Macropinocytosis: New Insights Into Its Underappreciated Role in Innate Immune Cell Surveillance

Introduction Macropinocytosis is the most effective way for cells to ingest large amounts of extracellular fluid. In some cell types macropinocytosis is a constitutive process: immature dendritic cells use it to sample soluble antigens (Sallusto et al., 1995) and Dictyostelium amoeba for nutrient uptake (Cardelli, 2001). Constitutive macropinocytosis is also observed in fibroblasts transformed with oncogenic v-Src or K-Ras (Amyere et al., 2000, 2002). Alternatively, macropinocytosis can be transiently induced by growth factors, such as epidermal growth factor or macrophage colony–stimulating factor (Racoosin and Swanson, 1989; West et al. 2000). The remodelling of the cytoskeleton that leads to macropinocytosis requires phosphatidylinositol-3-kinase (PI3K) activity at the plasma membrane (Araki et al., 1996; Rupper et al., 2001; Lindmo and Stenmark, 2006). Although the entire signaling sequence is incompletely understood, the GTPases Rac1 (West et al., 2000) and Cdc42 (Garrett et al., 2000), as well as p21-activated kinase 1 (PAK1; Dharmawardhane et al., 2000), are involved in actin polymerization, and CtBP1/BARS is required for macropinosome closure (Liberali et al., 2008). The activation of PI3K and the engagement of Rho family GTPases are common to a variety of actin-dependent processes such as phagocytosis and chemotaxis. Thus, treatment with inhibitors like wortmannin and Clostridium difficile toxin B effectively blocks these processes, as well as macropinocytosis. In contrast, macropinosome formation appears to be uniquely susceptible to inhibition by amiloride and its analogues, and this property has been extensively used as an identifying feature of macropinocytosis (West et al., 1989; Veithen et al., 1996; Meier et al., 2002). Amiloride, a guanidinium-containing pyrazine derivative, has been used extensively as an inhibitor of Na+/H+ exchangers (NHEs; Grinstein et al., 1989; Orlowski and Grinstein, 2004). However, amiloride is not a universal nor a specific inhibitor of NHE: the affinity of the different NHE isoforms for amiloride varies greatly and, importantly, the drug also inhibits conductive Na+ channels and Na+/Ca2+ exchangers (Alvarez de la Rosa et al., 2000; Masereel et al., 2003). To increase the potency and selectivity of NHE inhibitors several amiloride analogues have been synthesized, including ethylisopropylamiloride (EIPA; Masereel et al., 2003) and (3-methylsulphonyl-4-piperidinobenzoyl)guanidine methanesulphonate (HOE-694), which is specific for the NHE1 isoform (Counillon et al., 1993). How amiloride inhibits macropinocytosis remains unknown. To the extent that EIPA also blocks macropinocytosis, NHEs are likely to play a role in the process (Cosson et al., 1989; West et al., 1989), but the mechanism linking ion exchange and vacuole formation is not apparent. Three possible mechanisms can be contemplated: (1) uptake of Na+ by the exchangers may increase the intracellular solute concentration, driving osmotically obliged water and causing swelling that would favor the protrusion of macropinocytic pseudopods. Though the stoichiometric exchange of Na+ for H+ is osmotically neutral, extruded H+ are replaced from intracellular buffers, resulting in a net osmotic gain; (2) NHE could be acting indirectly by altering the cytosolic concentration of calcium, which has been suggested to regulate macropinocytosis (Falcone et al., 2006). Na+ delivered intracellularly in exchange for H+ can promote the uptake of calcium via Na+/Ca2+ exchange; (3) the effect of NHE on macropinocytosis may be mediated by changes in cytosolic pH. Stimulation of NHE by hormones or growth promoters has been shown to alkalinize the cytosol (Rothenberg et al., 1983; L’Allemain et al., 1984; Grinstein et al., 1985; Van Obberghen-Schilling et al., 1985). Conversely, inhibition of the antiporters impairs the ability of cells to eliminate H+ generated metabolically and can cause acidification (L’Allemain et al., 1984, 1985; Grinstein et al., 1985; Liaw et al., 1998). The changes in pH resulting from modulation of NHE activity could conceivably alter the signaling and/or cytoskeleton rearrangements required for macropinocytosis. We investigated the functional relationship between macropinocytosis and Na+/H+ exchange. Macropinocytosis was induced in A431 cells by EGF, and NHE activity was modulated pharmacologically and by ion substitution. Moreover, we measured the bulk cytosolic pH and the pH of the inner aspect of the plasma membrane during the course of macropinocytosis. Our results indicate that NHE1 activity is required to attain a critical H+ concentration in the immediate vicinity of the plasma membrane that promotes actin polymerization during macropinocytosis. Results Inhibition of macropinocytosis by NHE antagonists A431 cells, which have been used extensively to study macropinocytosis, were chosen to investigate the mechanism of action of amiloride and its analogues. As reported previously (West et al., 1989; Araki et al., 2007; Liberali et al., 2008), addition of EGF to serum-depleted A431 cells led to extensive membrane ruffling and uptake of extracellular medium, visualized as trapping of the fluid-phase marker tetramethylrhodamine (TMR)-dextran (Fig. 1 A). The ruffling, which was apparent by differential interference contrast (DIC) microscopy (Video 1), was associated with extensive actin recruitment, revealed by staining with labeled phalloidin. These effects were most noticeable in the cells at the periphery of the subconfluent islands (Fig. 1 A). The increases in fluid phase uptake and actin polymerization were obliterated by pretreatment with either latrunculin B or with the PI3K inhibitor LY294002, consistent with mediation by macropinocytosis (Fig. 1 A). Figure 1. Effect of inhibitors on macropinocytosis and NHE activity. (A) DIC (left) and TMR-dextran epifluorescence images (middle) of islands of A431 cells incubated in the absence (Untreated) or presence of EGF as detailed in Materials and methods. Arrowheads point to dextran-filled macropinosomes. After determination of macropinocytosis, cells were fixed and stained with rhodamine-phalloidin to visualize actin (left). Arrowheads point to the aspect of the cell not in contact with neighboring cells. Bar, 10 µm. (B) Quantification of macropinocytosis in control and HOE-694-treated cells. Data are means ± SE of ≥5 separate experiments. (C) Effect of 10 µM HOE-694 on Na+-induced recovery of pHc after an acid load. NHE activity initiated where indicated by reintroduction of Na+. Results are representative of 3–4 similar experiments. (D) Concentration dependence of the effect of HOE-694. NHE activity was measured as in C and rates were calculated from the slopes from Na+-induced pHc recovery curves. Data are means ± SE of three experiments. Where missing, error bars are smaller than symbol. As illustrated in Fig. 1 A, the prototypical NHE inhibitor amiloride effectively inhibited EGF-induced fluid phase uptake and actin polymerization. Because at the concentrations used to inhibit Na+/H+ exchange amiloride has been reported to affect several other pathways (Alvarez de la Rosa et al., 2000; Masereel et al., 2003), we also tested HOE-694, a more selective NHE antagonist. As shown in Fig. 1, A and B, 10 µM HOE-694 greatly depressed macropinocytic activity. Parallel experiments verified that, at this concentration, HOE-694 eliminated Na+/H+ exchange. NHE activity was measured as the rate of Na+-induced recovery of the cytosolic pH (pHc) from an acid load. Ratiometric determinations of pHc using seminaphthorhodafluor dye-5 (SNARF-5F) demonstrated that when Na+ was reintroduced to the medium the cells recovered rapidly from a cytosolic acidification imposed by an ammonium prepulse. In the presence of 10 µM HOE-694, however, this response was completely eliminated (Fig. 1 C). At the submicromolar doses found to inhibit exchange in A431 cells (Fig. 1 D) HOE-694 selectively inhibits NHE1, with negligible effects on other isoforms (Counillon et al., 1993). Fig. 1, C and D therefore suggest that NHE1 is the main, if not the sole isoform active in the plasma membrane of A431 cells. For this reason, and to minimize off-target effects, HOE-694 was the inhibitor of choice in subsequent experiments. Changes in pHc during macropinocytosis EGF is known to stimulate Na+/H+ exchange and is capable of elevating pHc (Moolenaar et al., 1983; Rothenberg et al., 1983; Yanaka et al., 2002). The resulting alkalinization has been implicated in the initiation of the proliferative effects of EGF (L’Allemain et al., 1984; L’Allemain and Pouyssegur, 1986) and may similarly be required for macropinocytosis. This notion was tested by measuring the pHc changes elicited by the growth factor in the presence and absence of HOE-694. As shown in Fig. 2 A, A431 cells stimulated with EGF underwent a rapid and sizable (≥0.3 unit) alkalinization. In contrast, a net acidification was observed when cells were treated with EGF in the presence of maximally inhibitory doses of HOE-694. The rapid acidification likely results from the generation of acid equivalents by metabolic pathways stimulated by the growth factor. This burst of acid generation is normally not apparent because it is outstripped by the vigorous H+ extrusion mediated by Na+/H+ exchange and is only detectable when unmasked by inhibition of NHE1. Figure 2. Effect of HOE-694 on EGF-induced changes in pH. (A) SNARF-5F fluorescence ratio measurement of pHc. Where indicated by arrow A431 cells were stimulated with EGF in the absence (Control) or presence of HOE-694. Data are means ± SE of 3–6 experiments. (B) Top: schematic of the structure of membrane-targeted SEpHluorin/mCherry chimaera used to measure pHsm. Bottom: confocal images of SEpHluorin (left) and mCherry fluorescence (right) in A431 cells. Bar, 10 µm. (C) Representative pHsm calibration curve. Cells transfected with membrane-targeted SEpHluorin/mCherry were incubated in the presence of K+/nigericin buffers of predetermined pH. Fluorescence intensities were measured and the ratio of SEpHluorin/mCherry fluorescence is plotted as a function of pH. (D) Comparison of pHc (SNARF5-F and soluble SEpHluorin/mCherry) vs. pHsm (membrane-targeted SEpHluorin/mCherry) in cells treated with EGF for 10 min in Na+ medium in the presence and absence of HOE-694 (10 µM). Data are means ± SE of 3–5 experiments. *, P 80% between pH 7.4 and 6.8, without discernible change in the extent of receptor phosphorylation. This implies that downstream signaling events must be responsible for most of the pH dependence of macropinocytosis. Figure 5. Effect of NHE inhibition and of cytosolic pH on EGF receptor autophosphorylation. (A) Immunoblot analysis of tyrosine phosphorylation (P-Tyr) of EGF-R (Mw 170 kD) in A431 cells incubated for 5 min with or without EGF in Na+-rich buffer, with HOE-694 in Na+-rich buffer or in NMG+-rich buffer. Blot is representative of four experiments. (B) Quantitation of the effect of HOE-694 or NMG+ on EGF-R autophosphorylation, obtained by scanning immunoblots like the one in A (black bars). Data are means ± SE of 4–7 experiments. The effect of the same agents/conditions on macropinocytosis is shown for comparison (open bars). (C) Quantification of EGF-R phosphorylation in cells stimulated in Na+-rich medium or clamped with nigericin/K+ at the indicated pH (black bars). Data are means ± SE of 3–4 experiments. Data were normalized to controls in Na+-rich medium; normalized macropinocytosis is shown for comparison (open bars). ***, P 3). This was especially important in thin parts of the cell where fluorescence was low. Actin-free barbed end assay Actin-free barbed ends were determined by a modification of previously described methods (Chan et al., 1998; Frantz et al., 2008). In short, serum-starved A431 cells on coverslips were incubated with or without EGF in Na+-rich or pHc-clamping buffer for 1 or 3 min. To inhibit Rho GTPases, cells were incubated in the presence of C. difficile toxin B (50 ng/ml) for 3 h before EGF stimulation. To label FBEs the cells were permeabilized for 15 s in a buffer (20 mM Hepes, 140 mM NaCl, 3 mM KCl, 2 mM MgCl2, 2 mM EGTA, 5 mM glucose, 1% BSA, and 0.5 mM ATP, pH 7.5) containing 0.04% saponin and 0.02 µg/µl rhodamine-labeled rabbit skeletal muscle actin. After 15 s the solution was diluted with a 3x volume of permeabilization buffer without saponin and rhodamine-actin, and incubation continued for 3 min followed by fixation. The extent of FBE formation was calculated by measuring fluorescence intensity of a band 0.3–0.5 µm wide at the edge of the cell (the edge of the protruding lamellipod in stimulated cells) and a band of the same width (∼0.5 µm) inside the cell. The fluorescence intensity is reported as the ratio of the fluorescence at the edge to that in the cytosol, and comparison between experiments was facilitated by normalizing to the cytosolic fluorescence. Other methods Samples for Western blotting were scraped off the substratum in the presence of protease inhibitors (Sigma-Aldrich), 1 mM PMSF, 1 mM Na3VO4, and 0.1 µM okadaic acid, subjected to SDS-PAGE, and transferred to nitrocellulose filters which were then blocked with 5% BSA or milk in TBS-Tween. The primary antibody dilutions used were 1:10,000 for anti-P-Tyr, 1:5,000 for anti-P-Akt, 1:10,000 for anti-GAPDH, 1:1,000 for anti-Rac1 and Cdc42, 1:1,000 for anti–phospho-cofilin, and 1:10,000 for anti-cofilin. After incubating with horseradish peroxidase–conjugated secondary antibody the chemiluminescence of the immunoreactive bands was quantified using the Fluorchem FC2 chemiluminescence system (Alpha Innotech). To visualize actin, cells were fixed with 4% paraformaldehyde, permeabilized, and stained with rhodamine-phalloidin. Online supplemental material Fig. S1 shows cofilin localization during macropinocytosis. Fig. S2 shows the effect of cytosolic pH on the localization of surface charge probes. Video 1 is a DIC illustration of the effect of pHc clamping on membrane protrusions during EGF stimulation. Video 2 depicts Rac1 FRET ratio in pHc-clamped, EGF-stimulated cells. Online supplemental material is available at http://www.jcb.org/cgi/content/full/jcb.200908086/DC1.

Introduction Nosocomial infections contribute $4.5 billion to annual healthcare costs in this country alone, with an estimated 2 million nosocomial infections occurring in the US annually, resulting in 99,000 deaths [1]. Many of these nosocomial infections are caused by Gram-negative pathogens, and interaction of these pathogens with the host is often mediated by secreted virulence factors. Bacteria have evolved mechanisms for the secretion of virulence factors into the host cell to alter host cell biology and enable bacterial colonization, and these mechanisms typically require that bacteria be in intimate contact with the host. For example, the Type III secretion system (T3SS) and Type IV secretion system (T4SS) deliver proteins directly into the host cytoplasm from an extracellular bacterial pathogen's cytoplasm [2] utilizing transport machines that act as macromolecular syringes [3]. Delivery of extracellular bacteria or bacterial products can also occur via endocytosis initially into the lumen of the host endocytic compartment, then movement to the host cytoplasm via lysis of the endocytic compartment or delivery of the proteins across the endocytic membrane via the Type III Secretion System (T3SS) [3]. For several decades, work by Beveridge's group has characterized bacterial-derived outer membrane vesicles (OMV) to be a novel secretion mechanism employed by bacteria to deliver various bacterial proteins and lipids into host cells, eliminating the need for bacterial contact with the host cell [4]–[7]. OMV are 50–200 nm proteoliposomes constitutively released from pathogenic and non-pathogenic species of Gram-negative bacteria [8],[9]. Biochemical and proteomic analyses have revealed that OMV are comprised of lipopolysaccharide, phospholipids, outer membrane proteins, and soluble periplasmic proteins [8],[9]. Many virulence factors that are periplasmic proteins are enriched in OMV, for example, Escherichia coli cytolysin A (ClyA), enterotoxigenic E. coli heat labile enterotoxin (LT), and Actinobacillus actinomycetemcomitans leukotoxin [10]–[12]. Beveridge's group and others have reported that some secreted virulence factors from P. aeruginosa, including β-lactamase, hemolytic phospholipase C, alkaline phosphatase, pro-elastase, hemolysin, and quorum sensing molecules, like N-(3-oxo-dodecanoyl) homoserine lactone and 2-heptyl-3-hydroxy-4-quinolone (PQS) [6],[7],[13],[14], are also associated with P. aeruginosa OMV [8],[9]. Whether these secreted virulence factors packaged in OMV are eventually delivered to the host and the mechanism by which this occurs is currently unknown. A recent study suggested that E. coli OMV fuse with lipid rafts in the host colonic epithelial cell, but the delivery and intracellular trafficking of the OMV cargo was not characterized [15]. Thus, we investigated the possibility that OMV deliver multiple secreted virulence factors into the host cell through a lipid raft-mediated pathway, eliminating the need for intimate contact of the pathogen with the host. Results Outer membrane vesicles deliver toxins to airway epithelial cells Based on reports that multiple virulence factors are packaged in OMV, we hypothesized that these virulence factors could be simultaneously delivered in a coordinated manner in OMV to the host cell by the microbe. We tracked four P. aeruginosa secreted factors, including alkaline phosphatase, β-lactamase, hemolytic phospholipase C, and Cif, previously reported to be packaged in OMV [6],[7],[13],[14],[16]. We chose these secreted virulence factors because they play important roles in host colonization, for example alkaline phosphatase promotes biofilm formation [17],[18], β-lactamase degrades host antimicrobial peptides, hemolytic phospholipase C is cytotoxic and promotes P. aeruginosa virulence [19], and Cif is a recently characterized toxin that inhibits CFTR-mediated chloride secretion in the airways [16] and thereby likely reduces mucociliary clearance. To purify OMV from bacterial products not packaged in OMV, like pilus, that may also elicit a host response, we modified a published protocol [14] utilizing high-speed differential centrifugation and density gradient fractionation to isolate OMV from an overnight P. aeruginosa culture supernatant. The Cif protein, as well as a protein present in the membrane of OMV, Omp85 [20], were identified in purified bacterial-derived OMV (Figure S1). When airway cells were treated with isolated and purified OMV for ten minutes all four OMV proteins examined were detected in host airway epithelial cell lysate (Figure 1A). By contrast, these virulence factors were not detected in lysates of control cells treated with vehicle (Figure 1A). Therefore, OMV deliver multiple virulence factors to host airway epithelial cells in the absence of bacteria, thus providing a mechanism for bacteria to alter host cell physiology without the need for intimate contact with the host. 10.1371/journal.ppat.1000382.g001 Figure 1 OMV deliver multiple toxins to induce cytotoxicity in airway epithelial cells. (A) Purified P. aeruginosa OMV deliver multiple virulence factors (Cif, alkaline phosphatase, β-lactamase, and PlcH) into the cytoplasm of airway epithelial cells. Western blot analysis of cell lysates of airway epithelial cells treated with purified OMV. CTRL, no OMV added. Δcif OMV applied to airway epithelial cells, cells lysed, and probed by Western blot analysis for Cif did not detect Cif in airway cell lysates (data not shown). (B) OMV elicit time-dependent cytotoxicity in host airway cells as determined by the CellTiter 96 AQueous One cytotoxicity assay. (C) OMV elicit time-dependent cytotoxicity in host airway cells that is not dependent on the presence of Cif, as determined by the CellTiter 96 AQueous One cytotoxicity assay. Δcif OMV, OMV purified from a strain of P. aeruginosa lacking the cif gene. Black bars, wild-type OMV (WT OMV); white bars, Δcif OMV. (D) Airway epithelial cells treated with lysed OMV components demonstrated a reduction in cellular cytotoxicity as determined by the CellTiter 96 AQueous One cytotoxicity assay. Data are presented as mean+/−SEM. n = 3, * p<0.05, OMV versus Control, intact OMV versus lysed OMV. To explore the significance of OMV in the delivery of virulence factors into the host cytoplasm, we examined the cytotoxic effect of P. aeruginosa OMV on host airway cells using the CellTiter 96 AQueous One cytotoxicity assay. OMV were cytotoxic after a delay of 8 hours (Figure 1B), although virulence factors could be detected in the cytoplasm of host cells after 10 minutes (Figure 1A). The time-dependent increase in cytotoxicity induced by OMV was not dependent on Cif expression in OMV, given that Δcif OMV did not produce a statistically significant difference in cytotoxicity compared to wild-type OMV (Figure 1C). To determine if intact OMV are required for cytotoxicity, purified OMV were lysed with 0.1 M EDTA and the lysate was applied to airway epithelial cells for 8 h (Figure S2, Figure 1D). This method was previously employed by Horstman et al. to effectively lyse E. coli OMV [21]. The lysed OMV did not have a cytotoxic effect on airway epithelial cells, demonstrating that cytotoxicity is mediated by virulence factors delivered into the host cell cytoplasm by bacterial-derived OMV (Figure 1D). In the next series of experiments we began to examine the mechanism whereby OMV deliver virulence factors into the cytoplasm of the host airway epithelial cell. Previously we reported that purified, recombinant Cif, a virulence factor secreted in OMV by P. aeruginosa, is necessary and sufficient to reduce apical membrane expression of CFTR and P-glycoprotein (Pgp) in human airway epithelial cells [16],[22], thus reducing mucociliary clearance and xenobiotic resistance of the host cells, respectively. In the current study we use Cif as a model protein to investigate how OMV deliver virulence factors into the cytoplasm of human airway epithelial cells. First, experiments were conducted to confirm our previous observation that Cif is secreted in purified OMV and second, to determine if Cif is an intravesicular component of OMV. Cif was detected in OMV derived from P. aeruginosa expressing the cif gene, but not in OMV derived from P. aeruginosa in which the cif gene was deleted (Figure S3). The inability of Proteinase K (0.1 µg/ml), which does not enter the lumen of OMV, to degrade OMV-associated Cif indicates that this virulence factor is an intravesicular component of OMV (Figure S3). Thus, these studies demonstrate that Cif maintains an intravesicular localization in purified OMV. Next, studies were conducted to determine if the Cif virulence factor packaged in OMV was functional when delivered to airway epithelial cells. Cif function was measured by examining the ability of Cif to reduce apical plasma membrane CFTR abundance in airway epithelial cells. Purified OMV containing Cif reduced apical plasma membrane CFTR in a time-dependent manner (Figure 2A), whereas purified OMV from P. aeruginosa deleted for the cif gene had no effect on CFTR membrane expression (Figure 2B). Taken together these studies confirm and extend our previous observations that OMV-packaged Cif reduces plasma membrane CFTR. 10.1371/journal.ppat.1000382.g002 Figure 2 OMV deliver functional Cif virulence factor to airway epithelial cells, across a mucus layer. (A) Purified OMV applied to the apical surface of airway epithelial cells reduce CFTR in the plasma membrane in a time-dependent manner, compared to the buffer control (Control), as measured by Western blot analysis. Overnight supernatant (PA14 O/N Sup), previously reported to decrease apical membrane CFTR, serves as a positive control [16]. (B) OMV purified from a P. aeruginosa Δcif mutant strain did not reduce apical membrane expression of CFTR, as measured by Western blot analysis. Black bar, wild-type OMV; striped bar, Δcif OMV. (C) Cif-containing OMV decrease apical membrane CFTR in airway epithelial cells (CFBE) and mucous-producing airway epithelial cells (Calu-3), as measured by Western blot analysis. Black bar, CFBE cells; striped bar, Calu-3 cells. Data are presented as mean+/−SEM. n = 3, * p<0.05 versus Control. If OMV modulate host physiology of lung cells without direct bacteria-host contact, we would predict that OMV secreted by bacteria should overcome barriers such as the mucus overlying human airway cells [23]. OMV containing Cif reduced CFTR in the apical plasma membrane of airway epithelial cells that have a thick layer of mucus on the apical surface (Calu-3 cells, Figure 2C). A delay in the Cif-mediated reduction of apical membrane CFTR abundance was observed in Calu-3 cells, compared to airway cells that lack a overlying mucus layer, suggesting the mucus only delays OMV from diffusing to host airway epithelial cells. Thus, OMV allow the long distance delivery of secreted bacterial factors to the host cell in the absence of direct bacteria-host contact. Host cell detergent-resistant membranes (aka, lipid rafts) are required for OMV fusion and toxin delivery We next explored the mechanism whereby OMV deliver bacterial proteins into the host cell using the Cif virulence factor as a model. Based on a recent study showing that Filipin III disrupts E. coli OMV association with host cells [15], we hypothesized that OMV deliver secreted bacterial proteins to host cells by fusing with lipid raft microdomains. Five minutes after addition of OMV to epithelial cells, Cif, as well as a protein documented to be associated with OMV, Omp85 [20], were detected in membrane lipid raft fractions (Figure 3A). The lipid raft (i.e. detergent insoluble membranes) fractions were separated with density gradient fractionation and characterized by labeling with the flotillin-1 antibody. The fusion of OMV with membrane rafts was observed visually by confocal microscopy using cholera toxin B subunit (labeled with FITC), a documented lipid raft marker [24], which co-localized with rhodamine-R18 labeled OMV five minutes after OMV were added to the apical side of airway cells (Figure 3B). The rhodamine-R18 dye is quenched when loaded in bilayer membranes at a high concentration and is subsequently dequenched, fluorescing in the red channel upon membrane fusion, which allows dilution of the probe and fluorescence detection. Pearson's correlation and Mander's overlap coefficients demonstrated a high degree of co-localization of cholera toxin B subunit and OMV (0.771+/−0.018 and 0.952+/−0.012 versus control levels of 0.153+/−0.026 and 0.259+/−0.026, respectively, p<0.0001). In further support that Cif-containing OMV fuse with lipid rafts, Cif co-immunoprecipitated with the glycosylphosphatidylinisotol-anchored protein p137 [25], a documented lipid raft-associated protein, from the lipid raft fraction of airway cells that had been treated with OMV for five minutes (Figure 3C). 10.1371/journal.ppat.1000382.g003 Figure 3 Cif-containing OMV fuse with the epithelial cell lipid raft microdomains. (A) Cif virulence factor and Omp85 localize to the lipid raft fraction. Omp85, a documented OMV protein, and Cif virulence factor co-fractionate with the lipid raft marker, flotillin-1, when raft and non-raft fractions are isolated from airway epithelial cells treated with OMV for 5 min, as measured by Western blot analysis. Right blot represents raft and non-raft fractions from the left blot, which are combined and the proteins precipitated to concentrate Cif signal to allow antibody detection. (B) Rhodamine R18-labeled OMV, which only fluoresce red upon fusion with the host plasma membrane, co-localize with the FITC-conjugated lipid raft marker cholera toxin B subunit (CtxB). All OMV co-localize with lipid rafts with the exception of five OMV that are indicated with small white arrows (see lower right panel). Scale bar equals 10 µm. (C) Cif virulence factor interacts with GPIp137 in the lipid raft fraction. Immunoprecipitation (IP) experiments demonstrate that the GPI-anchored protein p137 (GPIp137), a lipid raft marker, interacts with Cif in the lipid raft fractions of airway epithelial cells treated with wild-type OMV for 5 min. Serving as a negative control, cif mutant OMV (Δcif OMV)–treated airway epithelial cells cannot immunoprecipitate Cif and therefore do not demonstrate immunoprecipitation of GPIp137. 5% of total lysate run in Lysate lane. IB, immunoblot. Experiments repeated three times; representative blots/images are shown. To determine if host cell lipid raft microdomains are required for OMV fusion, the cholesterol-sequestering agent Filipin III complex was used to disrupt lipid raft domains and OMV fusion was assessed. The rhodamine-R18 dye was utilized to allow visualization and quantitation of OMV fusion with host cells. Rhodamine-R18 only fluoresces upon OMV fusion to host cells, thus an increase in fluorescence is interpreted as an increase in OMV fusion. Rhodamine-R18 labeled OMV applied to the apical membrane of airway epithelial cells produced a time-dependent increase in fluorescence (Figure 4A). In contrast, the fluorescence did not increase above background levels in samples containing only airway epithelial cells or only rhodamine-R18 labeled-OMV (Figure 4A). Filipin III eliminated the fusion of OMV with epithelial cells, indicated by a lack of fluorescence detected when compared to control epithelial cells (Figure 4B). Microscopy studies were confirmed by the quantitative, fluorescence-based assay, described in Figure 4A, which also demonstrated that OMV fusion to the host cell was blocked with Filipin III pretreatment of the host cells (Figure 4C). 10.1371/journal.ppat.1000382.g004 Figure 4 Disruption of lipid rafts blocks virulence factor delivery and function in airway epithelial cells. (A) Rhodamine-R18 labeled OMV were applied to the apical side of polarized airway epithelial cells, and fluorescence was measured over time as readout for OMV fusion. Rhodamine-R18 signal is quenched at high concentration in OMV, but fluorescence increases as the rhodamine probe is diluted by fusion of the OMV with the host plasma membrane. OMV and cells alone serve as negative controls where the fluorescence does not change over time (background fluorescence). Black square, OMV+cells; black triangle, OMV alone; black inverted triangle, cells alone. Data are presented as mean fluorescence intensity. (B,C) Visualization and quantitative assay demonstrating that disruption of lipid rafts with the cholesterol-sequestering reagent, filipin III complex (10 µg/ml), inhibits OMV fusion with airway epithelial cells. Rhodamine-R18 labeled OMV were applied to the apical side of polarized airway epithelial cells in the presence or absence of filipin III, and fluorescence was measured over time. Wheat germ agglutin (WGA-blue) is a marker of host epithelial cell membranes. Scale bars equal 10 µm. Black square, control; black triangle, filipin III. (D) Filipin III prevents delivery of Cif to host cell endosomes, as measured by cytoplasmic and endosomal fractionation and Western blot analysis. Rab5 GTPase and actin are endosomal and cytoplasmic markers, respectively. Cyto, cytoplasm; Endo, endosome. (E) Disruption of lipid rafts blocks Cif-mediated reduction in apical membrane CFTR expression, compared to buffer control (Control), as measured by Western blot analysis. “Centrifuge Sup” refers to the overnight culture supernatant of P. aeruginosa PA14 depleted of OMV, which serves as a negative control in this experiment. (F) Filipin III complex disruption of host cell lipid raft microdomains reduces the cytotoxic effect of OMV on host airway cells with 8 h OMV treatment, as measured with the CellTiter 96 AQueous One cytotoxicity assay. Data are presented as mean+/−SEM. n = 3, * p<0.05, OMV versus Control; # p<0.05, OMV versus OMV+Filipin III. We next tracked the Cif virulence factor biochemically to determine if lipid raft microdomains are required for virulence factor delivery and function in host cells. Five minutes after OMV are exposed to host airway epithelial cells, the Cif virulence factor is detected by Western blot analysis in the endosomal sub-fraction of the host cell lysate, but not in the cytoplasmic fraction (Figure 4D). The Filipin III complex prevented the appearance of Cif in the endosomal fraction of host cells (Figure 4D) and blocked the ability of Cif to reduce apical membrane CFTR (Figure 4E), demonstrating a requirement for the host lipid raft machinery for Cif delivery and function in host cells. Furthermore, disruption of lipid raft microdomains with Filipin III, and thus blocking OMV fusion with airway epithelial cells, reduced the cytotoxicity induced in the airway epithelial cells with 8 h of OMV treatment (Figure 4F). Thus, lipid raft microdomains are required for OMV-mediated delivery and function of secreted virulence factors in host cells. Actin cytoskeleton is required for OMV fusion, toxin entry, and function The actin cytoskeleton, in particular the neuronal WASP (N-WASP)–initiated actin assembly, is critical to the internalization of select lipid raft-associated cargo [26],[27]. Based on these previous studies, we investigated the role of the actin cytoskeleton, in general, and N-WASP-mediated cytoskeletal rearrangements specifically, in OMV fusion to the plasma membrane of the airway epithelial cell. Both cytochalasin D (an actin monomer-sequestering agent) and wiskostatin (an inhibitor of neuronal WASP (N-WASP) induced actin polymerization) disrupted the actin cytoskeleton in airway epithelial cells (Figure 5A), resulting in a loss of OMV fusion (Figure 5B,C), as measured by a reduction in OMV-dependent fluorescence in airway cells pretreated with cytochalasin D or wiskostatin. Thus, wiskostatin and cytochalasin blocked OMV fusion to human airway epithelial cells, demonstrating a need for N-WASP induced actin polymerization for OMV fusion to host cells. 10.1371/journal.ppat.1000382.g005 Figure 5 Intact N-WASP–induced actin cytoskeleton is required for OMV fusion and virulence factor delivery. (A) Wiskostatin and cytochalasin D disrupt actin cytoskeleton in airway cells. Alexa 647–conjugated phalloidin was used to label actin to monitor disruption of the actin cytoskeleton by two actin-disrupting agents, wiskostatin (Wisko, 10 µM for 30 min) and cytochalasin D (Cyto D, 2 µM for 45 min), by immunofluorescence microscopy. Scale bar equal to 20 µm. (B,C) Visualization and quantitative assay demonstrating that disruption of the actin cytoskeleton with either of the two actin-disrupting agents, wiskostatin (Wisko, 10 µM) or cytochalasin D (Cyto D, 2 µM), inhibits OMV fusion with airway epithelial cells. Rhodamine-R18 labeled OMV were applied to the apical side of polarized airway epithelial cells in the presence or absence of wiskostatin or cytochalasin D, and fluorescence was measured over time. Scale bar equals 10 µm (B). Representative images of three experiments revealing that OMV do not fuse with cells in which actin has been disrupted. WGA, wheat germ agglutin (blue) to label the cell surface. Black square, control; black triangle, Cytochalasin D; open circle, Wiskostatin. * p<0.05 (Control versus Cytochalasin D and Wiskostatin). (D) Wiskostatin or cytochalasin D treatment of airway epithelial cells prevents delivery of Cif virulence factor to endosomal fraction, as measured by cytoplasm and endosome fractionation and Western blot analysis. Rab5 GTPase and actin are endosomal and cytoplasmic markers, respectively. Cyto, cytoplasm; Endo, endosome. (E) Disruption of the actin cytoskeleton with wiskostatin blocks the Cif virulence factor–mediated reduction in CFTR apical membrane expression in airway epithelial cells, as measured by Western blot analysis. Data are presented as mean+/−SEM. n = 3, * p<0.05 versus Control. Wiskostatin and cytochalasin D were also utilized to determine if the actin cytoskeleton is required for OMV delivery of Cif to airway epithelial cells. Cytoplasmic and endosomal fractions were purified from airway epithelial cells pretreated with vehicle, cytochalasin D or wiskostatin in the presence or absence of Cif-containing OMV. In control cells Cif localized to the endosomal fraction, as described above, whereas cytochalasin D and wiskostatin (Figure 5D) blocked the entry of Cif into the endosomal and cytoplasmic fractions. Furthermore, wiskostatin pretreatment blocked the Cif toxin-mediated reduction of CFTR from the apical membrane of airway epithelial cells (Figure 5E). Because cytochalasin D changes the rate of CFTR endocytosis, we cannot assess the effects of this inhibitor on Cif virulence factor-mediated reduction of CFTR. In addition, the purified Cif virulence factor alone did not induce morphological changes to the actin cytoskeleton (data not shown). These results reveal that Cif does not alter the cytoskeleton and establishes the requirement for an intact actin cytoskeleton, specifically N-WASP-mediated actin polymerization, for OMV fusion and virulence factor delivery to the host airway epithelial cells. Cif toxin is localized to the cytoplasmic face of early endosomes after entry into host cell The data above strongly suggest that OMV deliver the Cif virulence factor into the interior of the host cell and allow this virulence factor to associate with an endosomal compartment (Figures 1A, 4D and 5D). To more precisely identify which endosomal compartment was the target of Cif, OMV were applied apically to airway cells for ten minutes, the airway cells were lysed and endosomes were purified by differential centrifugation. From the purified endosomal fraction, Cif co-immunoprecipitated with Rab5 GTPase, a marker of early endosomes, and the early endosomal antigen (EEA)-1 (Figure 6A). In contrast, Cif did not co-immunoprecipitate with Rab4 (a marker of sorting endosomes), Rab7 (a marker of late endosomes) or Rab11 (a marker of recycling endosomes) (Figure S4). Proteinase K, which does not degrade luminal endosomal proteins but can degrade proteins on the cytoplasmic face of endosomes, eliminated Cif from the endosomal fraction (Figure 6B). As expected, proteinase K did not affect the endosomal association of the transferrin receptor, a luminal endosomal protein that is resistant to proteinase K treatment [28]. However, the transferrin receptor was not resistant to proteinase K degradation in the presence of 0.1% Triton X-100, which disrupts the endosomal membrane and allows proteinase K access to luminal endosomal proteins (Figure 6B). These data reveal that OMV-delivered Cif is localized to the cytoplasmic face of the early endosomes after entry into the epithelial cell. 10.1371/journal.ppat.1000382.g006 Figure 6 Cif virulence factor is delivered to the cytoplasm and localizes to the cytoplasmic face of early endosomes. (A) Cif localizes to the early endosomal (Rab5 GTPase, early endosomal antigen (EEA)-1 labeled) compartment after entry into airway epithelial cells. Airway epithelial cells were treated with OMV for 10 min, cells lysed, and endosomes were purified. Cif was immunoprecipitated from the endosomal fraction, and Western blot analysis was performed for Rab5 GTPase and EEA-1, early endosomal markers. IgG IP is a non-immune control immunoprecipitation experiment. (B) The ability of proteinase K (PK) to degrade Cif from the early endosomal fraction reveals that the Cif virulence factor localizes to the cytoplasmic face of early endosomes, as measured by Western blot analysis. The transferrin receptor serves as a control for a luminal endosomal protein marker, which is only exposed to proteinase K after treatment with Triton X-100 (TX). (C) Cif entry into airway epithelial cells is not altered by disruption of endosomal acidification by NH4Cl (5 mM). Rab5 GTPase and actin are endosomal and cytoplasmic markers, respectively. (D) Entry of Cif virulence factor into airway cells is unaffected by inhibition of retrograde transport with Brefeldin A (1 µM), as measured by Western blot analysis. Rab5 GTPase and actin are endosomal and cytoplasmic markers, respectively. (E) Retrograde transport to the endoplasmic reticulum is not required for Cif to reduce plasma membrane CFTR in airway cells. Airway epithelial cells pretreated with Brefeldin A (1 µM), which inhibits retrograde transport, were treated with OMV. Brefeldin A had no effect on the ability of Cif to reduce plasma membrane CFTR, as measured by Western blot analysis. Data are presented as mean+/−SEM. n = 3, * p<0.05 versus Control. To determine if OMV-delivered Cif enters the host cell cytoplasm by penetrating the membrane of endosomal vesicles, cells were treated with ammonium chloride, a lysosomotropic drug that inhibits vesicle acidification, and thereby inhibits the movement of virulence factors from endosomal vesicles into the cytoplasm [3]. Ammonium chloride had no effect on the ability of Proteinase K to decrease the amount of Cif in the endosomal fractions (Figure 6C), indicating that the Cif virulence factor does not reach the cytoplasm via penetrating intracellular vesicular membranes. Some intracellular bacteria and virulence factors move through the retrograde pathway from endosomes, to the Golgi apparatus and endoplasmic reticulum, from which they enter the host cytoplasm. However, Brefeldin A, a pharmacologic inhibitor of retrograde transport, had no effect on the entry of the Cif into the airway cell and the appearance of Cif in the endosomal fraction (Figure 6D), or the Cif-mediated reduction in apical membrane CFTR abundance (Figure 6E). Thus, our data demonstrate that OMV deliver Cif directly to the host cytoplasm rather than requiring passage across an endosomal membrane or through the retrograde transport pathway. Interestingly, PlcH and alkaline phosphatase also localized to the endosomes after entry into the airway epithelial cells, whereas β-lactamase was detected in the cytoplasmic fraction, as determined by subcellular fractionation and Western blot analysis (data not shown). Thus, virulence factors with differing functions are distributed to different subcellular locations after entry into the host cytoplasm. OMVs are a physiological delivery mechanism for secreted virulence factors We propose that rather than secretion of virulence factors into the surrounding medium, OMV are a physiologically- and clinically-relevant mechanism utilized by Gram-negative bacterium, in particular P. aeruginosa, to deliver secreted products into the host cell. In support of this hypothesis, Cif packaged in OMV was 17,000-fold more effective than purified, recombinant Cif in reducing plasma membrane CFTR, with 3 ng of Cif in OMV- reducing plasma membrane CFTR expression as effectively as 50 µg of purified, recombinant Cif protein (Figure 7A). Cif was detected in lysates of airway epithelial cells exposed to OMV (3 ng Cif) and 50 µg of recombinant Cif (Figure 7B), but Cif was not detected in cells exposed to up to 10 ng of recombinant Cif, correlating the presence of the virulence factor inside the host cell with virulence factor function. Moreover, airway epithelial cells treated with lysed OMV (Figure S2) showed a dramatic reduction in the ability of the Cif toxin to reduce apical membrane CFTR, as compared to cells treated with intact OMV (Figure 7C). Therefore, OMV-mediated delivery of virulence factors to airway epithelial cells increases the efficacy of these virulence factors in altering host cell physiology. 10.1371/journal.ppat.1000382.g007 Figure 7 OMV required for virulence factor delivery to host airway cells. (A) OMV delivery enhances the ability of Cif to reduce plasma membrane CFTR in airway cells by 17,000-fold. Cif was applied to airway epithelial cells as 1 ng, 10 ng, 50 µg recombinant protein or 3 ng in OMV, and its effect on CFTR expression at the apical plasma membrane was measured by Western blot analysis. (B) Cif was detected in airway cells treated with 50 µg Cif or 3 ng Cif packaged in OMV. Airway epithelial cells were treated with vehicle (Control-CTRL), 1 ng, 10 ng, or 50 µg of recombinant Cif protein, or with 3 ng of Cif packaged in OMV (OMV). Cif was detected in the airway cell cytoplasm by Western blot analysis. The presence of Cif protein in the airway cell cytoplasm correlated with the ability of Cif to reduce plasma membrane CFTR (A). (C) Intact OMV required for reduction of apical membrane CFTR expression (i.e., Cif virulence factor function). Airway epithelial cells were treated with OMV or lysed OMV components, and airway epithelial cells were assayed for the Cif-mediated reduction in apical membrane CFTR, as measured by Western blot analysis. Data are presented as mean+/−SEM. n = 3, * p<0.05 versus Control. Discussion We have demonstrated that P. aeruginosa OMV deliver multiple virulence factors, simultaneously, into host airway epithelial cells via a mechanism of OMV fusion with host cell lipid raft machinery and trafficking via an N-WASP induced actin pathway to deliver OMV cargo directly to the host cytoplasm. The OMV-delivered Cif virulence factor is then localized to the cytoplasmic face of the early endosomal compartment (Figure 8). E. coli OMV association with host cells had previously been shown to be sensitive to Filipin III treatment, and thus was proposed to be lipid raft-dependent, but whether the OMV actually delivered cargo into the host cells and the mechanism by which this occurred was not characterized [15]. Fiocca et al. demonstrated that VacA packaged in OMV from H. pylori was internalized into a cytoplasmic vacuole in gastric epithelial cells, but did not investigate a mechanism [29]. We propose the first mechanism for the entry and intracellular fate of OMV-delivered bacterial virulence factors. 10.1371/journal.ppat.1000382.g008 Figure 8 Proposed model for P. aeruginosa OMV fusion with airway epithelial cells. OMV are released from P. aeruginosa, diffuse to the host cell plasma membrane, and subsequently fuse with host cell lipid raft microdomains. Virulence factors are subsequently released into the cytoplasm of the airway epithelial cell, through an actin-dependent pathway. Considering the pioneering work of Beveridge to characterize OMV and our current mechanistic studies, we propose that OMV-mediated virulence factor delivery should be considered for designation as a secretion system [4]–[7]. Like the T3SS, OMV can deliver bacterial proteins directly to the host cell cytoplasm without releasing the naked bacterial proteins into the extracellular environment where they could be degraded by secreted proteases [30]–[32]. OMV deliver fully-folded, enzymatically-active secreted virulence factors into host cells, ready for immediate action upon delivery. By delivering multiple, active OMV-packaged virulence factors, the pathogen may be able to impact the host on multiple levels. For example, simultaneously altering epithelial cell function by perturbing surfactant abundance or tight junction integrity, and the innate immune response to bacteria by stimulating pro-inflammatory cytokine production [14], [33]–[36]. Based on our studies in P. aeruginosa and published reports of OMV production by E. coli, H. pylori, A. actinomycetemcomitans, V. cholerae and N. meningitidis, it is likely that other bacteria package multiple secreted virulence factors in OMV for efficient transfer to host cells and thus, the studies proposed here likely represent a general strategy utilized by Gram-negative bacteria in their interactions with the host [10]–[12],[37],[38]. In contrast to known secretion systems, OMV-mediated direct delivery of bacterial proteins to the host can occur at a distance, and in the absence of bacteria, thus obviating the need for the pathogen to interact directly with the host cell to cause cellular cytotoxicity and alter host cell biology to promote colonization. Furthermore, OMV can deliver bacterial factors across host barriers, such as mucus layers. We believe that our work should prompt those studying bacterial pathogens to reconsider how secreted virulence factors impact host cells. That is, our data suggest that secreted virulence factors are not released individually into the surrounding milieu where they may randomly contact the surface of the host cell, but are released in a strategic manner, packaged with multiple virulence factors in OMV for coordinated delivery directly into the host cell cytoplasm. It is also possible that OMV provide a mechanism for delivering a concentrated bolus of virulence factors to the host, instead of individual toxins being delivered one at a time to the host cell. Moreover, OMV-mediated, long distance delivery of virulence factors might help explain observations such as, bacterial colonization of catheters causing systemic symptoms in kidney dialysis patients, ocular keratitis occurring in patients who do not have cultivatable pathogens, and the significant lung damage in cystic fibrosis, bronchiectasis, and chronic obstructrive pulmonary disease patients resulting from chronic infections with P. aeruginosa suspended in mucus above the airway epithelium. This mechanism of OMV-mediated protein secretion is reminiscent of the long distance delivery of signaling proteins between and among eukaryotic cells via exosomes [39], and may represent a general protein secretion strategy used by both pathogen and host. Materials and Methods Antibodies and reagents The antibodies used were: rabbit anti-Cif antibody (Covance Research Products, Denver, Pa [16]); rabbit anti-OprF antibody (a generous gift from Nobuhiko Nomura, Graduate School of Life and Environmental Sciences, University of Tsukuba); rabbit anti-pilus antibody (a generous gift from Michael Zegans, Dartmouth Medical School); goat anti-phospholipase C-H antibody (a generous gift from Michael Vasil); mouse anti-human CFTR C-terminus antibody (clone 24-1; R&D systems, Minneapolis, MN); mouse anti-CFTR antibody (clone M3A7; Upstate Biotechnology, Lake Placid, NY); mouse anti-EEA1 antibody, mouse anti-ezrin antibody, mouse anti-flotillin-1 antibody, mouse anti-Rab5 antibody, mouse anti-actin antibody (BD Biosciences, San Jose, CA); cholera toxin B subunit-FITC (Sigma-Aldrich, St. Louis, MO); rabbit anti-GPIp137 antibody (Abgent, San Diego, CA); Alexa 647-conjugated phalloidin (Molecular Probes, Carlsbad, CA); rabbit anti-Rab4 antibody, rabbit anti-Rab7 antibody (Santa Cruz Biotechnology, Santa Cruz, CA); mouse anti-Rab11 antibody, mouse anti-transferrin receptor antibody (Zymed, San Francisco, CA); mouse anti-β lactamase antibody (Novus Biologicals, Littleton, CO); rabbit anti-alkaline phosphatase antibody (GeneTex, Inc., San Antonio, TX) and horseradish peroxidase-conjugated goat anti-mouse and goat anti-rabbit secondary antibodies (Bio-Rad, Hercules, CA). Other reagents include: Filipin III complex, ammonium chloride, Optiprep, proteinase K, and cytochalasin D (Sigma-Aldrich), wiskostatin (Calbiochem, San Diego, CA), Triton X-100 (Bio-Rad, Hercules, CA). All antibodies and reagents were used at the concentrations recommended by the manufacturers or as indicated in the figure legends. Cell culture Two airway epithelial cell lines were studied to examine outer membrane vesicle fusion and toxin delivery to host epithelial cells. First, human bronchial epithelial CFBE cells (ΔF508/ΔF508) were stably transduced with WT-CFTR (generous gift from Dr. J. P. Clancy, University of Alabama at Birmingham, Birmingham, AL; hereafter referred to as airway epithelial cells) [40]. CFBE WT-CFTR cells were polarized on 24-mm transwell permeable supports (0.4-µm-pore size; Corning, Corning, NY) coated with vitrogen plating medium containing human fibronectin, as described previously [41]. Second, human airway epithelial cells (Calu-3) were obtained from the American Type Culture Collection (Manassas, VA) and polarized on 24-mm transwell permeable supports, as described previously [42]. Pseudomonas aeruginosa cultures Lysogeny broth (LB) was inoculated with P. aeruginosa strain UCBPP-PA14 (PA14) [43] and cultures were prepared as previously reported [16]. Outer membrane vesicle purification OMV were purified using a differential centrifugation and discontinuous Optiprep gradient protocol adapted from Bauman et al. [14] OMV were lysed, when noted, with 100 mM EDTA at 37°C for 60 minutes. Cell compartment fractionation To study the localization of the Cif toxin after OMV fusion with the airway epithelial cell, differential centrifugation and fractionation techniques were used to isolate cytosolic and early endosomal compartments. Early endosomes were isolated using a protocol adapted from Butterworth et al. [44]. Immunoprecipitation To characterize proteins interacting with the Cif toxin in lipid raft microdomains, Cif was immunoprecipitated from airway epithelial cell lipid raft fractions by methods described previously [45]. Detergent-resistant membrane fractionation To determine if OMV fuse with lipid raft microdomains of the host, detergent-resistant membranes were purified from airway epithelial cells that had been exposed to OMV. These studies were performed using a discontinuous Optiprep gradient in a protocol adapted from Pike et al. [46]. OMV fusion assay To monitor the fusion of OMV with airway epithelial cells, OMV were fluorescently labeled with a probe that fluoresces upon membrane fusion. OMV purified with the method described above were resuspended in labeling buffer (50 mM Na2CO3, 100 mM NaCl, pH 9.2). Rhodamine isothiocyanate B-R18 (Molecular Probes), which integrates in the membrane of the OMV, was added at a concentration of 1 mg/ml for 1 hour at 25°C, followed by ultracentrifugation at 52,000×g for 30 min at 4°C. Rhodamine isothiocyanate B-R18 fluorescence is quenched at high concentrations in bilayer membranes, and fluorescence is dequenched when the probe is diluted upon vesicle fusion. Subsequently, rhodamine labeled-OMV were resuspended in PBS (0.2 M NaCl) and pelleted at 52,000×g for 30 min a 4°C. After a final centrifugation step, the labeled-OMV were resuspended in 1 ml PBS (0.2 M NaCl) containing a protease inhibitor cocktail tablet (Complete Protease Inhibitor Tablet, Roche). Labeled-OMV were applied to the apical side of airway epithelial cells at 1∶4 dilution of labeled-OMV to Earle's Minimal Medium (MEM, Invitrogen) and fluorescence was detected over time as indicated on a fluorescent plate reader (Ex 570 nm; Em 595 nm). Fluorescence intensity was normalized for fluorescence detected by labeled-OMV in the absence of airway epithelial cells at the indicated time points. Confocal microscopy To visualize the fusion and localization of OMV with airway epithelial cells, rhodamine R18-labeled OMV (see OMV Fusion Assay method) were applied to the apical membrane of cells and confocal sections were captured over time. Airway epithelial cells were seeded at 0.1×106 on collagen-coated, glass-bottom MatTek dishes (MatTek, Ashland, MA) and grown for 6–7 days in culture at 37°C. For wheat germ agglutin (WGA, which labels the plasma membrane) studies, nonpermeabilized cells were incubated for 5 minutes with Alexa-647 WGA (1 µg/ml, 37°C; Molecular Probes) following 15-minute vesicle incubation. Z-stacks of all labeled cells were acquired with a Nikon Sweptfield confocal microscope (Apo TIRF 60× oil immersion 1.49 NA objective) fitted with a QuantEM:512sc camera (Photometrics, Tuscon, AZ) and Elements 2.2 software (Nikon, Inc.). For OMV fusion experiments, a single confocal section (0.4 µm) at the apical membrane of the airway epithelial cells is presented. Experiments were repeated three times, with five fields imaged for each experiment. Cell-surface biotinylations and Western blot analysis To examine the effect of OMV on the apical membrane expression of CFTR, cell surface biotinylation was performed as described in detail previously by our laboratory [41]. Protein band intensity was analyzed as described previously using NIH image software, version 1.63 (Wayne Rasband, NIH, USA; http://rsb.info.nih.gov). Cytotoxicity assay To determine if P. aeruginosa OMV are cytotoxic to airway cells, cells were incubated with OMV in serum-free media for the indicated time points. Cytotoxicity was measured using the CellTiter 96 AQueous One Solution Reagent (Promega, Madison, WI), according to the manufacturer's protocol. Data analysis and statistics Statistical analysis of the data was performed using Graphpad Prism version 4.0 for Macintosh (Graphpad, San Diego, CA). Means were compared using a Students t-test or one-way ANOVA, followed by a Tukey-Kramer post hoc test using a 95% confidence interval. Data are expressed as means+/−SEM. Accession numbers Cif (PA2934, NP 251624.1); PlcH (PA0844, YP 792433.1); alkaline phosphatase (PA3296, YP 789857.1); β-lactamase (PA1797, YP 791446.1); N-WASP (NP 003932); Omp85 (PA3648, YP 789516); GPIp137 (NP 005889). Supporting Information Figure S1 The Cif virulence factor is packaged in P. aeruginosa OMV. From an overnight P. aeruginosa PA14 culture, Optiprep density gradient centrifugation was utilized to purify OMV from the bacteria and possible contaminants, including pilus (PilA). Purified OMVs retrieved from fractions 2 and 3 were pooled for use in all experiments described. Experiment repeated three times; representative blot shown. (0.35 MB TIF) Click here for additional data file. Figure S2 EDTA effectively lyses OMV. EDTA (0.1 M) disrupted OMV membranes to allow proteinase K (PK)-mediated degradation of Cif, an intravesicular OMV component, as measured by Western blot analysis. Experiment repeated three times; representative blot shown. (0.18 MB TIF) Click here for additional data file. Figure S3 Cif virulence factor is an intravesicular OMV component. Isolated OMV treated with Proteinase K (PK: 100 µg/ml) for 1 h at 37°C to degrade proteins on the exterior of OMV. Δcif: OMV purified from a P. aeruginosa Δcif mutant strain. Experiment repeated three times; representative blot shown. (0.35 MB TIF) Click here for additional data file. Figure S4 Cif does not localize to Rab4, Rab7, or Rab11-labeled endosomes. Cif does not localize to the sorting endosomal (Rab4 GTPase-labeled), late endosomal (Rab7 GTPase-labeled), or recycling endosomal (Rab11 GTPase-labeled) compartments after entry into airway epithelial cells. Airway epithelial cells were treated with OMV for 10 min, cells lysed, and endosomes were purified. Cif was immunoprecipitated from the endosomal fraction and Western blot analysis was performed for Rab4, 7, and 11 GTPases. IgG IP is a non-immune control immunoprecipitation experiment. (0.60 MB TIF) Click here for additional data file.

Polymeric microspheres are extensively researched for applications in drug and vaccine delivery. However, upon administration into the body, microspheres are primarily cleared via phagocytosis by macrophages. Although numerous studies have reported on the biochemical pathways of phagocytosis, relatively little is known about the dependence of phagocytosis on particle size. Here, we investigate the previously unexplained dependence of phagocytosis on particle size. Rat alveolar macrophages and IgG-opsonized and non-opsonized polystyrene microspheres were used as model macrophages and drug delivery particles. Phagocytosis, attachment and internalization were measured by flow cytometry and time-lapse video microscopy. Particles possessing diameters of 2-3 microm exhibited maximal phagocytosis and attachment. Rate of internalization, however, was not affected significantly by particle size. Maximal attachment of 2-3 microm microspheres is hypothesized to originate from the characteristic features of membrane ruffles in macrophages. Elimination of ruffles via osmotic swelling nearly eliminated the peculiar size-dependence of phagocytosis. A simple mathematical model is presented to describe the dependence of phagocytosis on particle size. The dependence of phagocytosis on particle size originated primarily from the attachment step. These results reveal the importance of controlling drug delivery particle size distribution and selecting the size appropriate for avoiding or encouraging phagocytosis.