Invasion of endothelial cells by Neisseria meningitidis requires cortactin recruitment by a phosphoinositide-3-kinase/Rac1 signalling pathway triggered by the lipo-oligosaccharide

Cortactin has recently emerged as a potentially crucial regulator of actin assembly (Olazabal and Machesky, 2001; Weed and Parsons, 2001). This multidomain actin-binding protein can bind and activate the actin-related protein 2/3 (Arp2/3) complex that is a key component in the formation of branched networks of actin filaments at the cell cortex (Uruno et al., 2001; Weaver et al., 2001; Weaver et al., 2002; Weaver et al., 2003). Through this interaction, cortactin is involved in the formation of actin-rich structures such as lamellipodia or membrane ruffles, and it therefore plays a major role in cell motility. In most cell types, cortactin localizes to the perinuclear region and translocates to the cell cortex in response to many stimuli, including growth factors or engagement of adhesion molecules (Weed and Parsons, 2001). The activation of the Rac-1 GTPase seems required for cortactin translocation to the cortical actin network and its subsequent tyrosine phosphorylation in response to growth-factor stimulation (Head et al., 2003; Weed et al., 1998). Although cortactin was first discovered as a prominent substrate for the Src protein tyrosine kinase, the effect of tyrosine phosphorylation on its activity is still unclear. Removal of the C-terminal domain, including the Src phosphorylation site, does not affect the ability of cortactin to activate the Arp2/3 complex (Uruno et al., 2001), whereas expression of a cortactin mutated in all three Src tyrosine phosphorylation sites decreases the rate of endothelial-cell migration in culture (Huang et al., 1998) and the metastatic potential of breast-cancer cells (Li et al., 2001) through an unknown mechanism.

We have recently shown that Neisseria meningitidis, an extracellular human-specific pathogen responsible for septicaemia and meningitis (also referred to as meningococcus), promotes the recruitment and tyrosine phosphorylation of cortactin at its site of entry, suggesting that cortactin might play a role in the invasion process of these bacteria (Hoffmann et al., 2001). Invasion of N. meningitidis through host barriers involves a complex series of events that require the dynamic rearrangements of the actin cytoskeleton to allow the formation of membrane protrusions, membrane fusion and vesicular formation (Nassif et al., 2002). Virulent encapsulated bacteria first adhere to epithelial or endothelial target cells through their type-IV pili, which are long filamentous protein structures, and proliferate locally to form a colony at their site of attachment on the cell surface. When adhering to the apical membrane of epithelial cells, N. meningitidis induces the local elongation of microvilli towards the bacteria, leading to their engulfment and internalization (Merz et al., 1996; Pujol et al., 1997). Adhesion of N. meningitidis to endothelial cells promotes the local formation of membrane protrusions reminiscent of epithelial microvilli structures that surround bacteria and provoke their internalization within intracellular vacuoles (Eugene et al., 2002). This process was observed both in vitro and ex vivo, suggesting that it is essential for the crossing of human endothelium via a transcytosis pathway (Nassif et al., 2002).

The formation of membrane protrusions by encapsulated N. meningitidis stems from the organization of specific molecular complexes, referred to as cortical plaques, underneath bacterial colonies. Cortical plaques result from the recruitment of the molecular linkers ezrin and moesin, which cluster several membrane-integral proteins such as CD44 or ICAM-1, and from the localized polymerization of cortical actin (Eugene et al., 2002; Merz and So, 1997; Merz et al., 1999). Both RhoA and Cdc42 GTPases are crucial for the actin-cytoskeleton rearrangements induced by N. meningitidis in endothelial cells (Eugene et al., 2002). Moreover, we have shown that the pilus-mediated adhesion of encapsulated N. meningitidis onto human endothelial cells induces the clustering and tyrosine phosphorylation of the host-cell tyrosine-kinase receptor ErbB2 (Hoffmann et al., 2001). In turn, ErbB2 activates the Src protein tyrosine kinase and hence leads to tyrosine phosphorylation of cortactin. Although this signalling pathway appears to support bacterial entry into endothelial cells (Hoffmann et al., 2001), the mechanism involved in cortactin recruitment and the role of its tyrosine phosphorylation are still elusive.

The lipo-oligosaccharide (LOS) is one of the major surface molecules of neisseriae. It consists of two oligosaccharide chains attached to lipid A by the sequential action of glycosyl transferases and can be further modified by the addition of terminal sialic-acid moiety (Fig. 1). Meningococcal LOS is a well-known inflammatory mediator, owing to engagement of the lipid-A core with the macrophage CD14/Toll-like-receptor 4 (TLR4)/MD-2 receptor complex, leading to the production of proinflammatory cytokines and chemokines (Smirnova et al., 2003; Zughaier et al., 2004). Moreover, functional consequences of LOS sialylation on protection of N. meningitidis from the phagocytic activity of neutrophils, macrophages or dendritic cells are well documented (Estabrook et al., 1998; Moran et al., 1996; Unkmeir et al., 2002). Although live bacteria avoid phagocytosis, a role has been recently suggested for LOS production in the internalization of killed bacteria by specialized phagocytic cells. Indeed, interaction of sialylated LOS with sialic-acid-binding immunoglobulin-like lectins was shown to increase the uptake by macrophages of heat-killed bacteria (Jones et al., 2003), whereas the formation of complexes between LOS and the lipopolysaccharide-binding protein (LBP) seems to be required for rapid internalization by dendritic cells and their subsequent production of cytokines (Uronen-Hansson et al., 2004). However, the involvement of LOS in the internalization process of live N. meningitidis in non-phagocytic cells and the associated signalling events have not yet been explored.

In this study, we therefore investigated the contribution of LOS to the invasion of human endothelial cells by N. meningitidis. We provide evidence that LOS integrity is required to promote a phosphoinisitide-3-kinase (PI3K)/Rac1 co-stimulatory signal leading to the recruitment and tyrosine phosphorylation of cortactin at the bacterial adhesion site. Moreover, we show that phosphorylated cortactin plays a major role in the formation of actin structures allowing efficient bacterial invasion.

Antibodies against cortactin (p80/85) and p60^c-Src were purchased from Upstate Biotechnology. Anti-phosphotyrosine antibody (4G10) was from Meditech. Polyclonal antibody directed against ErbB2 was purchased from Santa Cruz Biotechnology. Polyclonal antibody directed against ezrin was provided by P. Mangeat (CNRS UMR 5539, Montpellier, France). Rhodamine-phalloidin and the PI3K inhibitor wortmannin were from Sigma.

2C43 (formerly clone 12) is a pilus-bearing encapsulated Opa^– variant of the serogroup-C meningococcal strain 8013 (Nassif et al., 1993; Pujol et al., 1999). ROU (group W135, ET37), is a pilus-bearing encapsulated isolate obtained from the cerebrospinal fluid of a 2-month-old human infant (Pron et al., 1997). Mutant strains used, kindly provided by V. Pelicic (Inserm U570, Paris, France), have been described previously (Geoffroy et al., 2003; Stojiljkovic et al., 1997). Bacterial strains were grown as described previously (Hoffmann et al., 2001).

Human bone-marrow endothelial cells (HBMECs) (Schweitzer et al., 1997) were kindly provided by B. B. Weksler (Weill Medical College, Cornell University, NY, USA) and were cultured in DMEM Glutamax (Life technologies) supplemented with 10% heat-inactivated foetal bovine serum, 7.5 μg ml^–1 endothelial-cell growth supplement (Sigma), 7 IU heparin, 10 mM Hepes (pH 7.4). Before experiments, N. meningitidis was grown overnight on GCB solid medium [4% GC medium base (Difco, MD, USA), 1% agar, 0.4% glucose, 0.2 mg ml^–1 thiamine, 0.0005% Fe(NO₃)₃.9H₂O, 0.01% L-glutamine] at 37°C under 5% CO₂. Several colonies were then selected and grown in DMEM Glutamax supplemented with 0.1% bovine serum albumin (BSA) for 2 hours and were finally diluted to approximately 10⁷ wild-type bacteria per ml inoculum or 10⁶ mutant bacteria per ml inoculum to obtain similar adhesion events. No differences were observed between the growth rates of the wild-type and mutant bacteria in suspension or at the endothelial-cell surface during the course of infection.

Infections were performed as follows. Confluent HBMEC monolayers were starved for 18 hours before infection in DMEM Glutamax supplemented with 0.1% BSA (starvation medium) to lower the basal level of tyrosine phosphorylation. HBMECs were then overlaid for 30 minutes with bacterial inoculum in starvation medium. Cells were then washed three times with starvation medium to remove nonadherent bacteria and infection was allowed to proceed for various periods of time. When indicated, cells were pretreated for 2 hours with 100 ng ml^–1 wortmannin before infection. Cells were then infected as above and the infection was allowed to proceed for 3 hours in the presence of the inhibitor. Wortmaninn treatment did not affect bacterial growth at the cell surface. For adhesion assays, cells grown to confluence in six-well plates were put in contact with bacterial inoculum in starvation medium for 30 minutes. Cell layers were extensively washed, collected by scraping into GCB liquid medium, serially diluted and spread onto GCB plates. Bacteria were grown overnight and colony-forming units were counted. Serial dilutions of bacterial inoculum were also spread onto GCB plates and the number of adherent bacteria in relation to the number of bacteria present in the inoculum was determined. To determine the number of internalized bacteria, cells were infected as above and the infection was allowed to proceed for 3 hours in the presence of wortmannin when indicated. Cells were then extensively washed, incubated with 150 μg ml^–1 gentamicin at 37°C for 1 hour, washed extensively before scraping and plating onto GCB plates. The number of internalized bacteria in relation to the number of adherent bacteria was determined. Each condition was performed in triplicate and each experiment was carried out four times independently.

Cells were washed once with cold PBS and lysed in a Nonidet P40-based buffer (1% Nonidet P40, 50 mM Tris-HCl pH 7.4, 250 mM NaCl, 5 mM EDTA, 1 mM CaCl₂, 1 mM MgCl₂, 2 mM NaVO₄, 5 mM NaF, 1 mM phenylmethylsulfonyl fluoride, 10 μg ml^–1 aprotinin, 10 μg ml^–1 leupeptin, 10 μg ml^–1 pepstatin). The insoluble fraction was removed by centrifugation and the cleared lysates were used for immunoprecipitation with specific antibodies. Precipitated proteins were separated by SDS-PAGE and transferred to nitrocellulose (Schleicher & Schuell). After blocking for 1 hour in PBS containing 1% BSA and 0.05% Tween 20, filters were probed overnight with specific antibodies. Proteins were visualized with peroxidase-coupled secondary antibody using the ECL system (Amersham). Src-autophosphorylation/kinase assays were performed as previously described (Hoffmann et al., 2001).

Vectors encoding full-length cortactin coupled to green fluorescent protein (GFP) or cortactin that had been mutated in its three Src tyrosine-phosphorylation sites and coupled to GFP, were kindly provided by X. Zhan (Holland Laboratory, Rockville, MD, USA). Vectors encoding wild-type or a dominant-negative form of Rac1 coupled to GFP were provided by J. Delon (Institut Cochin, Paris, France), the vector encoding the PH domain of the serine/threonine kinase Akt coupled to GFP was provided G. Bismuth (Institut Cochin, Paris, France) and the vector encoding the GTPase-binding domain of PAK1 coupled to GFP was provided by C. Gauthier-Rouvière (CRBM, Montpellier, France). HBMECs were transfected using the nucleofector system (Amaxa). Briefly, for optimal transfection of HBMECs, 10⁶ cells were suspended in 100 μl solution V (provided by Amaxa) in the presence of 1 μg DNA and subjected to electroporation using program U15 of the nucleofector system. Cells were plated at the same density as cultures that had reached confluency on Permanox coverslips (Costar) at confluence in complete medium for 24 hours before infection and fixation for immunofluorescence analysis.

HBMECs were grown on Permanox coverslips. After infection, cells were fixed in 4% paraformaldehyde for 10 minutes, washed three times with PBS and permeabilized with 0.2% Triton X-100 in PBS for 10 minutes. Cells were incubated for 30 minutes with 3% BSA in PBS and then for 2 hours with the primary antibodies. After three washes with PBS, cells were incubated for 1 hour with CY2-, CY3- or CY5-conjugated anti-mouse or anti-rabbit IgG (Jackson Immunochemicals). Labelled preparations were mounted in Mowiol and analysed with a confocal microscope (Biorad Laboratories). For quantification analysis, the frequency of recruitment of the different proteins at the entry site of wild-type, ΔrfaC or ΔlgtA bacterial strains was determined by counting 30-50 cells. Recruitment was scored when the accumulation of the proteins in a honeycomb lattice structure, defined as a cortical plaque, was clearly visible.

To determine the importance of an intact LOS structure on the ability of N. meningitidis to invade human endothelial cells, we have generated isogenic strains of the 2C43 and ROU strains of N. meningitidis that express LOS molecules lacking the lacto-N-neotetraose structure (Fig. 1A). By inactivating the gene encoding the heptosyltransferase I rfaC (ΔrfaC mutant strains), mutants containing only the lipid-A moiety and the two Kdo (3-deoxy-D-manno-2-octulosonic acid) moieties of the LOS were produced. Additional mutants expressing LOS molecules with a truncated oligosaccharide core were generated in the same way by inactivating the genes encoding the glycosyl transferase lgtA and lgtE (ΔlgtA and ΔlgtE mutants). Moreover, a mutant deleted in a gene unrelated to LOS biogenesis, encoding a membrane lipoprotein (ΔmtrC mutant) was used as control. In agreement with previous observations that LOS of neisseriae can produce an antiadhesive effect owing to the negatively charged carbohydrate sialic acid (Merz and So, 2000), we observed here that LOS truncation resulted in increased adherence of bacterial mutants to endothelial cells (Fig. 1B). In order to compare the number of internalized wild-type or mutant bacteria in relation to the number of adhering bacteria, our experiments were therefore performed with adjusted inocula to obtain similar adhesion events. We observed that the invasive abilities of the 2C43 ΔrfaC, ΔlgtA and ΔlgtE mutant strains were reduced by 50-60% compared with the wild-type strain, whereas deletion of mtrC did not reduce invasion (Fig. 1B). Similarly, the invasive ability of the ROU ΔrfaC strain was reduced by 60% in comparison with the ROU wild-type strain (not shown), strongly suggesting that the LOS efficiently increases the capacity of N. meningitidis to invade human endothelial cells.

N. meningitidis internalization into human endothelial cells depends on the dynamic assembly of actin filaments for the formation of membrane projections towards bacteria and their subsequent engulfment into intracellular vacuoles (Eugene et al., 2002). We therefore examined the reorganization of the actin cytoskeleton induced in cells infected by the ΔrfaC mutant strains. Ezrin, a molecular link between integral membrane proteins and actin cytoskeleton, which was previously shown to be recruited at the site of N. meningitidis adhesion, was normally recruited by ΔrfaC mutant strains (Fig. 2A), as well as the ezrin-binding transmembrane proteins CD44 and ICAM-1 (not shown). By contrast, cortical actin polymerization characteristic of the formation of cellular projections was much less frequently observed in cells infected by the ΔrfaC and ΔlgtA mutant strains than in cells infected by wild-type strains (Fig. 2D). Instead, a local formation of thick bundles of actin filaments was observed (Fig. 2A). These results indicate that a LOS-dependent signalling is essential for the development of the actin-rich cell projections that promote bacterial uptake.

We next examined the potential role of the LOS-induced signalling in the activation of the ErbB2-Src-cortactin pathway, which is required for an efficient bacterial entry into endothelial cells (Hoffmann et al., 2001). Interestingly, we observed that, although ΔrfaC and ΔlgtA mutants induced ErbB2 recruitment as efficiently as the wild-type strains (Fig. 2B), they induced cortactin recruitment poorly (Fig. 2C,D). We therefore analysed whether the mutant strains were defective in activating the ErbB2-Src-cortactin pathway (Fig. 3A). As previously shown, the infection of human endothelial cells by N. meningitidis wild-type strains induced the tyrosine phosphorylation of ErbB2 and the downstream activation of the tyrosine kinase Src. Infection of endothelial cells by the ΔrfaC mutant strains induced ErbB2 tyrosine phosphorylation and Src activity to a similar extent to the wild-type strains, but only weakly induced the tyrosine phosphorylation of cortactin, which was reduced by 60-70% (Fig. 3C). Cortactin phosphorylation was decreased to a similar extent in cells infected with ΔlgtA or ΔlgtE mutant strains, while being slightly increased by the ΔmtrC mutant compared with the 2C43 wild-type strain (Fig. 3B,C). These observations indicate that LOS is required to induce both recruitment and tyrosine phosphorylation of cortactin in cell projections surrounding the bacteria.

Cortactin is known to regulate the formation of dynamic actin networks, and so we investigated whether the induction of aberrant cortical actin polymerization by the bacterial mutant strains might be directly related to their deficiency in recruiting and phosphorylating cortactin. Actin polymerization induced by the 2C43 wild-type strain was analysed in endothelial cells expressing a GFP-tagged cortactin mutant deficient in all three tyrosine-phosphorylation sites (GFP-cortactin_Y/F) (Huang et al., 1998; Li et al., 2001), a GFP-tagged wild-type cortactin (GFP-cortactin) or GFP alone as control (Fig. 4). Infection of endothelial cells expressing GFP-cortactin led to the expected recruitment of both GFP-cortactin and ezrin, and to cortical actin polymerization associated with the formation of membrane protrusions. In cells expressing GFP-cortactin_Y/F, infection induced its recruitment, in agreement with our observation that cortactin phosphorylation was not required for its translocation to sites of cortical actin polymerization (Hoffmann et al., 2001). Moreover, whereas ezrin recruitment was not affected in these cells compared with control cells, the formation of thick bundles of actin filaments, reminiscent of those induced by the ΔrfaC and ΔlgtA mutant strains, was observed at the site of bacterial adhesion. This effect was exacerbated when GFP-cortactin_Y/F was expressed at a higher level, demonstrating that mutations preventing cortactin phosphorylation affect N.-meningitidis-induced actin-rich cell projections. In this condition, GFP-cortactin_Y/F was not recruited to bacterial entry sites, in line with the capacity of cortactin to bind to cortical actin but not to actin bundles. As a control, no recruitment of GFP alone was observed in infected endothelial cells, excluding the possibility that GFP was nonspecifically sequestered in the plasma membrane folds induced by N. meningitidis wild-type strains. Moreover, GFP expression did not affect ezrin recruitment or cortical actin reorganization induced by N. meningitidis.

Cortactin translocation to the cortical actin network in response to growth-factor stimulation is dependent on the activation of the small GTPase Rac1 (Head et al., 2003; Weed et al., 1998). We therefore assessed whether Rac1 GTPase was recruited during N. meningitidis interaction with endothelial cells. Cells expressing a GFP-tagged wild-type form of Rac1 (GFP-Rac1) were infected by the wild-type and mutant strains of N. meningitidis and its recruitment at bacterial entry sites was analysed (Fig. 5A). Interestingly, infection by wild-type strains induced a massive recruitment of GFP-Rac1, whereas GFP-Rac1 recruitment by ΔrfaC (Fig. 5A) or ΔlgtA mutant strains (not shown) was barely detected, suggesting that only the wild-type strains were able to recruit Rac1. Pull-down assays were performed to assess Rac1 activation directly, but the high level of Rac1 basal activity in uninfected cells prevented any clear evidence of Rac1 stimulation in infected cells. We therefore analysed and quantified the redistribution of a GFP-tagged form of the GTPase-binding domain of PAK1 (GFP-PAK1-PBD), a well-established downstream effector of both the Rac1 and the Cdc42 GTPases, which was shown to reflect the localization of activated Rac1 but not Cdc42 at the plasma membrane (Srinivasan et al., 2003). As shown in Fig. 5B,D, infection by wild-type strains induced a massive recruitment of GFP-PAK1-PBD at bacterial entry sites; by contrast, only a limited recruitment was observed in cells infected by either the ΔrfaC or ΔlgtA mutants, thus confirming a default in the activation of Rac1 by the bacterial mutant strains.

A key component in the activation of Rac1 by a wide range of membrane receptors is the PI3K that catalyses the production of plasma-membrane-associated phosphatidylinositol-(3,4,5)-trisphosphate [PtdIns(1,4,5)P₃] (Hall, 1998). In order to visualize the subcellular production of PtdIns(1,4,5)P₃, we therefore analysed the redistribution of a specific fluorescent probe containing the PtdIns(1,4,5)P₃-binding PH domain of Akt (GFP-Akt-PH), a well established PI3K effector (Gray et al., 1999). We observed that infection by the 2C43 wild-type strain induced the recruitment of GFP-Akt-PH (Fig. 5C,D), demonstrating the local activation of PI3K at bacterial entry sites. As expected, this recruitment was prevented in the presence of wortmannin, a selective inhibitor of PI3K (Fig. 5D). Moreover, wortmannin treatment prevented both GFP-Rac1 (not shown) and GFP-PAK1-PBD (Fig. 5D) recruitments induced by the wild-type strains, indicating that PI3K functions upstream of the activation of Rac1 GTPase. Interestingly, the recruitment of GFP-Akt-PH was substantially reduced in cells infected by the ΔrfaC or ΔlgtA mutant strain, indicating a default in the activation of PI3K by these bacterial mutants (Fig. 5C,D). Altogether, these results provide a strong evidence for the LOS-dependent activation by N. meningitidis of a PI3K/Rac1-GTPase signalling pathway.

We therefore asked whether the PI3K/Rac1 signalling pathway was involved in cortactin translocation upon N. meningitidis infection. For this purpose, cells were transfected with a GFP-tagged dominant-negative form of Rac1 (GFP-Rac1-DN) and infected by the 2C43 wild-type strain. We observed that infection induced a massive recruitment of GFP-Rac1-DN, accompanied by a drastic inhibition of cortactin recruitment and a robust actin reorganization into bundles of actin filaments, similar to the actin polymerization structures induced by the ΔrfaC and ΔlgtA mutant strains (Fig. 6A, top; Fig. 6B, grey bars). As a control, in the adjacent untransfected cells (Fig. 6A, bottom; Fig. 6B, white bars), cortactin recruitment induced by the bacterial wild-type strains was not affected and normal cortical actin polymerization associated with the formation of cell projections was still observed. Moreover, inhibition of PI3K activation by wortmannin was also accompanied by an inhibition of the recruitment and tyrosine phosphorylation of cortactin (Fig. 7A,B), and was associated with an extensive actin reorganization into bundles of actin filaments, whereas ezrin recruitment was not affected (Fig. 7A). These results therefore establish the involvement of the LOS-dependent PI3K/Rac1 signalling pathway in cortactin recruitment and in the formation of actin structures and cell projections associated with bacterial invasion.

The potential role of the PI3K/Rac1 signalling pathway in bacterial internalization into endothelial cells was therefore investigated. Cell treatment with wortmannin reduced the internalisation of the 2C43 wild-type strain by about 50%, to a level similar to that observed with ΔrfaC mutant strains (Fig. 7C), whereas it did not affect bacterial adhesion (not shown). Furthermore, in agreement with our conclusion that ΔrfaC mutant strains fail to activate the PI3K/Rac1 signalling pathway, wortmannin did not further reduce the internalization of these mutants. These experiments thus indicate that the LOS-dependent PI3K/Rac1 pathway is required for efficient internalization of N. meningitidis into endothelial cells.

Altogether, our results (summarized in Fig. 8) provide evidence that LOS plays a crucial role in meningococcal entry into endothelial cells by triggering a co-stimulatory signal that leads to PI3K and Rac1-GTPase activation. This signalling pathway is required for cortactin recruitment to the bacterial adhesion sites and its subsequent phosphorylation downstream of the ErbB2/Src pathway, contributing to the formation of actin structures and membrane protrusions associated with the internalization of N. meningitidis.

This study provides evidence that LOS integrity is required for efficient internalization of N. meningitidis into human endothelial cells. LOS integrity is necessary for the activation of a PI3K/Rac1-GTPase signalling pathway allowing cortactin recruitment and tyrosine phosphorylation at bacterial adhesion sites. Moreover, the absence of tyrosine phosphorylation of cortactin was correlated with the formation of bundles of actin filaments and the lack of efficient bacterial internalization. We therefore conclude from our experiments that cortactin controls the formation of the actin-rich cell projections that promote bacterial uptake.

Cortactin has been shown in several situations to distribute to sites of dynamic actin assembly, including lamellipodia, podosomes, invadopodia and bacterial entry sites (Bowden et al., 1999; Weed and Parsons, 2001; Wu and Parsons, 1993). In particular, cortactin seems to be involved differently in cytoskeleton rearrangements occurring during the interaction of various bacterial pathogens with host cells. For example, interaction of enteropathogenic Escherichia coli with epithelial cells results in cortactin recruitment to the bacterial adhesion site, where it seems to be required for F-actin accumulation in pedestal structures, without apparent changes in its tyrosine phosphorylation (Cantarelli et al., 2002). By contrast, invasion of epithelial cells by Shigella flexneri induces the tyrosine phosphorylation of cortactin by a Src-mediated signalling pathway (Dehio et al., 1995), whereas infection by Helicobacter pylori leads to cortactin dephosphorylation and actin rearrangement via Src inactivation (Selbach et al., 2003).

Because expression of cortactin mutant deficient in tyrosine phosphorylation drastically alters actin polymerization at N. meningitidis entry sites, we strongly suggest here that phosphorylated cortactin participate directly in the formation of the actin-rich cell projections that promote bacterial uptake. In vitro, cortactin can increase actin assembly by directly or indirectly stimulating the catalytic activity of the Arp2/3 complex (Uruno et al., 2001; Weaver et al., 2002) and by stabilizing actin filaments at a post-nucleation step (Weaver et al., 2001). However, it is not clear how tyrosine phosphorylation of cortactin modulates the architecture of the actin cytoskeleton, because it does not affect the ability of cortactin to bind F-actin and to activate the Arp2/3 complex. We observed here the formation of bundles of actin filaments in the absence of cortactin recruitment and/or phosphorylation at N. meningitidis entry sites, suggesting that phosphorylated cortactin might control the activity of an actin-bundling protein. These observations are consistent with previous investigations showing that dephosphorylation of cortactin appears to increase actin cross-linking in vitro (Huang et al., 1997). However the mechanism involved is still unclear and will be the subject of further studies.

The molecular mechanisms involved in cortactin translocation from cytosol to cortical actin network are also poorly described. We provide evidence here that cortactin recruitment at the site of formation of actin-rich cell protrusions promoted by N. meningitidis interaction with endothelial cells involves the activation of Rac1 GTPase. We further show that Rac1 is activated downstream of PI3K and production of PtdIns(1,4,5)P₃, a known essential factor for receptor-mediated activation of Rac1 in mammalian cells (Hall, 1998). We had previously shown that cortical actin polymerization induced by N. meningitidis was still observed in cells transfected with a dominant-negative form of Rac1, suggesting that Rac1 was unlikely to be involved in the formation of actin-rich cell protrusions (Eugene et al., 2002). We demonstrate here that, in the absence of Rac1 activation, N. meningitidis induces the formation of bundles of actin filaments that are clearly different from the actin-rich cell protrusions that promote bacterial uptake. Our results provide evidence that activation of the PI3K/Rac1 signalling pathway is required for cortactin recruitment and for N. meningitidis entry into endothelial cells. Activation of PI3K was previously reported in epithelial cells together with the Opa-dependent uptake of the related pathogen, non-piliated Neisseria gonorrhoeae, downstream of the interaction between the bacterial external membrane protein Opa and CEACAM cellular receptors (Booth et al., 2003). We conclude from these findings that piliated N. meningitidis and non-piliated strains of N. gonorrhoeae might similarly activate PI3K in host cells, albeit through different mechanisms. Interestingly, PI3K has also been implicated in the invasion of epithelial cells by several other bacterial pathogens, such as Listeria monocytogenes (Ireton et al., 1996), H. pylori (Kwok et al., 2002) and Rickettsia conorii (Martinez and Cossart, 2004). It is thus tempting to speculate, from the present study, that the PI3K/Rac1 signalling pathway is a general mechanism controlling cortactin translocation as well as cytoskeletal and membrane remodelling associated with cellular invasion by a range of bacterial pathogens.

Moreover, we noticed that a non-phosphorylatable cortactin mutant localized to the bacterial entry site, whereas inhibition of cortactin translocation by a PI3K inhibitor prevented its tyrosine phosphorylation; these findings indicate that cortactin translocation is required for its subsequent phosphorylation by Src. Interestingly, cortactin translocation in response to growth factor stimulation was also shown to require Rac1 activation and to be a prerequisite for its tyrosine phosphorylation (Head et al., 2003), suggesting that there are common mechanisms activated by both N. meningitidis and growth factors. By contrast, these mechanisms apparently differ from those involved during Shigella invasion, because it was recently shown that tyrosine phosphorylation of cortactin by Src precedes its relocalization to Shigella entry sites through interaction of phosphorylated cortactin with the SH2 domain of the adaptor protein Crk (Bougneres et al., 2004). Moreover, during Shigella invasion of epithelial cells, no actin foci developed in the absence of cortactin, and the expression of cortactin mutants deficient for binding to Arp2/3 drastically reduced actin polymerization, suggesting that cortactin is essential for the development of Shigella-induced actin-rich cell projections by directly participating in activation of the Arp2/3 complex (Bougneres et al., 2004). Although Shigella and N. meningitidis both exploit cortactin to modulate the architecture of the host actin cytoskeleton, they appear to use different strategies to recruit cortactin to their entry site, where it regulates actin polymerization and the formation of cellular protrusions by different mechanisms. These molecular differences might explain the diverse morphology of the cellular projections induced by these pathogens: Shigella promotes the formation of very large ruffles, whereas N. meningitidis triggers the formation of thin projections.

In the present study, we further demonstrate an essential role for LOS in the invasion of non-phagocytic cells by N. meningitidis, by showing that LOS is required for activation of the PI3K/Rac1 signalling pathway in endothelial cells, which leads to cortactin recruitment to bacterial adhesion sites. The addition of exogenous meningococcal LOS to endothelial cells infected by LOS mutant strains did not restore cortactin recruitment and phosphorylation (not shown). Similarly, addition of purified meningococcal LOS failed to restore the defective internalization of LOS-deficient bacteria into dendritic cells (Uronen-Hansson et al., 2004), suggesting the importance of the spatiotemporal engagement of membrane-bound LOS with bacterial and/or cellular partners for the association and internalization of N. meningitidis by both specialized and unspecialized phagocytic cells.

Although cortactin translocation is dependent on the activation of a PI3K/Rac1 signalling pathway promoted by the meningococcal LOS, the subsequent tyrosine phosphorylation of cortactin is dependent upon the ErbB2/Src kinase signalling pathway induced by the pilus-dependent adhesion of N. meningitidis. The formation of the actin-rich cell projections promoting efficient internalization of N. meningitidis therefore results from two converging co-stimulatory signalling pathways, triggered by pili and LOS, most likely via interactions with distinct receptors on the surface of endothelial cells. Interestingly, it appears that, in line with our observations on N. meningitidis, the Opa-independent internalization process of piliated strains of the related pathogen N. gonorrhoeae into epithelial cells also involves both pili and LOS interactions with specific cell-surface receptors (Song et al., 2000). Although the signalling events have not been fully identified, the elongation of epithelial microvilli induced by N. gonorrhoeae required the interaction of the lacto-N-neotetraose moiety of LOS with the cell-surface asialoglycoprotein receptor (ASGP-R) (Harvey et al., 2001; Harvey et al., 2002). However, ASGP-R could not account for the effect of N. meningitidis LOS on human endothelial cells, because it is not expressed by endothelial cells (not shown). Therefore, N. meningitidis and N. gonorrhoeae appear to use distinct LOS-dependent molecular mechanisms for invasion. The meningococcal LOS can interact with several receptors, including the serum lipopolysaccharide-binding protein LBP and the CD14/TLR4 receptor complex on human macrophages and other host cells. However, whereas unglycosylated lipid A is sufficient for binding to and activation of the CD14/TLR4 pathway (Zughaier et al., 2004), our observations that the glycosylated moiety of LOS is required to facilitate N. meningitidis invasion suggest the involvement of another receptor, the identification of which will be the subject of a future study.

In conclusion, our results provide evidence for a requirement for LOS in the activation by N. meningitidis of a PI3K/Rac1 signal-transduction pathway in endothelial cells that regulates cortical actin rearrangements and bacterial uptake. Moreover, our results highlight the role of cortactin in this process, at the crossroad of the pilus- and LOS-dependent co-signalling pathways, underlining the complex sequence of signalling events induced by N. meningitidis to elicit its uptake in non-phagocytic cells.

We thank B. B. Weksler (Weill Medical College, Cornell University, NY, USA) and V. Pelicic for kindly providing HBMECs and bacterial strains, X. Zhan (Holland Laboratory, Rockville, MD, USA), J. Delon, G. Bismuth (Institut Cochin, Paris, France) and C. Gauthier-Rouvière (CRBM, Montpellier, France) for kindly providing cortactin, Rac1, Akt and PAK constructs, P. Mangeat (CNRS UMR 5539, Montpellier, France) for kindly providing anti-ezrin antibody, and S. Cazaubon and F. Niedergang for critical reading of the manuscript. This work was supported by grants from the Centre National de la Recherche Scientifique (CNRS), the Institut National de la Santé et de la Recherche Médicale (INSERM), the Ministère de la Recherche et des Technologies and the Direction Générale des Armées. ML and IH were supported by a doctoral Allocation de Recherche du Ministère de la Recherche et des Technologies and a fellowship from the Fondation pour la Recherche Médicale.