Escherichia coli K1 internalization via caveolae requires caveolin-1 and protein kinase Calpha interaction in human brain microvascular endothelial cells.

The morbidity and mortality associated with Escherichia coli K1 meningitis during the neonatal period have remained significant over the last decade and are once again on the rise. Transcytosis of brain microvascular endothelial cells (BMEC) by E. coli within an endosome to avoid lysosomal fusion is crucial for dissemination into the central nervous system. Central to E. coli internalization of BMEC is the expression of OmpA (outer membrane protein A), which interacts with its receptor for the actin reorganization that leads to invasion. However, nothing is known about the nature of the signaling events for the formation of endosomes containing E. coli K1. We show here that E. coli K1 infection of human BMEC (HBMEC) results in activation of caveolin-1 for bacterial uptake via caveolae. The interaction of caveolin-1 with phosphorylated protein kinase Calpha (PKCalpha) at the E. coli attachment site is critical for the invasion of HBMEC. Optical sectioning of confocal images of infected HBMEC indicates continuing association of caveolin-1 with E. coli during transcytosis. Overexpression of a dominant-negative form of caveolin-1 containing mutations in the scaffolding domain blocked the interaction of phospho-PKCalpha with caveolin-1 and the E. coli invasion of HBMEC, but not actin cytoskeleton rearrangement or the phosphorylation of PKCalpha. The interaction of caveolin-1 with phospho-PKCalpha was completely abrogated in HBMEC overexpressing dominant-negative forms of either focal adhesion kinase or PKCalpha. Treatment of HBMEC with a cell-permeable peptide that represents the scaffolding domain, which was coupled to an antennapedia motif of a Drosophila transcription factor significantly blocked the interaction of caveolin-1 with phospho-PKCalpha and E. coli invasion. These results show that E. coli K1 internalizes HBMEC via caveolae and that the scaffolding domain of caveolin-1 plays a significant role in the formation of endosomes.

A broad range of pathogens have the capacity to induce their own uptake by host cells via classical endocytosis (1,2). Because the endocytic pathway mediated by caveolae appears to avoid fusion of lysosomes, pathogens often utilize this pathway to survive within the host cells. Caveolae are indentations in the plasma membrane thought to be involved in transcytosis, signal transduction, and uptake of membrane components and extracellular ligands. Although it is known that caveolae have the capacity to pinch off as endocytic vesicles and can internalize a variety of ligands, the process seems to be selective under normal culture conditions. Caveolae contain a distinct group of molecules, including cholesterol, a 22-kDa protein, caveolin-1 and various glycolipids, and glycosylphosphatidylinositol-anchored molecules (3)(4)(5)(6)(7)(8). Furthermore, receptors for various growth factors and hormones, including epidermal growth factor, platelet-derived growth factor, and insulin, have been localized in caveolae (9 -12). Caveolin-1 is assumed to take a hairpin-loop conformation in the lipid bilayer, thereby exposing both the N and C termini to the cytoplasmic surface (13). A stretch of amino acids referred to as the "scaffolding domain" within caveolin-1 interacts with many signaling proteins (3)(4)(5)13). Although the mechanism of caveola formation for transcytosis of small molecules has been well established, it is not clearly known how bacterial pathogens induce signaling events that lead to caveola formation during invasion of host cells.
One of the crucial events in Escherichia coli meningitis is the traversal of E. coli across the blood-brain barrier (BBB). 1 During this process, E. coli cells seek refuge in a safe compartment in human brain microvascular endothelial cells (HBMEC), which form a lining of the BBB. The BBB exhibits selective permeability mainly for transport of macromolecules, liquids, and nutrients between the blood and the brain. However, E. coli could exploit the mechanisms of transport of macromolecules to promote their entry. Using HBMEC culture as an in vitro model of the BBB, we have demonstrated that E. coli invasion is a complex, multifactorial process that involves important virulence factors like S-fimbriae, OmpA, IbeA, and IbeB (14 -18). However, our studies so far have suggested that OmpA interaction with endothelial cells via a gp96-like receptor on HBMEC is critical for invasion (19). OmpA-mediated E. coli invasion of HBMEC induces the phosphorylation of FAK and its interaction with phosphatidylinositol 3-kinase (20,21). We have further shown that the invading E. coli cells induce phosphorylation of PKC␣ in a phosphatidylinositol 3-kinasedependent manner and that PKC␣ is recruited to the plasma membrane, where it interacts with its substrate, MARCKS (myristoylated alanine-rich C kinase substrate) (22). These signaling events lead to the accumulation of actin beneath the E. coli entry site, which is required for the generation of lamel-lipodia (23). These protrusions of HBMEC enwrap the E. coli cell, which is progressively drawn into a compartment in the host cell. However, the nature of the compartment in which the E. coli cell resides in order to cross the cell is not known.
The interaction of phospho-PKC␣ with caveolin-1 has been shown to be crucial for the formation of caveolae, which contain a conserved consensus phosphorylation site for activated PKC␣ (24 -26). Thus, the increased PKC␣ activity that we observed during E. coli invasion could be directed toward initiation of the interaction with caveolin-1, which may play a key role in regulating the internalization of E. coli. Previous studies have shown that pathogens like E. coli expressing FimH antigen and Chlamydia trachomatis prefer caveola-dependent endocytosis to invade eukaryotic cells and actively recruit caveolin to the sites of bacterial entry (1,2,27). In addition, the internalization of SV40 via caveolae has been shown to trigger actin rearrangements and dynamin recruitment to the sites of entry (28 -32). These events appear to depend on the presence of cholesterol and on the activation of tyrosine kinases that phosphorylate proteins in caveolae. However, these studies used only general tyrosine kinase inhibitors and thus did not reveal the nature of specific kinases involved in caveola formation.
This report describes, for the first time, the role of PKC␣ in initiating caveolin-1-mediated uptake of E. coli into HBMEC via caveolae. During E. coli invasion, the activated PKC␣ migrated to the cell membrane and interacted with caveolin-1, which co-localized with condensed actin beneath the bacterial entry site. Interference of the PKC␣ interaction with caveolin-1, either by overexpression of a dominant-negative mutant form of caveolin-1 or by introduction of a peptide that represents the scaffolding domain into HBMEC, significantly reduced the invasion. In addition, we also demonstrate the association of caveolin-1 with internalized E. coli by confocal microscopy.

EXPERIMENTAL PROCEDURES
Bacteria-A rifampin-resistant mutant of E. coli K1 strain RS218 (serotype O18:K1:H7), E44, has been isolated from the cerebrospinal fluid of a neonate with meningitis and invades HBMEC in a cell culture model. E91 is a noninvasive derivative of E44 in which the ompA gene was disrupted (16). The bacteria were grown in brain heart infusion broth with appropriate antibiotics as necessary. All bacterial media were purchased from Difco.
Expression Plasmids and Antennapedia (AP) Peptides-Wild-type caveolin-1 (a HindIII/BamHI fragment (ϳ600 bp) containing cDNA for Myc-tagged canine caveolin-1) was blunt-ended and ligated in the sense orientation into the HpaI site of pCIS2. Mutant caveolin-1 in which F92A and V94A point mutations were introduced was cloned into the pCIS2 vector. The generation of these plasmids was described in detail previously (33). A peptide corresponding to the putative scaffolding domain of caveolin-1 (Cav, DGIWKASFTTFTVTKYWFYR, amino acids 82-101) or the scrambled control peptide (Cav-X, WGIDKAFFTTSTV-TYKWFRY) was synthesized as a fusion peptide with the C terminus of the AP internalization sequence (RQIKIWFQNRRMKWKK) (34). The peptides were purified and analyzed by reverse-phased high pressure liquid chromatography. The synthetic peptides (20 g each) were labeled using a fluoresceinamine labeling kit (Panvera, Madison, WI). The fluoresceinated peptides were purified by a Sephadex G-15 column.
HBMEC Culture Maintenance and Transfections-HBMEC were isolated and cultured as described previously (14). HBMEC cultures were maintained in RPMI 1640 medium containing 10% heat-inactivated fetal bovine serum, 10% NuSerum, 2 mM glutamine, 1 mM sodium pyruvate, 100 g/ml streptomycin, 100 units/ml penicillin, essential amino acids, and vitamins. HBMEC were transfected with mammalian expression vectors using LipofectAMINE. Briefly, DNA/LipofectAMINE in RPMI 1640 medium was added to 30 -40% confluent HBMEC monolayers. After 6 h of incubation at 37°C, the cells were washed with RPMI 1640 medium, and complete medium was added. The next day, the medium was replaced with medium containing G418, and the cells were maintained in the same medium until they were confluent. For introduction of AP-Cav and AP-Cav-X peptides into HBMEC, the peptides were dissolved in sterile water, added to the culture medium, and incubated with the cells for 6 h. The cells were washed before performing invasion assays.
E. coli Invasion Assays-Confluent HBMEC in 24-well plates were incubated with 1 ϫ 10 7 E. coli cells in experimental medium (1:1 mixture of Ham's F-12 and Medium 199 containing 5% heat-inactivated fetal bovine serum) for 90 min at 37°C. The monolayers were washed three times with RPMI 1640 medium and further incubated in experimental medium containing gentamycin (100 g/ml) for 1 h to kill extracellular bacteria. The monolayers were washed again and lysed with 0.5% Triton X-100. The intracellular bacteria were enumerated by plating on sheep blood-agar plates. In duplicate experiments, the total cell-associated bacteria were determined as described for invasion, except that the gentamycin step was omitted. In some experiments, HB-MEC were pretreated with various inhibitors for 30 min prior to the addition of bacteria. The effects of these inhibitors on HBMEC were assessed by the trypan blue exclusion method, and the effects on bacterial viability were tested by colony plate counting (16).
Preparation of Cytosolic and Membrane Fractions of HBMEC-For detection of PKC␣ activity, confluent monolayers of HBMEC grown on collagen-coated dishes (60-mm diameter) were washed with RPMI 1640 medium, and E. coli cells suspended in experimental medium were added. Following stimulation for varying periods of time (0, 5, 10, 15, and 30 min), the cells were rinsed twice in ice-cold phosphate-buffered saline and placed on ice. The cells in each 60-mm dish were harvested by scraping on ice into 2 ml of cell homogenization buffer consisting of 20 mM Tris (pH 7.5), 0.25 M sucrose, 10 mM EGTA, 2 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 10 g/ml leupeptin, and 2 mM dithiothreitol. The cells were subjected to mild sonication, and the cell lysates were used for the PepTag assay as described below. To obtain the membrane and cytosolic fractions, the cell lysates in the above buffer were initially centrifuged at 5000 ϫ g to remove the debris, followed by centrifugation at 100,000 ϫ g at 4°C for 45 min. The supernatant from this step was designated as the cytosolic fraction, and the pellet was designated as the membrane fraction. The procedures for immunoprecipitation, Western blotting, and scanning of the bands were described previously (22).
PepTag Assay for Nonradioactive Detection of PKC␣ Activity-The PepTag assay utilizes a brightly colored, fluorescent peptide substrate that is highly specific to PKC␣ (Promega). Phosphorylation by PKC␣ changes the net charge of the substrate from ϩ1 to 1, thereby allowing the phosphorylated and non-phosphorylated versions of the substrate to be separated on an agarose gel (0.8%). The phosphorylated species migrates toward the positive electrode, whereas the non-phosphorylated substrate migrates toward the negative electrode. The phosphorylated peptide in the band can then be visualized under UV light. For PKC␣ assay, HBMEC total lysates or membrane proteins (10 -25 g/10 l) were incubated with PKC reaction mixture (25 l) according to the manufacturer's protocol at 30°C for 30 min. The reactions were stopped by placing the tubes in a boiling water bath. After adding 80% glycerol (1 l), the samples were loaded onto an 0.8% agarose gel in 50 mM Tris-HCl (pH 8.0) and separated in the same buffer at 100 V for 15 min, and the bands were visualized under UV light and photographed.
Immunofluorescence Staining-HBMEC were grown in eight-well chamber slides coated with collagen and infected with E. coli as described above. The monolayers were then washed with phosphatebuffered saline and fixed in 2% paraformaldehyde for 15 min at room temperature. Subsequently, the monolayers were incubated with 5% normal goat serum in phosphate-buffered saline containing 1% Triton X-100 for 30 min and further incubated with primary antibody in the same mixture for 1 h at room temperature. The cells were then washed with phosphate-buffered saline and incubated with secondary antibodies conjugated to the fluorochromes Cy3 and fluorescein isothiocyanate, respectively, or rhodamine-phalloidin for 30 min at room temperature. The cells were washed again; the chambers were removed; and the slides were mounted in Vectashield anti-fade solution containing 4Ј,6diamidino-2-phenylindole. Cells were viewed under a Leica DMRA microscope with Plan-apochromat ϫ40/1.25 NA and ϫ63/1.40 NA oil immersion objective lenses. Images were acquired with a SkyVision-2/ VDS digital CCD (12-bit, 1280 ϫ 1024 pixels) camera in unbinned or 2 ϫ 2-binned models using EasyFISH software, saved as 16-bit monochrome, and merged as 24-bit RGB TIFF images (Applied Spectral Imaging Inc., Carlsbad, CA). Optical sectioning was carried out using a Leica TCS SP confocal laser scanning microscope, and images were analyzed using MetaMorph Version 5.0 software (Universal Imaging Corp., Downingtown, PA). The images were assembled and labeled by Adobe Photoshop Version 6.0.

Inhibitors of Caveola Formation Significantly Reduce E. coli
Invasion of HBMEC-Our previous studies showed that PKC␣ activated in E. coli invasion is critical for actin accumulation at the site of E. coli entry (23). Thus, we speculated that PKC␣ might interact with caveolin-1, one of the molecules that has a PKC activation site and is a constituent of caveolae. Thus, to examine the role of caveolae, E. coli invasion assays were performed following treatment of HBMEC with various concentrations of filipin, which is reported to cause disassembly of caveolae and enclosed receptors found in caveolae (35,36). Filipin was found to be effective in blocking the invasion of OmpA ϩ E. coli (E44) in a dose-dependent manner with a 50% inhibitory concentration of 2 M and 80% inhibition at 4 M ((1.12 Ϯ 0.03) ϫ 10 4 cfu/well for untreated HBMEC versus (0.2 Ϯ 0.05) ϫ 10 4 cfu/well for 4 M filipin; p Ͻ 0.001) (Fig. 1A). However, the total cell-associated bacteria (represented as binding) did not differ between untreated and filipin-treated cells, indicating that the inhibition was not due to inefficient binding of E. coli to HBMEC. In contrast, OmpA Ϫ E. coli (E91), which did not show significant invasion in untreated cells, also showed no inhibition of either total cell-associated bacteria or background invasion in filipin-treated HBMEC. Similarly, pretreatment of HBMEC with another inhibitor, cyclodextrin, an agent that inhibits caveola formation by sequestering plasma membrane cholesterol by inducing its efflux, also showed significant inhibitory effect on E. coli invasion. A dose of 2 mM exerted 40% inhibition, whereas a dose of 4 mM exerted a 75% inhibitory effect on invasion of HBMEC ((1.0 Ϯ 0.3) ϫ 10 4 cfu/well for untreated HBMEC versus (0.21 Ϯ 0.10) ϫ 10 4 cfu/well for cyclodextrin-treated HBMEC; p Ͻ 0.001) (Fig. 1B) without significant differences in the total cell-associated bacteria. These results suggest that caveolae may play a role in E. coli invasion.
Phosphorylated PKC␣ Binds to Caveolin-1 in E. coli Invasion-Having determined the importance of caveola formation in bacterial invasion, our next effort was to identify the activation of caveolin-1 during bacterial invasion. Because caveolin-1 activation requires interaction with phosphorylated PKC␣, immunoprecipitation studies were performed using anti-PKC␣ antibody from total cell lysates of HBMEC infected with either E44 or E91. Western blot analysis of the immune complexes using anti-phospho-PKC␣ and anti-caveolin-1 antibodies indicated an association of caveolin-1 with phospho-PKC␣ only in HBMEC infected with E44 (invasive strain) and not in the cells infected with E91 (noninvasive strain) ( Fig. 2A). The association of caveolin-1 with phospho-PKC␣ was observed within 5 min post-infection with E44 and peaked at 15 min, followed by a decline at 30 min. Densitometric analysis of caveolin-1 bands indicated a 3-fold increase in the association with phospho-PKC␣ when infected with E44 (Fig. 2B). In contrast, the noninvasive E. coli strain, despite containing similar levels of PKC␣ as revealed by blotting, showed neither phosphorylated species of PKC␣ nor caveolin-1. To rule out the possibility that the absence of caveolin-1 in E91 cell lysates is not due to low levels of caveolin-1, the total lysates of HBMEC infected with E44 and E91 were also immunoblotted with anticaveolin-1 antibody. The immunoblot showed equal quantities of caveolin-1 in cell lysates. These results suggest that phosphorylated PKC␣ interacts with caveolin-1 in E. coli invasion of HBMEC.
Our previous studies showed that FAK activation is important for PKC␣ phosphorylation (22); thus, we next examined whether blocking FAK activation would inhibit caveolin-1 association with phospho-PKC␣. Wild-type FAK-and FAK-related non-kinase (FRNK)-transfected HBMEC were infected with E44, and cell lysates were analyzed for caveolin-1 activation. FRNK has been shown to negatively regulate the function of HBMEC infected with either E44 or E91 for varying periods of time were immunoprecipitated (IP) with anti-PKC␣ antibody, and the immune complexes were separated by 10% SDS-PAGE. The proteins were then transferred to a nitrocellulose membrane; the upper portion of the blot (above 40 kDa) was immunoblotted with anti-phospho-PKC␣, and the lower portion with anti-caveolin-1 antibody. The upper portion was stripped with Restore Western Stripping buffer and reprobed with anti-PKC␣ antibody to verify equality of loading in all the lanes. In a separate experiment, total cell lysates (20 g) were subjected to Western blotting (WB) with anti-caveolin-1 antibody. B, the blots in A were scanned on a densitometer and plotted as the areas of the individual bands. C and D, total cell lysates of FAKor FRNK-transfected HBMEC or wild-type PKC (PKC-WT)-or PKC/ CAT-KR-transfected HBMEC, respectively, infected with E44 were subjected to immunoprecipitation with anti-PKC␣ antibody, followed by immunoblotting with the same antibody or anti-caveolin-1 antibody. The blot used for anti-PKC␣ antibody was stripped and reprobed with anti-phospho-PKC␣ antibody. In addition, the cell lysates were also subjected to immunoblotting with anti-caveolin-1 antibody.
of FAK (37). Our previous studies showed that overexpression of FRNK in HBMEC blocks E. coli cell-induced activation of FAK and subsequent actin rearrangements (20). As shown in Fig. 2C, immunoprecipitation with anti-PKC␣ antibody pulled down significant quantities of PKC␣ in both cell lysates. However, we observed significant phosphorylation of PKC␣ only in wild-type FAK-transfected HBMEC, and the pattern was similar to that previously reported for non-transfected and infected HBMEC (22). When probed with anti-caveolin-1 antibody, the immune complexes showed profound association with caveolin-1 in these cells. However, in FRNK-transfected HBMEC, neither phospho-PKC␣ nor the corresponding association of caveolin-1 was observed, suggesting that the interaction of PKC␣ with caveolin-1 is downstream of FAK activation.
To confirm this sequence of events, we further examined the association of caveolin-1 with phospho-PKC␣ in cell lysates of HBMEC overexpressing a dominant-negative mutant form of PKC␣. PKC/CAT-KR (a mutant in which the catalytic subunit of PKC␣ has been mutated)-and wild-type PKC-transfected HBMEC were infected with E44 for various time points, and the cell lysates were analyzed. As expected, wild-type PKCtransfected HBMEC showed significant phosphorylation of PKC␣ between 10 and 15 min post-infection, whereas PKC/ CAT-KR-transfected cells showed no activation at all (Fig. 2D). When stripped and reprobed with anti-caveolin-1 antibody, the blot showed association of caveolin-1 similar to that in nontransfected HBMEC, whereas there were no observable levels of associated caveolin-1 in PKC/CAT-KR-transfected HBMEC, suggesting that an intact and active PKC␣ is necessary for the association of caveolin-1. Lack of either phospho-PKC␣ or caveolin-1 was not due to unequal loading of the proteins, as the blot of cell lysates with anti-caveolin-1 antibody showed equal amounts of the protein. Taken together, these results suggest that E. coli cell-induced activation of caveolin-1 may be required for the formation of caveolae and is downstream of FAK and PKC␣.

Localization of Caveolin-1 at the E. coli Entry Site with Actin
and Phospho-PKC␣-Because E. coli invasion of HBMEC appears to be mediated by caveolae, we next examined the distribution of caveolin-1 by immunocytochemistry. Our previous studies showed that invading E. coli cells induce actin accumulation at the E. coli attachment site (23); thus, HBMEC infected with either E44 or E91 were fixed, permeabilized, and stained for both actin and caveolin-1. We observed several groups of bacteria attached to HBMEC, although the actin accumulation was observed only beneath select groups (Fig. 3,  A and B). We have previously shown that only E. coli cells that are entering the cell elicit actin condensation, whereas those merely attached to the surface do not (23). Furthermore, the co-localization of caveolin-1 with actin at the bacterial entry site was also observed (Fig. 3, C and D). In contrast, such a pattern was not present in E91-infected cells (Fig. 3, E-H), as would be expected, because E91 does not activate the signals necessary for either actin accumulation or caveolin-1 activation. Because immunoprecipitation studies revealed that phospho-PKC␣ interacted with caveolin-1, we also stained the infected HBMEC with anti-phospho-PKC␣ and anti-caveolin-1 antibodies. Strong accumulation of caveolin-1 beneath the E. coli entry site was observed (Fig. 3J), with caveolin-1 co-localization with phospho-PKC␣ in E44-infected cells (Fig. 3, K and L), but not in E91-infected cells (Fig. 3, M-P). Taken together, these results suggest that activated PKC␣ is probably recruited to the plasma membrane to interact with caveolin-1 to initiate the formation of caveolae and pulls the bacterium into the cytoplasm utilizing the force of the actin network.
Transcellular Association of Caveolin-1 with E. coli in HB-MEC-Transmission electron microscopy of E. coli invasion of HBMEC revealed that the bacterium resides in an endosome and crosses HBMEC with no signs of multiplication (23). In addition, we did not observe any lysosomal association with endosomes, indicating that E. coli cells somehow avoid lysosomal killing. The results described herein suggest that the en- dosomal compartment might contain caveolin-1, which was previously shown to be responsible for avoiding lysosomal fusion (13). Thus, to confirm the presence of caveolin-1 during the transcellular process, we utilized the confocal laser microscopy Z section technique. HBMEC monolayers were infected with E44 for 30 min, fixed, permeabilized, and stained for caveolin-1 with anti-caveolin-1 antibody followed by fluorescein isothiocyanate-conjugated secondary antibody. E. coli cells were labeled with either anti-K1 (capsular polysaccharide) or anti-Sfimbria antibody followed by Cy3-conjugated secondary antibody. We obtained ϳ45 Z sections of 0.3 m thickness from the top to the bottom of the cell; however, we have presented only five representative sections for each label. As shown in Fig. 4, several bacteria attached to HBMEC near the surface, and the corresponding caveolin-1 section revealed the distribution of caveolin-1 throughout the cells. A clear condensation of caveolin-1 around or beneath the bacteria in several places was observed (Fig. 4, A and F). Further sectioning showed that more bacteria were invading the cells especially at the center of monolayer. Interestingly, in this section, the density of caveolin-1 was slightly reduced throughout the cell, but more clear association was observed with the bacteria (Fig. 4, B and G). The sections that represent the middle of the cell (Fig. 4, D and I) showed a group of bacteria with significant association of caveolin-1. Other bacteria that showed accumulation of caveolin-1 around them at the top of the cell showed decreased association of caveolin-1, suggesting that these bacteria were still in the process of invasion at the plasma membrane. However, the bacteria at the center appeared to have already entered the cell. This group of bacteria continued to associate with caveolin-1 until reaching the bottom of the cell (Fig. 4, E-J). We also stained HBMEC infected with OmpA Ϫ E. coli in a similar fashion, but did not observe any bacteria invading the cells and correspondingly no association of caveolin-1 (data not shown).
In addition, an optical section obtained from the middle of the cell infected with OmpA ϩ E. coli from a different experiment was analyzed for the co-localization of caveolin-1 with OmpA ϩ E. coli using MetaMorph imaging software. As shown in Fig. 5A, the optical section contained several bacteria inside the cell. The corresponding caveolin-1 staining showed accumulation of caveolin-1 around the bacteria, although discontin- uously, but not in other areas where there were no bacteria (Fig. 5B). Overlay of these two images clearly showed that OmpA ϩ E. coli and caveolin-1 were co-localized (Fig. 5C). Orthogonal sections of this optical slice also showed that E. coli (red) and caveolin-1 (green) were co-localized (yellow) when viewed from the area where several bacteria had invaded (Fig.  5D). Moreover, the fluorescence intensities of caveolin-1 and OmpA ϩ E. coli were calculated from this optical section and converted into a line scan. The area of the section that was taken for these calculations is indicated with a blue line. In agreement with our above results that caveolin-1 accumulated at the bacterial entry site, we observed significant association of caveolin-1 (green) with E. coli (red) (peaks at positions 150 -175, 210 -225, and 250 -260). These results strongly suggest that caveolin-1 accumulates around the invading E. coli cells and is probably present within the endosomal membranes that are formed via caveolae.
Overexpression of a Dominant-negative Form of Caveolin-1 in HBMEC Blocks E. coli Invasion-Because caveolin-1 was found to be associated with phospho-PKC␣, we decided to analyze the importance of this association in E. coli invasion. Phospho-PKC␣ is reported to bind to a specific region of caveolin-1 known as the scaffolding domain. Thus, HBMEC were transfected with a mutant form of caveolin-1 in which two amino acids responsible for phospho-PKC␣ interaction in the scaffolding domain (Cav Ϫ /HBMEC) were mutated. We also transfected HBMEC with wild-type caveolin-1 (Cav ϩ /HBMEC) and pcDNA3 as positive and negative controls, respectively. Although efforts were initially directed toward stable transfection, we were unsuccessful in obtaining stable colonies. Thus, "ephémeré (French for short-lived) transfections" were done at 30 -40% confluence of HBMEC and continued in the presence of G418 until cultures were 90 -100% confluent. For each experiment, a portion of the transfected cells were stained with anti-Myc antibody to assess the efficiency of transfection, and we found that 50 -60% of the cells significantly expressed the Myc-tagged proteins (data not shown). Expression of mutant proteins was also verified by Western blot analysis for the expression of caveolin proteins. Total cell lysates of HBMEC and pcDNA3/HBMEC showed a band reactive to anti-caveolin-1 antibody, whereas Cav ϩ /HBMEC and Cav Ϫ /HBMEC showed two bands (Fig. 6A). The upper band was the Myctagged caveolin-1, which migrated slightly slower than that of native caveolin-1. When reprobed with anti-Myc antibody, the same blot revealed one band in both Cav ϩ /HBMEC and Cav Ϫ / HBMEC, suggesting that the caveolin-1 plasmid-transfected HBMEC express significant amounts of Myc-tagged caveolin-1. The equality of loading of total proteins in each lane was examined by blotting the cell lysates with anti-actin antibody. Invasion assays were then performed using these mutants. The results showed that E. coli invasion of Cav Ϫ /HBMEC was inhibited by Ͼ70% compared with either non-transfected or pcDNA3-transfected HBMEC ((1.2 Ϯ 0.2) ϫ 10 4 cfu/well for normal HBMEC and (1.1 Ϯ 0.3) ϫ 10 4 cfu/well for pcDNA3/HBMEC versus (0.35 Ϯ 0.1) ϫ 10 4 cfu/well for Cav Ϫ / HBMEC) (Fig. 6B). Interestingly, the E. coli invasion of Cav ϩ /HBMEC was observed to be 20% more compared with normal HBMEC ((1.5 Ϯ 0.3) ϫ 10 4 cfu/well for Cav ϩ /HBMEC). However, no differences in the extent of binding of E. coli to these cells were observed. OmpA Ϫ E. coli was also used in these invasion assays and showed no significant differences in the levels of background invasion compared with pcDNA3/HBMEC. These results suggest a definite role played by caveolin-1 in mediating E. coli entry into HBMEC.
To demonstrate that the decrease in the E. coli invasion of Cav Ϫ /HBMEC is due to the inability of phospho-PKC␣ to in-teract with caveolin-1, we first examined the activation of PKC␣ in these cells by a nonradioactive PepTag assay. Total cell lysates of both Cav ϩ /HBMEC and Cav Ϫ /HBMEC showed significant activation of PKC␣ between 10 and 15 min postinfection with E44 (Fig. 6C), suggesting that overexpression of mutant caveolin-1 did not affect E. coli cell-induced PKC␣ activation. Next, we examined whether PKC␣ interacts with caveolin-1 by immunoprecipitation. Concomitant with the Pep-Tag assay results, we observed significant amounts of phospho-PKC␣ in both Cav ϩ /HBMEC and Cav Ϫ /HBMEC (Fig. 6D). However, no association of caveolin-1 was found in Cav Ϫ /HB-MEC, whereas only Cav ϩ /HBMEC showed considerable amounts of caveolin-1 coprecipitated with phospho-PKC␣. When subjected to Western blotting with anti-caveolin-1 antibody, the total cell lysates of the transfected HBMEC indicated the presence of equal quantities of caveolin-1. This indicates that the absence of PKC␣ interaction with caveolin-1 might be responsible for the inhibition of E. coli invasion of HBMEC. In addition, this interaction could be crucial for the formation of caveolae.
Absence of Phospho-PKC␣ Interaction with Caveolin-1 at the Plasma Membrane in Cav Ϫ /HBMEC-Our previous studies showed that PKC␣ activated by invading E. coli cells translocates to the plasma membrane for further signaling events (22). One such event could be its interaction with caveolin-1, which is important for the formation of caveolae. Thus, we also examined whether the inhibition of E. coli invasion in Cav Ϫ /HBMEC is due to the inability of phospho-PKC␣ to interact with caveolin-1 at the plasma membrane. Membrane fractions of both Cav ϩ /HBMEC and Cav Ϫ /HBMEC infected with E44 were prepared and assessed for the activation of PKC␣ by PepTag In simultaneous experiments, the total cell-associated bacteria were also determined. The results are expressed as either relative binding or invasion, taking HBMEC values as 100%. The error bars represent the means Ϯ S.D. of four separate experiments performed in triplicate. C, total cell lysates (20 g of protein) of Cav ϩ /HBMEC and Cav Ϫ /HBMEC infected with E44 for varying periods of time were subjected to the PepTag assay for PKC␣ activity. D, ϳ200 g of total cell lysate proteins of Cav ϩ /HBMEC and Cav Ϫ /HBMEC infected with E44 were immunoprecipitated (IP) with anti-phospho-PKC␣ antibody. The immune complexes were subjected to Western blotting (WB) with antiphospho-PKC␣ and anti-caveolin-1 antibodies. In addition, the total cell lysates (20 g) were also subjected to immunoblotting with anti-caveolin-1 antibody. assay. As expected, both membrane fractions showed peak activation of PKC␣ at 15 min post-infection (Fig. 7A). The Cav ϩ /HBMEC membrane fractions showed activation even at 5 min post-infection. The membrane fractions were then immunoprecipitated with anti-phospho-PKC␣ antibody, followed by immunoblotting with the same antibody and anti-caveolin-1 antibody. We observed that a greater amount of phospho-PKC␣ had been recruited to the membrane fraction in Cav ϩ /HBMEC, with a corresponding increased interaction with caveolin-1 (Fig. 7B). Cav Ϫ /HBMEC showed normal levels of phospho-PKC␣ compared with non-transfected and infected HBMEC, but showed no association of caveolin-1. These results suggest that despite the translocation of activated PKC␣ to the plasma membrane, its inability to interact with caveolin-1 blocks E. coli invasion of Cav Ϫ /HBMEC.
Immunofluorescence studies on Cav ϩ /HBMEC infected with OmpA ϩ E. coli showed strong co-localization of caveolin at actin condensation sites (Fig. 8, A-D). In these cells, caveolin recruitment was much greater at the bacterial entry site compared with non-transfected HBMEC. Similarly, the density of PKC␣ recruitment was also much greater than in control cells, indicating that overexpression of caveolin may increase the interaction with PKC␣, thus increasing the formation of more caveolae necessary for invasion. This could be the reason for slightly increased E. coli invasion of Cav ϩ /HBMEC. In contrast, in Cav Ϫ /HBMEC, although there was significant phospho-PKC␣ and actin accumulation beneath the bacteria, there was no detectable accumulation of caveolin-1 (Fig. 8, I-P). Thus, it is clear that an intact scaffolding domain contributes to the association of caveolin-1 with PKC␣, thereby facilitating bacterial invasion.
Cell-permeable Scaffolding Peptide of Caveolin-1 Blocks E. coli Invasion-Because the binding of PKC␣ to the scaffolding peptide of caveolin-1 has been found to be crucial to the internalization of caveolae, we used a peptide that represents the caveolin-1 scaffolding domain (amino acid residues 82-101) and a scrambled peptide (control) to examine their effect on E. coli invasion. These peptides were synthesized with a 16amino acid portion of the AP homeodomain (RQIKIWFQNR-RMKWKK), a Drosophila transcription factor, at the N-termi- nal side of both the scaffolding peptide (AP-Cav) and the scrambled peptide (AP-Cav-X). The AP protein facilitates homogeneous uptake of peptides or oligonucleotides into cultured mammalian cells through a non-endocytic and non-degradative pathway (34). To analyze the efficiency of entry of the peptides into the cell, the peptides were labeled with fluoresceinamine and then added to the cells. These fluoresceinated peptides entered 60 -70% of the cells at a 2 M concentration within 6 h of incubation and distributed throughout the cell (Fig. 9, A and  B). HBMEC pretreated with AP-Cav showed significant inhibition of E. coli invasion compared with HBMEC pretreated with AP-Cav-X in a dose-dependent manner ((0.4 Ϯ 0.2) ϫ 10 4 cfu/well for AP-Cav versus (1.1 Ϯ 0.3) ϫ 10 4 cfu/well for AP-Cav-X at 4 M; p Ͻ 0.01) (Fig. 9C). In contrast, the total cell-associated bacteria did not differ significantly between AP-Cav-and AP-Cav-X-treated HBMEC, suggesting that lack of invasion is not due to the inability of bacteria to bind to the cells. Consistent with the ability of AP-Cav to inhibit the E. coli invasion, immunoprecipitation of the cell lysates of HBMEC treated with the scaffolding peptide with anti-phospho-PKC␣ antibody resulted in the inhibition of caveolin-1 association with phospho-PKC␣ (Fig. 9D). In contrast, AP-Cav-X-pretreated HBMEC showed an association of caveolin-1 with phospho-PKC␣ similar to that in untreated HBMEC. Lack of caveolin-1 association with phospho-PKC␣ is not due to the absence of caveolin-1 in these fractions, as the total cell lysates showed equal quantities of caveolin-1 when immunoblotted with anticaveolin-1 antibody. In addition, we also examined the association of actin and phospho-PKC␣ by immunocytochemistry. The results were similar to those obtained for Cav ϩ /HBMEC and Cav Ϫ /HBMEC (similar to Fig. 8), in which we observed accumulation of actin and phospho-PKC␣ at the sites of E. coli entry, but not co-localization of caveolin-1 in AP-Cav-pretreated HBMEC. This confirms our hypothesis that E. coli cells use caveolae as a mode of entry into HBMEC, thereby probably ensuring their survival by avoiding fusion with lysosomes. DISCUSSION E. coli invasion of HBMEC occurs by a zipper-like mechanism in which the host cell plasma membrane enwraps the invading bacteria and becomes an endosome (23). This mechanism requires E. coli cell-induced HBMEC cytoskeletal rearrangements for the accumulation of actin at the site of bacterial entry. Transcytosis of E. coli in the endosome occurs within 30 min post-infection without multiplication. In addition, there is no evidence of bacterial killing inside the endosome, which could be due to the ability of bacteria within the endosome to avoid the fusion of lysosomes. The nature of this endosome is not known to date. A few studies have attempted to establish the role of caveolae in the entry of pathogens such as E. coli expressing FimH and Chlamydia (1,2,27). However, the molecular events leading to the formation of caveolae for any pathogenic microorganism have not been described thus far.
Our findings underscore the critical nature of caveola formation in E. coli invasion of HBMEC via interaction of caveolin-1 with phospho-PKC␣. In addition, these data indicate that the integrity of cholesterol-enriched microdomains that are normally present in HBMEC must be necessary for invasion. This may explain why the drugs filipin and cyclodextrin, which are known to specifically disrupt raft microdomains, inhibit most E. coli entry. Because our previous studies showed that E. coli OmpA interacts with a 95-kDa gp96 like molecule for invasion of HBMEC (19), it is possible that the OmpA receptor might be enriched in caveolae during the invasion process. In agreement with this concept, we observed clustering of the OmpA receptor beneath the E. coli entry site by immunocytochemistry by 15 min post-infection. Receptors for insulin have been shown to cluster within caveolae and to interact with caveolin-1 to differentially modulate post-receptor signaling (33). Thus, it is reasonable to speculate that OmpA interacts with its receptor, gp96, which in turn interacts with caveolin-1 and clusters within the caveolae for further signaling necessary for E. coli invasion. However, caveolae are small relative to E. coli to be transported across the cell; therefore, it would be possible that several OmpA molecules on E. coli induce multiple raft domains after interaction with HBMEC receptors and that these are subsequently united to surround the bacterium.
Our studies show that E. coli invasion of HBMEC induces activation of PKC␣ by three folds in an OmpA-dependent manner (22). Activated PKC␣ then translocates to the plasma membrane for further signaling events. A major finding of this study is that the activated PKC␣ interacts with caveolin-1, a specific marker of caveolae. We also found by immunocytochemistry that PKC␣ activated by invading E. coli cells co-localizes with caveolin-1. In addition, these two molecules are also present at the actin condensation sites beneath E. coli attachment, reflecting possible signaling complex formation around the caveolae for efficient transcytosis of E. coli. However, the role of caveolin-1 in further signaling events is not clear at this point. Interestingly, overexpression of dominant-negative forms of either FAK (FRNK) or PKC␣ (PKC/CAT-KR) inhibited the association of phospho-PKC␣ with caveolin-1, indicating that both FAK and PKC␣ may be upstream of caveolin-1 interaction. Thus, targeting of phospho-PKC␣ to caveolae could be an important event in E. coli invasion. Several studies have previously shown that PKC␣ regulates the membrane invaginations and is enriched in caveolae, and our results are in agreement with these observations (4,25).
We further demonstrated that phospho-PKC␣ interacts at the scaffolding domain of caveolin-1, as overexpression of a dominant-negative form of caveolin-1 in which the amino acids in the scaffolding domain have been mutated significantly blocked E. coli invasion of HBMEC. Consistent with these In duplicate experiments, the total cell-associated bacteria (represented as binding) were determined. The peptide-treated and E44-infected HBMEC lysates were subjected to immunoprecipitation (IP) studies with anti-phospho-PKC␣ antibody. The immune complexes were probed with either anti-phospho-PKC␣ or anti-caveolin-1 antibody. In addition, total cell lysates were also subjected to Western blotting (WB) with anti-caveolin-1 antibody to verify the presence of caveolin-1 (D). findings, introduction of a cell-permeable peptide that represents the caveolin-1 scaffolding domain resulted in significant inhibition of E. coli invasion. In contrast, a scrambled peptide showed no such inhibitory activity. Bucci et al. (34) have successfully used this peptide both in vitro and in vivo for selective inhibition of acetylcholine-induced vasodilation and nitric oxide production. The uptake of this peptide by cells was rapid and independent of membrane fluidity and is a useful tool for studying the scaffolding domain-mediated signaling pathways. Indeed, in vitro biochemical studies have indicated that the caveolin-1 scaffolding domain regulates the activity of PKC␣ (3,4,6). In addition, overexpression of a dominant-negative form of caveolin-1 in HBMEC prevented accumulation of caveolin-1 at the bacterial entry site despite the recruitment of phospho-PKC␣ to the membrane. These results are in agreement with the immunoprecipitation studies in which phospho-PKC␣ present in the membrane fractions of Cav Ϫ /HBMEC showed no association with caveolin-1. These observations suggest that phospho-PKC␣ recruitment to the E. coli entry site does not require its interaction with caveolin-1, but is absolutely necessary for efficient E. coli invasion. Similarly, we observed actin accumulation at the bacterial attachment site despite absence of invasion. Because our previous studies showed that PKC␣ activation is important for actin accumulation beneath the bacteria (22), PKC␣ might play two differential roles in E. coli invasion: one in actin rearrangements and the other in caveolin-1 activation and effective caveola internalization. These results are in sharp contrast to those observed for SV40 internalization via caveolae, in which either actin condensation or tyrosine phosphorylation occurs only after SV40-induced caveola formation (31). As a control, we also used wild-type caveolin-1-transfected HBMEC in E. coli invasion assays, which showed a slight increase in invasion. Previous studies by other investigators have shown that overexpression of caveolin-1 induces the formation of caveolae in several cell types (3,5,13). Thus, the increase in E. coli invasion could be due to nonspecific trapping of bacteria in caveolae. Alternatively, overexpression of caveolin-1 increases the turnover of receptors to the cell surface, and these receptors are subsequently responsible for internalization of more E. coli.
The survival of E. coli in the endosome depends on the ability of the intracellular parasite to resist endosomal acidification and lysosomal fusion. Our studies show that during transcytosis, E. coli cells remain enclosed in the endosome containing caveolin-1, suggesting that caveola-mediated entry may play an important role in preventing lysosomal fusion. The caveolamediated SV40 and C. trachomatis entry pathways transport the microorganisms to the endoplasmic reticulum, a normal target for endocytic cargo, rather than the endosomal/lysosomal compartment (30 -32). However, caveolin-1 may not be the sole molecule responsible for the avoidance of lysosomal fusion, as shown for Chlamydia psittaci, in which early gene expression by the bacterium in the endosome is important (38). Despite poor understanding of the mechanisms that lead to lysosomal avoidance, the continuing association of caveolin with E. coli in caveola-mediated entry may have important implications in the crossing of the BBB.
In summary, we have shown that E. coli internalization of HBMEC occurs via caveolae. The scaffolding domain of caveolin-1 interacts with PKC␣ at the plasma membrane beneath the E. coli entry site. Because invasion by E. coli depends on OmpA interaction with its receptor, a gp96-like protein, identification of the role of caveolin-1 in bringing these receptor molecules into caveolae furthers our understanding of the mechanism of invasion.