Internalization of Pseudomonas aeruginosa Is Mediated by Lipid Rafts in Contact Lens–Wearing Rabbit and Cultured Human Corneal Epithelial Cells

Abstract

purpose. The internalization of Pseudomonas aeruginosa (PA) in nasal and tracheal epithelium has recently been shown to involve the formation of cholesterol- and sphingolipid-rich plasma membrane domains (lipid rafts). The purpose of this study was to investigate the role of lipid rafts in PA internalization by corneal epithelium in vivo, in vitro, and after contact lens wear.

methods. Lipid raft formation was evaluated in rabbit corneas with and without contact lens wear and a human corneal epithelial (hTCEpi) cell line before and after PA infection with cornea-pathogenic strains by staining with FITC-conjugated cholera toxin β-subunit, known to bind the lipid raft component GM1. Bacterial internalization was assessed by gentamicin survival assay. The role of lipid rafts in PA internalization was evaluated by pretreatment of hTCEpi cells with cholesterol metabolism inhibitors. The interaction of PA with lipid rafts was confirmed by flow cytometry.

results. Contact lens wear in rabbits induced lipid raft formation in occasional surface corneal epithelial cells. Subsequent PA exposure showed preferential binding to lipid raft–forming cells, leading to lipid raft aggregation and PA internalization. A similar sequence of lipid raft formation and PA internalization was also observed in hTCEpi for all PA strains. Internalization of all PA strains was blocked by three cholesterol metabolism inhibitors (P < 0.01). Flow cytometry showed an association of PA with rafts.

conclusions. These findings demonstrate that contact-lens–mediated PA internalization involves lipid raft formation. Also, hTCEpi cells may be used as an experimental model for studying further the molecular mechanism(s) of PA infection in the corneal epithelium.

Lipid rafts are glycosphingolipid- and cholesterol-enriched domains in the plasma membrane that move within the fluid lipid bilayer ¹ and are distinct from the remaining areas of the plasma membrane, which predominantly consist of phospholipids. These formations are characterized by a high melting temperature and resistance to solubilization in nonionic detergent at low temperature. Rafts can undergo macromolecular reorganization and form large membrane platforms, which have been implicated in various cellular functions, including membrane trafficking ¹ ; signal transduction, ^{2 3} particularly as defined for T cells and other leukocytes ^{4 5} ; and the regulation of integrin function. ⁶ Recent studies have also revealed that lipid rafts are involved in the internalization of pathogenic microorganisms into host cells. ^{7 8} Bacterial internalization mediated by lipid rafts in immune cells appears to involve an endocytic process, distinct from normal endosome–lysosome pathways, since the compartments formed by lipid-rich raft microdomains do not fuse with internal lysosomes. ^{7 8} Legionella pneumophila has been found to be internalized through a pathway involving the lipid raft component, ganglioside GM1. ⁹ Other studies have shown that cholesterol is essential for the internalization of Mycobacterium bovis into macrophages with subsequent intracellular survival. ^{10 11 12} Two components of lipid rafts, cholesterol and the ganglioside GM1, have recently been shown to be involved in the entry of Brucella suis into murine macrophages under nonopsonic conditions. ¹³  

Pseudomonas aeruginosa (PA) is a pathogenic Gram-negative bacterium that is the most frequent cause of severe corneal infection (infectious keratitis) associated with extended wear of soft contact lenses. ^{14 15} Results in one study have demonstrated that many clinical isolates of PA can invade and replicate within corneal epithelial cells in animal models of corneal infection. ¹⁶ Additional recent studies have further shown that PA infection of respiratory epithelial cells triggers the reorganization of lipid rafts to larger membrane platforms at the sites of bacterial and host cell interaction and that formation of membrane lipid rafts is necessary for bacterial internalization in mammalian nasal and tracheal epithelial cells. ¹⁷ Currently, however, it is not known whether lipid rafts are involved in the internalization of PA in the corneal epithelium after contact lens wear. 

The purpose of this study was to investigate the role of membrane lipid rafts in the internalization of infectious PA strains by corneal epithelial cells, using an established rabbit model of contact lens wear (in vivo) and immortalized human corneal epithelial cells, hTCEpi (in vitro). In this study contact lens wear in rabbits exposed a population of surface corneal epithelial cells that exhibited lipid rafts and preferentially bound and internalized PA. In addition, the disruption of lipid rafts using three inhibitors of cholesterol metabolism specifically blocked PA internalization in the hTCEpi cell line. Taken together, these findings provide important new information on the molecular mechanism of PA-induced bacterial keratitis, especially associated with contact lens wear. 

Materials and Methods

Bacteria

Three Pseudomonas aeruginosa (PA) strains, ATCC27853 (American Type Culture Collection, Rockville, MD), 6294, and 6487 were used for this study. Strain ATCC27853 is a noncorneal isolate that has been shown to be fully infectious in the cornea and has been used as a standard test organism in experimental corneal studies in rabbits ¹⁸ and humans. ¹⁹ In addition, ATCC27853 has been used in bacterial adherence and PA internalization studies of lipid raft formation in many cells and tissue systems. ¹⁷ Two infectious corneal isolates, strains 6294 and 6487, were kindly provided by Suzanne M. Fleiszig (University of California, Berkeley) and were characterized as invasive. ²⁰ Bacteria were cultured on Mueller Hinton II agar (BD Biosciences, Sparks, MD) for 17 hours at 37°C before use. For the internalization experiments, individual bacterial strains were suspended in phosphate-buffered saline (PBS). The bacterial concentration was adjusted to 1 × 10⁹ CFU/mL using a spectrophotometer. The bacterial suspension was then diluted with minimum essential medium (MEM; Invitrogen Corp., Carlsbad, CA) to a concentration appropriate for each experiment as indicated in the following sections. 

Animals

Thirteen New Zealand White rabbits (body weight, 2.5–3.5 kg) were used in the study. All rabbits were treated according to the ARVO Statement for the use of Animals in Ophthalmic and Vision Research. To facilitate contact lens retention, the nictitating membrane of both eyes was surgically excised under anesthesia (50 mg/kg ketamine and 5 mg/kg xylazine) and animals were allowed to recover for at least 1 week before contact lens wear. ¹⁸  

Lens Fitting

As previously reported, O₂-nontransmissible polymethylmethacrylate (PMMA) rigid contact lenses (Dk = 0) specially designed for the rabbit cornea (diameter, 14.0 mm; thickness, 0.15 mm) were used to induce maximum contact lens-associated hypoxic damage to the corneal epithelium. ¹⁸ The best-fitting base curve was selected after trial fitting with radii of 7.60, 7.80, 8.00, and 8.20 mm, by using fluorescein and blue light. PMMA contact lenses were fitted on the right eye of each rabbit; the left eye served as the control. After 24 hours of lens wear, animals were anesthetized and the contact lenses removed. The rabbits were killed humanely with pentobarbital (120 mg/kg). Both eyes were then enucleated immediately and used for subsequent studies. 

Human Corneal Epithelial Cell Culture

Human corneal epithelial cells immortalized with hTERT (hTCEpi) were prepared as described previously. ²¹ Briefly, human corneas were obtained from the Tissue Transplant Services Lion’s Eye Bank (Dallas, TX). Human corneal epithelial cells were isolated from the limbal region by dispase II digestion (1.2 U/mL) and grown in keratocyte growth medium (KGM)-2 (Clonetics-BioWhittaker, Inc., Walkersville, MD). Primary or first-passaged cells at 50% to 60% confluence were transfected with hTERT vectors containing a puromycin-resistant gene construct, as previously described. ²² Cells were selected with puromycin and then cloned. Cloned cells were maintained in KGM-2 medium containing 0.15 mM Ca²⁺ and were routinely passed at 50% to 60% confluence. 

For β-cholera toxin staining and the cytotoxicity assay, cells (5 × 10⁴) were grown on glass coverslips (12 mm in diameter) coated with collagen (Vitrogen; Cohesion Technologies Inc., Palo Alto, CA) in 24-well culture plates (Corning Inc., Corning, NY) using KGM-2 medium containing 0.15 mM Ca²⁺ (2.5 × 10⁴ cells/cm²). Thereafter, cells were incubated at 37°C in 5% CO₂ overnight and fed with antibiotic-free keratocyte basal medium (KBM)-2 (Clonetics-BioWhittaker, Inc.), followed by incubation for 24 hours at 37°C in 5% CO₂. For quantification of bacterial internalization, cells (1 × 10⁵) were seeded into 12-well culture plates (2.5 × 10⁴ cells/cm²; Corning Inc.) in KGM-2 and incubated at 37°C in 5% CO₂ for 2 days. 

PA Infection in Rabbit Cornea and Human Corneal Epithelial Cells

For rabbit corneas, normal and contact-lens wearing–eyes were enucleated and placed, corneal surface up, in 12-well culture plates (Corning, Inc.). As reported previously, ¹⁸ a silastic tube was placed on the eye that exposed the entire corneal surface up to the limbus. Test strains of PA (0.5 mL, 1 × 10⁸ CFU/mL) were then layered onto the corneal surface, and the culture plates were incubated at 37°C for 30 minutes. Eyes were then washed with MEM three times, followed by incubation at 37°C for an additional 30 minutes to allow bacterial internalization. 

For cultured human corneal epithelial cells, cells were washed with antibiotic-free MEM once and then infected with 2.5 × 10⁷ CFU (coverslips) or 5 × 10⁷ CFU of PA (12-well culture plate) for various times (cell-to-bacteria ratio, 1:500). ¹⁷ For quantification of bacterial internalization, after incubation with PA for 30 minutes, cells were washed with MEM three times, followed by incubation at 37°C for an additional 30 minutes to allow bacterial internalization. For β-cholera toxin staining, cells were continuously infected with PA for 0, 15, and 60 minutes. 

β-Cholera Toxin Staining

Lipid raft formation was assessed by staining with the FITC-conjugated β-subunit of cholera toxin (β-CT; Sigma-Aldrich St. Louis, MO) known to bind the lipid raft component, ganglioside GM1. ²³ After incubation with PA for the indicated times, corneas and cells were washed with cold MEM three times and fixed in 1% paraformaldehyde in MEM (pH 7.2) on ice for 30 minutes. After they were washed with cold MEM twice, corneas and cells were stained with 10 μg/mL β-CT for 60 minutes on ice and counterstained with propidium iodide (PI; 1 μg/mL) for 5 minutes at room temperature to visualize cell nuclei and bacteria. Corneas were then placed epithelium-down on a mylar Petri dish (Backhofer GmbH, Rütlingen, Germany) and examined with a laser scanning confocal microscope (model SP2; Leica, Heidelberg, Germany). Coverslips were placed upside down on glass slides and examined by laser scanning confocal microscopy. 

Quantification of Bacterial Internalization

After eyes and cells were infected with PA test strains, internalization of bacteria was quantified by gentamicin survival assay. Briefly, eyes and cells were exposed to the cell-impermeable antibiotic, gentamicin (200 μg/mL) in MEM at 37°C for 2 hours, to kill extracellular bacteria, and then were washed with MEM three times. Thereafter, from rabbit eyes, the entire corneal epithelium was collected by scraping with a Lance no. 10 surgical blade. The scraped rabbit epithelium or cultured cells were then incubated in 5 mg/mL saponin-MEM for 15 minutes at room temperature. The number of viable bacteria was subsequently quantified by serial dilution of cell lysate and culture on Mueller Hinton II agar plates. Colony forming units (CFU) were determined 20 hours after plating. 

Inhibition of PA Internalization in Cultured Corneal Epithelial Cells

We confirmed the role of lipid rafts in the internalization of three PA test strains by treating cultured epithelial cells with cholesterol-depleting (methyl-β-cyclodextrin [MβCD]) or -binding (nystatin, filipin) molecules that disrupt the function of lipid rafts. This method has been commonly used to confirm the involvement of lipid rafts (cholesterol) in various cellular functions, as well as PA internalization. ^{24 25 26}  

Cells were grown in 12-well plates and pretreated with various concentrations of each inhibitor in KGM-2 medium at 37°C for 60 (MβCD) or 30 (nystatin, filipin) minutes before infection. Cells were then infected with PA in the presence of the same inhibitors. ¹⁷ As a control, three empty wells were inoculated with the bacterial suspension. After infection for 30 minutes, cells were washed with MEM three times and incubated in inhibitors containing KGM-2 medium for an additional 30 minutes. Thereafter, bacterial internalization was quantified. At least three wells were used for each condition, and all experiments were repeated at least twice. 

Cytotoxicity of Cholesterol Metabolism Inhibitors

The cytotoxicity of cholesterol metabolism inhibitors was determined using a viability assay (Live/Dead Viability/Cytotoxicity Kit; Molecular Probes, Inc., Eugene, OR). Cells were incubated with various concentrations of each inhibitor at 37°C in 5% CO₂ for 2 hours (n = 3). After inhibitors were removed, cells were washed with MEM and stained with 2 μM Calcein AM and 4 μM ethidium homodimer (EthD-1) for 30 minutes at room temperature. Cells were washed with MEM and examined by fluorescence microscopy. To quantify cellular viability, we counted and averaged the number of live and dead cells in four randomly selected fields of each coverslip. 

The effects of cholesterol metabolism inhibitors on PA viability and PA growth were also investigated. PA (1 × 10⁸ cells/mL) was incubated with each inhibitor at 37°C for 1 hour. The PA viability was then measured by staining bacteria (Live/Dead BacLight Bacterial Viability Kit; Molecular Probes, Inc.) and counting live and dead bacteria through a microscope. Bacterial growth was also assayed by serial dilution of suspension and culture on agar plates. 

Flow Cytometry

To confirm the specific interaction of PA with lipid rafts, PA was complexed with lipid raft extracts from hTCEpi cells and FITC-conjugated β-CT, and then analyzed by flow cytometry (FACSCalibur; BD Biosciences, San Diego, CA). ¹⁷ Specifically, lipid rafts were extracted from hTCEpi cells (2 × 10⁶ cells) by first lysing the cells with 1% Triton X-100 in PBS containing 10 μg/mL aprotinin and 10 μg/mL leupeptin at 4°C for 30 minutes, followed by dilution with ice-cold PBS to 0.01% Triton X-100. Cell lysates were then centrifuged at 900g for 10 minutes at 4°C to remove nuclei and debris and then were incubated with PA suspension (10⁸ CFU) at 4°C for 1 hour with mild agitation. Bacteria were then collected by centrifugation, resuspended in PBS and incubated with FITC-conjugated β-CT. After the cells were washed with PBS three times, the FITC-conjugated β-CT/bacteria complex was analyzed by flow cytometry, using forward- and side-scatter gates to eliminate measurements from debris. PA or cell lysates alone incubated with FITC-conjugated β-CT were used as the negative control. 

Statistics

Differences between treatment groups were evaluated on computer by Student’s t-test, paired Student’s t-test, and one-way analysis of variance (ANOVA; SigmaStat for Windows; SPSS Inc., Chicago, IL). 

Results

β-Cholera Toxin Staining of Rabbit Corneas

In normal eyes, FITC-labeled β-CT showed no staining on the surface of central and peripheral corneal epithelial cells, with and without PA infection (Fig. 1A , data for PA-infected cells not shown). However, β-CT uniformly stained surface epithelial cells at the limbus and over the conjunctiva, with some cells showing more intense surface staining than others. After overnight PMMA lens wear, numerous superficial central corneal epithelial cells showed punctate β-CT staining suggesting formation of lipid rafts (Fig. 1B , central). In the peripheral cornea, fewer surface epithelial cells appeared to be stained with β-CT compared with the central cornea (Fig. 1B , peripheral). Within the limbus and conjunctiva, β-CT staining remained uniform except for occasional punctate staining regions within a few cells that appeared smaller and more localized to the cell periphery compared to the central corneal epithelium (Fig. 1B , conjunctiva and limbus, arrows). 

Staining of normal corneas after PA ATCC27853 or strain 6294 exposure for 1 hour failed to identify any binding of PA to either the corneal or limbal surface epithelial cells (Figs. 2A 2B) . Furthermore, no β-CT staining was detected in normal corneas, even after a 1-hour exposure (Fig. 2A) , whereas noninfected corneas from PMMA-lens–wearing eyes continued to show prominent punctuate staining with β-CT in isolated surface corneal epithelial cells (Fig. 2C , arrows) (data not shown for 6294). Subsequent infection with PA for 30 minutes of PMMA lens-wearing eyes showed preferential binding of PA to cells that stained with β-CT (Fig. 2D) ; however, initial PA binding was not confined to regions of β-CT staining but appeared to be diffuse over the cell surface (Fig. 2D , arrowheads). After 1 hour of infection, PMMA-lens–wearing eyes showed clustering of PA preferentially to regions that stained intensely for β-CT (Fig. 2Exy, arrow). In other regions of the cell that did not stain for β-CT, PA remained sparsely bound to the cell surface (Fig. 2Exy, arrowhead); however, the large β-CT-stained domains that included clusters of PA appeared to show evidence of bacterial internalization with extension of bacteria into the cell (Fig. 2Exz, arrow). Limbal epithelium from PMMA-lens–wearing eyes showed no binding of PA or aggregated β-CT staining after 1 hour of infection (Fig. 2F) , even though punctate regions over the cell surface stained with β-CT (arrowheads). 

Bacterial Internalization in Rabbit Corneas

In previous studies of PA internalization in the cornea, the nonionic detergent Triton X-100 (0.25% or 0.5%) was used to lyse corneal epithelial cells. ^{16 27} However, the PA strain used in this study was killed by Triton X-100 (data not shown). To avoid this problem, saponin (5 mg/mL) was used to ensure lysis of epithelial cells after PA internalization. Saponin was found not to affect the colony-forming ability of PA at the test concentration used (data not shown). 

Figure 3shows the number of internalized bacteria in rabbit corneal epithelium after 24-hour PMMA lens wear followed by 1-hour PA ATCC27853 infection. PMMA lens wear significantly increased bacterial internalization compared with control eyes (P = 0.025, paired Student’s t-test). These data support the interpretation that PMMA lens wear enhances the exposure of surface corneal epithelial cells capable of forming lipid rafts that bind and internalize PA through lipid raft aggregation, as shown in Figure 2 . 

β-Cholera Toxin Staining of Cultured Corneal Epithelial Cells

To investigate further the role of lipid rafts in PA internalization, hTCEpi cells were used as a model. In uninfected cells (Fig. 4A) , β-CT diffusely stained the entire plasma membrane surface, the cell–cell junctions Fig 4Axy, arrow) and intracellular transport vesicles (Fig. 4Axz, arrowheads). Infection with PA strain ATCC27853 for 15 minutes (Fig. 4B)showed colocalization of bacteria with focal regions of β-CT staining (Fig. 4B , arrows). XZ projection through the cell (Fig. 4B , between asterisks) showed internalization of PA at sites of β-CT aggregation (Fig. 4B , xz, arrowhead). After extended exposure to PA strain ATCC27853 (60 minutes; Fig. 4C ), β-CT stained larger focal regions of the plasma membrane that extended intracellularly (Fig. 4C , arrows) and contained multiple clusters of PA, suggesting bacterial internalization (Fig 4Cxz, arrowheads). In addition, β-CT-positive regions appeared to enlarge progressively and contained more bacteria by 60 minutes after PA infection (Fig. 4C) . Although a small number of bacteria were also observed in β-CT-negative regions, laser scanning confocal microscopy confirmed that these bacteria bound to the surface of cells and were not internalized. We observed similar colocalization of β-CT with bacteria and β-CT aggregation after infection with corneal isolate strains 6294 and 6487 (data not shown). Taken together, these results indicate that membrane lipid rafts are necessary for in vitro bacterial internalization by hTCEpi cells, in agreement with in vivo rabbit corneal epithelial studies. 

Inhibition of PA Internalization in Cultured Corneal Epithelial Cells

To confirm the role of lipid rafts in PA internalization, lipid raft formation was disrupted in each of the three test PA strains by using three inhibitors of cholesterol metabolism (MβCD, nystatin, and filipin). MβCD is effective in removing cholesterol from the plasma membrane of cells while allowing cells to retain their intact structure and viability. ^{24 28} Nystatin and filipin disrupt lipid raft function by precipitating cholesterol in the plasma membrane of the cell. ^{25 26}  

All three inhibitors blocked the formation of lipid rafts after 1 hour of exposure to PA (Fig. 4D , filipin; data for MβCD and nystatin not shown). PA appeared to remain bound to the surface of inhibitor-treated cells (Fig. 4Dxz, arrowhead) but no large lipid raft aggregations were identified. Quantification of PA internalization showed that all inhibitors significantly decreased the internalization of each test PA strain in a dose-dependent manner (Fig. 5) . MβCD blocked bacterial internalization of PA strain ATCC27853 by 66% at 5 mM (P < 0.001, one-way ANOVA), nystatin by 63% at 50 μg/mL (P < 0.001), and filipin by 98% at 2 μg/mL (P < 0.001). The inhibitor concentrations found to be most effective for blocking PA internalization were generally similar to those previously reported for other microorganisms. ^{12 26 29} A large number of bacteria were found to adhere to empty wells, but all were killed by gentamicin (data not shown). Taken together, these results indicate that intact lipid raft structures are necessary for the internalization of PA into hTCEpi cells. 

Cytotoxicity of Cholesterol Metabolism Inhibitors

We also investigated whether cholesterol metabolism inhibitors affect cellular and bacterial viability and PA growth. Nystatin and filipin did not affect the viability of cells at any concentrations we used in the study (Table 1) . MβCD had a slight but significant effect on hTCEpi cell viability, increasing cell death by approximately 5% compared with the control (P = 0.031, Student’s t-test). However, bacterial internalization was inhibited by 66% for ATCC27853, 89% for strain 6294, and 86% for strain 6487 at 5 mM MβCD (Fig. 5) , which cannot be explained by the decrease of cellular viability. In addition, the number of cells attached to the tissue culture plate did not decrease after treatment with each of the three inhibitors (data not shown). 

We also investigated the effect of inhibitors on bacterial viability and bacterial growth. None of the three inhibitors tested affected the bacterial viability and growth of three PA strains at the concentrations used (Table 1 ; data for strains 6294 and 6487 not shown). These results exclude the possibility that these drugs block bacterial internalization through effects on PA viability and growth. 

Interaction of PA with Lipid Rafts

Interaction of PA with lipid rafts was confirmed by formation of PA/lipid raft/β-CT complexes, followed by flow cytometry. For hTCEpi cell lysates alone and PA alone stained with β-CT used as the negative control, <3% and <2% of all particles were FITC positive, respectively (Fig. 6yellow trace and red trace, respectively). Exposure of PA to cell lysate significantly increased fluorescence intensity of bacteria (47% for ATCC27853 and 43% for strain 6294; P < 0.05 and P < 0.05, respectively), indicating the binding of lipid rafts to PA (Fig. 6 , blue trace). Comparable results were obtained with corneal isolate, strain 6294 (histogram not shown). 

Discussion

Previous clinical studies have shown that contact lens wear greatly increases the risk for PA infection in the cornea, ^{14 15} and that PA remains the most prevalent causative agent worldwide for contact-lens–related infectious keratitis over the past several decades. ^{15 30 31} In attempting to understand the complex pathophysiological mechanisms that mediate contact-lens–related PA corneal infections, both animal models and human clinical studies have shown that contact lens wear significantly increases the detectable binding of PA ATCC27853 to rabbit corneal epithelium ^{18 32} and to corneal epithelial cells exfoliated from patients exposed to contact lens use. ¹⁹ Furthermore, studies have established that PA ATCC27853 binding to corneal epithelial cells is inversely dependent on contact lens oxygen transmissibility (P < 0.01), ¹⁸ with the greatest binding observed with low-oxygen-permeable lens wear. Based on this past knowledge, we used an oxygen-nontransmissible PMMA rigid test lens (transmissibility, 0) to produce maximum hypoxic exposure of PA binding sites on the rabbit corneal epithelium in vivo and showed for the first time that in vivo overnight contact lens wear induces the appearance of a unique population of surface corneal epithelial cells capable of forming lipid rafts, irrespective of bacterial exposure, that preferentially bind PA. 

The results presented in these in vivo and in vitro studies also confirm and, for the first time extend to the corneal epithelium, previous reports establishing a requirement of lipid raft formation for the internalization of PA, as has recently been shown in human nasal epithelial cells (in vitro), murine tracheal epithelial cells (in vivo), Chang conjunctival epithelial (WI-38) cells, and murine lung fibroblasts. ¹⁷ Furthermore, in all cell and organ systems thus far studied, including the cornea, lipid raft disruption by cholesterol metabolism inhibitors prevents PA internalization in a concentration-dependent manner. The present study provided the important finding that rabbit corneal epithelium does not exhibit lipid raft formation as detected by β-CT binding in normal corneas irrespective of PA exposure. Only after PMMA lens wear does the central corneal epithelium show a punctate staining pattern with β-CT, suggestive of initial lipid raft formation. Furthermore, subsequent exposure of the lens-wearing corneal surface to PA then results in a dynamic process of PA internalization characterized by preferential PA binding to β-CT-stained cells. It should be noted that initial PA binding did not appear to be directly associated with sites of β-CT staining during the first 30 minutes, suggesting that lipid rafts induced by PMMA lens wear do not initially contain PA binding sites. However, continued infection from 30 minutes to 1 hour is associated with clustering of PA at sites of enlarged lipid raft platforms, suggesting that surface-bound PA initiates lipid raft aggregation leading to PA internalization. In support of this conclusion, PA clustering and internalization was not detected by direct observation at any sites other than those stained by β-CT. In addition, PA clustering at sites of lipid raft formation detected in cultured human corneal epithelial cells was effectively blocked by treatment with cholesterol metabolism inhibitors that are known to disrupt lipid rafts. Finally, PA internalization was significantly reduced or completely blocked by cholesterol metabolism inhibitors in cultured human corneal epithelial cells. Overall, these findings with three infectious strains of PA including two corneal isolates suggest that lipid-raft–mediated bacterial internalization is a common mechanism for infection among invasive PA strains. Although these data are supportive of the hypothesis that lipid rafts play a role in PA internalization, further work is necessary to identify more precisely and establish the role of the receptor molecules involved in PA binding and lipid raft formation. Nevertheless, it is important to reemphasize that the normal corneal epithelial surface did not appear to exhibit lipid rafts in vivo or bind PA unless first subjected to contact lens wear. 

An additional important finding was the observation that β-CT diffusely stained normal limbal corneal epithelium and adjacent conjunctival epithelium but did not bind PA ATCC27853, even after PMMA lens wear. This new finding suggests that although ganglioside GM-1-expressing membrane regions may form aggregated lipid rafts, neither limbal corneal or conjunctival epithelial cells appear able to bind PA. This latter observation is consistent with known clinical observations that infectious keratitis associated with contact lens wear targets the central and paracentral corneal epithelium and not the limbus or conjunctiva. Whether receptors for PA are absent or masked in limbal and adjacent conjunctival epithelial cells requires further study. 

It should be noted that although PA ATCC27853 internalization was observed only in association with β-CT-stained lipid rafts, a gentamicin survival assay data demonstrated that normal corneal epithelial cells that did not stain with β-CT had significant PA internalization—30% of that detected in PMMA lens-exposed eyes. Thus, although lipid raft formation was not observed in normal eyes, it is possible that lipid raft platforms may have disappeared during the 30-minute incubation conditions used in our current assay or were undetected because of their smaller size. In support of this possibility, previous studies with other cells have shown that lipid raft aggregation can occur as early as 5 to 10 minutes after PA exposure and can disappear over the next 30 minutes. ¹⁷ Conversely, PA in the normal cornea may be internalized through alternative pathway(s) other than lipid rafts detected by β-CT staining. Although the existence of an alternative pathway is speculative, it should be noted that PA binding and internalization in the normal versus contact-lens–exposed cornea may be different, since only the latter appears to lead to infection. 

It is also important to note that a discrepancy in β-CT staining pattern was observed between in vivo and in vitro studies. Specifically, β-CT staining was not observed in normal rabbit corneas before PMMA lens wear, whereas β-CT showed diffusely weak staining over the plasma membrane in non–air-lifted cultured hTCEpi cells. This difference may be explained by the fact that hTCEpi cells were cultured in KGM-2 medium containing 0.15 mM Ca²⁺, in which cells maintain a proliferative, basal epithelial phenotype and do not differentiate to a stratified squamous epithelial layer. ²¹ Thus, the apparent difference in β-CT staining between in vivo and in vitro is most likely related to the lack of complete differentiation of hTCEpi cells under submerged, low Ca²⁺ culture conditions. Further study is needed to establish whether cultured hTCEpi cells under differentiation conditions (i.e., high calcium), show a similar loss of plasma membrane lipid raft formation as observed in vivo. 

In general, formation of lipid rafts is believed to initiate signaling by concentrating a low number of membrane-associated proteins into larger, consolidated regions of high local concentration over time. Thus, localization of specific proteins to lipid rafts is often used to postulate that a unique candidate target protein is essential for initiating the signaling process. Recently, cystic fibrosis transmembrane conductance regulator protein (CFTR), a fairly ubiquitous chloride channel protein found apically in many surface epithelia, has been shown to localize in lipid rafts and to be essential for PA internalization in mammalian nasal, tracheal, lung, and corneal epithelial cells. ^{17 33} A portion of the first predicted extracellular loop of CFTR, composed of amino acids 108-117, recognizes the conserved outer core oligosaccharide of the lipopolysaccharide (LPS) in the outer membrane of PA, and subsequently PA is translocated into mouse respiratory epithelium ³⁴ as well as cultured rabbit and human corneal epithelial cells. ³⁵ Mutations in CFTR or LPS or associated flagellum assembly apparatus have also been shown to reduce PA internalization. ^{34 35 36 37 38} In addition, recent studies have shown that CD95 clusters in lipid rafts after PA infection and that this clustering is essential for host defense against PA infection. ^{17 39} Furthermore, studies of host proteins have shown that actin cytoskeleton and protein tyrosine kinase are also involved in PA internalization. ^{27 40} Additional studies have further suggested that specific phospholipids (phosphatidylserine [PS] and phosphatidylinositol [PI]) present in mucus or on the surface of the corneal epithelium may also function as PA receptors and contribute to selective bacterium–host interactions. ⁴¹ Taken together, all these findings suggest that PA binding to corneal epithelium is a complex process, in which the host protein(s) associated with subsequent bacterial internalization in the cornea remain incompletely characterized. 

Overall, the present study shows that lipid rafts appear to be necessary for the internalization of PA in corneal epithelial cells in vivo and in vitro and that PA infection appears to trigger the reorganization of lipid rafts into large platforms at the sites of cell–bacteria interaction. These large lipid raft platforms may concentrate as yet unrecognized binding receptors and intracellular signaling molecules involved in the cellular response to PA infection in addition to CFTR, PS, and PI. The study is also the first to report the lack of lipid rafts in the normal corneal epithelium with induction by contact lens wear but not by exposure to PA alone. This latter finding suggests a unique role of contact lens wear in the pathogenesis of contact lens-mediated PA infectious keratitis. 

Because lipid rafts can be separated from the other membrane components by ultracentrifugation in sucrose density gradients, ^{33 42} detailed analysis of isolated lipid rafts may provide additional information about other important molecular factors involved in PA internalization in corneal epithelial cells. Clearly, further study of lipid raft formation and the detailed analysis of lipid raft components will help in the understanding of the pathogenesis of PA bacterial keratitis. 

 Supported in part by a grant from the Center for Alternatives to Animal Testing, Baltimore, Maryland (JVJ); National Eye Institute Grant EY10738; and The Pearle Vision and Chilton Foundations, Dallas, Texas (HDC); Senior Scientist awards, Research to Prevent Blindness, Inc. (JVJ, HDC); an Olga Keith Weiss Scholar award (WMP); and unrestricted grants from Menicon Co., Ltd., Nagoya, Japan, and Research to Prevent Blindness, Inc.

 Submitted for publication May 14, 2004; revised August 27 and December 7, 2004; accepted December 20, 2004.

 Disclosure: Na. Yamamoto, Menicon Co., Ltd. (E); No. Yamamoto, Menicon Co., Ltd. (E); M.W. Petroll, None; H.D. Cavanagh, Menicon Co., Ltd. (C, F); J.V. Jester, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Figure 1.

 

Rabbit surface corneal epithelium from normal (A) and 24-hour PMMA-lens–wearing (B) eyes stained with β-CT (green) and PI (red). In normal eyes, β-CT showed no staining on the surface of central and peripheral corneal epithelial cells. However, β-CT uniformly stained surface epithelial cells at the limbus and conjunctiva. After overnight PMMA lens wear, numerous central corneal epithelial cells and a fewer peripheral epithelial cells showed punctate β-CT staining, suggesting formation of lipid rafts. Within the limbus and conjunctiva, β-CT staining remained uniform expect for occasional punctate staining regions within a few cells (B, conjunctiva and limbus, arrows).

Figure 2.

 

Normal (A, B) and PMMA-lens–wearing eyes (C–F) infected with PA ATCC27853 and stained with β-CT (green) and PI (red). (A) Central cornea and (B) limbal epithelium of a normal eye, 1 hour after PA infection; central cornea of a PMMA-lens–wearing eye (C) before, (D) 30 minutes after, and (E) 1 hour after PA infection; and (F) limbal epithelium of PMMA-lens–wearing eye, 1 hour after PA infection. β-CT stained the central cornea from only contact-lens–wearing eyes, irrespective of bacterial infection (C, arrows), and PA adhered preferentially to β-CT-stained cells (D, E, arrowheads) in the central cornea but not at the limbus (F). Furthermore, 1-hour PA infection resulted in large β-CT-stained plaques (E, arrow) containing aggregated PA that extended into the cells. (E xz): cross-section between asterisks in (E xy).

Figure 3.

 

Bacterial internalization in rabbit corneas after 24-hour PMMA lens wear. The internalization of PA ATCC27853 was significantly increased compared with that in the control eye (paired Student’s t-test).

Figure 4.

 

Cultured human corneal epithelial cells stained with β-CT (green) and counterstained with PI (red). (A) Control (noninfected); (B) infected with PA strain ATCC27853 for (B) 15 or (C) 60 minutes; (D) filipin-treated cells infected with PA strain ATCC27853 for 60 minutes. In uninfected cells (A), β-CT diffusely stained the plasma membrane surface, cell–cell junctions (A xy, arrow) and intracellular transport vesicles (A xz, arrowhead). After 15 (B) and 60 (C) minutes of PA infection, β-CT stained large focal regions (arrows) of the plasma membrane that extended intracellularly and contained PA (xz, arrowheads) suggesting internalization. In filipin-treated cells (D), the formation of lipid rafts and bacterial internalization were blocked. (A–D xz) cross-sectional slices between asterisks.

Figure 5.

 

Effect of cholesterol metabolism inhibitors on bacterial internalization in cultured corneal epithelial cells. Strains (A) ATCC27853, (B) 6294, and (C) 6487. All inhibitors significantly blocked the bacterial internalization in a dose-dependent manner (*P < 0.05; **P < 0.01; one-way ANOVA).

Table 1.

 

Effect of Cholesterol Metabolism Inhibitors on Cellular Viability and Bacterial (ATCC27853) Viability and Growth

	Cell Viability			PA Viability			PA Growth
		Avg.	SD	Avg.	SD	Avg.	SD
Control	98.3	0.5	98.1	1.7	100.0	10.1
MβCD (5 mM)	93.6*	2.3	98.8	0.9	106.4	15.1
Nystatin (50 μg/mL)	98.9	0.3	99.1	1.7	111.7	15.5
Filipin (2 μg/mL)	98.9	0.5	98.5	1.0	113.3	13.6

 Data are expressed as the mean percentage ± SD.

*   P = 0.031.

Figure 6.

 

Interaction of PA with lipid rafts: PA alone incubated with β-CT (red); cell lysate alone incubated with β-CT (yellow); and PA incubated with β-CT after exposure to cell lysate (blue). The table shows the percentage of PA that bound to GM1 (under the M2 bar, in histograms). Exposure of PA to cell lysate significantly increased fluorescence intensity, suggesting the binding of the raft component GM1 to PA. The flow cytometry plot is representative of results in at least three experiments.

The authors thank Erich Gulbins (University of Essen, Germany) for providing details of his flow cytometry protocol and for much helpful advice, and Christina Stevens (University of Texas Southwestern Medical Center) for assistance in performing the flow cytometry studies.