EPEC Recruits a Cdc42-Specific GEF, Frabin, To Facilitate PAK Activation and Host Cell Colonization

Enteropathogenic Escherichia coli (EPEC) is a leading cause of diarrhea in children, especially in the developing world. EPEC initiates infection by attaching to cells in the host intestine, triggering the formation of actin-rich “pedestal” structures directly beneath the adherent pathogen. These bacteria inject their own receptor into host cells, which upon binding to a protein on the pathogen surface triggers pedestal formation. Multiple other proteins are also delivered into the cells of the host intestine, which work together to hijack host signaling pathways to drive pedestal production. Here we show how EPEC hijacks a host protein, Frabin, which creates the conditions in the cell necessary for the pathogen to manipulate a specific pathway that promotes pedestal formation. This provides new insights into this essential early stage in disease caused by EPEC.

promoting the formation of microcolonies and potentially allowing spread to adjacent cells (7,8).
Both EPEC and EHEC utilize a type 3 secretion system (T3SS) to deliver a battery of virulence effector proteins into host cells in order to subvert the cellular signaling networks necessary to drive the cytoskeletal rearrangements underlying pedestal formation (9). We recently reported that one of these effectors, EspG, hijacks host p21-activated kinase (PAK) to facilitate pedestal formation and bacterial attachment (10). EspG was able to subvert PAK only in the presence of active Rho family small GTPases, which function to both concentrate PAK at the membrane and stimulate PAK activation (10).
Rho GTPases are master regulators of numerous eukaryotic signaling networks (11) and are consequently common targets for subversion by bacterial pathogens (12). For example, several bacteria release toxins that modify GTPases to bring about their permanent activation (13). Other pathogens deliver effectors that mimic host regulators of GTPase signaling, such as guanine nucleotide exchange factors (GEFs; activators) (14) and GTPase-activating proteins (GAPs; inactivators) (15). EPEC and EHEC are themselves known to encode several effectors that have GEF activity (16)(17)(18). However, whether these contribute to the GTPase activation required for EspG-dependent hijacking of PAK or whether some other host or pathogen factor is responsible is unknown. Here, we sought to determine how EPEC ensures a sufficient level of Rho GTPase activation in target host cells.

RESULTS
EPEC activates Rho GTPases. To determine the source of the active Rho GTPases required for EPEC to hijack host PAK (10), we first sought to determine whether the level of active GTPases changes in response to infection. EPEC-infected cell lysates were incubated with GST-PAK-PBD (GST stands for glutathione S-transferase, and PBD stands for protein-binding domain) beads, which specifically interact with the active form of Rac1 and Cdc42. EPEC infection resulted in significant GTPase activation, similar to that seen upon control infection with Salmonella ( Fig. 1A; see also Fig. S1A in the supplemental material). In contrast, infection with ΔescN EPEC, which lack a functional type 3 secretion system, triggered negligible GTPase activation, suggesting that an EPECdelivered effector is responsible for triggering activation.
The EPEC effector Map (mitochondrion-associated protein) is known to be a GEF for Cdc42, and indeed, it has been shown to trigger Cdc42-dependent filopodium formation upon EPEC infection (19,20). However, unlike ΔespG EPEC, Δmap EPEC showed no defect in either pedestal formation ( Fig. 1B and C) or attachment to cultured Hap1 cells (Fig. 1D). A double ΔespG Δmap strain displayed no additional defect in either phenotype compared to ΔespG EPEC, suggesting that Map is not required to activate the GTPases necessary for the hijacking of PAK and consequent pedestal formation. Indeed, in our cell culture infection model, Δmap EPEC activated Rho GTPases to a level equivalent to that of wild-type (WT) EPEC (ig. 1E, Fig. S1B). However, to our surprise, control ΔespG EPEC was unable to induce GTPase activation. Similar results were observed in multiple cell lines (Fig. S1C to E). In some cell types (Caco-2 and LoVo), the level of Cdc42 activation was lower in cells infected with Δmap EPEC than those infected with WT EPEC, suggesting that Map does play a role in Cdc42 activation. However, in all cases, this reduction was much smaller than that seen in cells infected with either ΔespG EPEC or the ΔespG Δmap strain in which Cdc42 activation was abolished. As EspG has no structural or sequence similarity to known GEFs and has never been reported to possess GEF activity, this suggests that EPEC may exploit a host cell GEF to induce the GTPase activity required for EspG-mediated hijacking of PAK and that seemingly EspG itself plays a key role in this process.
EPEC recruits a host Cdc42 GEF to induce GTPase activation. PAK function in cells requires Rho GTPases (21), but PAK can also itself promote activation of GTPases, for example via PIX (PAK-interacting exchange factor) (22), leading to amplification of PAK signaling. It is therefore possible that the EspG-dependent increase in active Rho GTPases upon EPEC infection is due to EspG hijacking the PAK cascade. To test this possibility, we performed a GTPase activation assay in HAP1 cells infected with ΔespG EPEC complemented with various derivatives of EspG ( Fig. 2A, Fig. S2A). Complementation with WT EspG restored the ability of ΔespG EPEC to activate Cdc42, whereas derivatives deficient in binding to Arf GTPases alone or to both Arf and PAK were unable to restore Rho activation. This is presumably because the interaction with Arf GTPases is required for correct localization of EspG. Mutations that disrupt the binding of EspG to either Rab GTPases (ΔR) or PAK (ΔP) did not affect Rho GTPase activation, suggesting that EspG can signal to Rho proteins independently of PAK, and therefore also PIX. In confirmation of this, Cdc42 was efficiently activated by EPEC in either ΔPAK1 cells, or ΔPAK1 cells treated with a chemical inhibitor of other class I PAK isoforms (FRAX486) (Fig. 2B, Fig. S2B). Immunoblot of the level of Cdc42 and Rac1 isolated by GST-GBD beads ("active") from lysates prepared from either uninfected Hap1 cells or those infected with WT Salmonella, WT EPEC, or ΔescN EPEC, as indicated. Also shown is the total level of Cdc42 and Rac1 (active and inactive) in the lysates ("total"). (B) Fluorescence microscopy images of actin pedestals formed on Hap1 cells by WT, ΔespG, Δmap, and ΔespG Δmap EPEC bacteria at 60 min postinfection. Actin (green) is stained with Alexa Fluor 488-phalloidin, and bacteria (red) are stained with anti-intimin antibody. Scale bar, 1 m. (C) Average pedestal length produced by EPEC strains as described above for panel B. Each bar represents the average of results from three separate experiments (100 to 200 cells for each replicate, 300 to 500 cells in total). (D) Quantification of the attachment of strains described above for panel B to WT Hap1 cells relative to that of WT EPEC (there are typically 6 or 7 WT EPEC bacteria per cell). Error bars indicate standard deviations (SD). ***, P Ͻ 0.001; ns, not significant (by one-way analysis of variance [ANOVA] followed by a post hoc Dunnett comparison) relative to WT EPEC attachment. (E) Immunoblot of the level of Cdc42 and Rac1 isolated by GST-GBD beads ("active") from lysates prepared from either uninfected Hap1 cells, or those infected with WT EPEC, ΔespG EPEC or Δmap EPEC, as indicated. Also shown is the total level of Cdc42 and Rac1 in the lysates ("total").
In order to identify the signaling network that leads to GTPase activation, we attempted to reconstitute EspG signaling in vitro using lipid bilayer-coated silica microspheres incubated in cell-free porcine brain extract. As we have previously shown, EspG anchored to microspheres coated with PC:PS (phosphatidylcholine:phosphatidylserine) bilayers fails to recruit any PAK, and it also fails to recruit any endogenous Cdc42 (Fig. 2C). Lipid bilayers designed to further mimic the site of action of EspG by including PIP2 (phosphatidylinositol 4,5-bisphosphate), known to be enriched at sites of EPEC attachment to host cells (23), recruited active, phosphorylated PAK (PAK-P) and a small amount of Cdc42. This was significantly enhanced by the presence of EspG, which leads to efficient recruitment of both Cdc42 and phosphorylated PAK (Fig. 2C). This confirms that EspG can trigger recruitment and activation of Cdc42 and suggests that membrane lipid composition is of critical importance in this EspG-mediated signaling to GTPases.
To attempt to identify how EspG signals to Rho GTPases, the proteins recruited from porcine brain extracts to PC:PS:PIP2-coated microspheres in the presence and absence of anchored EspG were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) (Fig. 2D) and identified by parallel mass spectrometry (see Table S1 in the supplemental material). The top hits are summarized in Fig. 2E. Of particular note, one of the proteins enriched in the presence of EspG was a Cdc42specific GEF, Frabin (also known as FGD4, FYVE, RhoGEF, and PH Domain Containing 4)  (24,25). The recruitment of Frabin was further confirmed using immunoblotting (Fig. 2F).
We next examined whether Frabin plays a role in EPEC attachment and pedestal formation. In contrast to WT Hap1 cells, when Frabin knockout Hap1 cells (ΔFrabin) were infected with WT EPEC, virtually no increase was seen in the level of active Cdc42 (Fig. 3A, Fig. S3A). The level of active, phosphorylated PAK induced by EPEC infection was also much lower in ΔFrabin cells than in WT cells (Fig. 3B, Fig. S3B). There was still a residual level of PAK phosphorylation triggered in ΔFrabin cells; as Frabin is a Cdc42-specific GEF, this may be due to Rac1 activation, triggered by a separate pathway. Consistent with this, treatment of ΔFrabin cells with a Rac1 inhibitor (EHT1864) completely abolished EPEC-induced phosphorylation of PAK, whereas PAK phosphorylation was only partially reduced in WT Hap1 cells (Fig. 3B, Fig. S3B). Pedestals formed by WT EPEC on ΔFrabin cells were less efficient in terms of actin recruitment and had a different morphology compared to those formed on control HAP1 cells (Fig. 3C). Both pedestal length (Fig. 3D) and bacterial attachment (Fig. 3E) were significantly impaired by either knockdown or knockout of Frabin. An equivalent reduction in EPEC attachment was observed in HeLa, Caco-2, and LoVo cells treated with Frabin small interfering RNA (siRNA) (Fig. S3C and D). As seen for PAK phosphorylation above, pedestal length and bacterial attachment showed a small further decrease upon inhibition of Rac1 in ΔFrabin cells and only a partial reduction in WT Hap1 cells ( Fig. 3D and E). Collectively, these data show that EPEC generates the level of active Cdc42 in infected cells required for subsequent hijacking of PAK via the host GEF Frabin.
The DH and PH domains of Frabin are essential for Rho GTPase activation by EPEC. Frabin consists of multiple functional domains (Fig. 4A) (26). At the N terminus is an actin-binding domain (ABD), followed by a Dbl homology domain (DH) that possesses GEF activity, a lipid-binding pleckstrin homology domain (PH), a phosphatidylinositol-binding Fab1, YOTB, Vac1, and EEA1 domain (FYVE), and finally, a second PH domain. To determine which functional domains of Frabin are required for EPEC pedestal formation, we generated a series of derivative versions corresponding to various combinations of the individual domains (Fig. 4A) and expressed these versions in ΔFrabin cells prior to infection with EPEC. The ABD, FYVE, and both PH domains could all be deleted without impacting the EPEC-triggered Cdc42 activation (Fig. 4B, Fig. S4A). Perhaps unsurprisingly, the minimum required for GTPase activation was the DH, i.e., the site of GEF activity. However, the DH domain was inactive when fused to the ABD alone (Fig. 4B, Fig. S4A; pFrabin with amino acids 1 to 392 [pFrabin ). This could be due to the ABD sequestering this construct to actin filaments, away from the plasma membrane where Rho GTPases are enriched. Indeed, Cdc42 activation was restored when a phospholipid-binding PH domain was also present (Fig. 4B,  Fig. S4A; pFrabin 1-523 ).
All of the Frabin constructs that were competent for EPEC-induced Cdc42 activation (Fig. 4B) were also able to promote PAK activation (Fig. 4C, Fig. S4B), with the exception of the DH domain alone (pFrabin 210-392 ). As PAK activation could be triggered by the DH-PH domain combination (pFrabin 210-523 ), this suggests that although the GEF domain alone is able to activate Cdc42 within the cell, the activation must occur at a specific subcellular location for PAK to be activated. Frabin constructs that were able to restore PAK activation in ΔFrabin cells were also able to restore pedestal formation (Fig. 4D) and bacterial attachment (Fig. 4E). Collectively, this demonstrates that Frabin is responsible for EPEC-induced Cdc42 activation, via its DH domain, and that the presence of the PH domain is required to ensure that this Cdc42 activation occurs in the correct location within the cell to allow PAK activation and consequent bacterial attachment to the host cell.
Presumably, once pedestals have been produced, the Frabin pathway may be switched off. Consistent with this, the levels of active Cdc42 are approximately 50% lower at 4 h compared to 1 h postinfection (Fig. S4C). Intriguingly, EPEC encodes an effector, EspH, that can inhibit certain mammalian GEFs (27,28). Transfection of HAP1 cells with a plasmid expressing EspH abolished EPEC-induced Cdc42 activation at all time points (Fig. 4F, Fig. S4C), suggesting it could play a role in downregulating the Frabin pathway following pedestal production.
PIP2 and Arf6 are required to localize Frabin. The first Frabin PH domain, required for EPEC to able to activate PAK, is known to bind to phosphatidylinositol 4,5bisphosphate (PIP2), though its binding specificity remains unclear (29). PIP2 was present in the lipid bilayer-coated microspheres used to identify Frabin as a target for EspG (Fig. 2), and it is known to be enriched at sites of EPEC pedestal formation (23). Consistent with the role of the PH domain, Frabin colocalized with PIP2, immediately below sites of EPEC attachment to cultured Hap1 cells (Fig. 5A). In the absence of EspG, although PIP2 was present at sites of EPEC attachment, Frabin was not recruited (Fig. 5A). When Hap1 cells were infected with EPEC that expresses Tir Y474A , a strain which attaches but is unable to generate actin pedestals, both PIP2 and frabin were still recruited to sites of bacterial adhesion (Fig. 5A), suggesting that actin polymerization is not required for Frabin recruitment by EspG.
As EspG is required for Frabin association with sites of EPEC attachment, we sought to determine whether EspG directly interacts with Frabin by reconstituting Frabin recruitment to lipid bilayers using purified recombinant proteins. Microspheres coated with a bilayer composed of phosphatidylcholine and phosphatidylinositol (PC:PI), and   (Fig. 5B). Recombinant Frabin could be efficiently recruited however when EspG and Arf6 QL were anchored to bilayers also containing PIP2, but not to control PIP2 microspheres in the absence of Arf6 QL and EspG. Surprisingly, Frabin was also recruited, though less efficiently, to control Arf6 QL PIP2 microspheres, though not to inactive, GDP-bound Arf6 (Fig. 5B).
It is therefore possible that active Arf6 and PIP2 are responsible for Frabin enrichment at sites of EPEC attachment and that, as EspG is known to prevent cellular Arf GAPs from inactivating Arfs (30), the role of EspG is to enhance the local level of active Arf. Alternatively, it is possible that Frabin is able to directly interact with either Arf6 or EspG. As Arf6 is required for anchoring EspG to lipid bilayers in our in vitro reconstitution system, we returned to cell infection to further test the role of EspG. When ΔespG EPEC was used to infect Hap1 cells transfected with EspG ΔA (a derivative incapable of binding Arfs and therefore unable to localize to the host cell membrane), Frabin was not significantly enriched at sites of bacterial attachment (Fig. 5C). However, when EspG ΔA is artificially targeted to the membrane by fusing to a lipid-binding PH domain (PH-EspG ΔA ), Frabin was efficiently recruited to sites of EPEC attachment. This suggests that EspG can enhance recruitment of Frabin without interacting with Arf6, and therefore does not act to simply protect active Arf6 from inactivation.

DISCUSSION
Collectively, the results presented here allow us to propose a model by which EPEC ensures that target cells contain the high levels of localized active Rho GTPases necessary for subversion of PAK signaling (Fig. 5D). Upon injection by the T3SS, EspG forms a complex with active, GTP-bound Arf6 at the plasma membrane. EspG binding sterically blocks cellular Arf GAPs from binding and inactivating Arf6. The initial attachment of EPEC leads to the accumulation of PIP2 directly beneath the adherent bacteria. The coincident PIP2 and EspG-Arf6 lead to enrichment of Frabin, triggering local activation of Cdc42. This in turn leads to activation of PAK, which is hijacked by EspG-Arf6 to promote the actin rearrangements necessary for pedestal formation and intimate EPEC attachment to the host cell.
Active Rho GTPases are a requirement for exploitation of PAK by EPEC (10). Rho GTPases are master regulators of multiple signaling pathways within eukaryotic cells, and consequently, their activation status is spatially and temporally dynamic (11). If EPEC encounters a cell in which these pathways are already active, which in cell culture may depend on specific cell line and growth conditions, the Frabin pathway may not be as important for pedestal generation. This possibly explains why previous studies have failed to identify a role for Rho GTPases in pedestal formation (31). However, when needed, by promoting Frabin activation directly beneath adherent bacteria, EPEC ensures that local PAK will always be in the GTPase-activated state required for EspG hijacking.
Little is known about the function of Frabin in healthy cells. Various mutations in the Frabin gene have been linked to the hereditary neuropathy Charcot-Marie-Tooth disease (32), and Frabin may play a role in myelin homeostasis in the peripheral nervous system (33). Frabin overexpression may also play a role in certain cancers (34). The presence of three potential lipid-binding domains (two PH domains and a FYVE domain), and an actin-binding domain suggest it is able to localize to various sites within the cell to perform multiple functions (29), and indeed, Frabin has been seen to localize to several actin-associated membrane structures (35). Here we show that the role of Frabin in EPEC pedestal formation requires only the DH (GEF) domain and the first PH domain. PH domains are found in many proteins and are known to target cellular membranes by binding to phosphoinositide lipids and proteins (36). Indeed, some PH domains are able to bind to both lipids and small GTPases, and they have been reported to act as coincidence detectors (37). For example, Fapp1 (fourphosphate adaptor protein 1) is targeted to membranes by its PH domain, which detects coincident Arf1 and phosphatidylinositol 4-phosphate (38).
Frabin enrichment at sites of EPEC attachment also seems to require two coincident signals. The first of these, PIP2, is known to be enriched directly beneath the adherent bacteria (23). Indeed, disruption of this PIP2 accumulation was shown previously to severely impair EPEC attachment and actin pedestal formation (23), which the results presented here suggest may be due at least in part to impaired Frabin localization. The second required signal for Frabin recruitment could be supplied by constitutively active Arf6 QL in vitro using purified proteins (Fig. 5B), but during cell infection EspG was necessary (Fig. 5A). As EspG is known to interact with Arfs and protect them from inactivation (30), the requirement for EspG could be due to it enhancing the level of active Arf present at the site of bacterial attachment. However, as a mutant of EspG unable to interact with Arfs could efficiently recruit Frabin when artificially targeted to the membrane, it is possible that EspG can directly interact with Frabin. Nevertheless, during infection with WT EPEC, EspG localizes to bacterial attachment sites by binding Arf6, and thus both potential Frabin interactors are present together and may well cooperate. It should also be noted that recruitment to membranes does not necessarily correlate with activation of Frabin's GEF activity (i.e., the DH domain), and the requirement of EspG for Cdc42 activation may be indicative of a direct role in activating Frabin.
The molecular details of the complex interactions regulating this pathway will require further study, but we have identified Frabin as a key component of the signaling complex assembled by EspG to ensure that conditions within the target cell are permissive for EPEC to form actin pedestals. This is the first report of Frabin being involved in bacterial pathogenesis. Previously, Frabin has been shown to be required for the entry of the intracellular parasite Cryptosporidium parvum into epithelial cells by promoting Cdc42 activation and consequent downstream actin rearrangements (39). Though the molecular details of this pathway remain unknown, as seen here with EPEC the recruitment of Frabin by C. parvum required the generation of phosphorylated phosphoinositide lipids in the target cell membrane. Epstein-Barr virus also subverts Frabin, leading to Cdc42 activation and enhanced motility of virus-infected cells (40). Rho GTPases such as Cdc42 are ubiquitous targets for pathogenic bacteria due to their central role in myriad cell signaling pathways (41). Likewise, many bacteria manipulate phosphoinositide lipids in target cell membranes (42). It is thus intriguing to speculate that Frabin may be an important component of the signaling networks subverted by other pathogens to drive the cytoskeletal changes underlying pathogenesis.
Where indicated, plasmids were introduced into mammalian cells using the Neon transfection system (Invitrogen) according to the manufacturer's instructions.
Attachment assay and pedestal quantification. Cells were infected as described previously (44). For pedestal quantification, cells were washed three times with phosphate-buffered saline (PBS), fixed, and stained with Alexa Fluor 488 phalloidin (Lifetech) to visualize actin and an anti-intimin antibody to visualize the bacteria. The numbers of actin pedestals per cell were then counted using fluorescence microscopy. For adhesion assays, after infection, cells were washed twice with PBS and then twice briefly with 200 mM glycine (pH 2), followed by a further two washes with PBS. Cells were fixed and stained as described above, and the number of bacteria still adherent was counted using microscopy.
The pedestal length was determined by drawing a line beneath the attached bacteria and measuring the length using ImageJ (FIJI) tool. A minimum of 150 attached bacteria were counted per condition to estimate the average pedestal length for a given condition. Pixel (fluorescence) intensity plots were generated using the "plot profile" tool of ImageJ (FIJI), and the lines were drawn through pedestal or pedestal-forming regions.
GST-PBD GTPase activation assay. Active GTPases were detected as described previously (45). Briefly, indicated cells were washed twice in ice-cold PBS and then lysed in radioimmunoprecipitation assay (RIPA) buffer (Sigma-Aldrich) supplemented with a cocktail of protease inhibitors. Cell lysates were subjected to centrifugation (13,000 ϫ g, 15 min, 4°C) to remove insoluble material. Equivalent amounts of cleared lysates were then incubated with GST-PBD bound to glutathione-Sepharose resin (GE Healthcare) at 4°C for 30 min. Resin was packed onto a column, extensively washed, and then bound Cdc42 and Rac1 were eluted using SDS-urea. The samples were then analyzed using SDS-PAGE and Western blotting.
In vitro pulldown assays. Pulldowns were performed as previously described (46). Briefly, silica microspheres (Bangs Laboratories) were coated with a bilayer composed of either phosphatidylcholine (PC) and phosphatidylserine (PS), at a molar ratio of 80:20 PC:PS, or phosphatidylcholine (PC), phosphatidylinositol (PI) and phosphatidylinositol 4,5-bisphosphate (PIP2), at a 48:48:4 molar ratio of PC:PI:PIP2. All lipids were purchased from Avanti Polar Lipids. Indicated proteins were anchored to these bilayers, prior to incubation in cell-free porcine brain extract. Following incubation for 15 min, bilayers were washed extensively and associated proteins were analyzed by SDS-PAGE. To identify recruited proteins, beads were washed with 50 mM ammonium bicarbonate (pH 8.5) supplemented with 0.5 mM dithiothreitol (DTT) and then were incubated with trypsin. The digested proteins were identified by electrospray ionization liquid chromatography mass spectrometry (MS) (Mass Spectrometry Service, Cambridge Centre for Proteomics, University of Cambridge, Cambridge, UK). The MS fragmentation data were used to search the National Center for Biotechnology Information (NCBI) database using the MASCOT search engine (www.matrixscience.com).
PAK activation assays. PAK activation was measured as previously described (10). Briefly, cells were cultured in 10-cm dishes and starved in serum-free IMDM overnight. The following day, cells were infected (90 min) or treated with appropriate drugs, as indicated, and then washed twice with PBS before being detached using a cell scraper and resuspended in SDS-urea. Samples were analyzed by SDS-PAGE and immunoblotting using appropriate antibodies. Immunoblots are representative of at least three separate repeats. Bands were visualized using a LI-COR Odyssey Fc imaging system, and band intensities were quantified using the LI-COR Image Studio software.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.

ACKNOWLEDGMENTS
We thank Rachael Stone for technical assistance. This work was funded by a Wellcome Trust Senior Investigator Award (101828/Z/ 13/Z), by the Medical Research Council (MR/L008122/1) and by the Cambridge Isaac Newton Trust.