Coordinated transient interaction of ZO‐1 and afadin is required for pedestal maturation induced by EspF from enteropathogenic Escherichia coli

Abstract Enteropathogenic Escherichia coli (EPEC) infection causes a histopathological lesion including recruitment of F‐actin beneath the attached bacteria and formation of actin‐rich pedestal‐like structures. Another important target of EPEC is the tight junction (TJ), and EspF induces displacement of TJ proteins and increased intestinal permeability. Previously, we determined that an EPEC strain lacking EspF did not cause TJ disruption; meanwhile, pedestals were located on the TJ and smaller than those induced by the wild‐type strain. Therefore, EspF could be playing an important role in both phenotypes. Here, using different cell models, we found that EspF was essential for pedestal maturation through ZO‐1 disassembly from TJ, leading to (a) ZO‐1 recruitment to the pedestal structure; no other main TJ proteins were required. Recruited ZO‐1 allowed the afadin recruitment. (b) Afadin recruitment caused an afadin–ZO‐1 transient interaction, like during TJ formation. (c) Afadin and ZO‐1 were segregated to the tip and the stem of pedestal, respectively, causing pedestal maturation. Initiation of these three discrete phases for pedestal maturation functionally and physically required EspF expression. Pedestal maturation process could help coordinate the epithelial actomyosin function by maintaining the actin‐rich column composing the pedestal structure and could be important in the dynamics of the pedestal movement on epithelial cells.

Besides Tir, other translocated effectors common to EPEC and related A/E pathogens are also encoded within LEE and injected into the host cell, including Map, EspH, EspF, EspG, and EspZ (Kanack, Crawford, Tatsuno, Karmali, & Kaper, 2005;Matsuzawa, Kuwae, & Abe, 2005;McNamara et al., 2001;Tu, Nisan, Yona, Hanski, & Rosenshine, 2003), as well as less conserved non-LEE effectors (Dahan et al., 2005;Li et al., 2006;Mundy et al., 2004;Tobe et al., 2006). These LEE or non-LEE effectors interfere with diverse cell functions. It is worthy to mention that EPEC 2343/69 does not harbor the E. coli secreted protein F in prophage U (EspFU) also termed TccP. EspFU is encoded in the O157 island, in contrast to LEE-encoded EspF (Campellone, Robbins, & Leong, 2004). Moreover, EspFU from canonical EHEC strains is 25% identical to EspF. EspFU displays a unique function because deletion of espFU impairs EHEC pedestal formation, whereas deletion of espF does not (Campellone et al., 2004;Garmendia et al., 2004), thus implying that these proteins have evolved for distinct cellular functions. Thus, unlike EspF, EspFU is recruited to the pedestal and is associated indirectly with Tir, since Tir from canonical EHEC strains (O157:H7) does not have the residue Y474 (Campellone et al., 2004). On the other hand, EspF is clearly involved with another important target of EPEC, the tight junction (TJ) complex, which leads to the displacement of several TJ proteins and increased permeability through the intestinal epithelium (Dean & Kenny, 2009). Besides the disruption of the epithelial barrier, EspF has been localized in multiple cellular compartments (including cytoplasm, mitochondria, nucleolus, and apical and lateral membranes) and interacts with at least 12 reported host proteins.
We have shown that EspF from EPEC E2348/69 has three almost identical proline-rich sequences, which can be recognized by class I SH3 domains, and three class III PDZ domain binding motifs (Peralta-Ramirez et al., 2008). In eukaryotic cells, these motifs are relevant for protein-protein interaction, that is, actin regulator proteins containing SH3 domains, and motifs interacting with PDZ domains present in scaffolding factors that recruit signaling molecules to cell junctions, including the zonula occludens-1 (ZO-1), ZO-2, and ZO-3 junctional proteins (Peralta-Ramirez et al., 2008). Thus, these EspF proline-rich motifs and PDZ domain binding motifs might be related to actin rearrangement and TJ disruption. In agreement with these in silico predictions, we also showed that after 2 hr of infection, EspF bound to the N-WASP and Arp2/3, as well as ZO-1 and ZO-2 proteins (Peralta-Ramirez et al., 2008). In fact, it has been shown that N-WASP regulates the apical junction complex homeostasis and that EspF exploits both N-WASP and SNX9 to disrupt intestinal barrier integrity during infection (Garber et al. 2017).
The actin cytoskeleton and the scaffold proteins are key for tight junctions' integrity. TJs are mainly composed of transmembrane proteins such as occludin, claudins, JAMs, and tricellulin, which are associated with the cytoplasmic plaque formed by ZO-1/2/3, connecting tight junction to the actin cytoskeleton, and cingulin and paracingulin connecting TJ to the microtubule network (Ugalde-Silva, Gonzalez-Lugo, & Navarro-Garcia, 2016). ZO-1 regulates the permeability through the modulation of the actin cytoskeleton (Van Itallie, Fanning, Bridges, & Anderson, 2009;Zihni, Mills, Matter, & Balda, 2016). F-actin is required for formation and maintenance of TJs and adherens junctions (AJs), and afadin, an F-actin binding protein localized at the AJs, regulates the formation of AJs and TJs. During the formation of AJs, afadin-nectin first recruits JAMs and then occludin and claudin through the interaction of afadin with ZO-1 for the formation of TJs (Sakakibara, Maruo, Miyata, Mizutani, & Takai, 2018).
In this context, we have found two interesting phenomena: An EPEC strain lacking EspF did not cause TJ disruption and recruitment of TJ proteins into the pedestal, and pedestals were smaller than those induced by the wild-type strain; the latter were located mainly on the TJ (Peralta-Ramirez et al., 2008). Thus, EspF interaction with host proteins induces the recruitment of junctional proteins into the pedestals, leading to the maturation of actin pedestals and paracellular permeability disruption. We speculated that the pedestal maturation could be important not only for the initial colonization but also for bacterial spreading by influencing the dynamics of the pedestal movement along and between epithelial cells. In order to understand how EspF could be influencing the pedestal maturation and its relationship with TJ proteins, we infected different cell lines as well as known polarized cells under the calcium switch assay and performed confocal microscopy, co-immunoprecipitation, and knockdown assays for understanding the role of the tight junction proteins and to decipher the mechanism involved in pedestal maturation induced by EspF-producing EPEC during epithelial cell infection.

| Disassembly of cell junction proteins is related to pedestal growth induced by EPEC
We previously found that EspF is involved in intercellular junction disassembly and ZO-1 recruitment to pedestals formed by REPEC E22 (Peralta-Ramirez et al., 2008). Thereby, we decided to investigate in detail the role of TJ proteins in the increase of the pedestal size. We used the model of calcium switch using MDCK cells.
In this model, at normal calcium concentration MDCK cells formed mature monolayers with assembled TJs, which were clearly detected using anti-ZO-1 antibodies. These antibodies decorated a continuous pattern of ZO-1 distribution along the cell periphery marking the cell-cell junction with the classical chicken wire staining ( Figure   A1a). But when calcium was removed from the cell medium, the TJs were disassembled and the junctional proteins were dispersed in the cytoplasm as detected by ZO-1 immunolabeling ( Figure A2b). Once calcium was added back to the medium, the cells initiated the formation of TJs ( Figure A1c) once again, as before calcium removal (see Figure A1a). Monolayers under these three modalities were also observed in z-slices, where clearly ZO-1 was detected at the TJ position in normal-calcium ( Figure A1a) and calcium-recovery ( Figure A1c) cells, but not in low-calcium cells, where ZO-1 was found dispersed in the cytoplasm ( Figure A1b Additionally, MDCK cells infected with any of the two strains under normal calcium concentration exhibited pedestals with slight recruited ZO-1 labeling, in contrast to epithelial cells of low transepithelial electrical resistance (TER) or without TER, such as Caco/B7 or HeLa cells (Hanajima-Ozawa et al., 2007). Interestingly, at normal calcium concentration, the pedestals formed by EPEC were significantly bigger (0.49 µm) than those formed by EPECΔespF (0.35 µm) ( Figure 1q). On the other hand, at low calcium concentration, the intercellular junctions of MDCK cells were disassembled and ZO-1 labeling was mainly detected in the cytoplasm. Infection with wildtype EPEC (Figure 1e-h) or EPECΔespF (Figure 1m-p) under low calcium concentration induced bigger pedestals than those in cells grown in medium with a normal calcium concentration, and the ZO-1 signal was also detected inside the pedestals (Figure 1h, p).
However, while labeling for both ZO-1 and F-actin coincided inside the pedestals of cells infected by wild-type EPEC (Figure 1h), this pattern occurred in less extension in cells infected by EPECΔespF ( Figure 1p) and the pedestals were formed mainly on the cell periphery ( Figure 1n vs. f). Nevertheless, disassembly of junctional proteins due to low calcium concentration caused that pedestals induced by either wild-type EPEC or EPECΔespF were of the same size, around 4 times higher (EPEC 2.09 µm and EPECΔespF 2.02 µm) than those formed in normal-calcium conditions (Figure 1q). Measurement of pedestals in Figure 1q was facilitated by using optical sections and confocal software; pedestals were easily detected as those shown in Figure 1r and Figure 1s. ZO-1 labeling in MDCK cells with restored normal calcium concentration was detected again in the continuous pattern along the restored intercellular junctions, being interrupted only where the bacteria were forming the pedestals (data not shown). Interestingly, pedestals formed by any of the strains in the condition of calcium recovery were reduced in size (0.5313 µm for EPEC and 0.4899 µm for EPECΔespF), and they were not statistically different ( Figure 1q). Taken together, these data indicate that EspF presence during the infection induces ZO-1 disassembly from TJs.
Furthermore, ZO-1 availability in the cytoplasm due to TJ disassembly (induced by EspF or by calcium switch) is important to increase the pedestal size.

| ZO-1 is necessary and enough to increase the pedestal size
To understand the role of the different TJ proteins, such as claudin, occludin, or tricellulin, we used the fibroblast line, L cells, which do not form intercellular junctions. L cells do not contain any of the intercellular junction proteins mentioned above but ZO-1 and can be used to transfect them for expressing other junctional proteins.
First, we determined ZO-1 distribution in noninfected semiconfluent L cells (80%). At this confluency (Figure 2a′-c′), ZO-1 was found mainly in the cytoplasm, but interestingly, accumulation of immunolabeled ZO-1 puncta was detected in some regions of cell-cell contacts (Figure 2a-c). Then, we infected L cells at this confluency with EPEC. Intriguingly, EPEC was able to form pedestals to which ZO-1 was recruited, despite the lack of expression of important TJ proteins in these cells, such as claudin and occludin. Interestingly, two different types of pedestal populations were observed: a minority of small pedestals formed at the cell surface and a majority of large pedestals. The latter were formed in the regions of cell-cell contacts, where a higher number of ZO-1 puncta accumulation was present (Figure 2d-f).
Since the cell-cell contact regions, enriched in ZO-1, promoted the pedestal growth, we decided to grow L cells at a confluency that would avoid the formation of cell-cell contacts. L cells were grown at 40% confluency (Figure 2g′-i′), and ZO-1 was immunodetected in EPEC-infected cells. At this confluency (subconfluent L cells), cells did not form contacts and ZO-1 puncta were not detected. In con- however, these pedestals were smaller than cell-cell contact pedestals observed for semiconfluent cells (Figure 2m). These data indicate that ZO-1 availability (disassembled from TJs) in the cytoplasm is important for inducing the growth of pedestals. On the other hand, since L cells do not express relevant junctional proteins such as claudin, occludin, ZO-2, and ZO-3, our data suggest that no other relevant proteins from the TJs might be required for this process.
These data also suggest that other proteins in the cell-cell contact that recruit ZO-1 to these sites could also be involved in the pedestal growth.

| EspF is required for sequential recruiting of ZO-1 and afadin to pedestals induced by EPEC
Afadin is a regulator of intercellular junction assembly in epithelial cells. Through its interaction with the TJ protein ZO-1 and adherent junction protein α-catenin, afadin regulates the assembly of TJs and AJs (Birukova et al., 2012). To determine the role of afadin, as a pos- These evidences support our previous observations, which revealed that infection with EPEC that lacked espF led to the formation of smaller pedestals than those in wild-type EPEC infections. Notably, F I G U R E 1 Disassembly of cell-cell junctions by using the calcium switch model or by EspF promotes the growth of pedestals. MDCK cells were infected with EPEC wild type (a-d) or EPECΔespF (i-l) at a MOI of 0.5 for 4 hr (normal calcium concentration). Two other groups of cells were infected: in cells kept in DMEM containing low concentration of calcium (TJ disassembly condition) for 2 hr, those with EPEC (e-h) or EPECΔespF (m-p), and in cells returned to normal concentration of calcium for 2 hr after incubation at low calcium concentration (calcium recovery) (see panel q). Cells were fixed, permeabilized, and stained with DAPI (for bacterial and nuclear DNA) and FAS (rhodaminephalloidin for F-actin). The stained cells were immunostained with a rabbit anti-ZO-1 polyclonal antibody followed by a secondary antibody, FITC-goat anti-rabbit IgG. Slides were analyzed and recorded by confocal microscopy (63X zoom 3). Bar: 20 µm. From each panel, sections of 0.5 μm were used to measure individual pedestals, 150 from three independent experiments (i.e., R and S are sections from panels d and h). (q) Pedestals were measured (µm) using the Leica Lite software, and data were plotted and analyzed using GraphPad Prism 6.0. All data from the different strains were compared with EPECΔespF using a one-way ANOVA test, n = 3 independent experiments. **p < .005, ****p < .0001 at 3 and 4 hr of infection, the afadin signal remained undetectable in the pedestal structures ( Figure A3i, m) while the recruited ZO-1 signal decorated the small actin pedestals ( Figure A3j-l and n-p) but with less signal intensity than that observed in wild-type EPEC-induced pedestals ( Figure A3j, n). Interestingly, there was a complete absence of colocalization between afadin and ZO-1 in the pedestals F I G U R E 2 ZO-1 is necessary and sufficient to cause an increase of pedestal size by EspF-expressing EPEC, without other main junctional proteins. L cells were grown at 80% (a-f) and 40% (g-l) of confluence. L cells without infection were used as mock cells at 80% (a-c, and at low magnification, a′-c′) and 40% (g-i, and at low magnification, g′-i′) of confluence. Cells at 80% (d-f) and 40% (j-l) of confluence were infected with EPEC wild type at a MOI of 0.5 for 4 hr. Cells were fixed, permeabilized, and stained with DAPI (blue) and FAS (red) as mentioned in Figure 1. The stained cells were immunostained with a rabbit anti-ZO-1 polyclonal antibody followed by a secondary antibody, FITC-goat anti-rabbit IgG. Slides were analyzed and recorded by confocal microscopy (63X zoom 3). Bar: 20 µm. Arrows point out ZO-1 puncta and arrowheads the size of pedestals in cell-cell interaction (blue) and cell surface region (cyan and these were smaller in size at all infection times analyzed with EPECΔespF. To acquire a better understanding of the kinetics of ZO-1 and afadin recruitment to the pedestal structures, we quantified the signal in- These data strongly suggest that EPEC can recruit first ZO-1 to the actin pedestals (about 1 hr) followed by afadin (about 2 hr) and that this sequential recruitment depends on EspF expression.
Based on our result that ZO-1 and afadin were transitory colocalized into the pedestals that peaked at 2 hr and gradually decreased until 4 hr of the infection by EPEC, we decided to quantify this phenomenon using Pearson's correlation coefficient analysis of cells infected by EPEC or EPECΔespF ( Figure 3v). In the case of pedestals formed by EPEC, no colocalization was detected at 45 min or 1 hr of infection; however, ZO-1 and afadin colocalization suddenly increased at 2 hr of infection. This colocalization decreased at 3 hr and went undetected at 4 hr of infection, coinciding with the separation of ZO-1 and afadin in the stem and tip pattern of the previously described large pedestals. In contrast, EPECΔespF-induced pedestals revealed a low colocalization coefficient value between afadin and ZO-1 throughout the total duration of the infection time course.
Taken together, these data strongly suggest that the process of pedestal growth and maturation comprises three phases: first, ZO-1 recruitment to the pedestal structures (small pedestals); then, afadin recruitment and colocalization with ZO-1 (medium pedestals); and finally, separation of these two proteins into the pedestal in a stem and tip pattern, respectively (big pedestals).
In order to correlate pedestal size with these phases of maturation, we measured the size of pedestals formed by EPEC and these slightly increased in size at 3 hr of infection (0.65 μm) and they remained at the same size at 4 hr. Comparing the these pedestals' size with the pedestal number in which only ZO-1 signal was detected (phase 1), at 45 min and 1 hr of infection, the pedestals induced by EPEC were smaller in size and the number of these pedestals was higher (around 5 and 15 pedestals per field) than that observed in cells infected with EPECΔespF; meanwhile, afadin was not recruited at these times ( Figure 4). Conversely, the number of pedestals induced at 2 and 3 hr of EPEC infection containing only ZO-1 was around two-to threefold lower than those formed by EPECΔespF. These data correlated with a greater number of EPEC-induced pedestals that displayed strong ZO-1/afadin colocalization (phase 2), which was not detected in EPECΔespF-induced pedestals. Interestingly, we detected a population of pedestals that lacked ZO-1/afadin colocalization at 3 hr of infection We then sought to corroborate whether a physical ZO-1/afadin interaction occurred in the pedestals and if EspF is participating in this process. L cells were mock-infected or infected with either EPEC or EPECΔespF for 1-hr increments until 4 hr in total, and cell lysates were obtained to conduct immunoprecipitation assays using anti-ZO-1 antibodies. Immunocomplexes were analyzed by Western blot using anti-ZO-1, anti-afadin, and anti-EspF antibodies. In total lysates, total concentration of ZO-1 and afadin did not change throughout the infection kinetics, whereas a time-dependent increase in the amount of EspF was observed during wild-type EPEC showed that afadin co-immunoprecipitated with ZO-1 from 1 hr of infection with maximum interaction peaks at 2 and 3 hr of infection, and this interaction pattern was not detected in the absence of EspF ( Figure 5c). These data support the results obtained in the colocalization analysis (see Figure 3i-l), in which the pedestals formed by EPEC had higher ZO-1/afadin colocalization at 2 hr of infection, followed by a lack of colocalization at 4 hr of infection (see Figure 3q-T).
This lack of colocalization was the result of ZO-1 and afadin segregation in the stem and tip pattern inside the pedestals, respectively.
Although it is evident that EspF is required for this interaction, surprisingly EspF was not detected in the immunocomplexes, despite being found in the total lysates, which indicates that EspF is involved in promoting the interaction between ZO-1 and afadin in a yet unknown mechanism.

| ZO-1 or afadin knockdown affects the pedestal growth induced by EPEC
In order to corroborate the role of ZO-1 and afadin on the pedestal maturation, we opted for a silencing approach of either of the F I G U R E 3 Afadin and ZO-1 transiently colocalize in the pedestal-like structure induced by EPEC. L cells were grown at 80% of confluence in mock conditions (a-d) and infected with EPEC at a MOI of 0.5 at different times: 1 (e-h), 2 (i-l), 3 (m-p), or 4 hr (q-t). Cells were fixed, permeabilized, and stained with DAPI (DNA, blue) (c, g, k, o, s) and FAS (F-actin, pseudocolored gray) (d, h, l, p, t). The stained cells were immunostained with a mouse anti-ZO-1 monoclonal antibody followed by a secondary antibody, CY5-donkey anti-mouse IgG (pseudocolored red) (b, f, j, n, r), and with a rabbit anti-afadin polyclonal antibody followed by a biotin-SP-conjugated AffiniPure goat antirabbit IgG and then fluorescein-conjugated streptavidin (green) (a, e, i, m, q). Slides were analyzed and recorded by confocal microscopy (63X zoom 3). Each panel is projecting a zoom for showing actin pedestals beneath adhered bacteria. Bar: 20 µm. (u) Recruitment of ZO-1 and afadin into the pedestal-like structure. Red (ZO-1) or green (afadin) pixels into each pedestal (150 in total) were measured using the Fiji 2.1.0 software (ImageJ) at the different infection times.
(v) Colocalization of ZO-1 and afadin into the pedestal-like structure. Pearson's correlation coefficient was performed using ImageJ, and data of colocalization by pedestal were plotted to compare both strains. The dotted line indicates the value (0.4) considered as less than moderate values of colocalization according to a Fuzzy Linguistic System. Data were compared using a one-way ANOVA test, n = 3 independent experiments. ***p < .0005; ****p < .0001 F I G U R E 4 EspF induces the recruitment of ZO-1 and afadin into the pedestals where their transitory interaction increases the pedestal size. Confocal microscopy images from kinetics of infection of L cells with EPEC or EPECΔespF at a MOI of 0.5 were further analyzed by the Leica Lite software using images as those in from Figure 3 and S3. Dynamic of localization of ZO-1 and afadin into the pedestal was analyzed by measuring the number of pedestals containing ZO-1 (red bars), ZO-1 and afadin colocalization (yellow bars), and ZO-1 and afadin delocalization: separated in the stem and tip pattern, respectively (orange bars). A representative image of the pedestals with each of these characteristics is shown below the graphic; insert: kinetics of pedestal growth during the infection of L cells with EPEC or EPECΔespF. The size of the pedestals was measured (µm) as mentioned before. Each infection time (EPEC and EPECΔespF) was compared using one-way ANOVA test, n = 3 independent experiments. ***p < .0005, ****p < .0001 proteins using two commercial previously tested knockdown methods: a plasmid expressing a shRNA for ZO-1 and a siRNA for afadin (Fanning Lab deposited in Addgene) (Yamamoto et al., 2015). Thus, ZO-1 silencing assays were achieved following the transfection of which were similar to those in GFP-transfected cells or mock cells F I G U R E 5 EspF is required for ZO-1 and afadin interaction. L cells at 80% confluence were infected with EPEC and EPECΔespF at a MOI of 5 for 1, 2, 3, and 4 hr. Infected cells were washed and lysed using RIPA buffer containing protease inhibitors (cOmplete ™ ).
Total proteins in the cell lysates were separated by SDS-PAGE (a) or used for co-immunoprecipitation assays using rabbit anti-ZO-1 antibodies (1 µg) and protein A-agarose; the immunocomplex was separated by SDS-PAGE (b). Both gels were transferred to PVDF membranes for analyzing by Western blot using antibodies against ZO-1, afadin, and EspF. (c) Densitometric analyses of protein bands co-immunoprecipitating with ZO-1. Protein bands were analyzed using the Fiji 2.1.0 software, and data were plotted and analyzed using GraphPad Prism 6.0 software. Precipitated proteins were normalized using the heavy chain of the antibodies against ZO-1 as protein load. All data were compared with mock cells using a oneway ANOVA test and Tukey test, n = 3 independent experiments. *p < .05; **p < .005 These data indicate that ZO-1 is essential for pedestal maturation and that its recruitment to the pedestal allows afadin recruitment to start a transient interaction leading to pedestal growth in enriched ZO-1 sites. These sites support the segregation of ZO-1 and afadin into the stem and tip of pedestals, respectively, to finalize the pedestal growth. Interestingly, data from the afadin knockdown cells suggest that afadin deficiency impedes the formation of ZO-1-rich sites at cell-cell contacts and that the relationship between ZO-1/afadin interaction dynamics and pedestal maturation is more complex.

| Models of polarized cells forming stable intercellular junction also reproduce the phases for pedestal maturation
To probe the ZO-1 and afadin dynamics in polarized cells with high and low transepithelial electrical resistances (Dukes, To trace the afadin redistribution during the infection by EPEC in HT-29 cells under normal-or low-calcium conditions, the HT-29 cells were infected by EPEC for 2, 6, and 10 hr or 2, 4, and 6 hr, respectively. In normal-calcium conditions and at 2 hr of infection, afadin and ZO-1 were excluded from intercellular junctions, but only ZO-1 was incipiently detected in the actin pedestals beneath adhered bacteria (Figure 9a (Figure 11d). At 4 hr of infection, pedestals at the monolayer edge recruited a high amount of ZO-1, and afadin recruitment was initiated but was not yet clearly detected into the F I G U R E 9 Intestinal epithelial cells reproduce all phenotypes detected in L and MDCK cell models: ZO-1 and afadin recruitment for pedestal maturation. Human epithelial cells, HT-29, were treated as MDCK in the calcium switch model. HT-29 cells under normal calcium concentration or low calcium concentration were infected with EPEC for 2, 6, and 10 hr or 2, 4, and 6 hr, respectively. After infection times, cells under the different conditions were fixed, permeabilized, and stained with DAPI (DNA, blue). The stained cells were immunostained with a rabbit anti-afadin polyclonal antibody followed by a biotin-SP-conjugated AffiniPure goat anti-rabbit IgG and then fluoresceinconjugated streptavidin (green), and a mouse anti-ZO-1 monoclonal antibody followed by a secondary antibody, CY5-donkey anti-mouse IgG (pseudocolored red). Slides were analyzed and recorded by confocal microscopy (63× zoom 3). Each panel is projecting a zoom for showing actin pedestals beneath adhered bacteria. Bar: 20 µm pedestals (Figure 11b). In contrast, the pedestals formed on the intercellular junctions at 4 hr of infection did not recruit much of these two proteins (Figure 11e), and the pedestals were smaller (0.66 vs. 1.4 μm) than those produced at the edge of the monolayer (Figure 11j). At 6 hr of infection, both ZO-1 and afadin were recruited in every pedestal formed at the edge of the monolayer with different maturation states: In some pedestals, there was colocalization; in other pedestals, both proteins were in the process of delocalization; and in lesser number, the stem and tip pattern was observed for ZO-1 and afadin, respectively (Figure 11c). In addition, these latter pedestals were larger (1.6 μm) than those formed at 2 and 4 hr of infection ( F I G U R E 1 0 ZO-1 and afadin colocalization is induced by EspF in intestinal epithelial cells  and in the classical polarized MDCK cells. (a-l) HT-29 cells were infected with EPEC (a-d), EPECΔespF (e-h), or EPECΔespF-pespF (i-l) for 10 hr. (m-x) MDCK cells were infected with EPEC (M-P), EPECΔespF (Q-T), or EPECΔespF-pespF (u-x) for 10 hr. Cells were fixed, permeabilized, and stained with DAPI (DNA, blue) (c, g, k, o, s, and w) and FAS (F-actin, pseudocolored gray) (d, h, l, p, t, and x). The stained cells were immunostained with a mouse anti-ZO-1 monoclonal antibody followed by a secondary antibody, CY5-donkey anti-mouse IgG (pseudocolored red) (b, f, j, n, r, and v), and with a rabbit anti-afadin polyclonal antibody followed by a biotin-SP-conjugated AffiniPure goat anti-rabbit IgG and then fluorescein-conjugated streptavidin (green) (a, e, i, m, q, and u). Slides were analyzed and recorded by confocal microscopy (63X zoom 3). Bar: 20 µm at this edge region (Figure 11j). In contrast, the pedestals formed at the same time on the intercellular junctions showed very low signals of these two proteins (Figure 11f), and these pedestals remained small (0.6 μm) as were those formed at 2 and 4 hr in this same region ( Figure 11j).
These data support the notion that the maturation process of actin pedestals depends on EspF and on the cytoplasmic availability of ZO-1 and afadin, due to their disassembly from intracellular junctions. This mechanism involves first a ZO-1/afadin interaction and second a separation of these proteins at both ends of the pedestals leading to the growth of these structures (Figure 11k). Indeed, the similar phenotype was observed in pedestals that were either present in the monolayer edges (see Figure 11c) or formed during the cell incubation at low calcium concentration (Figure 11i), a condition for which the processes previous to pedestal maturation were more evident (Figure 11g-h). In fact, these pedestals were slightly larger (1.9 vs. 1.5 μm) than those detected in the monolayer edges. This size difference was probably due to the higher availability of ZO-1 and afadin in conditions created by the calcium switch, which made the cell-cell contacts unfavorable, hence the disassembly of intercellular junctions generating more edges than at a monolayer front.

| DISCUSSION
EspF is a multifunctional protein that, once injected into epithelial cells, as an effector protein of the T3SS, is involved in subvert various cell processes in the cytoplasm as well as in some organelles (Holmes et al., 2010). EspF is associated with multiple functions, most of them into the cytosol of the host cell, several of which could be a consequence of intracellular junction disruption and/or cytoskeletal rear- TJs, or in cells where these proteins are highly disassembled. ZO-1 recruitment into pedestals is also required for the afadin recruitment to these structures, and thereby EspF will be also required.
Afadin recruitment is needed for a transient interaction with  and, at the end of this interaction, afadin is recruited to the tip of the pedestal and ZO-1 to the stem of the pedestal, and this separation leads to pedestal growth. Remarkably, in ZO-1 knockdown cells, the pedestals were significantly smaller and the number of pedestals decreased, whereas in afadin knockdown cells, the pedestals were also smaller, but the number of pedestals was similar to normal cells.
However, in these cells the pedestal distribution changed by favoring more pedestal formation along the cell surface than in cell-cell contacts.
An interesting finding by us and other groups is that EPEC expresses tropism toward intercellular junctions (Pedersen et al., 2017;Peralta-Ramirez et al., 2008;Ugalde-Silva et al., 2016). We found here that this tropism does not depend on EspF since both wild-type EPEC and an isogenic espF mutant bind to the intercellular junctions ( Figure 1).
However, the espF mutant is unable to cause discontinuity of ZO-1 along these intercellular junctions as the wild-type EPEC does. Both strains form pedestals; however, those induced by EPEC are larger than the ones induced by the espF mutant and this correlates with cellular ZO-1 redistribution from the TJs; in the case of cells treated with the espF mutant, the pedestals are formed on the intercellular junctions.
Since the pedestals are highly dynamic structure that allows the "surfing" of the bacteria along the cell (Shaner, Sanger, & Sanger, 2005), our data strongly suggest that TJ disassembly and the TJ proteins' availability along the cytoplasm could be allowing their easy recruitment and thereby the pedestal movement along the cells, which could be relevant for infection spreading. The lack of TJ disassembly could avoid the movement of the pedestals along the cells, since, when EspF is lacking, these pedestals are formed on the intracellular junctions and are seen arrested there. Thus, we are still working for trying to demonstrate that EspF-induced TJ disassembly and ZO-1-afadin-induced pedestal maturation (as shown here) could be required for pedestal movement.
Interestingly, even though EspF is not required for pedestal formation (McNamara et al., 2001), ZO-1 appears to be required for pedestal formation. The silencing of ZO-1 results in a decrease in pedestal number, and pedestals are formed where residual ZO-1 must be, particularly at sites of intercellular junctions. Moreover, these pedestals are smaller than those formed in ZO-1-enriched zones such as the cell-cell contacts or by ZO-1 disassembly, either by EspF from EPEC or by using the calcium switch assay. It is noteworthy that ZO-1 distribution is clearly homogenous along the pedestal structure when both EPEC infection and low-calcium condition are applicated to epithelial cells as compared with cells infected with the espF mutant. These data suggest that EspF could be supporting a better distribution of ZO-1 inside the pedestal.
In fact, ZO-1, as a protein connecting TJ membrane proteins to the actin cytoskeleton, exists in either stretched or folded conformations, ruled by actomyosin-dependent force, resulting in changes in the localization, stability, and downstream signaling of its interactors (Spadaro et al., 2017). Additionally, it has been shown that ZO-1 is incorporated within EPEC-induced F-actin bundles through its C-terminal prolinerich region (Hanajima-Ozawa et al., 2007).
The eukaryotic linear motifs of EspF (proline-rich sequences and the class III PDZ domain binding motifs) could be playing a role by interacting with actin binding proteins or scaffolding factors that recruit signaling molecules to cell junctions. Interestingly, EspF from EPEC 2348/69 (20.9 kDa) harbors three proline-rich motifs and five class III PDZ domain binding motifs, whereas EHEC O157:H7 EspF (26.2 kDa) harbors the same motifs exactly in the same positions, but it is additionally extended by an proline-rich motif and two PDZ domain binding motifs (Peralta-Ramirez et al., 2008). However, it is well known that EPEC forms highest pedestals than EHEC (Shaner, Sanger, & Sanger, 2005). Thus, EspF motifs could be interfering with ZO-1 and afadin, since ZO-1 contains one SH3 and three PDZ domains and afadin contains three proline-rich domains and one PDZ domain (Ooshio et al., 2010). Unlike EPEC, EHEC also expresses EspFU, which harbors five almost identical 47-residue repeats (R47) consisting of 21% proline, including 22 putative SH3-domain binding (PxxP) motifs. Thus, the C-termini of EspF and EspFU are quite divergent; however, their similarity is because both are rich in proline residues (Campellone et al., 2004). EspFU is 25% identical to EspF and much of the homology between them extends over the first 60-70 residues (40% identical), but it is in the N-terminus, which promotes type III translocation (Campellone et al., 2004). Unlike EspF, in the EspU R47 5 , the tandem PxxP motif is essential for the ability of EHEC to localize EspFU beneath bound bacteria and trigger the formation of an actin pedestal (Aitio et al., 2010). Thus, these proteins show different functions.
The relevant role of ZO-1 for pedestal structure maturation led us to hypothesize that tight junctional proteins could be forming a TJ-like complex between the eukaryotic membrane and the bacterial membranes. This hypothesis implies the participation of other proteins from the TJ such as transmembrane proteins, since ZO-1 and ZO-2 are cytoplasmic TJ proteins (Gumbiner, Lowenkopf, & Apatira, 1991). These latter proteins anchor actin filaments to membrane proteins through their C-terminal regions, and the N-terminal half of ZO-1 binds to the TJ membrane proteins such as claudins and JAM (Ebnet, Schulz, Meyer Zu Brickwedde, Pendl, & Vestweber, 2000;Itoh et al., 1999). In order to explore the role of these tight junctional proteins, we used L cell cultures. These cells lack claudin (Furuse, Sasaki, Fujimoto, & Tsukita, 1998), occludin (Saitou et al., 1997), ZO-2 and ZO-3 (Itoh et al., 1999), and JAM-B and JAM-C (Morris et al., 2006), but express ZO-1 (Itoh et al., 1993) and JAM-A (Morris et al., 2006). Our data clearly show that EPEC is able to form pedestals in L cells similar to those formed in epithelial cells. Furthermore, pedestals formed in the cell-cell contacts were larger than those already formed along the cells; in L cells, these cell-cell contacts were enriched in ZO-1. It has been reported that these nascent cell-cell contacts are primordial junctions where normally JAM-A, ZO-1, and PAR3-PAR6-aPKC complex are recruited (Zihni et al., 2016). In fact, we found that in L cells at 40% of confluence, the pedestals were smaller by avoiding these ZO-1-enriched zones in the cell-cell contact. All these data indicate that ZO-1, but no other main TJ protein, is required for pedestal formation and growth.
Moreover, unlike the primordial junctions, our results showed that JAM-A is not recruited into the pedestals.
Even though no other main TJ proteins appear to be participating in pedestal maturation during EPEC infection, we decided to explore for another protein partner associated with ZO-1 and in the pedestal maturation. Interestingly, L cells, which lack most of the intercellular junctional proteins, were still able to initiate these primordial junctions at the cell-cell contacts by enriching ZO-1 in these sites. Indeed, it has been reported that in cells lacking TJs such as fibroblasts and astrocytes, ZO-1 is localized at cell-cell contact sites with cadherin (Howarth, Hughes, & Stevenson, 1992;Itoh et al., 1993). Furthermore, afadin (Mandai et al., 1997) is localized with ZO-1 at cell-cell contact sites in these types of cells (Yamamoto et al., 1997). The same authors found that afadin is colocalized with ZO-1 at TJs of intestinal epithelial cells, whereas Ooshio et al. (2010) found that the formation of TJs in MDCK cells involves the interaction of afadin with ZO-1. Here, we found that afadin colocalized with ZO-1 at the pedestal structure, detected when cells were fixed with PFA 1% and then methanol-acetone, as used by Ooshio et al. (2010), instead of the classical fixation protocol using PFA 4%. Furthermore, it has been shown that afadin and ZO-1 interact through PR1-2 region of afadin recognizing the SH3 domain of ZO-1 before the formation of TJs, whereas during and after the formation of TJs, ZO-1 dissociates from afadin and associates with JAM-A (Ooshio et al., 2010). Interestingly and similarly, in an EspFdependent form, EPEC recruits sequentially ZO-1 and afadin at the pedestal structures, where both proteins interact before the pedestal growth. During the pedestal growth, ZO-1 is dissociated from afadin, whereas after the pedestal maturation, afadin is recruited at the tip and ZO-1 in the stem of the pedestals. In the TJ formation, ZO-1 and afadin are required for JAM recruitment at the nectin cell-cell contacts (Fukuhara et al., 2002), but in pedestal maturation, JAM was not recruited in the pedestals enriched in ZO-1 and afadin. In this way, these two main proteins necessary for the formation of TJs are required for pedestal maturation. Moreover, both are sequestrated in the pedestal structures, which must have strong consequences on the paracellular pathway of epithelia.
In fact, ZO-1 and afadin are relevant for both TJ formation and pedestal maturation and ZO-1 could also be important for pedestal formation. In the case of the intercellular junction, in ZO-1 knockdown cells, afadin is not recruited to the TJ but to the adherent junctions (Ooshio et al., 2010). We found that in ZO-1 knockdown cells, afadin is not recruited to the pedestal structure, and interestingly, the number of pedestals strongly decreased, suggesting that ZO-1 is also necessary for pedestal formation. These data also suggest that besides Tir, which is critical for F-actin recruitment to the pedestal, this process also requires ZO-1 and afadin that are actin binding proteins, which could work together for forming these intracellular columns composed of an actin-rich core region. In fact, it has been recently reported that multiple ZO-1-mediated interactions contribute to the coordination of epithelial actomyosin function and the genesis of unified apical surfaces. U5 and GuK domains of ZO-1 are necessary for proper apical surface assembly, including organization of microvilli and cortical F-actin (Odenwald et al., 2018). On the other hand, in afadin knockdown cells, ZO-1 is not recruited to the TJ and proteins of the adherent junction are not either (Ooshio et al., 2010). Recently, these data have been confirmed in afadin knockout cells and it was shown that F-actin-binding (FAB) domain of afadin is also required for the formation of TJs (Sakakibara et al., 2018). We found that afadin is critical for pedestal maturation, and without this protein, the pedestals are smaller, but unlike ZO-1, the lack of afadin did not decrease the number of pedestals per cell. Interestingly, the lack of afadin reallocates pedestal formation allowing that most of the pedestals were dispersed along the cells instead of in the cell-cell contacts.
This suggests an initial recruitment of afadin to the nectin cell-cell contacts, as in the TJ formation, and then, further afadin recruitment induced by ZO-1 occurs for the pedestal maturation process. Indeed,

during EPEC infection in polarized cells, such as MDCK and HT29
cells, afadin is disassembled before ZO-1 from the intercellular junctions, but afadin is recruited after ZO-1 at the pedestal structures.
In conclusion, we have shown that ZO-1 and afadin are recruited to the pedestal and they have a transient interaction leading to pedestal maturation. ZO-1-enriched zones are required for pedestal growth and for afadin recruitment. ZO-1 and afadin transiently interact inside the pedestal structures that end with their dissociation and recruitment of afadin in the tip of the pedestal and ZO-1 in the stem of these structures and finally leading to pedestal maturation.
We speculate that the strategic localization of these proteins supports the coordination of epithelial actomyosin function that allows to maintain the intracellular columns composed of an actin-rich core region, the pedestal structure. Additionally, this pedestal maturation could be also important for the dynamics of the pedestal movement along and between epithelial cells.

| espF deletion in EPEC E2348/69
To generate an espF deletion mutant of EPEC strain E2348/69, espF gene was replaced by a gene encoding kanamycin resistance by using the λ red recombinase system (Datsenko & Wanner, 2000). EPEC E2348/69, which was previously transformed with pKD46 expressing the λ red recombinase. Selection of EPEC∆espF::kam colonies was carried out as previously described (Datsenko & Wanner, 2000). espF deletion and kam insertion were verified by PCR. The EspF knockout was verified by Western blot using mouse specific antibodies against EspF.

| Infection assay
Overnight bacterial cultures grown in LB were diluted (1:20) and activated in DMEM without antibiotics or serum, and then incubated at 37°C until the mid-log phase of growth was achieved. MDCK II, L, and HT-29 cell lines were seeded on eight-well Lab-Tek chamber slides (Nalgene Nunc International, USA) or in 60-mm Petri dishes.
When cells reached a confluence of 95%, the monolayers were washed with antibiotic-and serum-free DMEM and then infected with bacteria in DMEM to a multiplicity of infection (MOI) of 0.5 or 5 and maintained for the indicated times in a humidified incubator at 37°C and an atmosphere at 5% CO 2 .  (Gonzalez-Mariscal et al., 1990), and starved in the same culture medium for 2 hr. For recovery condition, cells from low-calcium condition (after 2 hr) were washed with antibiotic-and serum-free DMEM containing normal-calcium condition and maintained in the same culture medium for 2 hr. For each condition, cells were infected with EPEC or EPEC∆espF for the indicated times.
Cells were washed with antibiotic-and serum-free DMEM and starved in the same culture medium for 2 hr. After that, cells were infected with EPEC, EPEC∆espF, or EPEC∆espF-pespF for the indicated times.

| Quantimetric analyses of confocal microscopy images
Pedestal size was measured by detecting the F-actin signal of these structures and using 150 pedestal structures from 5 field images for each experiment (n = 3) using LAS AF Lite (Leica Application Suite Advanced Fluorescence) software. Each pedestal was measured using 0.5-μm optical slices to allow individual pedestal measurement, and they were measured from the base of the pedestal (start of the stem) to the top of the pedestal (end of the tip), where the bacteria were adhered. Pedestal data were plotted as the mean size of each treatment and their standard error to the mean (means ± SEM).
Comparisons between groups were made using one-way analysis of variance (ANOVA) with Tukey's multiple comparison.
Protein recruitment into the pedestal structures was measured by detecting the fluorescence signal (ZO-1 in red and afadin in green), which was delimited as the region of interest (ROI) in 150 pedestals as mentioned above (n = 3 independent experiments), using the Fiji 2.1.0 (ImageJ) software.
Colocalization between ZO-1 and afadin in the pedestals was measured by delimiting the region of interest of pedestal structures in 150 pedestals, using the Fiji 2.1.0 (ImageJ) software. The Pearson correlation coefficients for ZO-1 and afadin colocalization in these structures were plotted as the means of these coefficients and their standard error to the mean (n = 3 independent experiments).
For pedestal quantification, the number of pedestals in the three maturation phases was quantified in at least 5 optical fields (63× zoom) for each group, using the LAS AF Lite software. The means ± SE of the pedestal number by optical field were plotted and compared as mentioned above (n = 3 independent experiments).

| Co-immunoprecipitation assays
L cells were seeded on 60-mm Petri dishes and incubated at 37°C in a humidified atmosphere at 5% CO 2 until 95% of confluence was reached.
Then, cells were infected as described above with EPEC or EPEC∆espF at a MOI of 5 for the indicated times. Proteins were co-immunopre-

| RNA interference
L cells were seeded on 35-mm Petri dishes at 50% of confluence in DMEM supplemented with 10% fetal bovine serum, 1% nonessential amino acids, 5 mM l-glutamine, penicillin (100 U/ml), and streptomycin (100 µg/ml) and incubated at 37°C in a humidified atmosphere at 5% CO 2 . After 18 hr, cells were washed two times with antibiotic-and serum-free advanced RPMI 1640 medium (Thermo Fisher Scientific) and incubated in the same medium for 1 hr. ZO-1 knockdown was performed by transiently transfected cells with a mixture containing

ACK N OWLED G M ENTS
We thank Lorenza Gonzalez-Mariscal for kindly providing MDCK and L cells, Lucia Chavez-Dueñas for her technical support, and Antony Boucard Jr. for invaluable critical review of the manuscript.
This work was supported by a grant from Consejo Nacional de Ciencia y Tecnología (CONACYT 221130) to FNG.

CO N FLI C T O F I NTE R E S T S
The authors confirm that this article content has no conflict of interest.

AUTH O R CO NTR I B UTI O N
FNG and PUS conceived and designed the study; performed analysis and interpretation of the data, and wrote the manuscript; and PUS performed the experiments.

E TH I C S S TATEM ENT
None required.

DATA AVA I L A B I L I T Y S TAT E M E N T
All data are provided in full in the Results section of this paper.

R E FE R E N C E S APPENDIX
F I G U R E A 1 A calcium switch model in MDCK cells to analyze assembly-disassembly-assembly of intercellular junctions. MDCK cells were seeded at 95% of confluency and then subjected to calcium switch assay: (a) MDCK cells were kept in DMEM containing normal concentration of calcium, without antibiotics and bovine fetal serum (BFS) for 6 hr. (b) MDCK cells were washed and kept in fresh DMEM containing low concentration of calcium, without antibiotics and bovine fetal serum (BFS) for 6 hr. (c) MDCK cells were treated as in panel (b) and then the medium with low concentration of calcium was removed and it was replaced by DMEM containing normal concentration of calcium and incubated for another 2 hr. Cells under the different conditions were fixed, permeabilized, and stained with DAPI to detect nuclear DNA and FAS (rhodamine-phalloidin) for F-actin fibers. The stained cells were immunostained with a rabbit anti-ZO-1 polyclonal antibody followed by a secondary antibody, FITC-goat anti-rabbit. Slides were analyzed and recorded by confocal microscopy (63× zoom 1). Each panel shows maximum projection and two z-sections (zx and zy). Bar: 20 µm

(a) (b) (c)
F I G U R E A 2 JAM is not recruited in the pedestal structure at 4 hr of infection when afadin and ZO-1 are dissociated into this structure. L or MDCK cells were infected with EPEC at a MOI of 0.5 for 4 hr. Cells were fixed, permeabilized, and stained with DAPI to detect bacterial and nuclear DNA (blue) and FAS (rhodamine-phalloidin) for F-actin (red). The stained cells were immunostained with a rabbit anti-JAM-A (a-f) or anti-afadin (g-i) polyclonal antibody followed by a biotin-SP-conjugated AffiniPure goat anti-rabbit IgG and then fluorescein-conjugated streptavidin (green). F I G U R E A 3 Afadin and ZO-1 do not colocalize in the pedestal-like structure induced by EPECΔespF. L cells were grown at 80% of confluence and infected with EPECΔespF at a MOI of 0.5 at different times 1 (a-d), 2 (e-h), 3 (i-l), or 4 hr (m-p). Cells were fixed, permeabilized, and stained with DAPI to detect bacterial and nuclear DNA (blue) (c, g, k, o) and FAS (rhodamine-phalloidin) for F-actin fibers (pseudocolored gray) (d, h, l, p). The stained cells were immunostained with a mouse anti-ZO-1 monoclonal antibody followed by a secondary antibody, CY5-donkey anti-mouse IgG (pseudocolored red) (b, f, j, n), and with a rabbit anti-afadin polyclonal antibody followed by a biotin-SP-conjugated AffiniPure goat anti-rabbit IgG and then fluorescein-conjugated streptavidin (green) (a, e, i, p). Slides were analyzed and recorded by confocal microscopy (63X zoom 3 F I G U R E A 5 The calcium switch model in MDCK and HT-29 cells for analyzing assembly-disassembly of afadin and ZO-1 from intercellular junctions. MDCK and HT-29 cells were seeded at 100% of confluency for 5 and 10 days, respectively, and then subjected to calcium switch assay: MDCK (a-c) and HT-29 (g-i) cells were kept in DMEM containing normal concentration of calcium, without antibiotics and bovine fetal serum (BFS) for 6 hr. MDCK (d-f) and HT-29 (j-l) cells were washed and kept in fresh DMEM containing low concentration of calcium, without antibiotics and bovine fetal serum (BFS) for 6 hr. Cells under the different conditions were fixed, permeabilized, and stained with DAPI to detect nuclear DNA (blue). The stained cells were immunostained with a rabbit anti-afadin polyclonal antibody followed by a biotin-SP-conjugated AffiniPure goat anti-rabbit IgG and then fluorescein-conjugated streptavidin (green), and a mouse anti-ZO-1 monoclonal antibody followed by a secondary antibody, CY5-donkey anti-mouse IgG (pseudocolored red). Slides were analyzed and recorded by confocal microscopy (63× zoom 1). Each panel shows maximum projection and a z-section (zx). Bar: 20 µm. Arrows point out the classical localization of ZO-1 and afadin in the intercellular junctions