Escherichia albertii, a novel human enteropathogen, colonizes rat enterocytes and translocates to extra-intestinal sites

Diarrhea is the second leading cause of death of children up to five years old in the developing countries. Among the etiological diarrheal agents are atypical enteropathogenic Escherichia coli (aEPEC), one of the diarrheagenic E. coli pathotypes that affects children and adults, even in developed countries. Currently, genotypic and biochemical approaches have helped to demonstrate that some strains classified as aEPEC are actually E. albertii, a recently recognized human enteropathogen. Studies on particular strains are necessary to explore their virulence potential in order to further understand the underlying mechanisms of E. albertii infections. Here we demonstrated for the first time that infection of fragments of rat intestinal mucosa is a useful tool to study the initial steps of E. albertii colonization. We also observed that an E. albertii strain can translocate from the intestinal lumen to Mesenteric Lymph Nodes and liver in a rat model. Based on our finding of bacterial translocation, we investigated how E. albertii might cross the intestinal epithelium by performing infections of M-like cells in vitro to identify the potential in vivo translocation route. Altogether, our approaches allowed us to draft a general E. albertii infection route from the colonization till the bacterial spreading in vivo.


Ethics statement
The protocols involving animal handling were approved by the Research Ethics Committee of UNIFESP, project license number 0342/09. "Comitê de Ética em Pesquisa da UNIFESP/ Hospital São Paulo" (CEP UNIFESP/HU-HSP) is in accordance with Good Clinical Practice (GCP) of the International Council for Harmonisation (ICH), formerly the International Conference on Harmonisation (ICH). Animals are handled under "Brazilian Guidelines For The Care And Use Of Animals In Educational Activities Or Scientific Research" standards that are in accordance with Brazilian Law 11.794/2008, which defined procedures to be employed in the scientific use of animals.

Bacterial strains
The invasive E. albertii 1551-2 strain (intimin subtype omicron) and its isogenic mutants obtained in previous studies by our group (Table 1) were statically cultured in Luria Bertani broth (LB) for 18 h at 37˚C. Antibiotics were added to select resistant strains as indicated in Table 1. The mutant strain 1551-2Δtir was constructed employing the one-step allelic exchange recombination method [35]. Primers containing a 40-bp region homologous to the 5' and 3' ends of the tir gene and a specific sequence for the zeocin (zeo) resistance-encoding gene (tir-zeo-F ATG CCT 1 ATT GGT AAT CTT GGT CAT AAT CCC AAT GTG AGT GGT CAT CGC TTG CAT TAG AAA GG and tir-zeo-R TTA AAC GAA ACG ATT GGA TCC CGG CAC TGG TGG GTT ATT CGA ATG ATG CAG AGA TGT AAG) were used to amplify the Zeo cassette [36]. Amplicons obtained in the PCR reaction were electroporated into competent wild type bacteria harboring the pKOBEG-Apra plasmid. The selection of recombinant bacteria were done on Zeo-containing LB agar plates (60 μg/mL), and the tir deletion in the isogenic mutant was confirmed by PCR. In addition, the loss of pKOBEG-Apra plasmid was confirmed by testing the mutant strain for apramycin susceptibility.

Adhesion and Invasion assays
Quantitative assessment of bacterial association and invasion was performed as described previously [24,41]. Briefly, differentiated Caco-2 cells were infected with 10 7 colony-forming units (CFU) of E. albertii strain 1551-2 and its isogenic mutants for 6 h. Thereafter, cell monolayers were washed three times with phosphate buffered saline (PBS). While one set of monolayercontaining wells was lysed in 1% Triton X-100 for 30 min at 37˚C, another set was treated with 100 μg/mL of gentamicin (Sigma, USA) for one hour at 37˚C, and then washed 5 times prior to lysis. Following cell lysis, bacteria were resuspended in PBS and quantified by plating serial dilutions onto MacConkey agar plates to obtain the total number of cell-associated bacteria and of intracellular bacteria. The invasion indexes were calculated as the percentage of the total number of cell-associated bacteria that were located in the intracellular compartment. Assays were carried out in triplicate, and the results from at least three independent experiments were expressed as the percentage of invasion (mean ± standard error).

Animals
Female Wistar-EPM rats,~3 months-old and weighting 200-250 g, were obtained from the Central Animal Facility of Universidade Federal de São Paulo (UNIFESP). After 14 days of environment adaptation, stool samples were collected for coproculture and E. coli recovered from each animal were screened for the presence of the eae gene, which encodes the adhesin intimin, by PCR (AE11 5'-CCCGGCACAAGCATAAGCTAA-3' and AE12 5'-ATGACT CATGCCAGCCGCTCA-3', generating a fragment of 917 bp [42]). This procedure was performed to avoid the use of experimental animals that were colonized by either E. coli or Citrobacter rodentium, a murine pathogen that also promotes AE lesion formation [43]. Prior to the assays, animals were fasted for 24 h with access to water.

In vivo organ culture (IVOC) bacterial colonization assay
For removal of ileum fragments, rats were held under anesthesia (pre-atropinization, induction of inhalation anesthesia with isoflurane and maintenance with intramuscular injection of 0.1 mL/100 g body weight ketamine + xylazine (4:1 1). After antisepsis, rats were subjected to median laparotomy for the collection of intestinal fragments of~0.5 cm 2 . Briefly, ileal segments were removed, sectioned longitudinally at its antimesenteric border and placed onto a sterile filter paper with its serous portion facing the filter. This procedure allowed the exposure of the entire apical surface of the mucosa to the bacterial inoculum. Fragments were kept in Dulbecco's Modified Eagle Medium (DMEM-Gibco Invitrogen, USA) supplemented with 10% fetal bovine serum (Gibco Invitrogen, USA) [44]. Fragments were infected with 10 10 CFU for 6 h of incubation (37˚C, 5% CO 2 ); fragments were then washed, macerated and suspended in PBS and plated in serial dilution onto MacConkey agar plates containing 20 μg/mL nalidixic acid [21] for quantification (calculation of the total number of mucosa-associated bacteria). Infected IVOC preparations were also fixed for electron microscopy analysis.

In vivo bacterial translocation assay
Animals were maintained under anesthesia (intramuscular injection of 0.1 mL/100 g body weight of ketamine and xylazine (4:1) during the entire procedure. Additional half dose of anesthetic was administered when necessary. Bacterial translocation (BT) was induced by a midline incision, oroduodenal cannulation, injection of 10 10 CFU/mL resuspended in 10 mL of saline through the catheter, and bacterial retention for a period of 2 h, within a portion between the duodenum and ileum, by means of ligatures [39]. The E. coli rat strain R6, which is devoid of the DEC virulence genes, such as the eae gene, was used as a BT-positive control strain [39], while the non-pathogenic E. coli strain HB101 was used as BT-negative control. Bacterial inoculation causes a transient dilation of the small bowel, which disappeared within a short period. Blood (1 mL), mesenteric lymph nodes (MLN), spleen and liver were then collected, weighed, macerated and suspended in PBS. Subsequently, bacterial colonies were enumerated after plating serial dilutions onto MacConkey agar plates containing 20 μg/mL nalidixic acid, to estimate the number of translocated bacteria. The results were expressed by mean log 10 values of CFU/g tissue.

M-like cell differentiation
M-like cells were obtained as previously described [45][46][47] with modifications. Briefly, Caco-2 cells (10 5 cells/filter) were seeded on the upper chamber of a Millicell filter (3.0-μm pore diameter, Millipore, USA) and kept in DMEM as described above for 10 days at 37˚C in an atmosphere of 5% CO 2 . The lower chamber was also filled with DMEM. During this incubation period, transmembrane electric resistance (TEER) was measured every two days using the Millicell 1 ERS (Electrical Resistance System, Millipore), until it reached~420 mΩ. Afterwards, Raji-B cells (10 6 cells/mL) were seeded at the Millicell lower chamber and cultured in RPMI-1640, as described above, for 6 days. In parallel, in some filters, Caco-2 cells were kept in monoculture for an additional 6 days (non-differentiated cells). Since galectin-9 is expressed on M cells but not on Caco-2 cell surface [47],

In vitro bacterial translocation assay
Bacterial suspensions (10 7 CFU) in DMEM (as described above, except for 1% antibiotics) were inoculated in the upper chamber of filters bearing either M-like/Caco-2 or Caco-2 cells only for 6 h. Filters were transferred to a well containing fresh medium (DMEM without antibiotics) every hour and the medium from the lower chamber was collected for bacterial quantification at 6 h [46]. In parallel, the transmembrane electric resistance (TEER) was measured. At the end of the infection period, monolayers were washed with PBS and fixed for microscopy.

Transmission electron microscopy (TEM)
Infected monolayers and ileum fragments were first fixed in 2% glutaraldehyde (EMS, USA) for at least 24 h at 4˚C. After primary fixation, cells and fragments were washed 3 times with PBS (10 min) and subjected to secondary fixation with 1% osmium tetroxide (EMS, USA) in 0.1 M sodium cacodylate buffer for 30 min. After being washed three times with distilled water, preparations were dehydrated through a graded ethanol series (50%, 75%, 85%, 95% and 100%), and propylene oxide (100%). Preparations were then gradually embedded in Araldite, which was allowed to polymerize for 24-48 h at 60˚C. Ultrathin sections were placed on Formvar (EMS, USA) coated 200 mesh copper grids and stained with 4% aqueous uranyl acetate (Merck, Germany) and Reynold's lead citrate (Merck, Germany). Grids were examined under TEM (LEO 906E-Zeiss, Germany) at 80 kV [48].

Statistical analysis
Differences in bacterial adherence, invasion percentages and translocation or differences in TEER of infected M-like cells were assessed for significance by using an unpaired, two-tailed t test (GraphPad Prism 4.0).

Intimin, Tir and T3SS are essential for invasion of human intestinal cells cultured in vitro
Strain 1551-2 had been previously evaluated regarding its ability to invade differentiated Caco-2 cells [24]. In this study, Caco-2 cells were infected with bacterial suspensions of the wild type or its isogenic mutant strains (Table 1). Compared to the wild type strain the adherence index of mutant strains was not altered (Fig 1A) while the invasion index decreased significantly (Fig 1B), except for fimA mutation that did not affect the adherence or invasion indexes (Fig 1A-1B). These results confirm that E. albertii 1551-2 invasion depends on intimin and/or proteins injected by the T3SS, such as Tir, but not on T1P. Besides that, as the T3SS mutant did not inject Tir into Caco-2 cells, it is possible that, in the absence of its receptor, the 1551-2 intimin might recognize another host cell membrane structure as site for adhesion, but not for invasion, as confirmed by results obtained in invasion assays with 1551-2Δtir strain. Complementation of T3SS mutant restored the invasion index to the wild type values (Fig 1B).

E. albertii 1551-2 colonizes rat enterocytes in in vitro organ culture (IVOC)
To evaluate whether E. albertii 1551-2 could colonize the rat intestinal mucosa, ileal fragments (approx. 0.5 cm 2 ) were individually infected with bacterial suspensions of the wild type or its isogenic mutant strains (Table 1). Methylene blue staining of the intestinal fragments was performed to confirm that all tissue layers were well preserved (S1 Fig). SEM images confirmed that the wild type strain strongly adhered to the intestinal mucosa (Fig 2A), whereas the T3SS-mutant comparatively showed a weaker adherence (Fig 2B). Noninfected fragments showed well-preserved bacterial-free brush borders (Fig 2C). Similarly to the wild type strain, the intimin, Tir and T1P mutants remained adherent to the intestinal mucosa (S2 Fig). Besides bacterial adherence, TEM images showed that the wild type strain  caused AE lesions with characteristic pedestals underneath adhered bacteria on the rat mucosal surface (Fig 2D). In contrast, the T3SS-translocon mutant failed to cause AE lesions ( Fig  2E), and non-infected fragments showed well-preserved bacterial-free brush borders (Fig 2F).
The number of CFU recovered from rat intestinal mucosa in vitro decreased significantly in the absence of the T3SS-translocon, while mutant strains deficient in intimin, Tir or T1P production, as well the T3SS mutant complemented strain, showed similar adherence levels in comparison with the wild type strain (Fig 2G).

E. albertii strain translocates across rat intestinal barrier in vivo
To reduce the number of animals utilized in the next approach, we selected the T3SS-translocon mutant for in vivo comparison with wild type strain based on results obtained with the IVOC infection assay. Our results demonstrated that E. albertii 1551-2 reached the liver, while the T3SS-translocon mutant was not recovered from this organ. These findings suggest that, as a consequence of the reduced adhesion of this mutant to the intestinal mucosa, as observed ex vivo, fewer bacteria were available to cross the intestinal barrier, reach and survive in the MLN (Fig 3).

E. albertii 1551-2 translocates across M-like cells
Considering our results in the BT assay described in Materials and Methods and that pathogens such as Shigella species use M cells to cross the intestinal barrier, we performed E. albertii infection of M-like cells in vitro to identify the potential BT route employed in vivo. Prior to infection, we confirmed the conversion of part of the Caco-2 cells to M-like cells as described elsewhere [47], by demonstrating the expression of galectin-9 on M-like cell surface but not on Caco-2 cells (S3 Fig). Moreover, cellular morphology alterations [45] were observed on M-like cells, such as a reduced number of microvilli, flattened apical surface and disorganized cytoplasm (Fig 4A), while fully differentiated Caco-2 cells displayed preserved brush borders (Fig 4B). The presence of M-like cells significantly increased bacterial translocation (Fig 4C) as compared to differentiated Caco-2 cells (Fig 4D).
For quantitative E. albertii 1551-2 translocation assessment, tEPEC prototype strain E2348/69 was used as control [46]. We demonstrated that E. albertii translocation through Mlike cells was significantly more effective than through differentiated Caco-2 cells (2,962.0 ±546.0 and 184.2±91.6, p = 0.0024, respectively) (Fig 5A), and as previously demonstrated [46], the presence of M-like cells did not increase the transcytosis of tEPEC E2348/69 in a significant manner as compared to differentiated Caco-2 cells (1.203±0.528 and 0.417±0.247, p = 0.1480, respectively). Additionally, E. albertii 1551-2 translocated through M-like cells more effectively than tEPEC E2348/69 (p = 0.033, Fig 5A). In order to exclude bacterial paracellular migration due to increased permeability as an invasion route, transepithelial electrical resistance was measured hourly during the infection period ( S3 Fig). Comparison between Mlike cells infected with the wild type or the T3SS mutant strains demonstrated a significant decrease in bacterial recovery (p = 0.0029, Fig 5B) with the latter strain, while complementation of the mutant strain restored its translocation capacity (p = 0.0418, Fig 5B and S4 Fig). Contrarily, non-significant differences between CFU recovered from T1P mutant and its complemented strains were observed with M-like cells (Fig 5B).

Discussion
Previous data from our laboratory showed that the 1551-2 strain invaded HeLa cells [21] with invasion being dependent on the intimin-Tir interaction, since the intimin mutant (1551-2eae::Kn) was non-invasive [21]. Later on, we demonstrated that, in contrast with the wild type 1551-2 strain that displayed a localized pattern of adherence (formation of compact bacterial clusters) in HeLa cells, its T3SS-mutant adhered weakly, while the intimin mutant adhered, showing a T3SS-dependent diffuse pattern of adherence [36]. In addition, Pacheco et al., 2014 [24] showed that the 1551-2 strain invades, persists and multiplies inside differentiated Caco-2 cells up to 48 h.
In this work, we demonstrated for the first time that intimin, Tir and T3SS are essential for invasion of enterocytes in vitro, since mutations in the corresponding genes abolished bacterial uptake. Bacterial adherence was preserved in mutants, including the T3SS mutant, which did not adhere on HeLa cells in a previous study [36]. This fact might be due to the interaction between either intimin or T1P and Caco-2 cell surface receptors. It has been previously demonstrated that Tir and Map, and EspF can induce tEPEC invasion of HeLa and Caco-2 cells, respectively [49,50].
We have previously shown that an aEPEC strain, 1711-4, is able to translocate across the rat gastrointestinal barrier and be isolated from the MLN, spleen and liver [51]. The mechanisms promoting this bacterial translocation, however, are unknown. Generally, studies on colonization and infection by enteropathogens are conducted with Caco-2 cells, but although this cell line mimicries enterocytes from the human small intestine, it does not represent the complex intestinal mucosa, since it is devoid of the mucosal layer and other intestinal cell types. It was demonstrated that EHEC [52] as well as tEPEC E2348/69 [44] colonize human IVOC. More recently, Etienne-Mesmin et al., [53] demonstrated that EHEC colonize and translocate into ileum fragments from mice, where Peyer's patches are available, but quantification was not performed. In the present study, we evaluated E. albertii capacity to colonize the rat intestinal mucosa in the IVOC model, to mimicry the first steps that lead to bacterial translocation from the intestinal lumen to the extra-intestinal sites demonstrated in vivo. We showed for the first time the interaction of E. albertii with rat intestinal mucosa ex vivo, which could be an alternative model to study AE-producing pathogens' interaction with more complex intestinal tissues. In this model, colonization was detected after 30 min of infection, and invasiveness was revealed after 2 h, when E. albertii 1551-2 could be found inside the enterocytes. Additionally, we demonstrated that E. albertii adherence to the rat IVOC depends on T3SS, as previously demonstrated in human IVOC for tEPEC E2348/69 [44], but not on intimin, Tir or T1P, since in the absence of these genes, bacterial adherence was qualitatively and quantitatively preserved. Thus, the use of this model may optimize the selection of potentially invasive strains to be tested in vivo, thus reducing the number of animals used to assess the fate of invasive E. albertii from the intestinal lumen to extra-intestinal sites.
We selected the T3SS mutant to compare to the wild type strain, since this mutant strain had previously shown a significantly reduced capacity to interact with the host epithelium in an ex vivo model, losing the capacity to invade cultured intestinal cells in vitro.
It has been reported that some E. albertii strains isolated from birds are able to adhere and to invade HEp-2 cells [54] and to reach the liver and spleen of one day-old chicks in vivo, possibly by disrupting the intestinal barrier, despite the minor intestinal mucosa alterations [54]. In this study, using an in vivo bacterial translocation assay in rats, we recovered the E. albertii 1551-2 strain in the MLN and liver but not spleen, while the T3SS mutant completely lost translocation capacity. It has been reported that T3SS-dependent effectors such as EspF, Map and NleA disrupt tight junctions that contribute to the integrity of the intestinal barrier [55][56][57]. In addition, some infectious processes can disturb the intestinal epithelium, for example, neutrophil migration during inflammation; this event promotes a transitory epithelial barrier destabilization, which exposes the basolateral side, either allowing enterocyte invasion [58] or offering an alternative route for bacterial translocation from the intestinal lumen to extraintestinal niches.
Based on our finding that E. albertii 1551-2 can reach the MLN and liver in vivo and that the invasion level through the basolateral surface is higher than at the apical surface of T84 cell monolayers [22], we investigated how E. albertii might cross the intestinal epithelium. It is well known that enteropathogens can reach basolateral receptors and promote enterocyte invasion in vivo by transcytosis through M cells [25,59]. According to Hase and coworkers [28], bacterial translocation depends on T1P-GP2 interaction, since isogenic mutant or non-T1P producer strains were unable to translocate through M-like cells. On the other hand, Inman and Cantey [60] described that a rabbit EPEC strain (RDEC-1) produced AE lesion on the M cell membrane, suggesting that AE lesions could prevent bacterial internalization, thus preventing transcytosis and antigen presentation, thereby delaying the immune response.
In this study, E. albertii 1551-2 translocation was significantly more effective through Mlike cells than Caco-2 cells only. This could not be observed with tEPEC as previously demonstrated by [46]. We also demonstrated that translocation depended on functional T3SS, and that T1P mutation did not compromise bacterial translocation, contrary to what was found by Hase et al., [28]. These differences could be due to allelic FimH alterations in T1P in different strains. Therefore, these data suggest that E. albertii 1551-2 may reach the enterocyte basolateral surface in vivo after M cell translocation. Etienne-Mesmin et al., [53] also found that EHEC O157:H7 and O113:H2 and their respective intimin and Shiga toxin mutants translocated more effectively through M-like cells in comparison with Caco-2 cells. Cieza et al., [61] reported that the translocation of adherent-invasive E. coli (AIEC) through M-like cells depends on IbeA (an invasin); however, E. albertii strain 1551-2 is devoid of the ibeA gene (not shown) and other invasion-related genes [62], reinforcing that the bacterial translocation ability of this E. albertii strain is due to intimin-Tir interaction.
Altogether, our results demonstrated for the first time that both ex vivo and in vivo bacterial infection of rat intestinal mucosa are useful models to study E. albertii interaction with the host. We also showed that E. albertii 1551-2 may also cross the intestinal mucosa in vivo possibly using M cells as a route to reach extra-intestinal organs.