhibitory concentrations (MICs) were determined using a broth microdilution method in 96-well microplates (Euroclone S.P.A., Milan, Italy – 96 well flat bottom). EAEC 17–2 wt strain and its derivatives 17-2 ΔacrA, 17-2 ΔacrB, 17-2 ΔacrB_pEAacrB and 17-2 ΔacrB_pEAacrB _D408A were grown in LB until an OD₆₀₀ of 0.5-0.7, corresponding to the exponential phase of growth. Each bacterial suspension was then diluted to OD₆₀₀ 0.02 in LB and 100 µL aliquots were inoculated in each micro-well into 100 μL of 2-fold serial dilutions of each antibiotic (erythromycin 0.5 µg/mL to 256 µg/mL, tetracycline 0.0313 µg/mL to 16 µg/mL). 1-(1-naphthyl-methyl)-piperazine (NMP) (Sigma-Aldrich, St. Louis, MO, USA) was dissolved in DMSO and tested from 1.25 µg/mL to 1280 µg/mL in BHI (Becton Dickinson, USA) medium ( Supplementary Figure S3 ). Bile salts (B8756, Sigma-Aldrich, St. Louis, MO, USA), consisting of an approximate 1:1 mixture of cholate and deoxycholate, were used at 0.16% (w/v). After an incubation of 18 h at 37°C, results were collected by measuring OD at 600 nm using a CLARIOstar plate reader (BMG LABTECH, Offenburg, Germany). The MIC value considered was the lowest drug concentration that exhibited complete inhibition of microbial growth. A 2-fold or greater decrease in MIC values was considered significant. The results are averages of at least three independent experiments.

Biofilm formation was assessed in E. coli strains using crystal violet (CV) staining, CFU quantification, and extracellular DNA (eDNA) measurement, as previously described (Sivori et al., 2024). 100 µL of bacterial suspension (10⁵ CFU/mL) were cultured in 96-well polystyrene plates at 37°C under static conditions for 24 h. Biofilm biomass was quantified by staining with 0.1% CV or 0.01% CV when combined with NMP, ethanol-acetone (4:1) solubilization, and absorbance measurement at 570 nm (Multiskan SkyHigh, Thermo Fisher Scientific, USA). Viable cell counts were determined by plating serial dilutions on MacConkey agar (BioMérieux, France). After biofilm formation, wells were washed, resuspended in sterile water, and scraped for CFU enumeration. eDNA quantification was performed by adding Tris-EDTA buffer (Sigma-Aldrich, St. Louis, MO, USA) and freshly prepared PicoGreen solution (Thermo Fisher Scientific, Waltham, MA, USA), followed by fluorescence measurement (excitation: 485 nm; emission: 535 nm; Wallace Victor 3, PerkinElmer). A lambda DNA (Thermo Fisher Scientific, Waltham, MA, USA) standard curve was used for quantification. Experiments were conducted in triplicate and repeated three times (Sivori et al., 2022).

For biofilm metabolic activity, bacterial suspensions were allowed to adhere for 5 h before removal of non-adherent cells. Resazurin solution (Promega, USA) was then added, and plates were incubated for an additional 20 h. Absorbance was measured at 570 nm using the Multiskan SkyHigh reader, with absorbance readings taken every 20 min (Sivori et al., 2022).

To characterize the protein component of the biofilm matrix, biofilms formed in BHI under static conditions for 20 h were stained with 100 µL of FilmTracer SYPRO Ruby (Invitrogen, Thermo Fisher Scientific, Waltham, MA, USA) for 30 min at room temperature. After two washes with sterile distilled water, fluorescence intensity was measured (excitation: 485 nm; emission: 625 nm; Wallace Victor 3, PerkinElmer) (Ravaioli et al., 2020). Amyloid fiber content was quantified using EbbaBiolight (Ebba Biotech AB, Sweden). 200 µL of bacterial cultures were incubated under static conditions at 37°C in BHI supplemented with 1,000-fold diluted dye. After 24 h, fluorescence was measured (excitation: 535 nm; emission: 660 nm) using the Wallace Victor 3 plate reader (Choong et al., 2016).

The HEp-2 cell line (CCL-23, American Type Culture Collection, Manassas, VA, USA) was grown in Dulbecco’s Modified Eagle Medium (DMEM) (Gibco, Thermo Fisher Scientific, Waltham, MA, USA) supplemented with 10% heat-inactivated fetal bovine serum (FBS) (Gibco, Thermo Fisher Scientific, Waltham, MA, USA), 0.05 I.U./mL penicillin and 0.05 I.U./mL streptomycin (Sigma-Aldrich, St. Louis, MO, USA). Cells were maintained at 37°C in a humidified atmosphere with 5% CO₂.

The adherence assay protocol was adapted from that of Cravioto et al., 1979. Briefly, HEp-2 cells at a concentration of 2 x 10⁵ cells/mL were seeded into 30 x 10 mm sterile plastic tissue culture Petri dishes (Nunc) into which a sterile rectangular coverslip had been placed. Cells were then incubated overnight and grown to 60% confluency. Two hours before bacterial infection, the culture medium was replaced with fresh DMEM. At the time of infection, HEp-2 cell monolayers were washed three times with 1X PBS and then overlaid with exponential cultures of EAEC strains diluted in DMEM containing 1% D-mannose, which was added to inhibit adhesion due to type I pili. This resulted in a final multiplicity of infection of 100. The infected cells were incubated for 24 h, after which they were carefully washed twice with 1X PBS and fixed with 4% paraformaldehyde for 20 min. Following fixation, the samples were washed three times with 1X PBS and permeabilized in 0.5% Triton X-100 for 10 min, then washed twice again. The samples were stained with 4’,6-diamino-2-phenylindole (DAPI, 0.5 µg/mL) for 5 min and washed twice. Finally, samples were stained with Evans Blue (0.01%) for 30 sec and then carefully washed with 1X PBS. Coverslips were mounted using 10 µL of a 1:1 PBS-glycerol solution. Images were acquired at the IBPM imaging platform https://www.imagingplatformibpmcnr.it with a Nikon Ti2 confocal spinning disk microscope (implemented with Crest X-Light V3 module from CrestOptics) equipped with the Kinetix sCMOS camera (Teledyne Photometrics), a 60× (immersion oil, NA 1.4) objective and CELESTA lasers (Lumencor). Image acquisition and analyses were performed using NIS-Elements AR software modules (Nikon). 2D image projections from z-stacks were created using the Maximum Intensity Projection (MIP) function. 3D reconstructions were created using the 3D Volume View function.

For quantitative adherence assays, after infection and 24 h incubation, HEp-2 monolayers were washed three times with 1X PBS. Each well was treated with 1% Triton X-100 for 20 min and serial dilutions of lysed sample were plated out for viable counting on LB, with antibiotics added where necessary. At least three independent assays were performed.

For C. elegans infection, bacterial lawns were prepared by spreading 30 μL of each overnight bacterial culture, corresponding to 1 × 10⁸ cells, onto Nematode Growth Medium (NGM) agar plates (35 mm). The plates were incubated at 37°C for 24 h to allow bacterial growth before being seeded with young adult nematodes grown at 16°C from a synchronized culture (Schifano et al., 2019). Plates seeded with the standard food bacterium E. coli OP50 served as controls. Each day, 80 worms per condition were transferred to new plates, plated with fresh bacterial cultures, and monitored under a stereomicroscope (Olympus SZ30, Olympus Optical Co. Ltd., Tokyo, Japan). The infections were performed at 25°C and the number of dead worms was recorded daily. A worm was considered dead if it failed to respond to touch with a platinum wire; dead worms stuck to the wall of the plate were excluded from the analysis. Mortality data from duplicate plates were pooled and subjected to survival analysis. The results shown are representative of repeated independent assays.

For intestinal lumen diameter analysis in C. elegans, 10 synchronized worms per condition, after 72 h of infection, were washed in M9 buffer, mounted onto 3% agarose pads containing 20 mM sodium azide and observed with a Zeiss Axiovert 25 microscope. Intestinal lumen diameter was measured by using Zeiss ZEN Microscopy Software 2011.

The statistical differences were assessed using GraphPad Prism 9.0, which calculated p-values using a one-way ANOVA or, when appropriate, a two-tailed student’s t-test. For C. elegans tests, survival data were analyzed using the Kaplan-Meier method. Survival curves were generated using GraphPad Prism 9.0, and statistical differences between groups were assessed using the log-rank (Mantel-Cox) test. Data from duplicate plates were pooled for analysis, and results are representative of at least three independent experiments. Statistical significance was set at p < 0.05.

Comparative in silico analysis revealed the presence of genetic loci encoding the 20 most relevant EPs associated with multidrug resistance in the EAEC 17–2 genome. The E. coli MG1655 genome was used as a reference (Nishino and Yamaguchi, 2001; Teelucksingh et al., 2022). Unlike other genera, such as Shigella, which underwent a strong gene decay in the evolutionary process from commensal to pathogen (Feng et al., 2011), we found that EAEC 17–2 retained the genes coding for all MDR EPs, including AcrAB. This analysis also highlighted the conservation of a high degree of sequence homology ( Table 1 ).

A set of EAEC 17–2 derivative mutants was generated to study the role of AcrAB in the physiology and infection process of EAEC. These include deletion mutants lacking the periplasmic adaptor protein AcrA or the inner membrane transporter AcrB ( Figure 1A ), as well as complemented strains such as the ΔacrB_pEAacrB, expressing the wt transporter, or the ΔacrB_pEAacrB _D408A, expressing an AcrB protein harbouring the aminoacidic substitution (D408A) that abolishes the efflux activity of the protein ( Supplementary Tables S1 , S2 ) (Wang-Kan et al., 2017). Since AcrB is a very abundant inner membrane protein, the use of this latter 17–2 derivative allowed us to analyse the AcrB efflux function without altering the membrane protein composition (Wang-Kan et al., 2017).

Growth behaviour of EAEC 17–2 wt strain and its derivatives was assessed in different culture conditions. As shown in Supplementary Figure S1 , the lack of either one of the acrAB genes does not affect the growth of 17–2 in LB or DMEM medium. In contrast, when growth was assessed in the presence of bile salts (cholate and deoxycholate) at physiological concentration (0.16% w/v) ( Figure 1B ), measurement of optical density over time showed that both the deletion of acrA and acrB significantly affected the resistance of 17–2 to bile salts, flattening its growth curve. This phenotype is fully restored in the 17-2 ΔacrB mutant transformed with a functional acrB gene. Complementation of the ΔacrB mutant with the plasmid containing the acrB _D408A allele produced a phenotype comparable to that of the ΔacrB mutant, confirming that the efflux activity and not the simple absence of an abundant inner membrane transporter is responsible for the hypersusceptibility to bile salts. Furthermore, in the 17-2 ΔacrB_pEAacrB _D408A strain, the growth curve appeared even more flattened than in 17–2 strains with ΔacrA and ΔacrB mutations. Considering the infidelity of the components forming the tripartite efflux pumps (Yamasaki et al., 2011), we can speculate that the AcrA protein produced in 17-2 ΔacrB_pEAacrB _D408A is sequestered by the mutated AcrB transporter to form a non-functional pump, thus preventing it from acting as a periplasmic adaptor for other efflux pumps that may provide partial resistance to bile salts (as in 17-2 ΔacrA and 17-2 ΔacrB). As expected, considering the role of AcrAB in MDR, we observed that the functionality of this MDR EP is relevant for resistance to erythromycin and tetracycline, two known substrates of AcrAB ( Figure 1C ).

Since the ability of EAEC 17–2 to form thick biofilm layers on the intestinal epithelium is a crucial part of its pathogenicity process (Estrada-Garcia et al., 2014), we evaluated the influence of AcrAB on various parameters characterizing bacterial biofilms.

An initial assessment of biofilm production was performed by counting viable cells ( Figures 2A, B ). The loss of the AcrB transporter caused a significant reduction in the biofilm component compared to the wt strain ( Figure 2A ), despite the total number of bacteria remaining unchanged after 24 h ( Figure 2B ). Consistent with these results, biomass production measured by the crystal violet assay was significantly lower in the ΔacrB mutant than in the wt strain ( Figure 2C ). Interestingly, ΔacrB biofilms contain less extracellular DNA (eDNA), a key structural component of the biofilm matrix ( Figure 2D ). The composition of the biofilm in terms of protein and amyloid fibres is comparable between wt and defective strains, as well as the metabolic activity, which is not affected at all by the lack of efflux via AcrAB ( Supplementary Figure S2 ). Silencing acrA had no significant impact on the biofilm parameters analysed, except for eDNA export, reinforcing the hypothesis that this adaptor protein can be easily replaced in the tripartite AcrAB-TolC pump assembly.

To confirm that the role of AcrB in biofilm production is closely linked to the extrusion mediated by this transporter protein, we used 1-(1-naphthyl-methyl)-piperazine (NMP), an AcrB efflux pump inhibitor that acts by mimicking the action of an EP substrate, thus preventing normal substrate extrusion (Bohnert and Kern, 2005). NMP treatment was carried out for 24 hours at 100 µg/mL, a concentration that does not affect the growth of EAEC 17-2 ( Supplementary Figure S3 ). Biomass measurement revealed that the inhibition of efflux via AcrB by NMP caused a drastic reduction of the biofilm formation ( Figure 2E ). Consistent with the results obtained with the ΔacrB mutant ( Figures 2A-D ), this strongly indicates that the efflux activity associated with AcrB is required to form a thick biofilm layer.

EAECs are recognized for their characteristic aggregative adherence pattern on HEp-2 cells. This pattern depends on both the self-agglutination of the bacterial cells with each other and their interaction with the epithelial cells (Nataro, 2005; Croxen and Finlay, 2010; Estrada-Garcia et al., 2014). To test whether AcrAB contributed to this phenotype, we performed cell adherence assays on the human epithelial cell line HEp-2. After 24 hours of infection with the 17–2 wt strain, its derivatives ΔacrA and ΔacrB, and the complemented strains ΔacrB_pEAacrB and ΔacrB_pEAacrB _D408A, samples were fixed and stained with DAPI to label bacterial and cellular DNA, and Evans Blue to stain the cell cytoplasm.

Confocal microscopy images presented in Figure 3A show 3D reconstructions of HEp-2 cells infected with strains 17–2 wt and ΔacrB. The wt strain exhibits the “stacked brick” adhesion pattern, with the bacterial cells massively adhering to each other as well as to the cell surface (left panel). Interestingly, the mutant lacking AcrB exhibits a completely different pattern (right panel). Indeed, compared to the wt strain, it appears incapable of forming large aggregates with each other and on the epithelial cell, with an increased dispersion in small clusters over the entire surface of the slide, clearly appreciable in the corresponding 2D panel shown in Figure 3B . The aggregative pattern is correctly restored in the complemented strain with a functional AcrB transporter ( Figure 3B , lower left panel). Conversely, the expression of the AcrB D408A mutated form was unable to functionally complement the defective adherence phenotype shown by ΔacrB ( Figure 3B , lower mid panel). In agreement with previous results, the strain lacking AcrA exhibits an adhesion phenotype similar to that of wt ( Figure 3B , lower right panel), indicating that this MDR EP may also function in the absence of this component, perhaps utilising other periplasmic adaptors.

Following 24 hours of infection, adherent bacteria were also quantified by viable counts ( Figure 3C ). The results demonstrated that the ΔacrB and ΔacrB_pEAacrB _D408A strains are drastically incapable of adherence, suggesting that a functional AcrB transporter is essential for the aggregative adhesion of EAEC.

Using the nematode C. elegans as a model to quantitatively evaluate differences in EAEC virulence (Hwang et al., 2010), we assessed the role of the AcrAB efflux pump in the virulence of EAEC 17–2 in vivo. Specifically, we fed a synchronous population of worms with either the wild-type strain 17–2 or its derivatives, compared to nematodes fed with a commensal E. coli strain (OP50), the standard diet. The number of dead worms was then recorded daily. In line with previous observations (Hwang et al., 2010; Bojer et al., 2012), we found that 17–2 wt is significantly more lethal to worms than OP50 ( Figure 4A ). However, the loss of AcrB transporter caused a significant reduction in pathogenicity for the nematode (p = 0.001), with a median lethal time (LT50) of 5 days for 17-2 ΔacrB compared to 3 days for 17–2 wt. In contrast, the 17-2 ΔacrA mutant appeared just as lethal as the wt strain. Notably, the 17–2 strain harbouring an efflux-defective AcrB protein (AcrB D408A) is even less virulent than the ΔacrB strain, with an LT50 similar to that of animals fed with OP50, likely due to its ability to sequester AcrA, making it unavailable for other transporters ( Figure 4B ).

The relevance of AcrB in the infection efficiency of C. elegans is evident from the evaluation of the worm gut colonization after three days of feeding with 17–2 wt, ΔacrA or ΔacrB strains ( Figure 4C ), by measuring the enlargement of the intestinal lumen, a classic marker of nematode stress due to infection by pathogenic bacteria. Indeed, the right panel of Figure 4C shows that the lumen diameter ranges between 25 and 35 µm when the worm is fed with the wt strain or its ΔacrA derivative. On the contrary, the lumen of the nematode fed with the ΔacrB strain appears quite comparable to that colonised by the non-pathogenic E. coli OP50, with a diameter of less than 10 µm. Altogether, these data confirm the previous observations on the role of AcrAB and highlight the relevance of the AcrB transporter in the pathogenicity in vivo.