Proteomic analysis of a detergent-resistant membrane skeleton from neutrophil plasma membranes.

Plasma membranes are organized into functional domains both by liquid-ordered packing into "lipid rafts," structures that resist Triton extraction, and by attachments to underlying cytoskeletal proteins in assemblies called "membrane skeletons." Although the actin cytoskeleton is implicated in many lipid raft-mediated signaling processes, little is known about the biochemical basis for actin involvement. We show here that a subset of plasma membrane skeleton proteins from bovine neutrophils co-isolates with cholesterol-rich, detergent-resistant membrane fragments (DRMs) that exhibit a relatively high buoyant density in sucrose (DRM-H; d approximately 1.16 g/ml). By using matrix-assisted laser desorption/ionization time of flight and tandem mass spectrometry, we identified 19 major DRM-H proteins. Membrane skeleton proteins include fodrin (nonerythroid spectrin), myosin-IIA, myosin-IG, alpha-actinin 1, alpha-actinin 4, vimentin, and the F-actin-binding protein, supervillin. Other DRM-H components include lipid raft-associated integral membrane proteins (stomatin, flotillin 1, and flotillin 2), extracellular surface-bound and glycophosphatidylinositol-anchored proteins (IgM, membrane-type 6 matrix metalloproteinase), and intracellular dually acylated signaling proteins (Lyn kinase, Galpha(i-2)). Consistent with cytoskeletal association, most DRM-H-associated flotillin 2, Lyn, and Galpha(i-2) also resist extraction with 0.1 m octyl glucoside. Supervillin, myosin-IG, and myosin-IIA resist extraction with 0.1 m sodium carbonate, a treatment that removes all detectable actin, suggesting that these cytoskeletal proteins are proximal to the DRM-H bilayer. Binding of supervillin to the DRM-H fragments is confirmed by co-immunoaffinity purification. In spreading neutrophils, supervillin localizes with F-actin in cell extensions and in discrete basal puncta that partially overlap with Galpha(i) staining. We suggest that the DRM-H fraction represents a membrane skeleton-associated subset of leukocyte signaling domains.

Plasma membranes in eukaryotic cells are organized into domains of specialized function. Many of these domains are relatively stable on the cell surface and involve a series of linkages between cytoplasmic cytoskeletal proteins, integral membrane proteins, and extracellular attachments to the substrate or to other cells. For example, intermediate filament anchorages to membranes at desmosomes and hemidesmosomes distribute mechanical stresses over epithelial surfaces and maintain the integrity of skin and other tissues (1,2). Plasma membrane domains anchored to actin filaments include spectrin-based meshworks, adsorptive cell surface microvilli, focal adhesions, and adherens junctions (3)(4)(5)(6)(7). In each of these domains, interactions between cytoskeletal proteins provide resistance against disruption by nonionic detergents, such as Triton X-100 and octyl glucoside, permitting the transmembrane protein assembly to be isolated as a "membrane skeleton" (8).
Lipid-lipid interactions also can organize membranes into detergent-resistant domains. Such lipid "rafts" are stabilized against disruption by cold Triton X-100 through liquid-ordered packing of component cholesterol, (glyco)sphingolipids, and gelphase glycerolipids (9,10). Lipid rafts contain resident integral membrane proteins, such as caveolin, stomatin, and flotillin (11), extracellular proteins with glycophosphatidylinositol (GPI) anchors containing long chain fatty acids (12), and cytoplasmic proteins modified by a dual myristoylation/palmitoylation motif or by palmitoylation on two or more cysteine residues (13)(14)(15). When analyzed in vivo by spectroscopic methods, plasma membrane lipid rafts consist of clusters of ϳ20 proteins and ϳ3500 sphingolipid molecules and are ϳ50 nm in diameter, a size well below the resolution limit of light microscopy (16,17). These dimensions can be increased into the submicron to micron range through aggregation of raft components, e.g. by receptor activation or addition of bivalent antibodies against raft-associated surface antigens (18 -20). After treatment of cells or membranes with cold Triton X-100, the coalescing detergent-resistant remnants of the lipid rafts are called detergent-resistant membranes (DRMs). 1 DRMs, which represent a subset of the endogenous raft lipids and proteins (21), can be recovered by flotation into low density fractions (DRM-L) from sucrose or Optiprep TM gradients (19).
Both cytoskeleton-stabilized membrane domains and lipid rafts provide sites at which localized signal transduction can occur (3,8,19). Selective recruitment of proteins from the cytoplasm or protein diffusion within the plane of the membrane can regulate signaling events, such as those downstream of growth factor addition, ion channel activation, or changes in cell adhesion. Because many signaling proteins, including heterotrimeric G␣ proteins and members of the Src family of protein tyrosine kinases, are dually acylated, lipid rafts have been proposed as platforms for the controlled interaction of substrates and signaling enzymes (19,22). Lipid rafts also have been proposed to play roles in membrane targeting and endocytotic trafficking (23). Cascades of recruitment of adaptor proteins and kinases to lipid rafts are well characterized during receptor activation in B lymphocytes, T lymphocytes, and mast cells (19). Compartmentalized signaling at lipid rafts may be especially important for hematopoietic cells, which form dynamic attachments to substrates and other cells and lack many types of stable cell surface adhesions. Although signaling downstream of lipid raft-associated adhesive stimuli requires filamentous actin (24,25), the role of F-actin in this process is unknown.
F-actin is known to be an important structural and functional component of plasma membranes from mammalian neutrophils (26). Neutrophils are the most abundant of the circulating leukocytes and are readily available in large quantities from bovine blood (27)(28)(29). They are terminally differentiated granulocytes that constitute the body's first line of defense against invading microbes (30). In response to chemotactic stimuli from pathogens or from other immune cells and/or mucosal cells, neutrophils rapidly migrate across capillary endothelia into body tissues (31). They engulf microbes coated with complement proteins and/or immunoglobulins and kill the ingested microorganisms by releasing proteases, antimicrobial proteins, and oxygen metabolites into phagolysosomal compartments. Although polarity is maintained by cytoplasmic microtubules (32), actin function in the membrane skeleton and/or cortical cytoskeleton is required for chemotactic signaling, cell movement, adhesion, and phagocytosis (33).
Despite its importance in early immune defense, the neutrophil membrane skeleton is largely uncharacterized. In agreement with previous observations of a spectrin-, actin-, and band 4.1-containing membrane skeleton in neutrophils (34,35), we have shown that a macromolecular complex containing fodrin (non-erythrocyte spectrin), actin, and non-muscle myosin-IIA exists in bovine neutrophil plasma membranes (28,36). We also have used F-actin blot overlays and immunoblotting to show that neutrophil plasma membranes contain ezrin, moesin, and a 205-kDa F-actin-binding protein that we named supervillin. Molecular cloning of supervillin cDNA indicated that this protein contains a carboxyl terminus with predicted similarities to villin and a novel amino terminus (36,37). The supervillin amino terminus contains functional nuclear targeting signals, sites for direct binding to F-actin, and regions that promote actin filament binding and bundling (38). Supervillin is a tightly bound peripheral membrane protein that is concentrated at sites of epithelial cell-cell adhesion (36). This protein also has been reported to be a minor constituent of lipid rafts from detergent-solubilized Jurkat T cells (39) and to co-purify with P2X 7 ATP receptors from embryonic kidney cells (40). Supervillin message levels are elevated in neutrophils stimulated with lipopolysaccharide from gingivitis-causing bacteria (41), implying a role during physiological activation of neutrophils.
Affinity-purified Anti-supervillin IgG-The HSV41 clone (37) containing human supervillin cDNA sequences (GenBank TM accession number AF051851) was digested with EcoRI and NotI and ligated into the pGEMEX-1 vector (Promega Corp., Madison, WI). A chimeric fusion protein consisting of 260 amino acids of T7 gene 10 and linker sequence plus the first 340 amino acids of human supervillin (H340) was expressed in BL21(DE3) bacteria after induction with isopropyl ␤-D-thiogalactopyranoside. The fusion protein was purified as inclusion bodies and used as an immunogen for the production of rabbit polyclonal antisera (Research Genetics, Huntsville, AL). Antibodies against the amino-terminal 340 amino acids of human supervillin (␣H340) were affinity-purified against a chimeric fusion protein of H340 and glutathione S-transferase (GST). The GST-H340 protein was expressed in bacteria from sequences encoded by the pGEX-6P-1 vector and HSV41 cDNA and was purified on glutathione-Sepharose TM , as described by the manufacturer (Amersham Biosciences). All manipulations were carried out at 4°C, unless stated otherwise. Purified GST-H340 (ϳ12 mg) was dialyzed against 20 mM sodium phosphate buffer (pH 7.6) and coupled to ϳ1.5 ml of CNBr-activated Sepharose 6MB (Sigma) for 24 h. The column was blocked with 0.2 M glycine ethyl ester (pH 8.0) for 24 h, washed with 40 ml of sodium phosphate buffer (pH 7.6), 2 ml of 4.5 M MgCl 2 , 10 ml TBS (25 mM Tris-HCl, 1 mM EDTA (pH 7.4)), and stored in TBS containing 0.1% thimerosal. The GST-H340 Sepharose beads were rotated overnight with 10-ml aliquots of specific rabbit antisera and washed sequentially at room temperature with the following: 1) ϳ150 ml TBS; 2) 6 ml of 1 M NaCl in TBS; 3) 4 ml of 2 M urea in TBS; and 4) 2 ml of 10 mM Tris-HCl (pH 7.5). High affinity antibodies were eluted with 4.5 M MgCl 2 , 72.5 mM Tris (pH 7.0) and collected in 1-ml fractions with equivalent volumes of 10 mM Tris-HCl (pH 7.5). Fractions containing affinity-purified IgG were pooled, repeatedly diluted with 10 mM Tris-HCl (pH 7.5), and re-concentrated to ϳ0.3 ml by centrifugation at 2000 ϫ g using Ultra-Free-10 TM ultrafiltrate concentrators (Millipore Corp., Bedford, MA). Concentrated antibody was clarified by centrifugation at 13,000 ϫ g for 5 min and stored in aliquots at Ϫ20°C with ϳ1 mg/ml bovine serum albumin and 50% (v/v) sterile glycerol. The specificity of ␣H340 antibody was confirmed by immunoblots of whole neutrophil extracts (Fig. 1A), and the suitability for staining of paraformaldehyde-fixed samples was demonstrated using immunolabeling experiments with COS-7 cells expressing a GFP-tagged bovine supervillin (not shown). H340 antigen was purified by cleavage from GST-H340 and chromatography on glutathione TM -Sepharose and DEAE-Sepharose, as recommended by the manufacturer (Amersham Biosciences).
Neutrophil Plasma Membranes-Neutrophils were treated with 5 mM diisopropylphosphofluoridate for 15 min, washed, and resuspended in relaxation buffer (28) (10 mM PIPES, 100 mM KCl, 3 mM NaCl, 3.5 mM MgCl 2 , 1 mM ATP, 1 mM EGTA, pH 7.3) containing a mix of protease inhibitors (1 mM phenylmethylsulfonyl fluoride, 0.4 M aprotinin, 10 M E-64, 20 M leupeptin, 10 M bestatin, 5 M N-acetyl-Leu-Leu-Met-al, 2 M pepstatin), and disrupted by nitrogen cavitation. All manipulations were carried out at 0 -4°C. Cavitates were centrifuged at 1000 ϫ g for 10 min to remove nuclei, and 10-ml aliquots of postnuclear supernatant were loaded onto 28 ml of 1.06 g/ml Percoll TM suspension in relaxation buffer. Initially, plasma membranes were separated from secretory vesicles, granules, and cytosol using Percoll TM step gradients, as described previously (28,36). As a faster alternative to manual pipetting, Percoll TM density gradients were generated by centrifugation. Density gradients were generated by centrifugation at 50,000 ϫ g for 30 min in a Beckman VTi50 vertical rotor. Neutrophil plasma membranes corresponded to the "␥ fraction" (ϳ1.04 g/ml) from these gradients. This fraction was enriched ϳ10-fold in cell surface (28) and plasma membrane (27,45) markers. After centrifugation at 140,000 ϫ g for 2 h to remove residual Percoll TM particles, plasma membranes were resuspended in relaxation buffer containing protease inhibitors, disrupted with 10 -20 strokes of a tight-fitting Teflon Dounce homogenizer, and stored in aliquots at Ϫ80°C.
Immunofluorescence Microscopy-Neutrophils suspended in HBSS were seeded onto coverslips that were coated with poly-L-lysine or zymosan-activated plasma for 15 min at 22°C (28). Zymosan-activated plasma was prepared by incubating bovine plasma (from the same batch of peripheral blood used to isolate the neutrophils) with 5 mg/ml zymosan at 37°C for 60 min and clarified by centrifugation at 100,000 ϫ g for 60 min. Non-adherent cells were removed, and adherent unactivated (poly-L-lysine) or activated (zymosan-activated plasma) cells were incubated at 37°C in 5% CO 2 for 30 -60 min in 10 mM potassium phosphate, 130 mM NaCl, 0.6 mM CaCl 2 , 1 mM MgCl 2 , 2.7 mM KCl, 0.18% glucose, pH 7.4. Neutrophils were fixed with 4% paraformaldehyde in PBS (pH 7.4) for 10 min, washed three times in PBS, and then permeabilized using 0.5% Triton X-100 for 10 min at 22°C. Exposed cytoplasmic surfaces of neutrophil plasma membranes were obtained by disrupting cells with a focused jet of 60 mM PIPES, 25 mM HEPES, 10 mM EGTA, 2 mM MgCl 2 (pH 6.9) (28, 46), delivered from a syringe with a 26-gauge needle at an oblique angle across the surface of the coverslip. Sheared cells were immediately fixed in 2% paraformaldehyde in the same buffer for ϳ15 min at 22°C. Fixed samples were washed 3 times for 5 min in PBS and blocked at 4°C overnight, or for up to 3 days, with 10% bovine serum, 0.02% sodium azide, 0.02% thimerosal in PBS (pH 7.4). Samples then were incubated 2-4 h with 3 g/ml affinity-purified ␣H340 or nonspecific rabbit IgG alone or together with mouse monoclonal anti-G␣ i-2 antibody at 2 g/ml in blocking solution. As an additional specificity control, aliquots of affinitypurified ␣H340 were incubated for 2 h with 10 g/ml purified H340 antigen before addition to coverslips. After three washes in PBS, samples were incubated with blocking solution containing 2 g/ml Alexa-Fluor 488-labeled F(abЈ) 2 fragment of goat anti-rabbit IgG and either 2 g/ml AlexaFluor 568-labeled F(abЈ) 2 fragment of goat anti-mouse IgG or 50 nM rhodamine phalloidin. At these antibody concentrations, no significant background was observed in the absence of primary antibody (not shown). After final washes, samples were mounted with VectaShield TM mounting medium (Vector Laboratories Inc., Burlingame, CA) and observed using an MC100 Axioskop epifluorescence microscope with filters for Texas Red and fluorescein (Carl Zeiss Inc., Thornwood, NY). Phase contrast and fluorescence micrographs were acquired with a model Retiga-1300 QImaging Camera and processed using Openlab imaging software (Improvision Inc., Lexington, MA). Confocal immunofluorescence images were recorded on a Bio-Rad MRC 1024 laser scanning confocal microscope equipped with a krypton/argon laser and LaserSharp version 3.2 software (Bio-Rad). No bleed-through fluorescence between different channels was detected in double-label experiments.
Electron Microscopy-Transmission electron microscopy of plasma membrane thin sections was carried out as detailed earlier (36). Negative staining of the DRM-H fraction was performed according to published methods (47), with minor modifications. Briefly, DRM-H fractions were deposited onto EM grids that had been coated with carbon and glow-discharged on the day of the experiment. Samples were allowed to adhere for 30 -60 s and stained for 30 s with 2% OsO 4 in 0.1 M cacodylate buffer (pH 7.4) and for 30 s with 2% uranyl acetate in PBS (pH 7.4). After each treatment, grids were rinsed with 6 -10 drops of PBS and drained using filter paper. Stained samples were air-dried and examined on a model EM-301 transmission electron microscope (Philips Electron Optics Inc., Rahway, NJ) at an accelerating voltage of 60 kV.
Octyl Glucoside and Sodium Carbonate Extractions-DRM-H preparations (600 g) were diluted 1:1 with relaxation buffer and pelleted by ultracentrifugation (100,000 ϫ g, 60 min, 4°C) using a Beckman TLA120.2 rotor. Pellets (100 g) were resuspended in 200 l of TEB, or 0.1 M octyl glucoside extraction buffer (0.1 M octyl glucoside, 25 mM Tris, 250 mM NaCl, 2.5 mM MgCl 2 , 1 mM EGTA, 1 mM ATP), or 0.1 M Na 2 CO 3 and extracted for 60 min on ice. As a positive control, the pellet was completely solubilized by incubation in 200 l of SDS at 70°C for 15 min. Samples were then re-centrifuged as above using a Beckman TLA120.1 rotor, and the pellets were resuspended to 200 l with TEB. Both supernatants (S) and resuspended pellets (P) were adjusted to final concentrations of 1% Triton X-100, 250 mM NaCl, 50 mM Tris-HCl (pH 7.5). Volumes equivalent to 100 l of the initial assay mixture were analyzed by SDS-PAGE, silver staining, and Western blotting.
Gel Electrophoresis and Immunoblotting-Protein or volume equivalents of Optiprep TM /sucrose density gradient fractions or DRM-H isolates/extracts were separated at 4°C on high resolution 100 ϫ 140 ϫ 1.5-mm 5-15% gradient polyacrylamide SDS gels (48). Coomassie Blue and silver staining were performed according to standard protocols (49). Molecular mass standards were prestained full range Rainbow markers (Amersham Biosciences). Separated proteins were documented by digitally scanning gels using a model GS-700 imaging densitometer (Bio-Rad). For immunoblotting, gels were electrotransferred to nitrocellulose membranes (Schleicher & Schuell) at 4°C. Blots were 1) blocked with 5% (w/v) nonfat dry milk in TBST (TBS containing 0.05% Tween 20) for 60 min, 2) incubated with the appropriate antibody for 2-4 h in this blocking solution, 3) washed 3 times for 5 min in TBST, 4) incubated for 1-2 h with alkaline phosphatase-or HRP-conjugated goat anti-rabbit or anti-mouse IgG in blocking solution, 5) washed 3 times for 5 min in TBST, and 6) visualized with BCIP/NBT Membrane Phosphatase Substrate (Kirkegaard & Perry Laboratories, Gaithersburg, MD) or Renaissance TM Western blot Chemiluminescence Reagent Plus (PerkinElmer Life Sciences), as appropriate. Results were quantified after exposure to BioMax MR Scientific Imaging Film (Eastman Kodak Co.) and/or digitally scanned, as above.
MALDI-TOF and MS/MS via MALDI Post Source Decay-For identification of DRM-H associated proteins, protein bands were excised from Coomassie Blue-stained gels with clean razor blades. Gel pieces (ϳ1.5 ϫ 1.5 ϫ 6 mm) were rinsed 3 times for 5 min in deionized water. In-gel tryptic digests were performed according to published protocols (50) with minor modifications, e.g. the addition of 0.01% n-octyl glucopyranoside in the digest buffer. Peptide digests were concentrated using ZipTip C18 micropipette tips (Millipore Corp.) according to recommended protocols, with the exception that the ion-pairing agent trifluoroacetic acid was replaced by 1% formic acid. MALDI-TOF and Post-Source-Decay (PSD) data were obtained on an AXIMA-CFR mass spectrometer using software version 2.0.1 (Kratos Analytical, Manchester, UK). Typically, mass spectrometric spectra were obtained at a laser power near threshold with pulsed extraction optimized for 2,000 Da; mass resolutions varied from 6,000 to 9,000 across the tryptic peptide mass range. PSD data were obtained at a laser power ϳ15% higher than threshold. Samples were co-crystallized with ␣-cyano-4-hydroxycinnamic acid (10 mg/ml in 1% formic acid, 50% acetonitrile) via the dried drop method. External calibration in MS mode was accomplished by a 3-point peptide mixture that covered the mass range typical of tryptic digests (600 -2,500 Da) and spotted locally to the sample of interest. PSD fragment ions were fitted to a generated curve, which was calibrated with PSD fragments from the synthetic peptide, P 14 R. Data base searches were performed using Protein Prospector version 3.4.1 (prospector.ucsf.edu). Monoisotopic peptide masses were searched against the NCBI nonredundant data base using the MS-Fit program and 50 -100 ppm mass tolerances. PSD fragments were searched against the NCBI nonredundant data base using MS-Tag and 0.2 Da as the parent tolerance and 0.5 Da as the fragment ion tolerance.
Immunoaffinity Purification-Preparative immunoisolations of supervillin DRM complexes were performed as described previously (36) and confirmed by analytical scale experiments, using M280 tosyl-activated or protein A-conjugated Dynabeads. Affinity-purified rabbit ␣H340 IgG and nonspecific control rabbit IgG (0.5-mg aliquots) in PBS (pH 7.5) were coupled to washed Dynabeads by end-over-end rotation overnight at 37°C (tosyl-activated) or 4°C (protein A-conjugated). Protein A-bound IgG was cross-linked using 20 mM dimethylpimelimidate in 0. In analytical experiments, aliquots of neutrophil plasma membranes (250 g of protein) were incubated for 1 h at 4°C in 0.5 ml of extraction buffer containing 1% Triton X-100 or 0.1 M octyl glucoside. The resulting detergent extracts were incubated for an additional 4 h at 4°C with Dynabeads (ϳ5 ϫ 10 7 beads/tube) containing covalently bound supervillin-specific ␣H340 or nonspecific rabbit IgG plus 1% bovine serum albumin (for silver staining) or 5% goat serum (for immunoblot analyses). Unbound material, including bovine serum albumin and goat serum proteins, was removed by washing the beads 3 times for 10 min in extraction buffer with 1% Triton X-100 or 0.1 M octyl glucoside, respectively. Immunoadsorbed material was magnetically separated, eluted in SDS sample buffer at 70°C for 10 min, and analyzed by SDS-PAGE, silver staining, and immunoblotting.

RESULTS
Neutrophil Membrane Actin and Supervillin-By using the ␣H340 antibody specific for supervillin (Fig. 1A), supervillin was co-localized with actin filaments at the peripheries of adherent neutrophils (Fig. 1B, e-h). Most unactivated neutrophils appeared round and roughly symmetrical, with evenly distributed cortical F-actin and supervillin (not shown). Activated neutrophils exhibited a spread, polarized morphology with supervillin and F-actin accumulations in ruffles and membrane protrusions (Fig. 1B, a-h). To examine signals from proteins closely associated with the plasma membrane, the ventral membrane skeleton was exposed by mechanical shear of spreading neutrophils (51,52). Membrane sheets representing the cytoplasmic faces of basal membranes contained discrete actin-and supervillin-associated puncta (Fig. 1B, m-p). These puncta were concentrated at or near the leading edges of lamellipodial extensions and occasionally were observed along radially projecting actin bundles. Staining also appeared as small patches within the interiors of the adherent membranes. No patches or puncta were observed with nonspecific rabbit IgG (Fig. 1B, b) or after preadsorption of the ␣H340 antibody with antigen (Fig. 1B, j), indicating specificity. These results suggest that supervillin is associated with a subset of the F-actin-rich neutrophil membrane skeleton that appears as discrete membrane-associated domains.
Supervillin-associated DRMs-We showed previously that Triton-insoluble remnants of the neutrophil plasma membrane contained F-actin and the actin-binding membrane skeleton proteins, fodrin, myosin-IIA, and supervillin (53). Resistance to solubilization by Triton X-100 suggested an association with the underlying cytoskeleton but was also consistent with an association with DRMs. We now report that, after Triton extraction of neutrophil plasma membranes, these membrane skeleton proteins exhibit low densities in Optiprep TM gradients and partition into fractions containing elevated ratios of cholesterol to protein (Fig. 2, A and B). These fractions are also enriched in proteins immunoreactive with antibodies against G␣ i-2 and Lyn kinase (Fig. 2C). All three of these properties are hallmarks of DRMs (10,16,54). The presence of at least two distinguishable membrane bands at the tops of the gradients ( Fig. 2A, arrows) suggested heterogeneity, such as that reported for DRMs isolated from lung (55,56), smooth muscle (57), and tissue culture cells (39,58).
Further purification of this light membrane fraction (Fig. 3A,  fraction 4)

. Supervillin and F-actin localizations in intact neutrophils and membrane sheets.
A, specificity of affinity-purified rabbit polyclonal IgG against supervillin (␣H340). Immunoblot of whole bovine neutrophil lysate (100 g) probed with ␣H340 and HRP-conjugated anti-rabbit secondary antibody. B, phase (a, e, i, and m) and immunofluorescence (b-d, f-h, j-l, and n-p) images of intact neutrophils (a-h) and the cytoplasmic surfaces of basal plasma membrane remnants exposed by mechanical shearing of adherent neutrophils (i-p). Cells and membrane sheets were stained with rhodamine-phalloidin (F-actin) and Alexa488-labeled goat F(abЈ) 2 (antibody) against nonspecific rabbit IgG (control), or affinity-purified ␣H340 IgG against supervillin (␣H340), or ␣H340 IgG that had been preincubated with 10 g/ml H340 antigen (pre-adsorbed). Supervillin localizes within F-actin-rich membrane extensions in polarized leukocytes (e-h) and as discrete ϳ200-nm patches in association with F-actin-rich lamellae in ventral membrane skeletons exposed by mechanical shear (m-p). Merged images (d, h, i, and p) show antibody stain in green, F-actin in red, and overlap as yellow. Scale bars ϭ 2 m. those reported for DRMs containing GPI-anchored proteins, heterotrimeric G-proteins, and Src family kinases (13,14,59). DRM-H fractions exhibited densities of 1.15 to 1.18 g/ml (Fig.  3A) and were highly enriched in polypeptides with the mobilities of fodrin and actin (Fig. 3B, arrowheads), membrane skeleton proteins reported to associate with DRMs (55,60,61). These and other polypeptides were consistently enriched (see below), along with supervillin (Fig. 3C), during the ϳ10 -20fold purification of DRM-H from total Triton-insoluble neutrophil plasma membranes (Fig. 3B). Conversely, the DRM-H fraction was depleted of numerous low molecular mass membrane proteins (Fig. 3B, lane 3), as well as of most glycosylated proteins recognized by the lectin, concanavalin A (data not shown).
As visualized by negative stain electron microscopy, the DRM-H fraction contained both membrane sheets and filamentous material (Fig. 3D). Cytoskeletal filaments reminiscent of those associated with neutrophil plasma membranes (Fig. 3D, a) also were apparent in micrographs of the DRM-H fraction (Fig. 3D, b). These filaments were associated with membrane fragments with surface areas of 0.05-0.5 m 2 that also contained a semi-regular array of particles with diameters of ϳ25 nm (Fig. 3D, c). Thus, the DRM-H fraction consisted of detergent-resistant membrane microdomains tightly associated with a discrete set of membrane skeleton proteins.
Abundant DRM-H Proteins-We used a mass spectrometric approach to identify 25 polypeptides in the 23 DRM-H bands that stained most intensely with Coomassie Blue (Fig. 4). These polypeptides were representative of 19 proteins or protein complexes, many of which are characteristic of liquidordered microdomains. For instance, the DRM-H proteins, stomatin, flotillin 1, and flotillin 2, are oligomeric integral membrane proteins found in lipid rafts from erythrocytes, brain, lung, and numerous other tissues (62)(63)(64). The DRM-H fraction also contained the Golgi-associated plant pathogenesis-related protein-1 (65), the dually acylated non-receptor tyrosine kinase Lyn, subunits of the heterotrimeric G i-2 protein, and the GPI-anchored membrane type 6 matrix metalloproteinase (MT-6-MMP, MMP-25, leukolysin). These membrane proteins are targeted to liquid-ordered microdomains through covalent modifications with saturated fatty acyl groups (12-14, 17, 59).
Membrane skeleton proteins constituted a second major category of proteins in the DRM-H fraction. We found heterodimeric fodrin (nonerythrocyte spectrins ␣II and ␤II), ␤-actin, and other F-actin binding proteins, including supervillin, ␣-actinin 1, ␣-actinin 4, non-muscle myosin-IIA heavy and light chains, and the heavy chain and calmodulin light chain of myosin-IG, an unconventional myosin originally identified as a minor histocompatibility antigen (66,67). The DRM-H fraction also contained the intermediate filament protein, vimentin, which binds fodrin (68) and which has been reported in other liquid-ordered microdomains (69). The IgM heavy and light chains may be attached extracellularly through a specific Fc receptor, which is reported to exist on bovine neutrophils (70 -72). Other proteins, such as clathrin, bactenecin (a positively charged neutrophil defensin), and the G18.1b polypeptide (which is known only as expressed sequence tags and a predicted open reading frame) are less easily categorized.
To confirm that these proteins were in the same structure as supervillin, a tightly bound neutrophil membrane skeleton protein (36), we immunoisolated supervillin-containing lipoprotein complexes from unfractionated detergent extracts of neutrophil plasma membranes (Fig. 5). Of the many proteins present in the total Triton X-100 extract of plasma membranes (Fig. 5A,  lane a), a specific subset of at least 15 silver-stained polypeptides (Fig. 5A, lane d, arrowheads and dashes) was recovered with beads coated with an affinity-purified IgG specific for supervillin (Fig. 5A, lane b). These 15 polypeptides were not observed in experiments with nonspecific rabbit IgG (Fig. 5A,  lane c), indicating specificity. The overall pattern and molecular masses of these polypeptides (Fig. 5A, arrowheads and  dashes) were very similar to the SDS-gel profile and sizes of cytoskeletal and signaling proteins identified by mass spectrometric analysis (Fig. 4A). The identities of eight of these major silver-stained polypeptides were confirmed by immunoblotting (Fig. 5B), using antibodies specific for ␣-fodrin, nonmuscle myosin-II, supervillin, Lyn kinase, flotillin 2, ␤-actin, G␣ i-2 , and myosin-I head domains. All eight proteins were specifically adsorbed to Dynabeads coated with antisupervillin IgG (Fig. 5B, ␣H340), as compared with control rabbit IgG (Fig. 5B, control). The other 10 polypeptides identified in Fig. 4 also may be present in the anti-supervillin pellets but could not be assayed by immunoblot analyses due to antibody unavailability, co-migration with IgG, and/or low molecular mass. Interestingly, similar results were obtained with both 1% Triton X-100, which does not extract lipid raft proteins (Fig. 5B, TEB), and with 100 mM octyl glucoside (Fig.  5B, OG), which does (58,73). Thus, most or all of the proteins identified as components of the DRM-H fraction are associated in the same lipoprotein complexes as supervillin, and these complexes may contain differentially stabilized lipid raft components.
Relative Extractabilities-To test the relative strengths of association between DRM-H proteins and lipids, we extracted the DRM-H fraction with 0.1 M octyl glucoside or 0.1 M sodium carbonate (pH 11.5). Supernatants and extracted membrane pellets were analyzed for the presence of component proteins (Fig. 6). We then organized the resulting immunoblots (Fig. 6, middle columns) from least extractable, i.e. most tightly bound (Fig. 6, top), to most extractable, i.e. most weakly Surface, extracellular cell surface-bound protein; Cytoskeleton, cytoskeletal protein; nd, not determined. Superscripts: †, details in Supplemental Material; 1, non-erythrocyte spectrin ␣II/␤II (fodrin) dimer/tetramer; 1*, ϳ150 kDa ␣-fodrin proteolytic degradation fragment; 2, non-muscle myosin-IIA heavy chain and regulatory light chain; 3, myosin-IG heavy and calmodulin regulatory light chain; 4, IgM heavy and light chain decamer; 5, heterotrimeric G i-2 protein ␣ and ␤ subunits. bound (Fig. 6, bottom). Sedimentation of proteins in 1% Triton-extraction buffer (Fig. 6, TEB, left column) or following membrane solubilization in hot 1% SDS (Fig. 6, SDS, right column) served as negative and positive controls, respectively. With octyl glucoside (Fig. 6, OG), we observed partial solubilization of Lyn and IgM, but relatively little dissociation of the raft proteins, flotillin 2 and G␣ i-2 . In extractions with sodium carbonate (Fig. 6, Na 2 CO 3 ), the integral (flotillin 2) and lipid-anchored (G␣ i-2 , Lyn) membrane proteins were retained in the pellets, as expected. Extractabilities of the membrane skeleton proteins varied from completely extracted (actin and ␣-actinin) and ϳ50% extracted (fodrin) to only ϳ20 -25% extracted (myosin-IIA and myosin-IG) and virtually unextracted (supervillin). Because a significant portion of supervillin can be extracted with 1 M KCl and 0.1 N NaOH (pH 13) (36), we conclude that it is a peripheral membrane protein that is very tightly associated with DRM-H lipids and/or integral membrane proteins. Surprisingly, myosin-IIA and myosin-IG are also relatively tightly associated with these membranes, remaining bound even after the removal of all detectable actin.
DRM-H Is a Subset of Neutrophil Lipid Domains-Sucrose density gradient profiles showed that DRM-H was characterized by increased amounts of membrane skeleton proteins, whereas most signaling proteins exhibited a biphasic distribution between DRM-L and DRM-H (Fig. 7, A and B). As assessed from immunoblot analyses of equal protein loads, all of the assayed proteins co-isolated to some degree with both DRM-L and DRM-H (Fig. 7B). The major difference between DRM-L and DRM-H was the higher percentages of fodrin, supervillin, myosin-IG, and actin that were associated with DRM-H fractions. DRM-H also contained significant amounts of myosin-IIA, Lyn kinase, the lipid raft marker flotillin 2, and G␣ i-2 immunoreactive proteins. Thus, after Triton extraction of neutrophil plasma membranes, there is both considerable overlap between the compositions of DRM-L and DRM-H and a characteristic enrichment of cytoskeletal proteins in association with DRM-H membranes. Increased association of cytoskeletal proteins correlates with increased sedimentability in sucrose density gradients (Fig. 7C).
In agreement with the observation of overlapping and distinct biochemical pools of supervillin and G␣ i proteins (Fig. 7B), we observed partial but significant overlap between the supervillin and G␣ i immunofluorescent signals in adherent membrane fragments from lysed neutrophils (Fig. 7D). This co-localization was especially prominent in puncta at the F-actin-rich membrane peripheries of actively spreading cells (Fig. 7D, a-c). By contrast, only occasional overlap of supervillin and G␣ i signals was observed in the centers of the membrane remnants (Fig. 7D, d-f). The extent of co-localization observed by immunofluorescence microscopy (Fig. 7D) was roughly consistent with the degree to which supervillin and G␣ i-2 immunoreactive proteins co-purified in the DRM-H sucrose density fractions (Fig. 7B). Taken together with the identification of G␣ i-2 as a major co-purifying protein (Fig. 4), these observations suggest that supervillin both co-localizes and co-purifies with a subset of G␣ i-2 -associated signaling domains. These results further suggest that the isolated DRM-L and DRM-H fractions represent different lateral domains within the neutrophil plasma membrane. DISCUSSION In this paper, we present proteomic and microscopic evidence for a DRM-associated membrane skeleton from neutrophil plasma membranes (Figs. 1-7). This lipoprotein complex is enriched in cholesterol (Fig. 2) and exhibits a higher buoyant density than do DRMs with less associated membrane skeleton material (Figs. 3 and 7, DRM-H versus DRM-L). Cytoskeletal proteins characteristically enriched in this fraction include fo-FIG. 6. Association of membrane skeleton and signaling proteins with DRM-H membranes. Purified DRM-H membrane fragments (Fig. 3, membrane fraction 6) were extracted with Triton extraction buffer (TEB; negative control), 100 mM octyl glucoside in extraction buffer (OG), 0.1 M sodium carbonate (pH 11.5) (Na 2 CO 3 ), or 1% hot SDS (SDS; positive control). Soluble (S) and insoluble (P) proteins were isolated at 150,000 ϫ g for 60 min. Volumes representing 100 l of the initial extraction solutions were separated by SDS-PAGE. Immunoblot strips were probed using antibody against membrane skeleton and signaling proteins, as indicated. Most signaling proteins, myosins, and supervillin remained insoluble, even under conditions (OG) that solubilize DRM proteins not stabilized by association with cytoskeletal proteins (73). Actin, ␣-actinin, and most of the fodrin were extracted by sodium carbonate, which removes loosely bound peripheral membrane proteins. drin, supervillin, myosin-IIA, myosin-IG, and ␣-actinins 1 and 4 (Figs. [3][4][5][6][7]. Immunoaffinity purification of supervillin-associated DRM complexes confirms that these proteins are in the same lipoprotein complexes (Fig. 5). Octyl glucoside extraction demonstrates that most of the hydrophobically associated lipid raft proteins in this complex are relatively stabilized against extraction by this detergent (Figs. 5 and 6). Carbonate extraction demonstrates that the peripheral membrane proteins, supervillin, myosin-IIA, and myosin-IG, are very tightly associated with DRM lipids and/or integral membrane proteins, remaining bound even in the absence of detectable actin (Fig.  6). As supervillin and G␣ i-2 immunoreactive proteins appear to preferentially co-localize in actin-rich areas of cells engaged in active spreading (Figs. 1 and 7D), the DRM-H fraction may represent a transiently induced, membrane skeleton-associated subset of neutrophil signaling domains (Fig. 8).
The buoyant density of the DRM-H fraction (d ϳ1.16 g/ml) is probably the reason it has not been described previously. DRM purification methods typically involve the collection of lipoprotein complexes with densities within the range of 1.02 to 1.16 (5-35% sucrose) (39,74), whereas Triton-insoluble membrane skeletons characteristically exhibit densities of 1.25-1.3 (ϳ50 -64% sucrose) (75,76). Presumably, attachment of DRM-associated proteins to the cytoskeleton causes an increased density due to the increased ratio of protein to lipid in the detergent-resistant fraction, resulting in equilibrium banding at intermediate sucrose concentrations. DRMs with densities in the range of our DRM-H fraction (d ϳ 1.15-1.18; 33-40% sucrose) are likely to sediment, or to be spread throughout, the 30 -35% sucrose step gradient used in many DRM purification methods (39,74). Similarly, the DRM-H fraction would not be detected in related lipoprotein complexes prepared in the absence of nonionic detergents (58,77,78) because these complexes are isolated after carbonate extraction and/or sonication, procedures that remove most associated cytoskeletal proteins (e.g. Fig. 6).
Although DRMs are commonly assumed to reflect the compositions and morphologies of lipid domains in vivo, these isolated structures may differ in detail from their endogenous counterparts (54,58,79). Also, the isolated DRM-H fraction is not necessarily homogeneous with respect to the relative ratios of the co-isolating proteins. For instance, the small amount of DRM-H clathrin (Fig. 4) may be caused by a few contaminating coated pits or it may represent a limited association between coated membrane and a subset of DRMs. Other components of the DRM-H fraction, especially proteins present in minor amounts, remain for future analysis. The determination of which proteins interact through direct interactions and which are co-associated through a common affinity for the gel-phase DRM lipids is similarly beyond the scope of this investigation. However, the silver-stained profile of proteins that co-sediment  (16). Approximate locations are shown for the gradient load (L), low density DRMs (DRM-L), coated pits (CP), high density DRMs (DRM-H), and pellet (P). C, densities (q) and apparent protein concentrations (OE) of the sucrose gradient fractions shown in A. D, confocal immunofluorescence images show partial co-localization of supervillin and G␣ i in peripheral micropatches of unroofed neutrophils. Neutrophils spread on poly-L-lysine-coated coverslips were mechanically sheared and stained with Alexa488-and Alexa568-labeled goat secondary antibodies against rabbit polyclonal anti-supervillin IgG and mouse monoclonal anti-G␣ i-2 IgG, respectively. Partial overlap (yellow) of the supervillin (green) and G␣ i (red) signals is apparent in the merged image. Scale bar ϭ 2 m. as the DRM-H fraction (Figs. 3, 4, and 7) is similar to the pattern of proteins that are co-purified using antibodies directed against supervillin (Fig. 5). This result suggests that supervillin directly or indirectly associates with lipoprotein complexes containing the other proteins in Fig. 4 and may therefore serve as a marker protein for DRM-H rafts in neutrophil plasma membranes.
Our results are consistent with electron microscopic observations of a subplasmalemmal filament complex associated with Triton X-100-extracted neutrophil ghosts (80) and plasma membrane remnants of sheared neutrophils (51,52). In the latter studies, high resolution scanning electron micrographs revealed that the cytoplasmic surfaces of adherent, spreading neutrophils contain interlocking networks of globular centers and radiating branched filaments. Morphologically similar networks are observed in cross-sections of purified plasma membranes (Fig. 3D, a) and in negatively stained DRM-H samples (Fig. 3D, b and c). Given our observations of myosin-IIA and myosin-IG bound to DRMs lacking detectable F-actin, these domains may represent regions where actin and myosin are recruited and organized at the membrane, as originally implied (52).
Our data also agree with the initial observation that neutrophils contain a fodrin-and actin-based, detergent-resistant membrane skeleton (35). This membrane skeleton has been implicated in the lateral organization of components of the superoxide-generating machinery (81) and in the segregation of chemoreceptors and G i-2 proteins during neutrophil stimulation (34,(82)(83)(84)(85). The partial superimposition of the supervillin and G␣ i-2 immunofluorescence signals (Fig. 7D) suggests that at least some of these areas of overlap may correspond to domains involved in the active signaling between chemotactic receptors and the actin cytoskeleton.
Actin, spectrin, or fodrin, and/or associated proteins have been co-purified with lipid raft components from other hema-topoietic cell types. For example, erythrocyte DRMs are enriched in stomatin, flotillins 1 and 2, ␤-actin, and a loosely bound member of the spectrin gene family (65). More recently, a complex mixture of DRMs from unstimulated Jurkat T lymphocytes was found to contain a subset of proteins reminiscent of those identified here (39). Among a total of ϳ70 DRM proteins, including markers for mitochondria, rough endoplasmic reticulum, the Golgi complex, and the nuclear envelope, were sequences representative of fodrin, myosin-IIA heavy and light chains, supervillin, myosin-IG (GenBank TM accession number AI089291), vimentin, ␤-actin, flotillin 2, G␣ i-2 , and the Srcrelated tyrosine kinase Lck. Although it is not known whether these proteins co-associate with each other in T cells, as opposed to being present in different DRMs, it is striking that 12 of the 25 polypeptides identified in our study are also present in total Jurkat T cell DRMs. Thus, lipid raft-associated membrane skeletons similar to DRM-H may be present in hematopoietic cells of different origin.
The neutrophil DRM-H fraction also exhibits some compositional similarities to the microdomains associated with the multichain receptor for IgE (Fc⑀R) in mast cells (86,87). Both types of domains are cholesterol-rich and contain bound immunoglobulin, actin, and Lyn kinase (19,88). The relatively tight association of IgM with DRM-H fragments (Fig. 6) is consistent with the reported presence of a specific receptor for IgM on bovine neutrophils (70 -72). In mouse and human, an ϳ70-kDa receptor that binds IgA as well as IgM (Fc␣/R) has been shown to mediate internalization of IgM-coated Staphylococcus aureus (89). Another Fc receptor exists as an ϳ60-kDa GPIlinked isoform in hematopoietic cells (90,91). Either type of Fc receptor would be hard to detect in the presence of the more abundant, co-electrophoresing IgM decamer, leaving open the question of whether the IgM that co-sediments with the neutrophil DRM-H fraction is bound through a specific Fc receptor.
Although not documented previously, the actin-independent association of myosins and other F-actin-binding proteins with DRMs is consistent with reports (92,93) indicating that lipid raft domains are intrinsically linked to active movements of the cortical actin cytoskeleton. After priming by receptor occupancy with IgE, for example, the Fc⑀R is associated with Lyn in lipid rafts on mast cells (86). Following receptor cross-linking by multivalent antigen, the structure and organization of these microdomains are changed both by actin-based lateral movements of signal transduction proteins within the plane of the membrane and by recruitment of signaling proteins from cytoplasmic stores (94,95). Similar mechanisms also target specific lipid raft components to the front and rear of polarized leukocytes (24,25,96) and to the immunological synapses of stimulated T cells (93,97).
We hypothesize that fodrin, myosin-IIA, myosin-IG, and supervillin are components of a mobile, actin-based membrane skeleton that regulates the organization and/or transport of cholesterol-rich signaling domains at the neutrophil plasma membrane (Fig. 8). Associations with membrane skeleton components also could differentially stabilize and/or localize these domains within the membrane. Whereas the intracellular localization of myosin-IG is unknown, fodrin and myosin-II both localize throughout the cortex, as well as at the plasma membrane (98,99). By contrast, plasma membrane-associated supervillin is observed as puncta or patches in many cell types (36,38), suggesting that supervillin may be a characteristic component of this type of lipid raft-associated membrane skeleton. The tight association between supervillin and DRM-H fragments, in conjunction with the high affinity binding of supervillin for F-actin and its ability to reorganize cytoskeletal These oligomeric proteins and tightly bound lipids are depicted here as a "lipid raft." Additional proteins typically associated with liquid-ordered domains are also present, including an extracellular GPI-anchored protein (MT6-MMP), and intracellular dually acylated signaling proteins (Lyn, G␣ i-2 ). Extracellular surface-bound IgM may be attached through a specific receptor. DRM-H rafts are characterized by a relatively high buoyant density in sucrose, due to their association with membrane skeleton proteins (fodrin, F-actin, ␣-actinin, vimentin, myosin-IIA, myosin-IG, and supervillin). DRM-bound supervillin, myosin-IG, and myosin-IIA resist extraction with 0.1 M sodium carbonate, a treatment that removes all detectable actin and ␣-actinin, suggesting that the unextracted cytoskeletal proteins are more proximal to the DRM-H bilayer.
structures (36,38), suggests that supervillin may be involved in recruitment of the actin-based cytoskeleton to such liquid-ordered regions of the plasma membrane. The tight association of myosin motors with the DRM-H fragments suggests the possibility of a role in membrane transport or directed motility at the cell surface.