Rhamnolipids Are Virulence Factors That Promote Early Infiltration of Primary Human Airway Epithelia by Pseudomonas aeruginosa

Airway epithelia.Biopsies of noninvolved nasal mucosa were performed on 35 male and 18 female patients (age range, 21 to 60 years) according to the guidelines of the Ethical Committee for Clinical Studies of the Geneva State Hospital (authorization no. 04/019). Informed consent was obtained from all patients. Epithelial cells were dispersed from the biopsy specimens (29) and plated at a density of 5 × 10⁵ cells/cm² onto 0.6-cm² collagen-coated filters (Millipore, Molsheim, France). Cells were cultured in Dulbecco's modified Eagle medium (DMEM)-F12 (Invitrogen, Basel, Switzerland) supplemented with 2% Ultroser G (Biosepra, Ciphergen Biosystems, Cergy-Saint-Christophe, France), 100 U/ml penicillin, and 100 mg/ml streptomycin. After 1 day, filters were taken at the air-liquid interface for 2 to 3 weeks. Twenty-four hours prior to the infection assay, penicillin and streptomycin were removed from the culture medium.

Pseudomonas aeruginosa.The P. aeruginosa strains, listed in Table 1, were transformed with plasmid pIAX2 to express the gene coding for green fluorescent protein (GFP) (a gift from I. Attree, CEA-Grenoble, France) (12). Bacteria were inoculated in Luria-Bertani (LB) medium overnight at 37°C, diluted in LB, and grown to an optical density at 600 nm of 0.5, under which conditions all strains grew similarly. Supernatants of an overnight culture of wild-type or mutated P. aeruginosa strains grown in DMEM-F12-HEPES were centrifuged, filtered on a 0.22-μm-pore-size filter, and adjusted to a pH of 7.5.

The rhlA promoter was fused to the GFP gene by replacing the X2 promoter of the pIAX2 plasmid with the rhlA promoter of plasmid pECP60 (42), between the SmaI and BamHI sites of pIAX2, to generate the pZC1 vector. Strains listed in Table 2 were electroporated with pZC1. Promoter activity was further analyzed in LB and DMEM-F12-HEPES culture media by assessing the fluorescence of GFP.

The rhlA mutation in the PT712 strains was complemented as follows. The HindIII-EcoRI fragment of plasmid pJPP6, containing the rhlABRI operon and the pheC gene (Jim Pearson, unpublished data), was ligated into the HindIII/EcoRI sites of pBluescript II SK(+) to generate plasmid pAKRHL. The EcoRI fragment of this plasmid, containing the rhlABRI operon, was ligated into the EcoRI site of the pIAX2 plasmid (12). The resulting pZC6 plasmid, which allows for expression of the rhl operon under the control of its own promoter, constitutively expresses the GFP gene under the control of the X2 promoter. Transformation of strain PT712 with pAKRHL and pZC6 generated strains PT1323 and MZ10, respectively. The promoter activity for rhlABRI was assessed on SW blue plates as described below, while expression of GFP was observed by fluorescence.

Elastase and rhamnolipid production.Elastase production was measured by an elastin Congo red assay, as previously described (52). Rhamnolipid production was measured on SW blue plates by inoculating strains in M9-based agar plates (36) supplemented with 0.2% (vol/vol) glycerol, 2 mM MgSO₄, 5 mM KNO₃ (instead of NH₄Cl), 0.0005% (vol/vol) methylene blue, and 0.02% (vol/vol) cetyltrimethylammonium bromide (46). Plates were incubated at 37°C for 24 h and then kept for at least 48 h at room temperature until a blue halo appeared around colonies. For quantitative assays, rhamnolipids were extracted from supernatants of PAO1 cultures grown in M9 minimal medium supplemented with 2% glycerol, 2 mM MgSO₄, 0.05% glutamate (instead of NH₄Cl), and 0.05% Casamino Acids. After ether extraction, rhamnolipids were quantified by the orcinol procedure (42). Purified rhamnolipids JBR515 were also obtained from the Jeneil Company (Saukville, Wis.) and diluted in DMEM-F12-HEPES immediately before use.

Measurement of the epithelial barrier.The TER of the reconstituted epithelia was assessed using a Millicel ERS Volt-ohm meter (World Precision Instruments, New Haven, CT). Paracellular permeability was monitored after apical addition of 1μ Ci/ml [³H]inulin in the presence or absence of 150μ g/ml rhamnolipids. At the indicated times, aliquots of the apical and basolateral media were sampled and counted.

Antibodies.Rabbit polyclonal antibodies to claudin-1, occludin, and ZO-1 were purchased from Zymed Laboratories (San Francisco, Calif.), rabbit polyclonal antibodies to ezrin from Upstate (Lake Placid, N.Y.), mouse monoclonal antibodies to mucin 5AC 1 from NeoMarker (Fremont, Calif.), mouse monoclonal anti-human cystic fibrosis transmembrane conductance regulator (CFTR) from R&D Systems, Inc. (Minneapolis, Minn.), horseradish peroxidase-conjugated anti-rabbit or anti-mouse antibodies from Bio-Rad (Reinach, Switzerland), and fluorescein isothiocyanate (FITC)-, Texas Red-, and Cy5-conjugated anti-mouse or anti-rabbit antibodies and Alexa Fluor 488, Texas Red, and Cy5 phalloidin from Molecular Probes (Leiden, The Netherlands). The mouse monoclonal antibody to JAM1 was given by M. Aurrand-Lions,Geneva, Switzerland. GM1 ganglioside was detected with horseradish peroxidase-conjugated cholera toxin B (Sigma). Fluorescent l-rhamnose and rhamnolipids were generated by coupling the diol group of the sugar moiety with 5-(4,6-dichlorotriazinyl) aminofluorescein (Molecular Probes) and purifying (Eurogentec, Seraing, Belgium).

Experimental treatments.The apical surfaces of reconstituted epithelia, featuring similar transepithelial resistances, were exposed for 10 min to 16 h to one of the following conditions: (i) no treatment, (ii) P. aeruginosa bacteria washed off the culture medium, (iii) supernatants of P. aeruginosa cultures, (iv) 15 to 150 μg/ml purified rhamnolipids, (v) purified rhamnolipids followed by Pseudomonas washed off the culture medium, (vi) purified rhamnolipids (unlabeled or FITC labeled) followed by 0.5 μm carboxylate-modified fluorospheres (Molecular Probes, Leiden, The Netherlands), or (vii) unlabeled or FITC-labeled l-rhamnose (Fluka). In each set of experiments, epithelia showing comparable TERs were exposed in parallel to several of these conditions. The experiments were stopped by extensive washing of the epithelia in DMEM-F12-HEPES, and metabolically active cells were evaluated using an MTT[1-4,5-dimethylthiazol-2-yl)-3,5-diphenylformazan] assay (Sigma). Cytotoxicity and viability were also assessed using a LIVE/DEAD viability/cytotoxicity assay kit (Molecular Probes, Leiden, The Netherlands). To this end, epithelia were stained with 4 μM calcein and 2 μM ethidium for 40 min at 37°C. Filters were then cut off and mounted for live confocal microscopy analysis.

Immunostaining.The reconstituted epithelia were fixed in 4% paraformaldehyde, permeabilized in 0.1% saponin, and incubated for 1 h with one of the primary antibodies listed above diluted in phosphate-buffered saline containing 1% bovine serum albumin. After a wash, the tissues were incubated again for 30 min with an appropriate secondary antibody as per standard protocols (58). Filters were cut off from the culture inserts, mounted in Vectorshield-DAPI (4′,6′-diamidino-2-phenylindole) (Vector Laboratories) between glass coverslips, and observed with an LSM 510 confocal microscope (Zeiss). For convenience and consistency of image representations, some of the immunostaining was captured with the green (488 nm) and red (543 nm) channels and appears red and green, respectively, in some of the figures.

For live imaging, filters of control epithelia and of epithelia exposed to FITC-labeled rhamnolipids were cut off and mounted in culture media containing 1-(4-trimethylammoniumphenyl)-6-phenyl-1,3,5-hexatriene-p-toluenesulfate (TMA-DPH; Molecular Probes, Leiden, The Netherlands), a cationic linear polyene that is readily incorporated into the plasma membranes of host cells. Time lapse video microscopy was performed using a Hamamatsu high-resolution black/white charge-coupled device camera coupled to Openlab software.

Electron microscopy.Tissues were fixed in 2.5% glutaraldehyde in 0.1 M phosphate buffer (pH 7.4) and processed for either conventional or freeze fracture electron microscopy as described previously (8). Sections and replicas were photographed with a Philips CM10 microscope (Eindhoven, The Netherlands). Quantitative analysis of TJ was carried out on photographs of 60 to 100 cells per condition. The area, length, and width encompassed by TJ fibrils were measured at a magnification of× 34,000, using an Acecad graphic tablet connected to Quantimet 500 software (Leica). Numbers of strands and loose ends of fibrils were also scored.

Statistics.Values were expressed as means± standard errors of the means (SEMs) and were compared by analysis of variance and t tests for independent variables using SPSS software (SPSS Inc., Chicago, Ill.).

Differentiated respiratory epithelia are reconstituted with cells of human nasal mucosa.After a 2- to 3-week culture at the air-liquid interface, cells of nasal mucosa formed epithelia showing a TER of 1,000 to 2,500 Ω · cm² and resembling a native respiratory epithelium, i.e., comprising basal, goblet, and ciliated cells (Fig. 1A). These cells were immunostained for both ezrin and CFTR (Fig. 1B). The polarization of goblet and ciliated cells correlated with the presence of continuous TJ belts, which freeze fracture electron microscopy revealed between the basolateral and apical domains of the cell membranes (Fig. 1C). Immunofluorescence showed that these belts were made of claudin-1, occludin, ZO-1, and JAM-1 and that they uninterruptedly surrounded each cell (Fig. 1D).

• Open in new tab
• Download powerpoint

FIG. 1.

Tissues reconstituted at the air-liquid interface show features of native airway epithelia. (A) After a 2- to 3-week culture, human nasal epithelial cells adhered to each other via junctional complexes (inset) and differentiated into either basal, ciliated, or goblet cells. (B) Fully polarized reconstituted epithelia expressed ezrin (green, left panel), a marker of ciliated cells, and CFTR (green, right panel) on the apical surface. (C) Freeze fracture revealed that TJ formed regular and uninterrupted belts comprising at least five fibrillar strands. (D) Immunofluorescence showed that these junctions contained both JAM-1 (green) and ZO-1 (red) proteins, which were often colocalized (yellow) and continuously surrounded each cell. Bars, 3 μm (A), 20 μm (B), 100 nm (inset), 180 nm (C), 20 μm (D). f, filter; g, goblet cell; c, ciliated cell; b, basal cell; J, junctional complex.

P. aeruginosa expressing the Rhl quorum-sensing system alters the epithelial barrier.When the apical surface of reconstituted epithelia was exposed to PAO1 in the exponential-growth phase that had been washed off its supernatant and added at a ratio of 1 to 20 bacteria per host cell, we observed no invasion or infiltration of the tissue for up to 8 h (Fig. 2A). In contrast, when the infection was prolonged to 16 to 24 h, we observed that GFP-PAO1 efficiently infiltrated the epithelium (Fig. 2A). When the epithelial surface was exposed to a filtered (i.e., devoid of bacteria) supernatant of an overnight PAO1 culture, a decrease in TER was observed which was sustained throughout the duration of the experiment (Fig. 2B). During the first 4 h, this reduction in TER was not paralleled by a decrease in the viability of epithelial cells, as evaluated by the MTT assay (data not shown). These results suggest that diffusible toxins released by a high density of bacteria, not the TTSS (10), are responsible for the alteration of the epithelial barrier.

• Open in new tab
• Download powerpoint

FIG. 2.

Bacterial invasion of the reconstituted epithelia. (A) Addition of GFP-PAO1 failed to result in adhesion of the bacteria to the apical surface of the epithelia for up to 8 h (top left panel, en face and profile views). In contrast, when the exposure was prolonged to 16 h to allow for activation of the QS systems, bacterial infiltration was evident (top right panel, en face and profile views). Strains deficient in the two QS systems (PT531) or in rhlA (PT712), as well as strain PAK, failed to infiltrate the reconstituted epithelium (profile views). In contrast, epithelia were readily invaded by the cystic fibrosis strain CHA, the TTSS-deficient strain CHAexsA, and the lasR-deficient strain PT498. The red immunostaining of actin delineates the cell profiles. Bar, 20μ m. (B) TER significantly decreased as a function of time after exposure of the apical surface of epithelia to pathogen-free supernatants of P. aeruginosa strain PAO-1, as well as to those of strain PT712, provided the latter was supplemented with purified rhamnolipids. This decrease was also observed with supernatants of the mutated P. aeruginosa strains PT531 and PT712 but was less drastic than the effects observed with supernatants containing rhamnolipids (PAO1 and PT712 plus purified rhamnolipids). Data are means ± SEMs from four independent experiments.** , P < 0.01;*** , P < 0.001 (versus control value).

To identify the relevant factors under the control of the quorum-sensing systems, the epithelia were exposed to strain PT531, which was derived from PAO1 and is deficient in both the las and rhl quorum-sensing systems (Table 1). After 16 h, strain PT531 multiplied to high titers (approximately 10⁹ bacteria per ml) in the apical medium without inducing cell damage or infiltrating the epithelia (Fig. 2A). Similar observations were made with the rhlR mutant PT462 (Table 1). In contrast, invasion was detected with the lasR mutant PT498 (Fig. 2A).

Since these results showed that mutations affecting the rhl QS system impaired the infiltration of airway epithelia by P. aeruginosa, we next tested whether rhamnolipids, the synthesis of which depends mainly on the rhl QS system, were involved in bacterial invasion. We observed that after an overnight period, the reconstituted epithelium was not susceptible to infection by the rhlA mutant PT712, which is specifically impaired in rhamnolipid synthesis (Fig. 2A; Table 1). Since these results suggested that the activity of the rhl quorum-sensing system and the production of rhamnolipids were necessary and sufficient to promote infiltration by P. aeruginosa, we compared the effects on TER of filtered supernatants from strains deficient in both the las and rhl QS systems (PT531) or only in the rhlA gene (PT712). Even though a slight decrease in TER with time was observed for PT531 and PT712, neither strain induced the large TER drop observed after the epithelia were exposed to the supernatant from wild-type strain PAO1 (Fig. 2B). However, addition to the PT712 supernatant of purified rhamnolipids isolated from the PAO1 supernatant induced a drop in TER comparable to that induced by the PAO1 supernatant alone (Fig. 2B). Furthermore, when strain PT712 was transformed with plasmid pAKRHL, carrying the entire rhlABRI operon, and plasmid pZC6, carrying GFP under the control of the X2 promoter, we observed that the defective rhamnolipid expression was complemented, as evaluated on a blue SW plate (Fig. 3A), and that the complemented strains were able to infiltrate the epithelium (Table 2). These results suggest that rhamnolipids cause P. aeruginosa to invade respiratory epithelium by modulating the permeability of the tissue.

• Open in new tab
• Download powerpoint

FIG. 3.

Production of rhamnolipids, but not elastase, is needed to promote bacterial infiltration. (A) (Left) PAO1 and the mucoid strain CHA secreted detectable levels of rhamnolipids, as visualized by the blue halo in the plate assay. In contrast, strains PT712, PAK, and PT531 did not release detectable levels of these virulence factors. Complementation of the rhlA mutation by plasmids pAKRHL and pZC6 is sufficient to release detectable levels of rhamnolipids. (Right) Elastase was produced at high levels by the PT712 mutants and to a lesser extent by PAO1. In contrast, strains PAK, PT531, and CHA did not release detectable levels of elastase in the media used for the invasion test. **, P < 0.01;*** , P < 0.001 (versus PAO1 value). Numbers of independent experiments are given on the right. (B) Pseudomonas strains MZ2 and MZ6 were derived by transforming strains PAO1 and PT712, respectively, with pZC1 (rhlA::gfp). Strain MZ2, which produced rhamnolipids, infiltrated the epithelium (upper panel). In contrast, strain MZ6, which was deficient in rhamnolipid production, was not detected within the epithelium (lower panels). Bar, 20μ m.

Elastase, another factor encoded by las and rhl, has been previously suggested to promote opening of the paracellular route (4, 2). To test its implications in our model, we measured the elastolytic activity in the supernatant of strain PT712. We found that this rhamnolipid-deficient strain, which did not invade the reconstituted epithelium, secreted elastase at even higher levels than the invasive strain PAO1 (Fig. 3A; Table 2). We further observed that the cytotoxic Pseudomonas aeruginosa isolate CHA from a cystic fibrosis patient (13) produced elastase levels lower than those produced by both strains PAO1 and PT712 (Fig. 3A). However, we observed that both CHA and its isogenic mutant CHAexsA, which lacks the type III secretion system (12, 25), infiltrated the epithelia (Fig. 2A; Table 2). In view of previous reports, we further analyzed the behavior of strain PAK (31, 20) and observed that it did not invade the epithelium (Fig. 2A), consistent with the finding that it also did not produce rhamnolipids under our culture conditions (Fig. 3A). The results show that invasion of P. aeruginosa does not correlate with elastolytic activity and type III secretion, at least during the early stages of infection.

To test whether rhamnolipids were actually produced under the conditions we used for the invasion assay, the promoter of the rhlA gene, which codes for rhamnosyltransferase, was fused to GFP to generate plasmid pZC1, which was used to transform PAO1 and PT712, yielding strains MZ2 and MZ6, respectively. We found that the expression of the rhlA promoter was activated in both strains as soon as the bacteria reached a density of >10⁹/ml, indicating that the rhl quorum-sensing system and the rhlA pathways were properly activated, irrespective of the culture medium tested (data not shown). However, when reconstituted epithelia were exposed overnight to exponentially grown Pseudomonas strains MZ2 and MZ6, we observed that only the rhamnolipid-secreting strain MZ2 infiltrated the epithelium (Fig. 3B). Taken together, the results show that the actual release of rhamnolipids is essential for epithelial infiltration (Table 2).

Purified rhamnolipids alter the epithelial barrier without affecting cell viability.To determine whether exogenous rhamnolipids could reproduce the effects of P. aeruginosa or supernatants of high-density bacterial cultures, we applied purified rhamnolipids to the apical surface of the epithelia. We observed that rhamnolipids resulted in a rapid reduction in TER, which was dose and time dependent. Thus, whereas 15 μg/ml rhamnolipids did not significantly alter the values of TER after 3 h (846± 6 Ω · cm² [n = 4]) compared to those observed for untreated controls (996 ± 48Ω · cm²), 50 μg/ml rhamnolipids decreased the TER within 30 min, reaching less than 10% of the control value (P < 0.001) after 360 min (Fig. 4A). In the presence of 150 μg/ml rhamnolipids, such a drop in TER was observed within 10 min (163 ± 83 Ω · cm² [n = 5]) (Fig. 4A).

• Open in new tab
• Download powerpoint

FIG. 4.

Purified rhamnolipids decrease the permeability of epithelia and promote their invasion by P. aeruginosa without altering cell viability. (A) TER was not altered by 15 μg/ml rhamnolipids but was markedly decreased by ≥50-μg/ml concentrations of these factors. The rapidity of this change increased with the concentration of rhamnolipids. (B) After treatment with 150μ g/ml rhamnolipids, the permeability of the reconstituted epithelia to [³H]inulin also increased significantly as a function of time. (C) Under these conditions, the viability of epithelial cells was not affected, as evaluated by the MTT assay. (D) Staining with the LIVE/DEAD viability and cytotoxicity kit revealed that 99.5% of rhamnolipid-treated epithelial cells incorporated calcein (green), indicating cell viability (middle panel), a proportion similar to that observed in control untreated epithelia (top panel). In contrast, treatment with 0.1% saponin for 1 h induced 90% cell death, as indicated by the staining with ethidium bromide (red) (lower panel). Bar, 20μ m. (E) Addition of GFP-PAO1 to epithelia previously exposed for 60 min to 150 μg/ml rhamnolipids resulted in adhesion of the fluorescent bacteria to the surfaces of epithelial cells (upper left panel) and in invasion by numerous pathogens (lower left panel). The red immunostaining of actin, used to delineate the cell periphery, indicates that a minority of epithelial cells were in contact with P. aeruginosa and suggests that the bacteria did not enter the cells. Quantitative analysis (right panel) confirmed that, under the conditions we used, 10% of the cells contacted about 6 bacteria (stippled bars). **, P < 0. 01;*** , P < 0.001 (versus control value [solid bars]).

To determine whether this drop was paralleled by a change in epithelial permeability, we next studied the flux of [³H]inulin. We observed that, compared to the control value of 1.2%, the transepithelial passage of this extracellular marker was increased to 4.5% and 25% after 1-h and 4-h exposures of the epithelium to 150μ g/ml rhamnolipids, respectively (Fig. 4B), in spite of unchanged viability of the epithelial cells, as assessed by the MTT (Fig. 4C) and LIVE/DEAD cytotoxicity/viability (Fig. 4D) tests. Under these conditions, the proportion of dead cells (0.5%) was like that observed in untreated epithelia, showing that cell death could not account for the large alterations in permeability of rhamnolipid-treated epithelia.

Consistent with these alterations, we observed that within 30 min, GFP-PAO1 infiltrated epithelia that had previously been exposed for 60 min to 150 μg/ml rhamnolipids (Fig. 4E). Under these conditions, an average of 6 bacteria contacted 9% of the cells, compared to control values of 1.2 bacteria on 1.7% of the cells (31 filters were scored from 19 independent experiments). The number of bacteria reaching the basal surface of the epithelia was evaluated by plating the bacteria onto agar plates after extensive washing of the epithelia followed by hypotonic lysis. We observed that 7.4% of the wild-type bacteria that were applied on top of the epithelia had infiltrated (data not shown). Similar infiltration of rhamnolipid-exposed epithelia was also observed with Pseudomonas strains PAK, PT531, and PT712, which otherwise did not infiltrate our control epithelia even when applied at high densities (Table 2), as well as with inert carboxylate-modified microspheres (data not shown). These data indicate that once the paracellular pathway was made accessible, no further active process was required for epithelial invasion.

To identify the type of cells affected by rhamnolipids, we monitored the passage of GFP-P. aeruginosa through epithelia that had been exposed either to purified rhamnolipids or to supernatants of overnight cultures of PAO1. In both cases, we observed that the fluorescent bacteria infiltrated the epithelia at sites where the immunolabeling of ezrin was displaced from the apical (control group) to the basolateral (rhamnolipid-exposed group) membrane (Fig. 5A). Double immunolabeling for ezrin and MUC5AC showed that most of these cells did not express mucins and had a ciliated phenotype (Fig. 5B). These findings, together with the loss of cilia, which was observed after rhamnolipid exposure (Fig. 5A, C, and D), indicated that PAO1 infiltrated the epithelium close to cells featuring an altered polarity. To determine whether this infiltration occurred through the transcellular or the paracellular pathway, epithelia exposed to Pseudomonas for 16 to 24 h were examined by electron microscopy. Irrespective of the invading strain (PAO1 or CHA), the bacteria were seen exclusively within intercellular spaces throughout the duration of the experiment (Fig. 5C). However, after several hours of infection, a few necrotic cell profiles were observed, over which numerous P. aeruginosa bacteria were concentrated (Fig. 5C). The finding of similar profiles in epithelia exposed to rhamnolipids, to permit the infiltration of the PT712 and PT531 bacteria, suggests that this limited cell necrosis was not caused by factors secreted under the control of the quorum-sensing systems. Comparable observations were made in epithelia exposed to rhamnolipids and then to PAO1 for as long as 5 h (Fig. 5D).

• Open in new tab
• Download powerpoint

FIG. 5.

P. aeruginosa infiltrates the paracellular spaces in between ezrin-positive cells. (A) Profile views of control (upper left panel) and rhamnolipid-treated (lower left panel) epithelial cells immunolabeled for F-actin (blue staining) and ezrin (red staining). After exposure to rhamnolipids, the ezrin labeling decreased in the apical membrane, where it became patchy, and appeared in the basolateral domain of the cell membrane (arrows, lower left panel), indicating loss of cell polarity. Epithelia with regular, apical ezrin staining were not invaded by PAO1 (upper right panel). The bacteria (green) infiltrated intercellular spaces at sites where ezrin was observed in the basolateral membranes (arrows, lower right panel). (B) Immunostaining for goblet (mucin) (blue) and ciliated (ezrin) (red) cells revealed that, after an overnight infection by PAO1, Pseudomonas (green) was found mainly close to ezrin-positive cells. (C) After an overnight infection, electron microscopy showed that all PAO1 bacteria (arrowheads) were within the paracellular spaces. However, a few necrotic cell profiles (N), over which numerous P. aeruginosa bacteria were concentrated, were observed. (D) Shortly after exposure to rhamnolipids, PAO1 (arrowheads) was found in the paracellular spaces between ultrastructurally normal cells (upper panel). However, after 5 h, a few necrotic cell profiles were observed (lower panel) where P. aeruginosa bacteria were concentrated. Loss of polarity was evident upon paracellular infiltration. Bars, 20μ m (A and B) and 5 μm (C and D). F, filter; AD, apical domain.

Rhamnolipids localize within the lateral membranes of epithelial cells.To assess how rhamnolipids altered the epithelial barrier, we coupled fluorescein to their carbohydrate moiety and applied either the labeled molecules or fluorescein-tagged l-rhamnose to the apical surface of the epithelia. Labeled rhamnolipids drastically reduced TER, as did the unlabeled molecules (Fig. 6A). This effect, which was not observed with either unlabeled or FITC-labeled rhamnose, was reversible after washout (Fig. 6B). At early time points, fluorescent rhamnolipids were uniformly distributed within the apical membranes of the ezrin-positive cells (Fig. 6C). After 60 to 120 min, the rhamnolipids featured a sparser and patchier distribution in the apical membranes and became concentrated in the basolateral membranes (Fig. 6C). The incorporation of FITC-labeled rhamnolipids into the host membranes was confirmed by live imaging of treated epithelial cells stained with TMA-DPH (data not shown). This incorporation appears to depend on the alkyl chains of rhamnolipids, inasmuch as no labeled l-rhamnose bounded or was internalized by epithelial cells (data not shown).

• Open in new tab
• Download powerpoint

FIG. 6.

Rhamnolipids initially bind to the apical membrane and progressively enter the basolateral membrane. (A) Apical treatment of the reconstituted epithelia with FITC-rhamnolipids resulted, within 60 min, in a drastic drop in the TER, which was comparable to that caused by unlabeled rhamnolipids. In contrast, addition of labeled or unlabeled l-rhamnose did not alter the transepithelial resistance. (B) The loss of transepithelial resistance induced by rhamnolipids was rapidly rescued after the molecules were washed off from the apical surface of the epithelia. (C) Initially, labeled rhamnolipids were found associated with the apical membranes of epithelial cells. With time, the apical labeling (top left panel) decreased and the labeling of the basolateral membranes of the reconstituted epithelia increased (lower left and right panels). Bar, 20 μm. Actin (red) was labeled by phalloidin Texas Red to delineate the cells.

In several systems, rhamnose promotes the binding of polysaccharides to specific receptors (49). To assess whether rhamnose promoted the insertion of rhamnolipids within the membranes of epithelial cells, we tried to compete for this insertion by using 0.3 to 2 mM l-rhamnose for 1 to 16 h. We found that l-rhamnose prevented neither the drastic drop in TER (Fig. 7A) nor the bacterial infiltration (Fig. 7B) induced by rhamnolipids.

• Open in new tab
• Download powerpoint

FIG. 7.

l-Rhamnose cannot prevent the rhamnolipid-induced alterations of the epithelial barrier. (A) Apical pretreatment of epithelia with 2 to 0.3 mM l-rhamnose did not protect the epithelia against the significant reduction in TER that is caused by rhamnolipids. (B) Pretreatment of epithelia with 4 mM l-rhamnose also did not protect the epithelia against an overnight invasion by Pseudomonas GFP-PAO1. Actin was immunolabeled by Texas Red phalloidin to delineate the cells. Bar, 20μ m.

Rhamnolipids alter tight-junction architecture.The opening of the paracellular route caused by rhamnolipids implies alterations of TJ. Consistent with this view, significant alterations in TJ morphology were seen, which increased with time after rhamnolipid treatment (Fig. 8A). Thus, whereas TJ of control cells formed regular and uninterrupted belts, made on average of five interconnected strands of fibrils (Fig. 8A), morphometric analysis revealed that the number of strands and the area of TJ belts were significantly reduced after a 30- to 60-min treatment with 150 μg/ml rhamnolipids (Fig. 8). After 4 h, TJ belts comprised fibrils that were often interrupted and either partially or fully disconnected from the TJ belt, resulting in an increase in the membrane area occupied by TJ fibrils, in spite of a reduced number of strands (Fig. 8).

• Open in new tab
• Download powerpoint

FIG. 8.

Rhamnolipids alter tight-junction architecture. (A) Control epithelial cells (time zero) featured uninterrupted belts of TJ fibrils, running in parallel, which separated the basolateral domain (BLD) of the cell membrane from the apical domain (AD). This organization was progressively altered as a function of time after rhamnolipid treatment. Thus, by 120 min, TJ belts showed reduced numbers of strands. At later time points (≥240 min), TJ belts featured strands with loose ends (arrowheads) or strands that encircled domains of the cell membrane (asterisks). At this time point, TJ belts were interrupted and no longer separated the apical and basolateral membrane domains (bottom panel). Bar, 200 nm. (B) Quantitative analysis revealed that the number of TJ strands decreased with time after rhamnolipid treatment, whereas the number of fibrils showing loose ends, i.e., not connected to other fibrils, increased. As a result of these changes, the area occupied by TJ fibrils was rapidly reduced after rhamnolipid treatment but returned to control levels within 4 h. Data are means ± SEMs for the number of measurements (one measurement per TJ belt) given at the bottoms of the bars.** , P < 0. 01;*** , P < 0.001 (versus control value).