Biochemical and Functional Characterization of Membrane Blebs Purified from Neisseria meningitidis Serogroup B*

Studies with purified aggregates of endotoxin have revealed the importance of lipopolysaccharide-binding protein (LBP)-dependent extraction and transfer of individual endotoxin molecules to CD14 in Toll-like receptor 4 (TLR4)-dependent cell activation. Endotoxin is normally embedded in the outer membrane of intact Gram-negative bacteria and shed membrane vesicles (“blebs”). However, the ability of LBP and CD14 to efficiently promote TLR4-dependent cell activation by membrane-associated endotoxin has not been studied extensively. In this study, we used an acetate auxotroph of Neisseria meningitidis serogroup B to facilitate metabolic labeling of bacterial endotoxin and compared interactions of purified endotoxin aggregates and of membrane-associated endotoxin with LBP, CD14, and endotoxin-responsive cells. The endotoxin, phospholipid, and protein composition of the recovered blebs indicate that the blebs derive from the bacterial outer membrane. Proteomic analysis revealed an unusual enrichment in highly cationic (pI > 9) proteins. Both purified endotoxin aggregates and blebs activate monocytes and endothelial cells in a LBP-, CD14-, and TLR4/MD-2-dependent fashion, but the blebs were 3-10-fold less potent when normalized for the amount of endotoxin added. Differences in potency correlated with differences in efficiency of LBP-dependent delivery to and extraction of endotoxin by CD14. Both membrane phospholipids and endotoxin are extracted by LBP/soluble CD14 (sCD14) treatment, but only endotoxin·sCD14 reacts with MD-2 and activates cells. These findings indicate that the proinflammatory potency of endotoxin may be regulated not only by the intrinsic structural properties of endotoxin but also by its association with neighboring molecules in the outer membrane.

Cells of the innate immune system provide a first line of defense against invading microorganisms. The ability to sense invading microorganisms via pattern recognition receptors that recognize conserved microbial structures (e.g. Gram-negative bacterial lipopolysaccharides; endotoxin) is linked to induction of an inflammatory response that, in turn, leads to mobilization of host defenses needed to eliminate the invading microbes (1)(2)(3). However, excessive inflammation can play an important role in the pathology of many invasive infections, dramatically illustrated in fulminant meningococcemia. In that disease, the severity of inflammation and disease correlates with levels of circulating meningococcal endotoxin (lipo-oligosaccharide (LOS)) 2 (4,5).
Potent cell activation by endotoxin depends on sequential interactions of endotoxin with several extracellular and/or cell surface host proteins: LBP, CD14, MD-2, and TLR4 (6 -11). This pathway has been defined with purified aggregates of endotoxin. However, endotoxin is normally an integral component of the Gram-negative bacterial outer membrane, either as part of intact bacteria or of released membrane "blebs" (12). Membrane blebs are released constitutively from many growing Gram-negative bacteria (13)(14)(15)(16)(17). Release may be increased when bacteria are exposed to stressful conditions such as antibiotic or serum (complement) treatment or nutrient deprivation (18,19). Although it has not yet been possible to quantitatively monitor the release and persistence of membrane blebs in vivo, bleb-like particles have been detected in plasma from a patient with meningococcemia (20). The ability of membrane blebs to induce host cell proinflammatory responses has been studied (21)(22)(23)(24) but not nearly as extensively as purified endotoxin has been examined.
This study focuses on the comparison of endotoxin activity as part of intact membrane blebs and as purified aggregates. To compare quantitatively the bioactivity of endotoxin in these two different forms, we have used an acetate auxotroph of Neisseria meningitidis serogroup B, strain NMB (NMB-ACE1) (25), to isolate metabolically labeled ([ 14 C]acetate) membrane blebs and aggregates of purified endotoxin (pLOS agg ). Our findings confirm the prominent role of endotoxin in the proinflammatory activity of the membrane blebs and indicate that the activity of membrane-associated endotoxin may be dampened by interactions with neighboring bacterial outer membrane molecules. Proteomic analysis has revealed that these membranes are unusually enriched in highly cationic proteins, which possibly are needed to reduce potential electrostatic repulsion between densely packed, polyanionic endotoxin molecules.
Gel Sieving Chromatography-Columns of acryl HR S500 (1.5 ϫ 18 cm) were pre-equilibrated in HEPES-buffered (10 mM, pH 7.4) Hanks' balanced salt solution with divalent cations (HBSSϩ) plus 0.1% HSA. Chromatographed samples were applied in 0.2-0.5 ml of column buffer. Fractions (1 ml) were collected at a flow rate of 0.5 ml/min at room temperature. Aliquots of the collected 14 C-labeled material were analyzed by liquid scintillation spectroscopy using a Beckman LS liquid scintillation counter. Total recoveries were Ͼ70%. To prevent contamination of purified LOS or bleb preparations, all solutions were pyrogenfree and sterile-filtered, and the glass columns and connecting tubing were either autoclaved or washed extensively with 70% ethanol. After chromatography, fractions selected for use in bioassays were pooled and passed through sterile syringe filters (0.22-m pore size) with greater than 90% recovery of 14 C-labeled material in the sterile filtrate. Fractions were stored under sterile conditions at 4°C for more than 3 months with no detectable changes in chromatographic or functional properties.
Sephacryl S200 chromatography was performed on an AKTA FPLC system using a 1.6 ϫ 40 or 1.6 ϫ 70 cm column with a flow rate of 0.5 ml/min in HEPES-buffered HBSSϩ; 1-ml fractions were collected and 14 C cpm was measured by liquid scintillation spectroscopy. The column was calibrated with Bio-Rad gel filtration standards that include thyroglobulin ␥-globulin, ovalbumin, myoglobin, and vitamin B 12 .
For preparative generation of 14 C-LOS⅐sCD14 from purified 14 C-LOS and from membrane blebs, aggregates and blebs containing 250 ng/ml LOS were preincubated with 300 ng/ml LBP and 5 g/ml sCD14 in a total volume of 1 ml of HEPES-buffered HBSSϩ containing 0.1% albumin for 90 min at 37°C and products resolved by Sephacryl S200 chromatography.
PL Analysis-14 C-Labeled blebs, before and after treatment with purified PLA 2 , were extracted by Bligh-Dyer procedure and 14 C-PL and 14 C-labeled breakdown products were recovered in the chloroform phase and resolved by TLC on silica gel G using a solvent system of chloroform/methanol/water/acetic acid (65:25:4:1 by volume) (26). Resolved radiolabeled lipids were detected and quantified by image analysis using a tritium screen that permitted quantitation of as little as 200 cpm and identified by co-migration with 14 C-labeled lipid standards. Quantitation was done using ImageQuant software from Amersham Biosciences.
Fatty Acid Analysis-To determine the 14 C-fatty acid composition of recovered 14 C-labeled lipids, the lipids were treated sequentially with 4 N HCl and 4 N NaOH at 90°C to release ester-and amide-linked fatty acid from the parent 14 C-labeled lipids (26). After this treatment, the sample was subject to Bligh-Dyer extraction, and the 14 C-labeled fatty acids were recovered in the chloroform phase. Individual 14 C-labeled fatty acids were resolved by reverse-phase TLC (0.2-mm HPTLC, RP-18; Merck) using acetonitrile/acetic acid (1:1, v/v) as the solvent system (25) and identified by co-migration with authentic fatty acid standards. The relative amounts of individual fatty acids were determined by image analysis as indicated above.
Thin-section Electron Microscopy-Purified blebs were dried on poly-L-lysine-coated coverslips followed by treatment with 2.5% glutaraldehyde in 0.1 M sodium cacodylate buffer, pH 7.2, for 1 h at 4°C. After three buffer rinses, the fixed sample was incubated for 1 h with 1% OsO 4 , 1.5% potassium ferricyanide in cacodylate buffer. The sample was dehydrated with graded amounts of ethanol series and then infiltrated with Eponate 12 epoxy resin (Ted Pella, Redding, CA). Sections (100 m) of the cured resin blocks were cut using a Reichert Ultracut E microtome. Post-section staining with uranyl acetate and Reynold's lead acetate was performed. Thin-section electron microscopy images were recorded using a Hitachi H07000 TEM.
SDS-PAGE and Mass Spectrometry Analysis-Membrane blebs derived from ϳ10 10 bacteria were concentrated either by membrane ultrafiltration or by trichloroacetic acid precipitation before treatment with SDS-PAGE sample buffer and electrophoresis in a 12% acrylamide gel. After electrophoresis, protein bands were stained with Coomassie Blue, excised manually, and collected in a 96-well plate. In-gel digestion was performed using the automated ProGest system (Genomic Solutions, Ann Arbor, MI). Gel pieces were destained and dehydrated with acetonitrile followed by reduction with 10 mM dithiothreitol (60°C, 30 min) and alkylation with 100 mM iodoacetamide (37°C, 45 min). Tryptic digestion was performed at 37°C for 4 h with 125-200 ng of sequencing grade trypsin (Promega, Madison, WI). Tryptic peptides were extracted from the gel pieces with an aqueous formic acid solution. Initial spectra of the excised bands were acquired using matrix-assisted laser desorption ionization mass spectrometry (MALDI-MS) on an Applied Biosystems Voyager DE-STR (Foster City, CA) mass spectrometer operating in the positive ion reflector mode under delayed extraction conditions (200 ns delay time, with a grid voltage of 66.5% of full acceleration voltage (25 kV)). Peptides were mixed with an equal volume of ␣-cyano-4-hydroxycinnamic acid matrix (Agilent Technologies, Palo Alto, CA), and 1 l was loaded onto the MALDI-MS target. Mass spectra were acquired, averaged (typically 100 laser shots), and externally calibrated with a standard peptide mixture consisting of angiotensin I, and ACTH fragments 1-7, 18 -39, and 7-38 (Bachem, Torrance, CA). Peptide mixtures obtained after proteolysis were further analyzed by reverse-phase HPLC connected to a quadrupole orthogonal time-of-flight mass spectrometer (QSTAR Pulsar i, MDS Sciex, Concord, Canada). Samples were diluted (1:1, v/v) in 0.1% formic acid before loading onto the HPLC system. Separation of peptides was performed on-line using the Ultimate nano binary HPLC system fitted with a Famos micro auto-sampler using a Switchos micro-column switching module (Dionex/LC Packings, Sunnyvale, CA), with first a Micro guard column (CapTrap "green" for peptide binding/washing, 0.5-l bed volume, Michrom, Auburn, CA) and second an analytical PepMap C18 nano-column (75 m inner diameter, 15 cm length, Dionex/LC Packings). Peptides were loaded onto the guard column and washed with loading solvent (0.05% formic acid in H 2 O; flow rate, 20 l/min for 5 min). Loading and elution of peptides onto the analytical column was performed using solvent A (0.05% formic acid in 98% H 2 O, 2% acetonitrile) and solvent B (0.05% formic acid in 2% H 2 O, 98% acetonitrile) at a flow rate of 300 nl/min and a gradient of 2% solvent B (from 0 to 5 min) and then 2-70% solvent B (from 5 to 55 min). Peptides were eluted directly into a QSTAR equipped with a Proxeon Biosystems (Odense, Denmark) nanospray ion source. Mass spectra (ESI-MS) and tandem mass spectra (ESI-MS/MS) were acquired in the positive ion mode with a nanospray needle voltage of 2300 V. Selected ions were fragmented in a collision cell, using nitrogen as the collision gas, and were analyzed in the orthogonal time of flight. The mass window for the precursor ion selection of the quadrupole mass analyzer was generally set to Ϯ 1 m/z. Data were collected in the advanced information-dependent acquisition mode, which allowed automated data acquisition throughout the liquid chromatography MS gradient. Spectra were externally calibrated using MS/MS fragment ions of a renin peptide standard (His immonium ion with m/z at 110.0713 and b 8 ion with m/z at 1028.5312) providing a mass accuracy of Յ50 ppm.
SDS-PAGE of 14 C-LOS purified from intact bacteria and from purified membrane blebs was carried out in 4 -20% gradient gels. 14 C-LOS was detected by radioautography.
Data Base Searches of Mass Spectrometry Data-The Protein Prospector MS-Fit program (prospector.ucsf.edu), developed at the University of California, San Francisco, was utilized for data base searches of the MALDI-MS data using the following search parameters: trypsin digest, molecular weight 1000 -100,000; NCBI data base, all species or Neisseria species; maximum of two missed cleavages, minimum number of peptides required to match set at 4. A licensed in-house version of the Mascot search engine (Matrix Science, London, UK) was utilized for data base searches of the ESI-MS/MS data using the following search parameters: NCBI data base, all species or bacterial species; trypsin digestion, maximum of 2 missed cleavages, peptide and MS/MS tolerance of Ϯ0.2 Da. Information-dependent acquisition (IDA) and MS/MS centroid parameters were set as follows: height percent of 50%, merge distance of 0.02 atomic mass unit, and threshold of 0.1% of the highest peak. The Mascot search engine uses a probability-based MOWSE (molecular weight search) scoring program to determine significant peptide matches. The Mascot Daemon application (Matrix Science) was used to manage files and perform batch searches.
Immunocapture-Monomeric 14 C-LOS⅐sCD14 complexes were identified by immunocapture (7) using a mAb to CD14 (18E12) that binds to CD14 outside of the lipopolysaccharide-binding site. This mAb and an isotype-matched control were diluted in pH 9.6 bicarbonate buffer, applied to Nunc-Immuno Maxisorp plates (1 g/well), and incubated overnight at 25°C. The wells were rinsed with HBSSϩ, HEPES, 1% HSA three times and then incubated for 2 h with the same buffer for blocking. 14 C-LOS-containing samples were then added and incubated overnight with shaking at 25°C. The supernatant was removed and an aliquot taken to measure unbound 14 C-LOS. Captured material was eluted with 2% SDS by incubation for 15 min at 37°C and was counted by liquid scintillation spectroscopy.
Human Cells-HUVEC were routinely cultured on collagen-coated plasticware (Costar, Cambridge, MA) at 37°C, 5% CO 2 , and 95% relative humidity in endothelial basal medium supplemented with 5% fetal bovine serum, 12 g/ml bovine brain extract, 10 ng/ml human endothelial growth factor, 1 g/ml hydrocortisone, and 50 g/ml gentamicin. Cells were subcultured and grown to confluence (ϳ4 -5 days). Cell monolayers were then washed twice with warm HBSS to remove traces of serum before adding experimental media. Experiments were done with cells between passages 2 and 6.
Human peripheral blood mononuclear cells (PBMC) were obtained from heparinized venous blood from healthy volunteers, with informed consent, as described (28) and stored on ice in sterile pyrogen-free HBSS supplemented with 10 mM D-glucose until use (Ͻ60 min). Purified PBMC contained ϳ80% lymphocytes, 20% monocytes, and Ͻ2% polymorphonuclear leukocytes.
Assay of HUVEC Activation-Cells in 48-well tissue culture plates were incubated for 20 h at 37°C, 5% CO 2 , and 95% humidity in Dulbecco's minimal essential medium and 0.1% HSA with various concentrations of 14 C-LOS or blebs (or purified 14 C-LOS⅐sCD14) Ϯ LBP and sCD14 as appropriate. Activation of HUVEC was monitored by measuring accumulation of extracellular interleukin-8 by enzyme-linked immunosorbent assay as described previously (29). The role of TLR4 in LOS or bleb-induced HUVEC activation was tested by preincubation of the cells for 15 min at 4°C with neutralizing anti-TLR4 (2891; Ref. 9) mAb (10 g/ml) or an irrelevant isotype-matched control mAb.
Assay of PBMC Activation-Purified human PBMC (ϳ500,000 total cells) were incubated for 60 min at 37°C in 96-well Optiplates with or without 14 C-LOS or blebs (ϮLBP), as indicated, in a total volume of 0.2 ml containing HEPES-buffered HBSS (pH 7.4) containing 0.1% HSA and 10 Ϫ4 M lucigenin. An oxidative response of PBMC (monocytes) was measured indirectly as lucigenin-enhanced chemiluminescence using a LUCY1 version VI.5 luminometer (Anthos Labtec Instruments) as previously described (30). The roles of mCD14 and TLR4 in LOS or blebinduced PBMC chemiluminescence was tested by preincubation of the cells for 15 min at 4°C with neutralizing anti-CD14 (MY-4) or anti-TLR4 (2891) mAb (10 g/ml) or an irrelevant isotype-matched control mAb.
Assay of Cell Association of 14 C-LOS or Blebs-Freshly isolated PBMC were resuspended at a concentration of 4 ϫ 10 6 cells/ml in RPMI supplemented with 20 mM HEPES and 1% HSA. The HSA was filtered through a 0.45-m syringe filter just before use. The cells were incubated with 14 C-LOS or blebs (ϮLBP), as indicated, in a total volume of 1.0 ml in small Teflon wells for 1 h at 37°C on an orbital shaker in a 5% CO 2 incubator. After incubation, the cell suspensions were placed on ice for 30 min to facilitate detachment of monocytes from the surface of the wells. The cell suspension was then removed from the wells and the wells washed with HBSSϩ supplemented with 20 mM HEPES and 1% HSA. The cell suspension and wash were combined in one polypropylene tube and centrifuged at 500 ϫ g for 5 min at 4°C in a swinging bucket rotor. The cell pellet was washed once with 1 ml of HBSSϩ, 20 mM HEPES, 1% HSA, resuspended in 200 l of 5% SDS, 10 mM EDTA, and boiled for 10 min. Radioactivity in the reaction medium, cell washes, NOVEMBER 18, 2005 • VOLUME 280 • NUMBER 46 and cell lysate was counted in a Beckman scintillation counter. Total recovery of radioactivity was typically Ͼ80%.

Isolation of Intact Membrane Blebs and Purified Endotoxin
Aggregates-Metabolically labeled ( 14 C-acetate) membrane blebs and purified endotoxin aggregates were isolated from an acetate auxotroph of NMB as outlined in Fig. 1. We did not use detergents, chaotropic agents, or shearing to increase release of blebs or ultracentrifugation or ammonium sulfate precipitation to concentrate the membranes in an effort to minimize alteration of the physical state of the recovered membrane blebs. No differences in either the apparent physical state or bioactivity were observed in blebs isolated directly from bacterial conditioned medium by gel sieving or after concentration of the medium by ultrafiltration, as needed in preparation of samples from large cultures. The yield of membrane blebs was greatest when harvested from late logarithmic/early stationary phase cultures as judged by 14 C-LOS recovery. Approximately 10% of the total 14 C-LOS in the bacterial suspension was recovered in the blebs.
Physical Characterization of Purified 14 C-Labeled Membrane Blebs-Sephacryl S500 chromatography showed that the membrane bleb preparations were nearly homogeneous chromatographically, eluting slightly after pLOS agg ( Fig. 2A). Transmission electron microscopy revealed that the blebs existed predominantly as vesicles with an average diameter of 40 -80 nm (Fig. 2B). The recovered membrane blebs retained these properties during long term storage at Ϫ80°C (at least 1 year) and after one freeze/thaw.
Lipid Analysis of Purified 14 C-Labeled Membrane Blebs-Approximately 80% of the incorporated [ 14 C]acetate in the recovered blebs was in fatty acids of complex lipids, mainly PL and LOS. There were little or no free fatty acids present in the blebs (Ͻ2% of total lipid-associated 14 C cpm). Metabolic labeling for several generations of growth ensured that the distribution of 14  . This is closely similar to the endotoxin/PL ratio measured in isolated outer membranes from E. coli and Salmonella enterica serovar Typhimurium (33).
Analysis of the 14 C-PL present in the purified blebs (versus intact bacteria) showed that the blebs are enriched in phosphatidylethanolamine relative to phosphatidylglycerol (Fig. 3B, TABLE ONE). This is also characteristic of the bacterial outer membrane (34). Thus, both the LOS/PL ratio and PL composition indicate that the recovered blebs derive from the bacterial outer membrane. Because 80% of the total 14 C cpm in blebs is in fatty acids, ϳ50% of which derive from LOS, ϳ40% of the total 14 C cpm in blebs is 14 C-LOS. Therefore, 14 C-labeled blebs containing 1500 total cpm contain 600 cpm of 14 C-LOS. In preparations of maximum specific radioactivity, this corresponds to 1 ng of LOS, representing the amount of LOS derived from 1-2 ϫ 10 5 bacteria or outer membrane equivalents (35,36). LOS purified from intact bacteria (pLOS agg ) and from blebs (LOS (blebs)) were indistinguishable as judged by fatty acid composition (TABLE ONE) and by electrophoretic migration during SDS-PAGE (Fig. 3C).
Added purified human Group IIA PLA 2 degraded, at maximum, 25-30% of the total PL of the membrane blebs (Fig. 3C). Because the activity of this enzyme is generally restricted to PL within the outer leaflet of membrane bilayers (37), these findings suggest that up to 30% of the PL within the blebs is present in the outer leaflet of this membrane. PLA 2 treatment did not produce gross alteration of the physical properties of the blebs, as judged by gel sieving chromatography (data not shown).
Proteomic Analysis of Membrane Blebs-To provide a much more complete compositional analysis of the purified blebs, we also performed a proteomic analysis. To provide enough material for analysis, 200 -400-ml cultures were used. Additional manipulation and time needed for concentration of the conditioned medium and of the purified blebs did not alter the protein profile of the recovered blebs, as judged by one-dimensional (SDS-PAGE) electrophoresis (data not shown).
Initial attempts to use two-dimensional gel electrophoresis for resolution of bleb proteins for proteomic analysis revealed a relatively high abundance of very cationic proteins (pI Ͼ 9) that were incompletely resolved (data not shown). Because the one-dimensional gels suggested a relatively small subset of the total bacterial proteins in the blebs, we chose to rely on this method to optimize recovery of cationic membrane proteins. Twenty-five bands were visible by Coomassie Blue staining;  this protein pattern was seen in each of several independent preparations of blebs ( Fig. 4; data not shown).
To identify the proteins present, mass spectrometry analysis was performed. Each of the 25 bands was excised from the gel and subjected to proteolysis with trypsin. Initial peptide mass data were used to search the Neisseria genome using the protein Prospector search engine. These searches indicated that, in the majority of the bands, each consisted of a mixture of proteins. Therefore, ESI-MS/MS analyses were performed to more definitively identify the proteins present. Mascot searches of the MS/MS data identified 1-9 proteins/band (TABLE TWO). In toto, 48 unique proteins 3 were identified with M r ranging from 8,800 to 160,000. Of those proteins for which the subcellular localization is known, nearly 75% (27 of 37) are known outer membrane proteins (e.g. opacity proteins, porins, pilin proteins, iron-binding proteins, adhesins). Most of the remaining proteins probably derive from the periplasm and/or sites of close outer membrane/inner membrane apposition (e.g. thiol disulfide interchange protein, disulfide oxidoreductase). The presence of acetate kinase, a cytosolic enzyme, in bleb preparations probably reflects high cellular levels of this enzyme in acetate auxotrophs dependent on utilization of dietary acetate. A majority of these proteins are highly basic with a predicted pI Ͼ 9 (TABLE TWO;       Scores are reported as Ϫ10Log 10 (P), where P is the probability that the observed match is a random event. Scores listed are significant matches (p Ͻ 0.05). For this data set, total ion scores in the range of Ͼ38 -46 were determined to be significant matches. b Proteins identified based on one peptide match were evaluated by hand to confirm the match. same bacterial suspension offered the opportunity to quantitatively compare their ability to activate endotoxin-responsive host cells. Fig. 6A and TABLE THREE show that both pLOS agg and blebs activated cultured endothelial cells in an LBP-, sCD14-, and TLR4/MD-2-dependent fashion. However, comparison of cell activation (Fig. 6A, IL-8; tissue factor, data not shown) over a broad dosage range of pLOS agg and blebs, normalized for LOS content, revealed that the blebs were significantly less effective at cell activation than the purified LOS aggregates at all doses examined (Fig. 6B). These differences in potency were seen at all time points examined (up to 24 h; data not shown) and at doses of sCD14 that were limiting or in molar excess to LOS (Fig. 6C). The LBP dose dependence was more complex, reflecting the dual roles of LBP in CD14/MD-2/TLR4-dependent cell activation by endotoxin, i.e. promoting activation at low LBP concentrations and inhibiting at higher LBP concentrations (38 -41). A bell-shaped curve of LBP dose dependence was observed with both pLOS agg and purified blebs, but LBP con-centrations needed to maximally promote cell activation were substantially lower for pLOS agg than for blebs (Fig. 6D).

Differences in LBP-and sCD14-dependent Activation of Endothelial Cells by pLOS agg and Blebs Parallel Differences in Formation of Monomeric LOS⅐sCD14
Complex-Maximal potency of LBP-and sCD14-dependent cell activation of endothelial cells by endotoxin requires formation of monomeric endotoxin⅐sCD14 complex (7,9). Thus, differences in cell activation by pLOS agg and by membrane blebs could reflect differences in the efficiency of LBP/sCD14-dependent extraction and transfer of endotoxin to sCD14 from purified endotoxin aggregates versus purified membranes. To test this hypothesis, we monitored by Sephacryl S200 chromatography the accumulation of the LOS⅐sCD14 complex (M r ϳ60,000) after incubation of pLOS agg and membrane blebs with LBP and sCD14. As shown in Fig. 7, A and B, both the rate and extent of accumulation of radiolabeled M r ϳ60,000 species was greater from pLOS agg than from membrane blebs. These products were not generated from either pLOS agg or blebs when LBP or sCD14 was added alone (data not shown (7,25). The product derived from pLOS agg is LOS⅐sCD14 (7). This was confirmed by fatty acid analysis (Fig. 7C) and immunocapture ( Fig. 7D) with an anti-CD14 mAb that binds CD14 outside of the endotoxin binding site of CD14 (7,42). In contrast, less than half of the M r ϳ60,000 14 C-labeled product(s) derived from the blebs appeared, by these criteria, to be LOS⅐sCD14. The presence of extracted 14 C-phospholipid in this fraction was demonstrated by Bligh-Dyer extraction and TLC (Fig. 7E). Incubation of the bleb-derived M r ϳ60,000 species with MD-2 resulted in quantitative transfer of LOS to MD-2, whereas the complexes containing PL were unaffected by incubation with MD-2 (Fig. 7, F and G). Thus, the M r ϳ60,000 14 C-labeled product(s) generated by LBP/sCD14 treatment of the blebs appeared to be a mixture of 14 C-LOS⅐sCD14 and 14 C-PL⅐"X" (possibly sCD14 or LBP). Activation of synthesis and secretion of interleukin-8 from HUVEC was induced exclusively by the M r ϳ60,000 species; material recovered in the void volume (V o ) derived from either pLOS agg or membrane blebs was inactive (data not shown). When normalized for LOS (i.e. LOS⅐sCD14) content, the activity of the recovered M r ϳ60,000 species derived from pLOS agg and from membrane blebs was essentially the FIGURE 5. Dot plot of proteins from the total N. meningitidis proteome and those identified from purified membrane blebs. The total N. meningitidis, strain MC58, predicted proteome was retrieved from the TIGR website (www.tigr.org). The proteins from this total proteome (black circles) and the identified bleb proteins from this study (white squares) were plotted on a graph based on their predicted pI values and their predicted molecular sizes.  A, B, D). Results shown represent the mean Ϯ S.E. of at least three independent determinations. same (Fig. 7H). Thus, differences in LBP/sCD14-dependent activation of HUVEC by pLOS agg and membrane blebs can apparently be attributed to differences in extraction and transfer of endotoxin to sCD14. There were no differences in either the fatty acid composition or electrophoretic (SDS-PAGE) migration of LOS recovered in LOS⅐CD14 or (MD-2) complexes from the LOS that remained in the blebs (void volume peak in Fig. 7B) after LBP/sCD14 treatment (data not shown).
Membrane CD14-dependent Cell Activation by pLOS agg and Membrane Blebs-Many cells highly responsive to endotoxin contain mCD14 (6,11). In these cells, cell activation is not dependent on sCD14. We therefore compared the ability of pLOS agg and membrane blebs to activate monocytes in mixed PBMC preparations. Both pLOS agg and blebs activated monocytes, as manifested by lucigenin-enhanced chemiluminescence, in an LBP (Fig. 8, A and B), mCD14 (Fig. 8C) and largely TLR4 (MD-2) (TABLE THREE)-dependent manner. As seen with endothelial cells, the potency of pLOS agg toward monocytes was ϳ10-fold greater than that of the blebs. Differences in potency paral-leled differences in LBP/mCD14-dependent delivery of endotoxin from pLOS agg and blebs to the cells (Fig. 8D).

DISCUSSION
We have described a simple and highly reproducible method to recover and isolate bacterial membrane blebs. We have used an organism (N. meningitidis) known to bleb copiously to facilitate recovery of material for more complete structural and functional analyses. At the maximum, nearly 20% of total outer membrane LOS in the whole bacterial suspension was recovered in the form of purified blebs, corresponding closely to the yield of meningococcal blebs originally reported by DeVoe and Gilchrist (13). This yield contrasts markedly with the much lower yield typically obtained from E. coli, S. enterica serovar Typhimurium, etc. (16,17), and is consistent with the greater propensity of the meningococci to release endotoxin in the form of membrane blebs. The molecular and cellular bases of these differences in blebbing are unknown.
Both lipid and protein compositional analyses suggest that, as expected, the blebs represent predominantly outer membrane and further attest to the purity of these membrane-bound vesicles. All of the known major outer membrane proteins (porins, RmpM, Opa, pilin, IgA protease, siderophores, and others) were identified in the blebs. Several periplasmic and/or peptidogylcan-associated proteins were also present (e.g. EF-Tu, thiol disulfide interchange protein (DsbC), putative disulfide oxidoreductase (DsbA), ␥-glutamyl-transpeptidase). These proteins may be trapped within the lumen of the blebs when shed outer membranes reseal (16). These properties are consistent qualitatively with the characteristics of blebs/vesicles recovered from other Gramnegative bacterial species (16). However, whereas recovered extracellular membrane fragments from S enterica serovar Typhimurium suggest a relatively selective release of lipopolysaccharide (14), the LOS/PL con-

Effect of anti-TLR4 mAb on cell activation by pLOS agg and by blebs
LBP-and CD14-dependent activation of monocytes and of HUVEC by pLOS agg and by membrane blebs was carried out as described in the legends for Figs. 6 and 8 and under "Experimental Procedures." Cells were pretreated with Ϯanti-TLR4 mAb (10 g/ml) for 30 min at 37°C before the addition of pLOS agg or blebs (5 ng of LOS/ml) ϩ LBP (100 ng/ml) and sCD14 (250 ng/ml; HUVEC samples only). Results shown represent the mean Ϯ S.E. of three or more independent determinations. Bacterial Outer Membrane Blebs NOVEMBER 18, 2005 • VOLUME 280 • NUMBER 46 tent of the meningococcal blebs suggest that these membrane vesicles are more representative of the outer membrane as a whole. The unexpected presence of a significant subpopulation of PL within the outer leaflet of the bleb membrane, as deduced from PLA 2 sensitivity (Fig. 3C), may be because of the increased curvature of the released blebs in comparison with that of the intact outer membrane, promoting migration of PL that normally reside within the inner leaflet of the outer membrane (33). However, the extreme lipid asymmetry observed in the outer membrane of E. coli and Salmonella typhimurium may not be representative of all Gram-negative bacteria (43). A striking finding was the remarkable enrichment in highly basic proteins within the recovered outer membrane blebs. The calculated average pI (without weighting for differences in abundance of individual proteins) was 8.49 for the blebs as opposed to 7.35 for the total meningococcal proteome (Fig. 5). Clusters of basic residues, but not overall basicity, are common in integral membrane proteins (44). The abundance of highly cationic proteins within the outer membrane blebs may instead reflect the abundance of polyanionic endotoxins within the outer leaflet of the Gram-negative bacterial outer membrane, creating a structural pressure for expression of membrane-associated counterions (12). Some proteomic analyses of other Gram-negative bacterial outer membrane preparations have also revealed an enrichment of cationic proteins (45), but others have not (46). It is not clear whether the origin of these differences is biologic (i.e. reflecting genotypic and/or phenotypic differences in outer membrane proteins among different Gram-negative bacteria) or technical. Integral membrane and cationic proteins typically represent some of the most significant technical challenges in proteomic analysis (47). Their high representation in our analysis suggests that the approach we used (one-dimensional SDS-PAGE followed by in situ proteolytic fragmentation, extraction, and mass spectrometry) was very effective.

Source of LOS % Inhibition by anti-TLR4
An important goal of this study was to quantitatively compare the "proinflammatory" activity of purified endotoxin to membrane-associated endotoxin. We have addressed this question by utilizing a rigorously controlled comparison of purified endotoxin and blebs from the same bacteria. We chose cell types (HUVEC and monocytes) and cellular responses that were know to be highly responsive to purified endotoxin so that responses to the membrane blebs would be driven largely by LOS, facilitating direct comparison of the activity of purified and membrane-associated endotoxin. The fact that the responses induced by purified LOS and membrane blebs were similarly dependent on added LBP, CD14, and TLR4 suggests that, as hoped, the effects of the membrane blebs we observed were due largely to the LOS within these vesicles. This was most convincingly demonstrated in the case of activation of HUVEC in which LBP/sCD14-dependent activation by purified LOS and by membrane blebs each traced to the generation of monomeric (M r ϳ60,000) LOS⅐sCD14 complex (Fig. 7).
These findings strongly suggest that the molecular requirements for TLR4-dependent (endothelial) cell activation defined by use of purified endotoxin also apply to membrane-embedded endotoxin. However, as shown previously (24), the potency of endotoxin (LOS) in blebs is lower than that of LOS in purified aggregates. There is no observable difference in the chemical composition of LOS in purified aggregates and in membrane blebs (TABLE ONE; Fig. 3C). Thus, the reduced potency of endotoxin in blebs is not due to differences in endotoxin per se but rather to its presentation, as modulated apparently by the presence of neighboring outer membrane molecules. The differences in endotoxin potency were apparent for both sCD14-and mCD14-dependent cell activation and paralleled the differences in efficiency of LBP-dependent delivery of endotoxin to CD14. In experiments with HUVEC, in which it was possible to independently vary LBP or sCD14 concentrations, differences in LBP (Fig. 6D) but not sCD14 (Fig. 6C) requirements for cell activation were observed. LBP binds less avidly to intact Gram-negative bacteria than to purified endotoxin aggregates (30). Thus, differences in LBP binding to purified LOS and to LOS-rich membrane blebs (not yet measured) could account for the differences seen. Initial LBP-endotoxin interactions likely involve cationic residues within LBP and anionic moieties clustered at and near the lipid A region of endotoxin (3,48,49). Abundant, highly cationic proteins in the outer membrane blebs (e.g. porins) also interact with endotoxin during and after assembly in the outer membrane (33, 50 -53) and may thereby impede LBP binding and LBP-catalyzed extraction and transfer of individual endotoxin molecules to CD14 that is important for cell activation. Alternatively, the presence of competing "substrate" (e.g. PL) for LBP in the blebs might account for the reduced efficiency of LBP-dependent endotoxin extraction. One may anticipate an even lower efficiency of LBP-dependent extraction of endotoxin from intact bacteria, making the copious release of endotoxin-rich blebs by meningococci an important factor in the potent proinflammatory effects of acute meningococcal infection.
An interesting by-product of these studies was the demonstration that LBP/sCD14 treatment of the membrane blebs produced extraction of bacterial membrane phospholipids as well as LOS (Fig. 7, C and E). The ratio of PL to LOS in recovered M r ϳ60,000 products was similar to that of the blebs, suggesting similar efficiency of extraction of PL and LOS, i.e. no intrinsic specificity of LBP/sCD14-dependent action on PL and LOS. However, it should be noted that phospholipids recovered within the M r ϳ60,000 products were enriched in phosphatidylglycerol relative to phosphatidylethanolamine (Fig. 7E, PG and PE, respectively), consistent with the reported preference of LBP for interaction with acidic versus zwitterionic lipids (54,55). In studies with purified dispersions of phospholipids, LBP-dependent transfer of (acidic) phospholipids to (s)CD14 has been demonstrated (56). The co-elution of products containing 14 C-phospholipids with 14 C-LOS⅐sCD14 (Fig. 7, B and C) is compatible with formation of 14 C-PL⅐sCD14 complexes, but we have not yet been able to demonstrate this directly. The fact that an anti-CD14 monoclonal antibody that can capture 14 C-LOS⅐sCD14 could not capture the 14 C-PL-containing complexes (Fig. 7D) could either mean that PL binding to CD14, in contrast to LOS, blocks the epitope recog- Interactions of blebs with LBP and CD14 cause extraction of some membrane PL and endotoxin (LOS) and formation of monomeric LOS⅐sCD14 and perhaps PL⅐CD14 complexes. Only LOS⅐sCD14 reacts with MD-2 to cause TLR4-dependent cell activation. The release or degradation (see Fig. 3D) of membrane lipids, especially when in close proximity to host cell membrane, may facilitate translocation of integral bacterial outer membrane proteins (e.g. porins) into host cell membrane. See "Discussion" for additional details. nized by this antibody or that phospholipid extracted by LBP/sCD14 from membrane blebs is bound to a different protein, perhaps even LBP itself. A monomeric phospholipid⅐LBP complex would not be resolved under our chromatographic conditions from LOS⅐sCD14.
Whatever the nature of this complex, the demonstration that LBP/ sCD14 can extract both LOS and phospholipids may have important implications for the fate of integral bacterial outer membrane proteins such as the porins. These proteins from Neisseria have been shown to translocate to host cell membranes and as a result to modify host cell properties (57)(58)(59)(60). How these integral membrane proteins are able to translocate from the bacterial outer membrane to host cell membranes is unknown. We speculate (Fig. 9) that depletion of bacterial lipids surrounding these outer membrane proteins by LBP/CD14 and/or Group IIA PLA 2 action may create physical circumstances favorable for transfer of the bacterial proteins from bacterial to host membranes. This may be most likely to occur in cells containing mCD14, where extraction of bacterial membrane lipids would occur in immediate juxtaposition to the host cell membrane. Such a gradual disruption of the integrity of these bacterial membrane vesicles may favor translocation of other bacterial membrane proteins as well as release of soluble toxins, etc., trapped within the lumen of vesicle (61,62), thus providing a novel LBP/(s)CD14-dependent mechanism of delivery of bioactive bacterial products to host targets.