Characterization of Pseudomonas aeruginosa-Induced MDCK Cell ...

INFECTION AND IMMUNITY, Apr. 1995, p. 1541–1551 0019-9567/95/$04.0010 Copyright q 1995, American Society for Microbiology

Vol. 63, No. 4

Characterization of Pseudomonas aeruginosa-Induced MDCK Cell Injury: Glycosylation-Defective Host Cells Are Resistant to Bacterial Killing GERARD APODACA,1* MORGANE BOMSEL,1† RAGNAR LINDSTEDT,1‡ JOANNE ENGEL,2 DARA FRANK,3 KEITH E. MOSTOV,1,4,5 AND JEANINE WIENER-KRONISH2,6 Department of Anatomy,1 Department of Biochemistry and Biophysics,4 Department of Anesthesiology,6 Department of Medicine,2 and Cardiovascular Research Institute,5 University of California, San Francisco, San Francisco, California 94143, and Department of Microbiology, Medical College of Wisconsin, Milwaukee, Wisconsin 532263 Received 16 September 1994/Returned for modification 24 October 1994/Accepted 8 December 1994

As a model for bacterium-induced epithelial cell injury, we have studied the interaction of Pseudomonas aeruginosa with polarized Madin-Darby canine kidney (MDCK) cells grown on filters. Following an initial period of bacterial adhesion, foci of injured host cells, which consisted of a central region of cell debris, surrounded by cells that were permeable and apparently necrotic, were formed. Host cell death was quantified by measuring the increased permeability of the monolayer to the macromolecular tracer [14C]inulin. Using this MDCK model system, we have identified bacterial and host cell factors necessary for the host cell damage. The ability of P. aeruginosa to cause MDCK cell damage was independent of elastase or exotoxin A production. In contrast, bacteria with a mutation in the regulatory locus exsA (which are deficient in exoenzyme S production) neither bound to nor caused host cell injury. MDCK cells with defects in cell surface glycosylation were resistant to cell injury, indicating that bacteria may require host cell glycolipids and/or glycoproteins as points of adhesion to cause subsequent host cell injury. sible for the lung epithelial injury, we used parental strains of P. aeruginosa with mutations in exoproducts thought to be important in P. aeruginosa pathogenesis. We found that the parental strains of P. aeruginosa that produced exoenzyme S caused more alveolar epithelial injury than the strains that did not produce this toxin (23). Exoenzyme S is a toxin that ADP ribosylates certain small GTP-binding proteins (10–12). A bacterium with a regulatory mutation that blocks production of exoenzyme S (exsA::V) and possibly of other proteins coregulated with exoenzyme S did not cause acute injury of the alveolar epithelium (23). Other isogenic P. aeruginosa strains, including ones that did not produce elastase or exotoxin A (a protein synthesis inhibitor), did not influence the alveolar epithelial injury produced by the parent bacteria (23). These results suggest that exoenzyme S or other molecules coregulated with exoenzyme S are important contributors to acute lung injury. Previous in vitro studies have focused almost entirely on the adhesion of P. aeruginosa to mammalian cells and have largely not examined other (presumably subsequent) processes, such as injury of host cells (3, 15, 18, 22, 26, 27, 31, 33–36, 40–42, 45). To investigate the events subsequent to adhesion, we have analyzed the interaction between P. aeruginosa and MadinDarby canine kidney (MDCK) cells. These ciliated cells have been widely used for studies in epithelial cell biology, because when grown on permeable filters they form a well-polarized epithelial monolayer, essentially reconstituting a simple epithelial tissue. We reasoned that filter-grown MDCK cells might therefore be a useful in vitro model for mimicking the interaction of P. aeruginosa with epithelial cells in vivo. We found that adding P. aeruginosa to the apical surface of MDCK cells resulted in killing of the host cells and concomitant loss of the tight epithelial barrier to diffusion. We examined a variety of strains of P. aeruginosa and discovered that there was a strong correlation between the virulence of various strains of bacteria in the animal models of acute pneumonia and the in vitro

Pseudomonas aeruginosa is among the most virulent opportunistic pathogens of humans, colonizing and invading injured epithelial surfaces (including the lungs of intubated patients, corneal abrasions, and injured skin of burn patients). Seventyfive percent of all intensive care unit patients are colonized by this organism, and not surprisingly, P. aeruginosa is the leading cause of nosocomial pneumonia (7, 19). In addition, a principal cause of mortality among cystic fibrosis patients is colonization of their lungs by mucoid strains of P. aeruginosa. The treatment of P. aeruginosa infections is difficult, in part because antibiotic resistance is common and becoming more widespread in this organism (7, 19). A better understanding of the mechanisms of Pseudomonas virulence could lead to new strategies for treatment. To investigate the interaction between P. aeruginosa and the lung, we have previously developed a quantitative model of acute lung injury (44). Live P. aeruginosa organisms were instilled via the trachea into the lungs of experimental animals, including sheep, rabbits, or rats (23, 30, 43). The bacteria caused a dose- and time-dependent injury to the lung, which was manifested by (i) increased bidirectional fluid and protein movement across the alveolar barrier, (ii) histologic evidence of damaged and necrotic alveolar cells, and (iii) dissemination of bacteria beyond the air spaces of the lung (30). Different parental strains of P. aeruginosa varied in their ability to cause acute injury when instilled in the lungs of anesthetized rabbits (strain PA103 . strain PAK . strain PAO1). To define the exoproducts of P. aeruginosa that were respon-

* Corresponding author. Mailing address: Department of Anatomy, S-1372, University of California, San Francisco, CA 94143-0452. Phone: (415) 476-6047. Fax: (415) 476-4845. † Present address: Institut Cochin de Genetique moleculaire, Paris, France. ‡ Present address: Department of Immunology, Scripps Institute, La Jolla, CA. 1541

1542

APODACA ET AL.

killing of the MDCK cells. Thus, the MDCK system is a valid in vitro model for important aspects of P. aeruginosa pneumonia. Finally, we found that two mutant strains of MDCK cells that are defective in glycosylation were much less susceptible to killing by P. aeruginosa, presumably because host cell glycolipids and/or glycoproteins are important in the adhesion of the bacteria to the MDCK cells. MATERIALS AND METHODS Bacteria and growth conditions. Three parental strains of P. aeruginosa were utilized. Strain PA103 has been described elsewhere (20). Strain PAK was generously provided by Stephen Lory (University of Washington, Seattle), and strain PAO1 was generously provided by Barbara Iglewski (University of Rochester, Rochester, N.Y.). Two isogenic strains of PA103 were also investigated. PA103tox::V (generously provided by Barbara Iglewski) harbors a mutation in the structural gene for exotoxin A (32), and PA103exsA::V (14) is a regulatory mutant defective in production of the exoenzyme S toxin. P. aeruginosa parental and mutant strains were stored as bacterial stocks at 2708C in 10% sterile skim milk solutions. The bacterial stocks were inoculated into a deferrated dialysate of trypticase soy broth supplemented with 10 mM nitrilotriacetic acid (Sigma Chemical Co.), 1% glycerol, and 100 mM monosodium glutamate and grown at 338C for 13 h in a shaking incubator (29). At this time the cells were in early stationary phase. Cultures were centrifuged at 8,500 3 g for 10 min, and the bacterial pellet was washed once with minimal essential medium (Sigma, St. Louis, Mo.) supplemented with 0.6% (wt/vol) bovine serum albumin (MEM/ BSA). Following centrifugation, the bacteria were resuspended in MEM/BSA and diluted to the appropriate number of CFU per milliliter and 200 ml of the suspension was added to the apical surface of host cells cultured on 12-mmdiameter Transwells (Costar, Cambridge, Mass.). The number of bacteria added to the cells was confirmed by serial dilution of the bacterial suspension and subsequent culture. MDCK cell culture. MDCK strain II cells were maintained in MEM (obtained from the University of California, San Francisco Cell Culture Facility) supplemented with 10% fetal bovine serum (Hyclone, Logan, Utah), 100 U of penicillin per ml, and 100 mg of streptomycin per ml in 5% CO2–95% air. Cells were routinely maintained on 10-cm-diameter culture dishes and were split 1:10 and passaged once weekly. For experiments, cells were cultured on 12-mm-diameter, 0.4-mm-pore-size Transwells as described previously (6) and used 3 or 4 days postculture. The RCAr cell line was kindly provided by E. Rodriguez-Boulan (Cornell University, New York, N.Y.), and the ConAr cells were provided by M. Lisanti (Whitehead Institute, Cambridge, Mass.). These cell lines were cultured as described above for the wild-type MDCK cell line. Fixation and fluorescent labeling of cells. Samples were fixed with paraformaldehyde by using a pH shift protocol and stained with propidium iodide or primary and secondary antibodies as described previously (2). The mouse hybridoma RR1 was cultured as described previously (16), and undiluted hybridoma supernatant was mixed with saponin (Sigma) to a final concentration of 0.025% and used to label E cadherin. RR1 was subsequently detected by using 10 mg of goat anti-mouse antibodies conjugated to Texas Red (Jackson ImmunoResearch Laboratories, West Grove, Pa.) per ml. Mouse monoclonal antibody specific for vimentin (Sigma, mouse clone V-9) was diluted 1:50 in MEM/BSA, and detected with 10 mg of goat anti-mouse antibodies conjugated to tetramethyl rhodamine isocyanate (TRITC). Scanning laser confocal analysis of fluorescently labeled cells. The samples were analyzed by using a krypton-argon laser coupled with a BioRad MRC600 confocal head, attached to an Optiphot II Nikon microscope with a 103 or Plan Apo 603 1.4 N.A. objective lens. The samples were scanned for Texas Red emission by using the K1 and K2 filter blocks. Collection parameters were as follows: zoom of 1.0 to 3.0, 0.5 s per scan, five frames per image, Kalman filter, motor step size of 1.0 mm, and diaphragm closed or set at 1/3 open. The data were analyzed by using Comos software. The images were converted to tagged information file format, and the contrast levels of the images were adjusted by using the Photoshop program (Adobe Co., Mountain View, Calif.) on a Macintosh IIci (Apple, Cupertino, Calif.). The contrast-corrected images were imported into Pagemaker (Aldus Corporation, Seattle, Wash.) and printed from an Agfa 9800 imagesetter at 2,400 dots per inch with a line screen of 150 lines per inch. Ultrastructural analysis. Bacterium-infected cells were fixed with 2.0% (vol/ vol) glutaraldehyde in a solution containing 200 mM Na cacodylate (pH 7.4), 1 mM CaCl2, and 0.5 mM MgCl2 for 30 min at room temperature. Samples were rinsed with 200 mM Na cacodylate (pH 7.4) and treated with a solution containing 1% OsO4 (wt/vol), 200 mM Na cacodylate (pH 7.4), and 1% (wt/vol) K4Fe(CN)6 for 90 min at 48C. After several rinses with H2O, the samples were stained en bloc overnight with 0.5% (wt/vol) uranyl acetate in H2O. Filters were dehydrated in a graded series of ethanol, embedded in the epoxy resin LX-112 (Ladd, Burlington, Vt.), and sectioned with a diamond knife (Diatome, Fort Washington, Pa.). Sections, pale gold in color, were mounted on Butvar-coated nickel grids, contrasted with lead citrate and uranyl acetate, and viewed at 60 kV in a Zeiss (Germany) EM-10 electron microscope.

INFECT. IMMUN. Preparation of bacterial culture supernatants, bacterial cell lysates, and killed bacteria. To investigate the toxicity of bacterial supernatants, a 100-ml culture of PA103 was centrifuged for 20 min at 25,000 3 g in a refrigerated RC-5 centrifuge (Sorvall) at 48C. The bacterial pellet was discarded, and the supernatant was collected. Because the bacterial culture broth itself caused cell damage, the proteins in the supernatant were precipitated with 75% ammonium sulfate. The resulting pellet was carefully collected and dialyzed in 4 ml of MEM/BSA supplemented with 100 U of penicillin per ml and 100 mg of streptomycin per ml. Two hundred microliters of this suspension (containing approximately 200 mg of bacterial proteins) was added to the apical surface of MDCK cells cultured on 12-mm-diameter Transwells. Additionally, a culture of PA103 grown overnight was collected, washed as described above, and PA103 (109 CFU) was grown for 6 h in MEM/BSA in the absence or presence of MDCK cells. The medium was centrifuged for 10 min at 25,000 3 g and filtered twice through a 0.2-mm-pore-size filter (Costar). When cultured, no bacteria were detected in this filtered medium. Two hundred microliters of this medium was applied to the apical surface of MDCK cells for assessment of bacterial production of toxic products before and after contact with host cells. Lysates of PA103 bacteria were made as follows. Portions (10 ml) of an early-stationary-phase culture of PA103 bacteria were harvested by centrifugation at 10,000 3 g for 10 min at 48C and washed twice in 10 ml of ice-cold 50 mM Tris-HCl buffer, pH 7.2. The cell pellet was suspended in 2 ml of cold lysis buffer (50 mM Tris-HCl, 20 mM EDTA, 10 mM benzamidine, 1 mg of leupeptin per ml) and lysed by passage through a French pressure cell. Unbroken cells and debris were removed by centrifugation at 14,000 3 g for 15 min at 48C, and the supernatant fraction was stored at 2708C until use in tissue culture assays for toxicity. Cell lysates were diluted 1:30 in 200 ml of MEM/BSA and added to MDCK apical surfaces. Lysis buffer alone had no effect on monolayer permeability. To heat kill PA103 cells, bacteria (109 CFU) were placed in a water bath at 608C for 60 min, resuspended in 200 ml of MEM/BSA, and added to cells for 6 h at 378C. Cultures of the heat-killed bacteria confirmed that no viable bacteria persisted after this treatment. Quantitation of Pseudomonas-induced host cell injury. [14C]inulin (ICN, Irvine, Calif.) (3.15 mCi/mg, 1 mCi/ml) was used to quantitate the extent of bacterium-induced damage to MDCK monolayers. The apical surfaces of MDCK cell monolayers were exposed to a bacterial inoculum, culture supernatant, or cell lysate for a given interval, and then the cells were washed three times in phosphate-buffered saline containing 0.9 mM CaCl2 and 0.5 mM MgCl2 to remove free bacteria or bacterial products. This wash was sufficient to remove the majority of nonadherent negative-control bacteria. [14C]inulin, (2 mCi), diluted in 500 ml of MEM/BSA, was added to the apical chamber of the Transwell and the cells were incubated for 1 h at 378C. The amount of [14C]inulin that remained in the apical well, traversed the filter to the basal well, or remained cell associated was quantitated in a liquid scintillation counter. The extent of monolayer damage was quantified by dividing the amount of [14C]inulin found in the basal chamber by the total [14C]inulin added to each sample. In the absence of cells, 38.0 6 1.0% of apically added [14C]inulin was found in the basal well of the Transwell. In contrast, in Transwells with a confluent monolayer of cells, only 0.10 6 0.03% of apically added [14C]inulin was collected in the basal well. Assessment of bacterial factors leading to cell cytotoxicity. To assess the effect of bacterial protein, DNA, and RNA syntheses on cell cytotoxicity, 5 3 106 CFU of PA103 per ml was treated with 30 mg of chloramphenicol per ml, 20 mg of naladixic acid per ml, or 32 mg of rifampin per ml for 15 min, and then 200 ml of the bacterial suspension was applied to the apical surface of MDCK cells for 3 h. The appropriate drug was included throughout the 3-h incubation period. Alternatively, 5 3 106 CFU of PA103 per ml was cooled to 48C, 200 ml was placed on the apical surface of MDCK cells bathed in ice-cold media, and the cells were kept at 48C for 3 h. Drug or cold treatment of MDCK cells in the absence of bacteria had no effect on host cell viability or permeability of the MDCK monolayer to [14C]inulin.

RESULTS P. aeruginosa induces epithelial cell injury. Initially, we determined what effect P. aeruginosa would have when allowed direct contact with the apical surface of MDCK cells. Bacteria (200 ml of 107 CFU of P. aeruginosa PA103 per ml; multiplicity of infection (MOI) of 1:1) were added to the apical surface of 2 3 106 MDCK cells cultured on 12-mm-diameter Transwell inserts and incubated at 378C for various amounts of time. The cells were gently washed to remove nonadherent bacteria and fixed, and the DNAs of both the MDCK cells and bacteria were stained with the fluorescent dye propidium iodide. Scanning laser confocal microscopy was used to analyze the threedimensional architecture of infected-cell monolayers. Following 1.5 to 2.0 h of incubation at 378C, aggregates of

VOL. 63, 1995

bacteria were found associated with the apical pole of one or more adjacent cells. A series of optical sections taken from a typical focus of bacterial adherence are shown in Fig. 1A to C. The bacteria (marked with an arrowhead in Fig. 1A) were found predominantly associated at the apex of the centralmost cell in Fig. 1A, but changes in adjacent cells were also observed. The nuclei of the three affected cells (indicated by arrows in Fig. 1A) were abnormal, were characteristically pyknotic or vacuolated, and stained brightly with the dye. Often, affected cells were swollen and their cytoplasm rose above the surrounding monolayer. The nuclei of adjacent cells in Fig. 1A to C were largely normal. The affected host cells were apparently randomly distributed across the monolayer. It is possible that these cells express increased amounts of a particular ligand for bacterial adhesion or in some other way differ from the great majority of MDCK cells which were not associated with bacteria. As the incubation period was increased, the adjacent cells become involved and the zone of cell death radiated outward. Figure 1D is a typical example of a monolayer of MDCK cells following a 3-h addition of bacteria. In this low-power overview of the monolayer, numerous foci of infection which contained a dark central region surrounded by adjacent pyknotic nuclei were observed (examples indicated by arrows). Identical results were observed if bacteria were allowed to attach for 1.5 h, nonadherent bacteria were removed by washing, and the incubation continued for 2 h at 378C (data not shown). Presumably, there is an initial phase of approximately 1.5 h in which bacteria bind to the cell surface, and these bound bacteria are sufficient to cause subsequent cell injury. Fig. 1E to G are a series of optical sections taken from the foci of infection enclosed by the rectangle in Fig. 1D. These samples were additionally stained with an antibody to E cadherin to highlight the borders of adjacent cells. Scattered bacteria were found at the apical region of the MDCK cells (Fig. 1E). The nuclei underlying these cells were pyknotic and stained brightly with propidium iodide (Fig. 1F), and bacteria were found interspersed with the affected cells. The dark central core was devoid of nuclei and by electron microscopic (EM) analysis was found to contain remnants of cell membranes and debris (data not shown). The largest accumulation of bacteria was usually found between the bottom of the cell monolayer and the filter (Fig. 1G). The border between the affected cells in the foci and the surrounding normal cells was marked by how far the bacteria had spread below the monolayer. Note that bacteria were always associated with the foci of host cell death. Macrophages undergo apoptosis when infected with Shigella flexneri (46), suggesting that one mechanism for bacterial killing of mammalian cells is apoptosis. Signs of apoptosis include condensation of both the chromatin and cytoplasm, disappearance of microvilli, and presence of pedunculated protuberances on the cell surface that give rise to apoptotic bodies (21). As the pyknotic nuclei we have observed in our immunofluorescence analysis are reminiscent of apoptotic changes seen in other cells, we analyzed bacterium-infected MDCK cells for ultrastructural evidence of apoptosis. MDCK cells were incubated with bacteria for 3 h at 378C, fixed, and prepared for EM analysis. Sections were cut through a focus of bacterial infection, similar to that shown in Fig. 1E to G. Figure 2A is an electron micrograph at the border between unaffected host cells and the adjacent cells in a typical focus of host cell death. The cell at the right of this figure was largely normal, while the cell to the left was devoid of the normal electron density imparted by the cytoplasm, and appeared to be permeable. As we observed in our confocal analysis, bacteria were found underlying this affected host cell. The

PSEUDOMONAS-MEDIATED MDCK CELL INJURY

1543

cells near the center of the lesion were also permeable, their nuclei were often constricted, and their nuclear contents often appeared to be pulled away from the nuclear membrane (see arrowheads in Fig. 2B). This may explain, in part, the pyknotic appearance of the nuclei we observed by light microscopy. Bacteria were seen underlying these affected cells. The absence of normal cytoplasm and lack of chromatin condensation in infected cells are more consistent with cells undergoing necrotic rather than apoptotic death. In addition, we could not detect the characteristic ladder of DNA fragments found in many cells as they undergo apoptosis (data not shown). Finally, the anucleate central regions of larger lesions seen by confocal analysis were largely devoid of nuclei and typically contained cell debris. With time (4 to 5 h), these individual foci merged with one another, forming increasingly larger zones of cell death. To confirm that the cells at the perimeter of these larger lesions had also been made permeable, we have used the following technique. Bacteria were added to the apical surface of the cells and incubated 4 h at 378C, and nonadherent bacteria were removed by washing. Mouse antibodies against vimentin, an intracellular intermediate filament protein, were added to the monolayer without prior permeabilization. The cells were then fixed, the DNA was stained with the nuclear dye 49,6-diamidino-2-phenylindole (DAPI), and the antivimentin antibodies were detected with goat anti-mouse-TRITC. Antivimentin antibodies do not enter nonpermeable cells, and staining is rarely observed in untreated monolayers. In contrast, each of the foci of infection described above contained a central region devoid of nuclei surrounded by a zone of antivimentin staining, indicating that the cells at the periphery of each of these larger foci were permeable (Fig. 3). This observation is consistent with the permeable cells we observed in our EM analysis. Note that the zone of permeability was seen around all foci regardless of their size, and bacteria were always seen associated with these permeable cells (data not shown). Bacterium-treated MDCK cell monolayers are permeable to [14C]inulin. To quantitate the extent of cell damage induced by P. aeruginosa, we have measured the amount of apically added [14C]inulin that diffuses into the basal chamber of the Transwell following a specified interval of bacterial addition. Inulin is a 5,000-Da polysaccharide molecule that does not normally pass through the tight junctions of the cell; 0.10% 6 0.03% of apically added [14C]inulin is found in the basal medium of untreated MDCK cells following 1 h of incubation at 378C. At low doses of bacteria (105 CFU/ml, MOI of 1:100), we observed no morphological evidence of bacterial adhesion or host cell injury (data not shown), and there was no effect on monolayer permeability to [14C]inulin (Fig. 4). However, as the dose of bacteria was increased, there was a corresponding increase in the permeability of the monolayer to [14C]inulin (Fig. 4). Larger inocula of PA103 (.107 CFU/ml) caused the same pattern of cell damage we described above in our morphological analysis using 107 CFU of bacteria per ml. However, extensive cell damage was observed more quickly (data not shown), and there was a correspondingly larger amount of [14C]inulin that traversed the monolayer (Fig. 4). These data confirmed that the amount of [14C]inulin detected basally reflected the extent of host cell damage we observed in our morphological analysis and was therefore a reliable measure of bacterium-induced host cell injury. Contact between metabolically active bacteria and MDCK cells is necessary for host cell cytotoxicity. To ascertain whether a soluble bacterial factor was involved in causing host cell injury, several experiments were performed. First, proteins from the supernatants of early-stationary-phase cultures of

1544

APODACA ET AL.

INFECT. IMMUN.

FIG. 1. Confocal analysis of P. aeruginosa binding to MDCK cells and subsequent host cell injury. Two hundred microliters of bacteria (107 CFU of PA103 per ml, MOI of 1:1) was added to the apical surface of filter-grown MDCK cells for 1.5 h (A to C) or 3 h (D to G) at 378C. At the end of the incubation period, the cells were washed to remove nonadherent bacteria, fixed, and stained with propidium iodide. Cells in panels D to G were additionally stained with an antibody against E cadherin. (A) Projection (sum) of four 1.0-mm sections, obtained with a scanning laser confocal microscope, from the apical pole of the cell. Use of projections gives one an overview of the distribution of a particular marker in a single image. The nuclei of affected cells are indicated by arrows, and the aggregate of bacteria is indicated by an arrowhead. (B and C) Single optical sections 1 and 3 mm below the last section of panel A, respectively. (D) Low-power overview of the cell monolayer. Examples of foci of cell death are indicated by arrows. The area enclosed within the box is magnified in panels E to G. Sections from the apical pole of these cells (E), from the lateral surface of the cells at the level of the nucleus (F), and from the basal portion of the cells below the nucleus (G), are shown. The small rods are bacteria. Bar 5 10 mm in panels A to C, 100 mm in panel D, and 25 mm in panels E to G.

VOL. 63, 1995


1545

FIG. 2. EM analysis of a focus of bacterial infection. Bacteria (200 ml of 107 CFU of strain PA103 per ml, MOI of 1:1) were added to the apical surface of filter-grown MDCK cells for 3 h at 378C. At the end of the incubation period, the cells were washed to remove nonadherent bacteria, fixed with glutaraldehyde, and processed for EM analysis. Sections were cut through a focus of cell death. Examples of cells at the periphery (A) and near the center (B) of a typical lesion are shown. Bacteria underlying the host cells (small arrows), regions of the cell nucleus where nuclear material has pulled away from the nuclear envelope (arrowheads), and nuclei (N) are shown. Bar 5 2.5 mm.

1546

APODACA ET AL.

INFECT. IMMUN.

FIG. 3. Detection of permeable cells at the periphery of bacterium-induced host cell lesions. Bacteria (200 ml of 107 CFU of strain PA103 per ml, MOI of 1:1) were added to the apical surface of filter-grown MDCK cells for 4.0 h at 378C. At the end of the incubation period, the cells were washed to remove nonadherent bacteria, and incubated with mouse antivimentin immunoglobulin G for 30 min at 378C. The cells were fixed, mouse antibody was detected with goat anti-mouse-TRITC, and DNA was stained with the nuclear dye DAPI. Large foci of infection were observed. The cells at the periphery of these lesions were permeable and reactive to the antivimentin antibody. Bar 5 100 mm.

PA103 were precipitated with ammonium sulfate and resuspended in MDCK cell culture medium. These proteins caused no injury when added to the MDCK monolayers, as assessed by the [14C]inulin assay (Table 1). Second, early-stationaryphase PA103 were inoculated into the cell culture medium alone or on MDCK cells for 6 h. The medium was collected, centrifuged to remove bacteria, and filtered twice through a 0.2-mm-pore-size filter. The fluid obtained from these filtered supernatants was applied to the apical well of MDCK monolayers for another 6 h. These supernatants did not cause cell injury (Table 1). Finally, heat-killed PA103 or cell lysates of PA103 were added to cell culture medium and applied to the apical well of MDCK Transwells. These preparations did not cause cell injury, as assessed by the [14C]inulin assay (Table 1). Therefore, we could not cause cytotoxicity of the MDCK cells with any soluble product produced by the bacteria or by the application of bacterial components. In addition, we have determined if bacteria must be metabolically active to cause host cell death. We found that bacteria (200 ml of 5 3 106 CFU/ml, MOI of 1:2) treated with chloramphenicol, a protein synthesis inhibitor, were unable to cause host cell damage (Table 2). Rifampin, an RNA synthesis inhibitor, inhibited bacterium-induced cell injury as well, but not to the same extent as treatment with chloramphenical (Table 2). In contrast, an inhibitor of DNA synthesis, naladixic acid, had little effect on cytotoxicity (Table 2). Cooling the cells to 48C also blocked bacterium-induced cell injury (Table 2). Our

FIG. 4. Quantitation of MDCK monolayer permeability following incubation with increasing doses of bacteria. Two hundred microliters of PA103 bacteria (1 3 105 to 2 3 108 CFU/ml) were added to the apical surface of 2 3 106 MDCK cells and incubated for 3 h at 378C. The cells were washed, and [14C]inulin was added to the apical compartment of the Transwell. Permeability of the monolayers was subsequently assessed by quantitating the percentage of apically added [14C]inulin found in the basal well of the Transwell following a 60-min incubation at 378C. The data are means 6 standard deviations of three experiments.

VOL. 63, 1995


TABLE 1. Permeability of MDCK monolayers following treatment with bacterial lysates, heat-killed bacteria, or bacterial supernatantsa

TABLE 3. Effects of the three parental strains of P. aeruginosa on bacterium-induced host cell injury

Treatment

NH4SO4-precipitated proteins ................................... MEM/BSA conditioned by PA103 bacteria............. MEM/BSA conditioned by PA103 bacteria in the presence of MDCK cells ........................................ Heat-killed bacteria..................................................... Bacterial lysate............................................................. Untreated culture medium......................................... Live PA103 bacteria (107 cfu/ml) added apically for 6 h at 378C .........................................................

Permeability (%) (mean 6 SD; n 5 3)

0.9 6 0.5 0.16 6 0.05 0.18 6 0.03 0.15 6 0.09 0.24 6 0.05 0.10 6 0.03 30.4 6 3.4

a Bacterial lysates, supernatants, and heat-killed bacteria were prepared as described in Materials and Methods and added in a 200-ml volume to the apical surface of filter-grown MDCK cells for 6 h at 378C. In control reactions, cells were incubated with untreated culture medium or with 200 ml of 107 CFU of live PA103 bacteria per ml. The permeability of the monolayers was subsequently assessed by quantitating the percentage of apically added [14C]inulin found in the basal well of the Transwell following a 60-min incubation at 378C. In the absence of MDCK cells, 38.0% 6 1.0% of the [14C]inulin was found in the basal chamber of the Transwell.

results suggest that bacteria, actively synthesizing RNA and proteins, are necessary for host cell injury to occur. Comparison of effects of bacterial strains and exoproducts on bacterium-induced host cell injury. Previously, we have observed that various parental strains of P. aeruginosa differed in their ability to cause acute lung injury in vivo (23, 44). We performed similar experiments by using our MDCK model system. The three parental strains (PA103, PAK, and PAO1) were all seen to adhere to the apical surface of the epithelial cells (data not shown). As in vivo, the parental strain PA103 caused the most cell injury, whereas less cell injury was seen after the application of equal inocula of parental strain PAK or PAO1 (Table 3) (23, 30). The difference between the strains was most obvious at early time points (3 h) but became less dramatic as the incubation was extended to 6 h. We extended our analysis to include isogenic strains containing mutations in one or more pseudomonal exoproducts. A PA103 strain with a structural gene mutation in exotoxin A (32), which inhibits protein synthesis by ADP-ribosylating elongation factor 2, caused cell injury and the same increase in permeability of the cell monolayer as the PA103 wild-type parental strain did (Table 4). Therefore, exotoxin A does not appear to be necessary for bacterium-induced host cell injury

TABLE 2. Effects of bacterial protein, DNA, and RNA synthesis inhibitors on host cell cytotoxicitya Treatment

None.............................................................................. Chloramphenicol ......................................................... Rifampin....................................................................... Naladixic acid............................................................... Addition at 48C............................................................

Permeability (%) (mean 6 SD; n 5 3)

4.5 6 1.2 ,0.1 1.8 6 0.4 3.2 6 0.2 ,0.1

Two hundred microliters of PA103 (5 3 106 CFU/ml) was treated with 30 mg of chloramphenicol per ml, 20 mg of naladixic acid per ml, or 32 mg of rifampin per ml for 15 min and then applied to the apical surface of MDCK cells for 3 h (MOI of 1:2). The appropriate drug was included throughout the 3-h incubation period. Alternatively, 5 3 106 CFU of PA103 per ml were cooled to 48C and placed on the apical surface of MDCK cells bathed in ice-cold medium and kept at 48C for 3 h. In control reactions, cells were incubated with untreated PA103 bacteria. The permeability of the monolayers was subsequently assessed by quantitating the percentage of apically added [14C]inulin found in the basal well of the Transwell following 60 min of incubation at 378C. a

Incubation time (h)

3 6

1547

% Inulin traversing the monolayer (mean 6 SD; n 5 3) after addition of P. aeruginosa straina: PA103

PAK

PAO1

2.3 6 0.4 30.4 6 3.4

0.10 6 0.06 22.4 6 5.8

0.10 6 0.04 13.2 6 5.8

a Percentage of apically added [14C]inulin traversing monolayers of MDCK cells following 3 or 6 h of application of 200 ml of 107 CFU (MOI of 5:1) of P. aeruginosa parental strains PA103, PAK, and PAO1 per ml.

in vitro. A PA103 strain with a regulatory mutation in exoenzyme S production (PA103exsA::V) (14) failed to cause an increase in the permeability of the cell monolayer following 3 h of incubation at 378C (Table 4). Even when the dose of these bacteria was increased to 108 or 109 CFU/ml, we observed no increase in the permeability of the monolayer to [14C]inulin (0.15% 6 0.20% and 0.03% 6 0.02%, respectively). In fact, this regulatory mutant does not adhere to host cells when fixed and analyzed by the morphological techniques described in the legend to Fig. 1 (data not shown). This lack of adherence probably accounts for the inability of this regulatory mutant to cause cytotoxicity both in vitro and in vivo. We note that PA103 neither produces elastase nor is motile, suggesting these features may not be necessary for the bacterium-induced cell injury we have observed (23). RCAr and ConAr MDCK cells are resistant in killing by PA103. Several strains of P. aeruginosa have been shown to bind to glycolipids including asialo GM-1, lactosyl-ceramide, and glycolipids that contain sialic acid or the GalNacb1-4Gal disaccharide (1, 3, 22, 36, 38, 41). We have found that PA103 agglutinates latex beads coated with either Galb1-4GlcNac or Galb1-3GlcNac (data not shown). To test the hypothesis that changes in the glycosylation of the mammalian host cell would affect bacterium-cell interactions, we have used mutant MDCK cell lines that have altered cell surface glycosylation. The RCAr cell line is resistant to the toxic lectin ricin (5, 25). This strain is deficient in the transporter that imports UDP-Gal into the lumen of the Golgi apparatus and has an approximately 80% decrease in the amount of galactose present on cell surface glycoproteins and glycolipids. In addition, the ConAr cell line is resistant to the lectin concanavalin A (17). While the exact defect in its glycosylation pathway is not known, there are alterations in the mannose core of its N-linked carbohydrates, and the distribution of its glycophosphatidylinositol-anchored proteins. We tested these mutant MDCK cell lines for sensitivity to PA103-mediated cell killing. When fixed and processed for immunofluorescence analysis, bacteria (200 ml of 107 CFU/ml;

TABLE 4. Effects of parental and mutant strains of P. aeruginosa on bacterium-induced host cell injury Incubation time (h)

3 6

% Inulin traversing the monolayer (mean 6 SD; n 5 3) after addition of P. aeruginosa straina: PA103

PA103tox::V

PA103exsA::V

2.3 6 0.4 30.0 6 2.3

5.4 6 0.7 26.0 6 1.5

0.13 6 0.03 3.5 6 2.8

a Percentage of apically added [14C]inulin traversing monolayers of MDCK cells following 3 or 6 h of application of 200 ml of 107 CFU (MOI of 1:1) of P. aeruginosa parental strain PA103 or mutant strains PA103tox::V and PA103exsA::V per ml.

1548

APODACA ET AL.

INFECT. IMMUN.

FIG. 5. Glycosylation-defective MDCK cells are resistant to cell killing. Bacteria (200 ml of 107 CFU of strain PA103 per ml, MOI of 1:1) were added to the apical surface of filter-grown RCAr (A and B) or ConAr (C and D) cells for 3 h at 378C. At the end of the incubation period, the cells were washed, fixed, and stained with propidium iodide. Cells in panels A and B were additionally stained with an antibody against E cadherin. A low-power overview of the cells is shown in panels A and C. Only one focus of infection was observed in panel A and is boxed. This area is magnified in panel B, which is a single optical section taken at the level of the nucleus. No foci of infection were found in panel C. The boxed area is magnified in panel D, which confirms that the nuclei are normal. Bar 5 100 mm in panels A and C and 25 mm in panels B and D.

MOI of 1:1) were rarely found attached to the RCAr or ConAr cells following 1 to 2 h of incubation at 378C (data not shown), and when the incubation was extended, only limited cell injury was observed. A rare focus of cell injury was occasionally observed with the RCAr cells. For example, in Fig. 5A only one bacterial focus, enclosed by the box, was observed in this typical low-power field. An optical section through this focus is shown in Fig. 5B. In contrast, greater than 20 foci of infection were observed when wild-type MDCK cells were used in the identical experiment (Fig. 1D). The ConAr cells were particularly resistant to cell injury, and in a typical low-power field, no foci of infection were observed (Fig. 5C). In addition, neither adherent bacteria nor foci of host cell death were found when the monolayer was observed at higher magnifications (for example, see Fig. 5D). Finally, morphological evidence of cell injury was observed in these mutant host cell lines if the dose of bacteria was increased to .108 CFU/ml. To quantify how resistant glycosylation-defective host cells are to bacterium-induced cell killing, we performed the [14C]inulin permeability assay. Wild-type, RCAr, and ConAr host cell monolayers were incubated with increasing numbers of bacteria. As shown in Fig. 6, low doses of bacteria did not alter the permeability of the host cell monolayers. However, as the dose was increased to between 5 3 106 to 2 3 108 CFU/ml, the wild-type MDCK cell monolayers were consistently more

permeable to [14C]inulin than the RCAr and ConAr cell lines. When the dose of bacteria was increased to .5 3 108 CFU/ml, there was little difference between the susceptibility of host cells to Pseudomonas-induced cell death (data not shown). It is possible that at these higher doses other mechanisms of bacterium-host cell adhesion come into play that allow for cytotoxicity. Our results with the RCAr cells suggested that cell surface galactosylated proteins or lipids may be important in allowing bacterium-induced cytotoxicity of MDCK cells. Presumably, these macromolecules are important in adhesion as we did not observe bacteria bound to these mutant host cell lines. To test this hypothesis, we have attempted to block binding of bacteria to wild-type host cells. PA103 (107 CFU/ml, MOI of 1:1) were pretreated with 20 mM N-acetyllactosamine (Galb1-4GlcNac) for 45 min at 378C and then applied to MDCK cells in the continued presence of the carbohydrate for 3 h. The treated bacteria did cause somewhat less cell injury (74% of the damaged caused by untreated PA103 bacteria), but the same bacteria pretreated with a control sugar, 20 mM cellubiose, also caused less cellular injury (84% of the damage caused by PA103 bacteria); therefore, lactosamine pretreatment did not efficiently prevent cell killing, presumably because bacterial adhesion was not efficiently blocked. This phenomenon could reflect the need for multivalent, high-affinity competitors to


VOL. 63, 1995

FIG. 6. Quantitation of MDCK, RCAr, and ConAr monolayer permeability following incubation with increasing doses of bacteria. Two hundred microliters of PA103 bacteria (1 3 105 to 2 3 108 CFU/ml) was added to the apical surface of 2 3 106 MDCK, RCAr, or ConAr cells and incubated for 3 h at 378C. The cells were washed, and [14C]inulin was added to the apical compartment of the Transwell. The permeability of the monolayers was subsequently assessed by quantitating the percentage of apically added [14C]inulin found in the basal well of the Transwell following 60 min of incubation at 378C. The data are means 6 standard deviations of three experiments.

reduce adhesion or the presence of additional adhesion mechanisms. DISCUSSION As a model for bacterium-induced epithelial cell injury, we have studied the interaction of P. aeruginosa with MDCK cell monolayers. When added to the apical surface of MDCK cells, the bacteria rapidly caused host cell death and loss of the integrity of the monolayer. This cell death was easily and reliably quantitated by measuring the increased permeability of the monolayer to the macromolecular tracer [14C]inulin. MDCK cells with defects in cell surface glycosylation were resistant to cell injury, and bacteria with a mutation in the regulatory locus exsA were unable to induce cell death. Morphologically, the first step in host cell injury was the attachment of bacteria to scattered MDCK cells. These nidi of infection almost invariably involved aggregates of bacteria, consistent with previous observations that P. aeruginosa forms clusters when attached to airway cells in vivo and in vitro (4, 18, 24). In addition, other bacteria are known to coaggregate as they adhere to one another and the host cell surface. We do not know if these aggregates of P. aeruginosa were formed prior to host cell attachment or if they formed subsequent to binding of the MDCK cells. The adherence of P. aeruginosa to MDCK cells is most probably mediated by ‘‘adhesins’’ present on the bacterial surface and ‘‘receptor’’ molecules present on the host cell surface. The best described pseudomonal adhesins include pili (9, 34, 37, 45) and the anionic exopolysaccharide alginate that coats mucoid strains of P. aeruginosa (8, 13, 18, 27, 35). However, the PA103 strain used in this study was not mucoid, so this adhesin may not be involved in the processes we have described. Possible host receptors for P. aeruginosa include cholesterol and cholesterol esters present on the host

1549

cell plasma membrane (40), host cell lectins (28), and host cell mucins (42). In addition, it is well established that various strains of P. aeruginosa bind to host glycolipids including asialo GM-1 and GM-2, lactosyl-ceramide, and glycolipids that contain sialic acid or the GalNacb1-4Gal disaccharide (1, 3, 22, 36, 38, 41). Adhesion of P. aeruginosa to MDCK glycolipids and/or glycoproteins is probably necessary for cell killing in our model system as well. P. aeruginosa PA103 did not bind to RCAr or ConAr MDCK cells when examined by using immunofluorescence, and as a result, these cell lines were resistant to killing. The RCAr cells have a well-described defect in adding galactose to glycolipids and glycoproteins (5, 25), and P. aeruginosa binding to MDCK cells may require galactose moieties. We tested this hypothesis by adding bacteria, pretreated with lactosamine, to wild-type MDCK cells. There was some decrease in cell killing but not to the same extent observed with the galactose-deficient RCAr cells. One possible explanation is that the lactosamine disaccharide is not effective because a multivalent, high-affinity competitor is necessary to reduce bacterial adhesion. Alternatively, there may be other glycolipid or glycoprotein host cell receptors, in addition to galactosylated ones, which are absent from RCAr cells but required for bacterial binding to wild-type MDCK cells. Our results with the ConAr cells further suggest that host cell glycolipids or glycoproteins may be important receptors for bacterial adhesins, since bacteria did not bind to this cell line either. The defect in the ConAr cells may be more severe than that present in the RCAr cells, because the absence of mannose residues would result in the inability to form more-complex N-linked carbohydrate structures. This could explain, in part, the greater resistance to killing seen when using ConAr cells versus the RCAr ones. We should note that Rostand and Esko (40) have used glycosylation-defective CHO cell lines and found their isolates of P. aeruginosa bound independently of several typical host cell ‘‘receptors’’ widely thought to be important in Pseudomonas adhesion including glycoproteins, glycolipids, or proteoglycans. The differences between their results and other studies, including our own, may reflect the different bacterial strains used in these investigations. Following the initial period of bacterial adhesion, foci of cell death were formed throughout the infected monolayer. By immunofluorescence and EM analysis, these foci were seen to consist of a central region of cell debris, surrounded by cells that were permeable and apparently necrotic. These foci gradually enlarged centrifugally until the whole monolayer was involved. It is possible that once the integrity of the monolayer is breached at the initial center of the focus, the bacteria gain access to the basolateral surface of adjacent cells, which enables the bacteria to spread rapidly. This rapid spread could be a result of bacterial division or the acquisition of motility induced by host cell contact. Because bacteria were always seen underlying the basal surface of the permeabilized host cells but not under the surrounding normal cells, they may be chiefly responsible for the later stages of host cell injury we have observed. The bacterial factors that mediate MDCK cell death are presently unknown. However, direct interaction with metabolically active bacteria was necessary for cell killing; culture supernatants, killed bacteria, or homogenized bacteria were not sufficient. The nature of these toxic factors are unknown, but exotoxin A is not involved, and elastase is not produced by PA103 (23). We have no evidence that the bacteria invaded host cells as part of the killing process, and P. aeruginosa is usually not considered an invasive bacteria. We did observe bacteria inside cells, but these were generally cells that had

1550

APODACA ET AL.

been made permeable and were dead. However, we cannot rule out that the initial nidus of infection is the result of bacterial invasion of the host cell and subsequent cell death which allows bacteria access to the basolateral surface of the cells. Most likely, direct contact between bacteria and MDCK cells allows for the concentrated delivery of bacterial products, i.e., toxins, to the host cell. The mechanism of delivery could be similar to the mechanism whereby a Yersinia organism, in close apposition to its host cell, injects YopE into mammalian cells (39). Interestingly, YopE shares homology with exoenzyme S (13a). Host cells had the appearance of being made permeable with streptolysin O (1a), and a similar pore-forming toxin may be involved in Pseudomonas-mediated cell toxicity. We are presently using genetic approaches to identify the bacterial products necessary to induce cell death. Previously, we have studied the ability of a variety of strains of P. aeruginosa to cause lung damage in whole-animal models. In general, we have found a strong correlation with the virulence in lungs and MDCK cell killing in vitro. For example, strain PA103 was most virulent in vivo and was most able to cause MDCK killing in vitro (23). Conversely, the PAO1 strain was relatively nonvirulent in vivo and caused less damage than strain PA103 in vitro. This correlation extended to mutant strains of P. aeruginosa that were defective in several previously identified virulence factors. For instance, exotoxin A and elastase were not of major importance in virulence in vivo or cell death in vitro. However, a regulatory mutation that blocked production of exoenzyme S, exsA::V, greatly decreased virulence in vivo and cell death in vitro (23). Our in vitro results suggest that the diminished virulence of this bacterium in vivo may reflect its decreased adhesion to host cells. Unfortunately, we do not yet have strains that are defective in the structural genes for the 49- and 53-kDa isoforms of exoenzyme S, so we cannot determine if exoenzyme S is directly involved in the killing we have observed. Recently, we have found that addition of an antibody to exoenzyme S does not prevent host cell cytotoxicity (42a). This result could reflect the fact that either exoenzyme S is unimportant in host cell killing or that the mechanism of exoenzyme S delivery (e.g., by a mechanism similar to YopE injection) is unaffected by antibody addition. Ideally, one would like to use human lung epithelia to perform the experiments we have described. Unfortunately, cell lines that maintain the features of alveolar type 1 and type 2 cells are currently not available. Primary cell cultures of rat alveolar type 2 cells exist but they rapidly change phenotype in culture and require the sacrifice of a large number of animals. We have performed a limited number of experiments using these cells and found that PA103 kills them in a manner identical to that of MDCK cells (42a). Although MDCK cells are not identical to the epithelia normally found in lung, they are a useful model system that accurately reproduces at least four important aspects of acute pneumonia caused by P. aeruginosa in vivo. (i) The MDCK monolayer becomes permeable to fluid movement, which appears to correspond to the bidirectional fluid movement observed in injured lungs in vivo. (ii) The MDCK cells die by necrosis, while alveolar cell death is observed histologically in lungs. (iii) The bacteria penetrate the MDCK monolayer and appear to move under cells that are dead or dying, while in vivo the bacteria disseminate beyond the air spaces. (iv) The ability of various strains of P. aeruginosa to cause damage to the MDCK cells correlates well with their virulence in lungs. This cultured cell system has several major advantages over working with whole animals. First, it is much quicker, easier, less costly, more reproducible, and avoids use of live animals. Second, the MDCK cells are a relatively homogeneous popu-

INFECT. IMMUN.

lation of one cell type. Third, other factors, e.g., host immune responses or airway clearance, are eliminated. Fourth, mutant MDCK cell lines can be used. In this study, we have used such mutants to analyze the functional consequences of defective host cell glycosylation on bacterial adhesion and killing. A major focus of studies on bacterial pathogenesis in recent years has been the bacterial genes that are involved in these processes. However, the mammalian genes involved have been much less studied, and this may be a fertile approach for future research. Use of the MDCK model will be useful in the future analysis of host and bacterial factors necessary for bacterial adherence and host cell death in pseudomonal infections. ACKNOWLEDGMENTS We thank L. Katz and R. Shanks for their excellent technical support and C. Baumeister for help in preparing the manuscript. In addition, we thank E. Rodriguez-Boulan and M. Lisanti for providing the RCAr and ConAr MDCK cell lines. This work was done during the tenure of an American Heart Association Minority Development Award to G.A.; a gift from the Lucille P. Marky Charitable Trust to J.E., who is a Lucille P. Marky Fellow; a gift from the Lucille P. Marky Charitable Trust to UCSF; a grant from the Cancer Research Institute; scholarships from the Charles Culpeper Foundation and the Edward Mallinckrodt Foundation; and an American Heart Association Wyeth-Ayerst Established Investigator Award to K.M., who is an established investigator of the American Heart Association; an Academic Senate Award; a grant from the National Heart and Lung Institute (RO1 NHLBI-HL49810); a Cystic Fibrosis Foundation Grant in Aid; and a grant from the American Lung Association to J.W.-K., who is an established investigator of the American Lung Association; and a grant from the NAIAD (R29-A131665) to D.F. REFERENCES 1. Andersson, B., J. Dahmen, T. Frejd, H. Leffler, G. Magnusson, G. Noori, and C. Svanborg-Eden. 1983. Identification of an active disaccharide unit of a glycoconjugate receptor for pneumococci attaching to human pharyngeal epithelial cells. J. Exp. Med. 158:559–570. 1a.Apodaca, G. Unpublished observations. 2. Apodaca, G., L. Katz, and K. Mostov. 1994. Receptor-mediated transcytosis of IgA in MDCK cells is via apical recycling endosomes. J. Cell Biol. 125: 67–86. 3. Baker, N., G. C. Hansson, H. Leffler, G. Riise, and C. Svanborg-Ede´n. 1990. Glycosphingolipid receptors for Pseudomonas aeruginosa. Infect. Immun. 58:2361–2366. 4. Baltimore, R. S., C. D. C. Christie, and G. J. Walker Smith. 1989. Immunohistopathologic localization of Pseudomonas aeruginosa in lungs from patients with cystic fibrosis. Am. Rev. Respir. Dis. 140:1650–1661. 5. Brandli, A. W., G. C. Hansson, E. Rodriguez-Boulan, and K. Simons. 1988. A polarized epithelial cell mutant deficient in translocation of UDP-galactose into the golgi complex. J. Biol. Chem. 263:16283–16290. 6. Breitfeld, P. P., J. E. Casanova, J. M. Harris, N. E. Simister, and K. E. Mostov. 1989. Expression and analysis of the polymeric immunoglobulin receptor in Madin-Darby canine kidney cells using retroviral vectors. Methods Cell Biol. 32:329–337. 7. Bryan, C. S., and K. L. Reynolds. 1984. Bacteremic nosocomial pneumonia. Analysis of 172 single episodes from one metropolitan area. Am. Rev. Respir. Dis. 129:668–671. 8. Ceri, H., H. A. I. McArthur, and C. Whitfield. 1986. Association of alginate from Pseudomonas aeruginosa with two forms of heparin-binding lectin isolated from rat lung. Infect. Immun. 51:1–5. 9. Chi, E., T. Mehl, D. Nunn, and S. Lory. 1991. Interaction of Pseudomonas aeruginosa with A549 pneumocyte cells. Infect. Immun. 59:822–828. 10. Coburn, J., and D. M. Gill. 1991. ADP-ribosylation of p21ras and related proteins by Pseudomonas aeruginosa exoenzyme S. Infect. Immun. 59:4259– 4262. 11. Coburn, J., A. V. Kane, L. Feig, and D. M. Gill. 1991. Pseudomonas aeruginosa exoenzyme S requires a eukaryotic protein for ADP-ribosyltransferase activity. J. Biol. Chem. 266:6438–6446. 12. Coburn, J., R. T. Wyatt, B. H. Iglewski, and D. M. Gill. 1989. Several GTP-binding proteins, including p21c-H-ras, are preferred substrates of Pseudomonas aeruginosa exoenzyme S. J. Biol. Chem. 264:9004–9008. 13. Doig, P., N. R. Smith, T. Todd, and R. T. Irvin. 1987. Characterization of the binding of Pseudomonas aeruginosa alginate to human epithelial cells. Infect. Immun. 55:1517–1522.

VOL. 63, 1995 13a.Frank, D. Unpublished observations. 14. Frank, D. W., G. Nair, and H. P. Schweizer. 1994. Construction and characterization of chromosomal insertional mutations of the Pseudomonas aeruginosa exoenzyme S transregulatory locus. Infect. Immun. 62:554–563. 15. Grant, M. M., M. S. Niederman, M. A. Poehlman, and A. M. Fein. 1991. Characterization of Pseudomonas aeruginosa adherence to cultured hamster tracheal epithelial cells. Am. J. Respir. Cell Mol. Biol. 5:563–570. 16. Gumbiner, B., and K. Simons. 1986. A functional assay for proteins involved in establishing an epithelial barrier: identification of a uvomorulin-like polypeptide. J. Cell Biol. 102:457–468. 17. Hannan, L. A., M. P. Lisanti, E. Rodriguez-Boulan, and M. Edidin. 1993. Correctly sorted molecules of a GPI-anchored protein are clustered and immobile when they arrive at the apical surface of MDCK cells. J. Cell Biol. 120:353–358. 18. Hata, J. S., and R. B. Fick. 1991. Airway adherence of Pseudomonas aeruginosa: mucoexopolysaccharide binding to human and bovine airway proteins. J. Lab. Clin. Med. 117:410–422. 19. Horan, T. C., J. W. White, W. R. Jarvis, T. G. Emori, D. H. Culver, V. P. Munn, C. Thornsberry, D. R. Olson, and J. M. Hughes. 1986. Nosocomial infection surveillance 1984. CDC Surveillance Summary 32:1SS–16SS. 20. Johnson, J., R. L. Warren, and A. A. Branstrom. 1991. Effects of FP2 and a mercury resistance plasmid from Pseudomonas aeruginosa PA103 on exoenzyme production. J. Clin. Microbiol. 29:940–944. 21. Kerr, J. F. R., J. Searle, B. V. Harmon, and C. J. Bishop. 1987. Apoptosis, p. 93–128. In C. S. Potten (ed.), Perspectives on mammalian cell death. Oxford University Press, Oxford. 22. Krivan, H. C., V. Ginsburg, and D. D. Roberts. 1988. Pseudomonas aeruginosa and Pseudomonas cepacia isolated from cystic fibrosis patients bind specifically to gangliotetraosylceramide (Asialo GM1) and gangliotriaosylceramide (Asialo GM2). Arch. Biochem. Biophys. 260:493–496. 23. Kudoh, I., J. P. Wiener-Kronish, S. Hashimoto, J. F. Pittet, and D. Frank. 1994. Exoproduct secretions of P. aeruginosa strains influence severity of alveolar epithelial injury. Am. J. Physiol. Lung Cell. Mol. Physiol. 267:551– 556. 24. Lam, J., R. Chan, K. Lam, and J. W. Costerton. 1980. Production of mucoid microcolonies by Pseudomonas aeruginosa within infected lungs in cystic fibrosis. Infect. Immun. 28:546–556. 25. Le Bivic, A., M. Garcia, and E. Rodriguez-Boulan. 1993. Ricin-resistant Madin-Darby canine kidney cells missort a major endogenous apical sialoglycoprotein. J. Biol. Chem. 268:6909–6916. 26. Marcus, H., A. Austria, and N. R. Baker. 1989. Adherence of Pseudomonas aeruginosa to tracheal epithelium. Infect. Immun. 57:1050–1053. 27. Marcus, H., and N. R. Baker. 1985. Quantitation of adherence of mucoid and nonmucoid Pseudomonas aeruginosa to hamster tracheal epithelium. Infect. Immun. 47:723–729. 28. McArthur, H. A., and H. Ceri. 1983. Interaction of a rat lung lectin with the exopolysaccharides of Pseudomonas aeruginosa. Infect. Immun. 42:574–578. 29. Nicas, T. I., and B. H. Iglewski. 1984. Isolation and characterization of transposon-induced mutants of Pseudomonas aeruginosa deficient in production of exoenzyme S. Infect. Immun. 45:470–474. 30. Pittet, J. F., J. P. Wiener-Kronish, M. C. McElroy, H. G. Folkesson, and M. A. Matthay. 1994. Stimulation of lung epithelial liquid clearance by


1551

endogenous release of catecholamines in septic shock in anesthetized rats. J. Clin. Invest. 94:663–671. 31. Plotkowski, M. C., M. Chevillard, D. Pierrot, D. Altemayer, J. M. Zahm, G. Colliot, and E. Puchelle. 1991. Differential adhesion of Pseudomonas aeruginosa to human respiratory epithelial cells in primary culture. J. Clin. Invest. 87:2018–2028. 32. Prentki, P., and H. M. Krisch. 1984. In vitro insertional mutagenesis with a selectable DNA fragment. Gene 29:303–313. 33. Ramphal, R., C. Guay, and G. B. Pier. 1987. Pseudomonas aeruginosa adhesins for tracheobronchial mucin. Infect. Immun. 55:600–603. 34. Ramphal, R., L. Koo, K. S. Ishimoto, P. A. Totten, J. C. Lara, and S. Lory. 1991. Adhesion of Pseudomonas aeruginosa pilin-deficient mutants to mucin. Infect. Immun. 59:1307–1311. 35. Ramphal, R., and M. Pyle. 1983. Adherence of mucoid and nonmucoid Pseudomonas aeruginosa to acid-injured tracheal epithelium. Infect. Immun. 41:345–351. 36. Ramphal, R., and M. Pyle. 1983. Evidence for mucins and sialic acid as receptors for Pseudomonas aeruginosa in the lower respiratory tract. Infect. Immun. 41:339–344. 37. Ramphal, R., J. C. Sadoff, M. Pyle, and J. D. Silipigni. 1984. Role of pili in the adherence of Pseudomonas aeruginosa to injured tracheal epithelium. Infect. Immun. 44:38–40. 38. Rosenstein, I. J., C. T. Yuen, M. S. Stoll, and T. Feizi. 1992. Differences in the binding specificities of Pseudomonas aeruginosa M35 and Escherichia coli C600 for lipid-linked oligosaccharides with lactose-related core regions. Infect. Immun. 60:5078–5084. 39. Rosqvist, R., K.-E. Magnusson, and H. Wolf-Watz. 1994. Target cell contact triggers expression and polarized transfer of Yersinia YopE cytotoxin into mammalian cells. EMBO J. 14:964–972. 40. Rostand, K. S., and J. D. Esko. 1993. Cholesterol and cholesterol esters: host receptors for Pseudomonas aeruginosa adherence. J. Biol. Chem. 268:24053– 24059. 41. Saiman, L., and A. Prince. 1993. Pseudomonas aeruginosa pili bind to AsialoGM1 which is increased on the surface of cystic fibrosis epithelial cells. J. Clin. Invest. 92:1875–1880. 42. Simpson, D. A., R. Ramphal, and S. Lory. 1992. Genetic analysis of Pseudomonas aeruginosa adherence: distinct genetic loci control attachment to epithelial cells and mucins. Infect. Immun. 60:3771–3779. 42a.Wiener-Kronish, J. Unpublished observations. 43. Wiener-Kronish, J. P., K. H. Albertine, and M. A. Matthay. 1991. Differential responses of the endothelial and epithelial barriers of the lung in sheep to Escherichia coli endotoxin. J. Clin. Invest. 88:864–875. 44. Wiener-Kronish, J. P., T. Sakuma, I. Kudoh, J. F. Pittet, D. Frank, L. G. Dobbs, M. L. Vasil, and M. A. Matthay. 1993. Alveolar epithelial injury and pleural empyema in acute P. aeruginosa pneumonia in anesthetized rabbits. J. Appl. Physiol. 75:1661–1669. 45. Woods, D. E., D. C. Straus, W. G. J. Johanson, V. K. Berry, and J. A. Bass. 1980. Role of pili in adherence of Pseudomonas aeruginosa to mammalian buccal epithelial cells. Infect. Immun. 29:1146–1151. 46. Zychlinsky, A. 1993. Programmed cell death in infectious diseases. Trends Microbiol. 1:114–117.