Lipid rafts regulate lipopolysaccharide-induced activation of Cdc42 and inflammatory functions of the human neutrophil.

Lipid rafts are cholesterol-rich membrane microdomains that are thought to act as coordinated signaling platforms by regulating dynamic, agonist-induced translocation of signaling proteins. They have been described to play a role in multiple prototypical cascades, among them the lipopolysaccharide pathway, and to host multiple signaling proteins, including kinases and low molecular weight G-proteins. Here we report lipopolysaccharide-induced activation of the Rho family GTPase Cdc42, and we show its activation in the human neutrophil to be mediated by a p38 mitogen-activated protein kinase-dependent mechanism. Subcellular fractionation reveals that lipopolysaccharide induces translocation of Cdc42 to lipid rafts, where it and p38 are both found to be activated. By contrast, lipopolysaccharide causes translocation of Rac from the polymorphonuclear leukocyte (PMN) rafts and does not induce its activation. With the use of methyl-beta-cyclodextrin, a cholesterol-depleting agent that reversibly disrupts rafts, we confirm an important regulatory role for rafts in the activation state of p38 and Cdc42 and in the Rho GTPase-dependent functions superoxide anion production and actin polymerization. Methyl-beta-cyclodextrin induces activation of p38 and Cdc42, but not Rac, in the nonstimulated PMN, yet inhibits subsequent lipopolysaccharide-induced activation of p38 and Cdc42. In parallel, methyl-beta-cyclodextrin primes the human PMN for subsequent superoxide release triggered by the formylated bacterial tripeptide formyl-Met-Leu-Phe, and induces actin polymerization in a subcellular distribution distinct from that induced by lipopolysaccharide. In sum, these findings provide evidence for an important regulatory role of cholesterol in both transmission of the lipopolysaccharide signal and the inflammatory phenotype of the human neutrophil.

Lipid rafts are cholesterol-and glycosphingolipid-rich membrane microdomains that are thought to function as coordinated signaling platforms in cells through their regulation of dynamic, agonist-induced translocation of specific signaling proteins (1,2). A variety of signaling components, including receptors (e.g. epidermal growth factor, platelet-derived growth factor, and endothelin receptors), kinases (e.g. Lyn, Syk, and protein kinase C), phosphatases (e.g. Syp), and both heterotrimeric and low molecular weight G-proteins (e.g. Rac, Cdc42, and Rho) have been localized to rafts (1,2). Raft compartmentalization is thought to play an important role in regulating the activation state of, and dynamic interactions among, proteins. A centrally important implication of localization of signaling events to lipid rafts is that such molecular interactions, like the physicochemical organization of rafts themselves, are sensitive to alteration of cellular cholesterol content. By exploiting this property of rafts, multiple investigators (3)(4)(5) have used cyclodextrins (cyclic oligomers of glucose that can be used both to extract cholesterol from intact, living cells, as well as to deliver cholesterol to them) as a tool to perturb rafts physically and functionally by cholesterol depletion, and thereby to uncover their native regulatory influence upon cellular signaling phenomena. The mechanism of cyclodextrin-induced effects upon the cell can be confirmed as cholesterol depletion either by co-incubating cells with cyclodextrin and cholesterol (thereby preventing depletion) or by using the cyclodextrin-cholesterol complex as a vehicle to sequentially cholesterol-replete cells that have been depleted by cyclodextrin. Repletion is thought to restore raft structure and function (5).
Although the best established role for rafts as signaling platforms may have been demonstrated in T and B cell receptor activation (6,7), recent reports indicate lipopolysaccharide (LPS) 1 induced interactions between Toll-like Receptor 4 (TLR4) and other proteins in monocyte rafts (8,9). This discovery (made possible by the ability to isolate lipid rafts by exploiting their properties of Triton insolubility and relative buoyant density) has ignited interest in using rafts as a window to discover novel signaling interactions in LPS signaling (8,10). In this light, Rho family GTPases (i.e. Cdc42, Rac, and Rho) have recently been independently reported to be localized to rafts in ECV304 cells (11), to undergo PDGF-induced translocation into fibroblast rafts (Rac and Rho) (12), and to play a role in LPS signaling in an endothelial cell line (13) and in fibroblasts (14). By analogy, Rac1 and Cdc42 have also been re-ported to be activated and dynamically recruited to TLR2 in THP-1 cells stimulated with Staphylococcus aureus (15). To our knowledge there have been no reports describing LPS-induced activation of Rho GTPases in a human primary immune cell nor in confirming a clear role for lipid rafts in the regulation of these proteins.
The Rho GTPases Rac and Cdc42 have been shown in the neutrophil (PMN) to regulate innate immune response functions, including superoxide anion (O 2 . ) generation, actin polymerization, and chemotaxis (16,17). Therefore, we sought in the present study to investigate whether LPS activates Rac and/or Cdc42 in the human PMN and, in parallel, to discern whether rafts exert any regulatory influence upon their activation or upon Rho GTPase-dependent functions. Here we report that LPS activates Cdc42 via a p38 mitogen-activated protein kinase (MAPK)-dependent mechanism and induces recruitment of Cdc42 into rafts, where it and p38 are both found to be activated. Furthermore, we demonstrate a clear role for lipid rafts in regulation of the activation state of p38 and Cdc42 and in regulation of Rho GTPase-dependent functions, specifically O 2 . generation and actin polymerization. These findings shed new light upon the basic mechanisms underlying the innate immune response, and raise new and interesting questions about the role of cholesterol in regulating the inflammatory phenotype of the human PMN.
Antibodies-Mouse anti-Rac antibody was purchased from Upstate Cell Signaling Solutions (Lake Placid, NY). Rabbit anti-Cdc42 and rabbit anti-p38 were purchased from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA). Mouse anti-moesin antibodies were purchased from BD Biosciences. Mouse anti-CD14 antibody was purchased from Coulter Corp. (Hialeah, FL). Rabbit anti-phospho-p38 antibody specific for the activation residues of p38 (Thr-180/Tyr-182) was purchased from Cell Signaling (Beverly, MA). Mouse anti-flotillin-1 antibody was purchased from BD Biosciences.
Cdc42 and Rac Activation Assays-Cdc42 and Rac activation levels in LPS-treated PMNs were assayed as described (19). Briefly, stimulated cells were lysed in 500 l of ice-cold 1ϫ MLB buffer (Upstate Biotechnology, Inc.) supplemented with 1 mM AEBSF, 5 g/ml leupeptin, 5 g/ml aprotinin, 1 mM NaF, and 1 mM Na 3 VO 4 and clarified by centrifugation (15,000 rpm, 10 min, 4°C). 5 g of PBD-GST-agarose was then added to the supernatants, and they were incubated with continuous rotation (60 min, 4°C). Prior to incubation with PBD-GST, non-LPS-exposed negative and positive control supernatants were incubated with 1 mM GDP and 100 M GTP␥S, respectively (15 min, 30°C), in the presence of 10 mM EDTA, after which they were immediately placed on ice and treated with 60 mM MgCl 2 . PBD-GST-agarose affinity precipitates were washed three times with 1ϫ MLB buffer, eluted in 2ϫ Laemmli reducing buffer (5 min, 100°C), electrophoresed on a 12% SDS-polyacrylamide gel, transferred to nitrocellulose, and then probed with anti-Rac and anti-Cdc42 antibodies. In some experiments, sucrose density gradient fractions of PMN lysates (described further below) were similarly assayed for active Cdc42.
Lipid Raft Isolation-PMNs (400 ϫ 10 6 ) were resuspended at a concentration of 20 ϫ 10 6 /ml in RPMI 1640 supplemented with 10 mM Hepes (pH 7.6) and 1% human heat-inactivated platelet-poor plasma, and either left untreated or stimulated as described. Following stimulation, rafts were isolated by using a modification of the method of Cheng et al. (20). Cells were centrifuged (15,000 rpm, 20 s) and resuspended in 1 ml of ice-cold lysis buffer (25 mM MES (pH 6.5), 150 mM NaCl, 1% Triton X-100, 1 mM AEBSF, 1 mM NaF; 1 mM Na 3 VO 4 ; 5 g/ml leupeptin; 5 g/ml aprotinin). PMNs were then sonicated for 60 s on ice. High density insoluble debris was removed by centrifugation (1000 ϫ g, 10 min, 4°C). The supernatant (950 l) was added to 2 ml of 80% sucrose in MBS (25 mM MES (pH 6.5), 150 mM NaCl), gently mixed by vortexing, and then centrifuged through a 10 -40% continuous sucrose gradient using an SW41 rotor (20 h, 37,000 rpm, 4°C). Following centrifugation, 10 1-ml fractions were harvested from the bottom by puncturing the tube with an 18-gauge needle. Protein concentrations in each fraction were determined by the Bradford Protein Assay (21), and 10 g of protein from each fraction was electrophoresed on 12% SDSpolyacrylamide gels, transferred to nitrocellulose, and probed with antibodies for proteins of interest. Raft-containing fractions were identified by measurement of alkaline phosphatase activity (see below), as well as by immunoblotting for the raft-specific protein flotillin-1 (22,23).
Alkaline Phosphatase Assay-Sucrose density gradient fractions were measured for enzymatic activity of alkaline phosphatase (a raftspecific, glycosylphosphatidylinositol-linked membrane protein (24,25)) to identify rafts (26). Briefly, 20 l of each fraction was added to separate wells of a 96-well flat-bottom plate and then mixed with 200 l of reaction buffer (5 mM p-nitrophenyl phosphate in 100 mM 2-amino-2-methyl-1-propanol, pH 10.0). Reaction mixtures were incubated (15 min, 37°C), and the absorbance was then read in a microplate reader at 405 nm.
Cellular Cholesterol Depletion and Repletion-PMNs were cholesterol-depleted with m␤CD using a modification of Ilangumaran and Hoessli (27). Briefly, cells were resuspended at 5 ϫ 10 6 /ml in Krebs-Ringer phosphate buffer supplemented with 0.2% dextrose and 0.25% delipidated bovine serum albumin, incubated with 5 mM m␤CD (15 min, 37°C), washed, and then resuspended. In order to demonstrate the raft specificity of m␤CD, cholesterol repletion of depleted cells (hereafter, "sequential repletion") was performed by incubation with m␤CD-100 M cholesterol complex (see below) (15 min, 37°C), followed by washing and resuspension. As a control for cholesterol repletion, depleted cells were incubated with m␤CD-100 M pregnenolone, a complex ineffective in cholesterol delivery. As an alternative control to sequential cholesterol depletion and repletion, untreated cells were treated with a concentration of cholesterol (10 M) complexed to m␤CD that was confirmed to maintain cellular cholesterol levels constant (hereafter, "iso-repletion," see under "Results"). 10 mM stock solutions of m␤CDcholesterol and m␤CD-pregnenolone complexes were formed by dissolving 30 mg of cholesterol or 24.7 mg of pregnenolone, respectively, in 400 l of 2:1 methanol/chloroform, and then dissolving these dropwise in 11 ml of 69.4 mM m␤CD in PBS preheated to 80°C. This solution was freeze-dried overnight, and working stock solutions were made in water.
Cellular Cholesterol and Phospholipid Quantitation-A mass spectroscopic method was developed for cellular cholesterol quantitation. To determine cholesterol content, 2 parts of MeOH was added to samples (1 ml). An aliquot of this sample (10%) was taken for analysis, and stable isotope-labeled d 6 -cholesterol (50 ng, Cambridge Isotope Laboratories, Inc., Andover, MA) was then added to this aliquot. After mixing, samples were extracted with 1 ml of iso-octane. A portion of the extract was dried under a stream of N 2 , and trimethylsilyl derivatives were prepared by adding 50 l each of bis(trimethylsilyl)trifluoroacetamide and acetonitrile. The vials were then capped and heated at 60°C for 15 min. Analysis of the samples was performed by a Finnigan Trace gas chromatography/mass spectrometry system (Thermo Finnigan, Thousand Oaks, CA), using a 15-m ZB-l column from Phenomenex (Torrance, CA), by injecting 1% of the extract. The mass spectrometer was operated in the electron impact mode. Data were acquired by selected ion monitoring, where m/z 368 was monitored for cholesterol and m/z 374 was monitored for the trimethylsilyl ether derivative of d 6 -cholesterol. Peak areas of each analyte were measured, and the ratio of the area of the cholesterol-derived ion to that from the internal standard was calculated. In separate experiments, 50 ng of the same internal standard solution of d 6 -cholesterol was added to known amounts of cholesterol in solution (reference standards) ranging from 640 pg to 2 g. The cholesterol was then extracted, derivatized, and analyzed as described above. When the 640-pg reference standard was analyzed (corresponding to 6.4 pg injected on-column) a signal-to-noise for m/z 368 was found to be 38. After the ratio of m/z 368 to m/z 374 was plotted against the amount of the reference standard divided by the amount of internal standard (50 ng), the resulting curve was found to be linear (y ϭ 0.9114 ϩ 0.0654) with a correlation coefficient (r 2 ) of 0.999. This curve was used to calculate the amount of cholesterol in each sample which ranged between 500 and 1400 ng/sample, all within the range of this standard curve.
To quantitate total phospholipid content, lipids were extracted using the method of Bligh and Dyer (28). The concentration of total phosphate within the extracted fractions was then determined as a surrogate measure of phospholipid concentration (29).
Assay of Superoxide Anion Release-Release of superoxide anion was determined in triplicate experiments using methods described previously (30).
Assay of Actin Polymerization by Flow Cytometry-Actin assembly was determined in triplicate by NBD-phallacidin staining, as described previously (31).

LPS Activates Cdc42 by a p38-dependent Mechanism-
Whereas an important role for Rho GTPases in such prototypical inflammatory responses of the PMN as O 2 . generation and actin polymerization has been demonstrated previously (16,17), to our knowledge, there have been no reports of activation of Rho GTPases by LPS in the human PMN. Given that LPS activates only the p38 MAPK arm of the MAPK superfamily in the human PMN in suspension (33), and that p38 plays an important role in O 2 . generation (18), we further queried whether p38 may lie upstream of Cdc42. In order to test these hypotheses, affinity precipitation was performed upon lysates of LPS-exposed PMNs using a recombinant PBD-GST fusion protein (19), which recognizes only the active, GTP-bound form of Cdc42 and Rac. Immunoblotting of the affinity precipitates indicates that LPS induces a time-dependent increase in the abundance of GTP-bound Cdc42 first detected after 5 min (Fig.  1A) and observed for 40 min (data not shown). LPS-induced activation of Cdc42 is inhibited by preincubation with the p38 inhibitor SB203580. No consistent effect upon LPS-induced Cdc42 activation was noted with piceatannol, a tyrosine kinase inhibitor, nor wortmannin, a phosphatidylinositol 3-kinase inhibitor (data not shown). Because Cdc42 activation by TLR2 agonists such as S. aureus has been reported previously in THP-1 cells (15), and concern has been expressed over the possibility of TLR2 agonists contaminating commercial LPS preparations (34), we tested whether peptidoglycan, a TLR2 agonist, would activate Cdc42 in the human PMN. Similar to S. aureus in THP-1 cells (15), peptidoglycan induced a rapid and transient activation of Cdc42 in the human PMN, with activation noted at 1-2 min and a return to base line before 5 min (data not shown). The much more delayed and prolonged time course of Cdc42 activation noted with LPS ( Fig. 1) indicates that TLR2 agonist contaminants (e.g. bacterial lipoproteins) are very unlikely to explain the phenomena noted with LPS in this study. LPS does not induce activation of Rac in the human PMN (Fig. 1A). Whereas no Rac activation was noted with LPS exposures ranging between 1 and 30 min (data not shown), rapid and transient Rac activation was observed after fMLP treatment (Fig. 1B). Of note, Rac activation in the human PMN has been reported previously not only with fMLP but also with platelet-activating factor, phorbol myristate acetate, and leukotriene B 4 (35, 36).

FIG. 1. Effect of LPS on Cdc42 and Rac activation.
A, human PMNs (20 ϫ 10 6 /ml) were preincubated with 10 M SB203580 or 0.1% Me 2 SO carrier, treated with 100 ng/ml of E. coli 0111:B4 LPS for the times indicated, lysed, and then incubated with 5 g of PBD-GST-agarose (4°C, 60 min). Precipitates were eluted, run on a 12% SDS-PAGE gel, transferred to nitrocellulose, and probed with either mouse anti-Rac or rabbit anti-Cdc42 antibody. Before incubation with PBD-GST, non-LPS-treated negative and positive control PMN lysates were treated with 1 mM GDP and 100 M GTP␥S, respectively. Lysates from all conditions were probed for total Cdc42 and Rac. The results shown are representative of five separate experiments. B, using the same methods as in A, activation of Rac by 10 M fMLP was assayed. A representative experiment of three consecutive experiments is shown.
LPS Induces Recruitment of Active Cdc42 to, and Exclusion of Rac from, Lipid Rafts-To test the possibility that LPS activates Cdc42 in human PMN rafts, PMNs were left untreated or exposed to LPS for 15 min. Following this, the cells were lysed, and 10 contiguous sucrose density gradient fractions were collected. Equal protein loads from all fractions were assayed for GTP-Cdc42 by incubation with PBD-GST-agarose, followed by elution, SDS-PAGE, transfer to nitrocellulose, and probing with anti-Cdc42. All fractions were also probed for total Cdc42 and were assayed for the presence of rafts by measurement of alkaline phosphatase activity and by immunoblotting with anti-flotillin-1 antibody. Of interest, despite the fact that the vast majority of cellular Cdc42 is localized to non-raft fractions in both nonstimulated and LPS-treated PMNs, Cdc42 is selectively activated in raft fractions, as identified by flotillin-1 ( Fig. 2A). This suggests that LPS activates a very minor, raft-localized, subfraction of total cellular Cdc42 to high specific activity. The small degree of Cdc42 activation in nonstimulated cells is also localized to rafts ( Fig. 2A). LPS alters neither the profile of alkaline phosphatase activity nor flotillin-1 abundance among sucrose density fractions ( Fig. 2A).
To confirm that LPS induces dynamic recruitment of total Cdc42 to lipid rafts, as has been demonstrated previously (37) for other signaling proteins, the lipid raft microdomain (collected from fractions 8 and 9, corresponding to Fig. 2A) was isolated from human PMNs exposed to LPS for varying durations ranging from 2 to 20 min. A 10-fold higher protein load (10 g) than that immunoblotted from each fraction in Fig. 2A   FIG. 2. A, LPS induces Cdc42 activation in PMN rafts. PMNs (200 ϫ 10 6 ) were left untreated or incubated with 100 ng/ml E. coli 0111:B4 LPS (37°C, 15 min) and lysed, and then 10 sucrose density gradient fractions of the lysates were collected as described under "Experimental Procedures." Twenty g of protein from all fractions (Bradford protein assay (21)) were tested for GTP-Cdc42 by incubation with 5 g of PBD-GST-agarose (4°C, 60 min) followed by SDS-PAGE, transfer to nitrocellulose, and probing with anti-Cdc42 antibody. Equal protein loads (1 g) from all fractions were also probed for total Cdc42 and flotillin-1. Fractions were tested for alkaline phosphatase (AP) activity (see "Experimental Procedures"). B, LPS induces translocation of signaling proteins to and from rafts. PMNs were left untreated or incubated with 100 ng/ml E. coli 0111:B4 LPS for the times indicated, and rafts were isolated (see "Experimental Procedures"). Rafts from each time point were identified by alkaline phosphatase activity in two contiguous low density fractions (fractions 8 and 9), and 10 g of protein from these fractions was electrophoresed on 12% SDS-polyacrylamide gels, transferred to nitrocellulose, and probed with antibodies of interest. C, LPS activates signaling proteins in rafts. Raft fractions from LPS-exposed PMNs were probed with anti-PO 4 -p38. To confirm a time course of Cdc42 activation in lipid rafts, PBD-GST-agarose precipitates from raft fractions were probed with anti-Cdc42. The results shown are representative of three separate experiments. was immunoblotted for Cdc42 and for other selected signaling proteins (Fig. 2B). As shown, LPS induces recruitment of Cdc42 to rafts, with a modest increase first detectable at 5 min. By contrast, Rac clearly diminishes in abundance in rafts after 10 min of LPS exposure. The cytoskeletal linker protein, moesin, reported previously to activate Cdc42 and other Rho GTPases by promoting their release from the Rho GDP dissociation inhibitor (38), is located within nonstimulated PMN lipid rafts and appears modestly to increase in rafts after LPS exposure.
In agreement with reports in other cell types (9,39), we find the LPS receptor component, CD14, to be localized within resting PMN rafts; moreover, it appears to increase, albeit transiently, upon LPS exposure (Fig. 2B). In addition, p38 MAPK is located within lipid rafts and is phosphorylated in rafts after LPS exposure (Fig. 2C). Immunoblotting of non-raft sucrose density gradient fractions also reveals the presence of p38 and LPSinduced PO 4 -p38 outside rafts (data not shown).
To confirm a time course of LPS-induced activation of Cdc42 in rafts, PBD-GST affinity precipitates from contiguous raft fractions of PMNs exposed to LPS for 0, 5, or 10 min were immunoblotted for Cdc42 (Fig. 2C). As shown, Cdc42 is activated in rafts after 5 min of LPS exposure, corresponding to the first detectable LPS-induced increase in total Cdc42 within PMN rafts (Fig. 2B), and increases further in activation level at 10 min.
m␤CD Extracts Cholesterol Selectively from the Lipid Raft Microdomain of Human Neutrophil Membranes-Cyclodextrins are cyclic oligomers of glucose that extract cholesterol from intact, living cells in a reversible, time-and concentrationdependent fashion (40). Their utility lies in the fact that cyclodextrin-mediated cellular cholesterol depletion is thought to disrupt lipid rafts physically and functionally (41), thereby providing clues to the native regulatory influence of intact rafts upon cellular phenomena. Moreover, the specificity for raft integrity of cyclodextrin-induced effects can be confirmed by using the cyclodextrin-cholesterol complex as a vehicle subsequently to "replete" cellular cholesterol in cyclodextrin-"depleted" cells, thereby restoring raft integrity and function (5). Because several different cyclodextrin compounds and protocols have been applied in the literature, and the specificity of cyclodextrins for raft versus non-raft cholesterol is a matter of some dispute (4,27), we preferred to confirm the specificity of our cholesterol depletion protocol for PMN raft microdomains. Rather than quantitating absolute cholesterol levels in membrane fractions as others have done (5), we normalized cholesterol to membrane phospholipid content in an attempt to verify the specificity of cyclodextrin for cholesterol extraction. As depicted in Fig. 3A, our m␤CD depletion protocol (5 mM m␤CD, 15 min, 37°C) extracts cholesterol specifically from rafts. Nevertheless, our m␤CD protocol did not noticeably alter the distribution of CD55, flotillin, or alkaline phosphatase activity within sucrose density gradients of resting PMNs (data not shown, see "Discussion").
Like other investigators, we tested the reversibility of effects induced by cholesterol depletion by performing cholesterol repletion on depleted cells. This sequential repletion (see below and under "Experimental Procedures") is thought largely to restore raft structure and function. However, we surmised that certain physiologic effects of m␤CD-mediated cholesterol depletion described below (e.g. actin polymerization) would be unlikely to be reversed upon subsequent cholesterol repletion, despite effective re-ordering of rafts. Therefore, we sought to confirm a dose of cholesterol complexed to m␤CD for primary treatment of cells that could serve as a control for the raft specificity of m␤CD by maintaining cellular cholesterol at a constant level, what we have termed iso-repletion. As depicted in Fig. 3B, 10 M cholesterol complexed with m␤CD was confirmed to maintain base-line cellular cholesterol levels. Progressive increases in the amount of cholesterol complexed to m␤CD yield progressively higher deliveries of cholesterol to the cell (data not shown). In sum, these studies confirm the selectivity of the m␤CD depletion protocol for raft cholesterol.
Cholesterol Depletion Activates p38 and Cdc42-Given our observations of the raft localization of p38, Cdc42, and Rac, and of LPS-induced translocation of the latter two ( Fig. 2A), we queried whether rafts might exert a regulatory influence upon the activation state of these signaling proteins. To address this question, we performed immunoblotting upon lysates (to quan- titate active phospho-p38) and PBD-GST affinity precipitates (to quantitate GTP-Cdc42 and-Rac) of PMNs that were cholesterol-depleted with m␤CD or-iso-repleted with the m␤CD-10 M cholesterol complex. As depicted in Fig. 4A, cholesterol depletion induces phosphorylation of p38 and activation of Cdc42. Neither of these effects occurs in cells subjected to cholesterol iso-repletion, confirming that the m␤CD effect is through cholesterol depletion. These findings suggest the following: 1) a role for rafts in maintenance of p38 and Cdc42 in their inactive state in nonstimulated PMNs, and 2) that cholesterol depletion is one mechanism for activation of both of these proteins. Most intriguingly, preincubation of PMNs with the p38 inhibitor SB203580 before cholesterol depletion partially inhibits m␤CD-induced Cdc42 activation, confirming an intermediary, albeit partial, role for p38 in this phenomenon (Fig. 4B). By contrast, Rac is not activated by m␤CD (Fig. 4A). The m␤CD preparation used was confirmed to have no detectable LPS contamination by Limulus amebocyte lysate assay (sensitivity 0.06 enzyme units/ml Food and Drug Administration endotoxin).
CD14 and other members of the LPS receptor complex have been reported to reside within the lipid raft microdomain of monocytes (9, 39). We therefore further queried whether cholesterol depletion would inhibit subsequent LPS-induced acti- LPS is an effective priming agent for subsequent O 2 . release triggered by such agents as the formylated bacterial tripeptide fMLP (42). Given these premises, and our observation of LPSinduced translocation of Rac out of rafts (Fig. 2), we queried whether rafts might exert a regulatory influence upon O 2 . generation in the human PMN. We therefore tested whether m␤CD has any priming effect upon subsequent O 2 . production triggered by fMLP (30). As depicted in Fig. 5, cholesterol depletion with m␤CD was effective by itself in priming the PMN for subsequent O 2 . production triggered by both fMLP and cytochalasin D, a commonly used PMN priming agent (43). Whereas this priming effect of m␤CD upon fMLP-induced O 2 .
production is in agreement with one previous report (44), to our knowledge ours is the first report to confirm the reversibility of this priming effect with sequential cholesterol repletion (Fig.  5), but not with the pregnenolone control. Reversal of priming was also observed with sequential treatment with 5-cholesten-3-one, an alternative molecule effective in cellular cholesterol repletion (Fig. 5).
Cholesterol Depletion Induces Actin Polymerization-Because Cdc42 is thought to play an important role in actin remodeling (45) and is regulated by lipid rafts (Fig. 4), we queried whether rafts might regulate actin polymerization in the human PMN. As shown in Fig. 6A, cholesterol depletion is, by itself, a potent stimulus for actin polymerization in the human PMN, inducing actin assembly to a magnitude comparable with LPS. We thought it unlikely that subsequent cholesterol repletion of the m␤CD-treated cells would reverse Factin formation and thereby serve as a suitable control to demonstrate the raft specificity of the m␤CD effect. In fact, sequential repletion of cholesterol-depleted cells was noted to further increase F-actin content beyond that measured with m␤CD treatment (data not shown), suggesting that perturbations of cellular cholesterol content upward and downward are both sufficient to induce actin polymerization. Most intriguingly, in this vein, we noted a parallel increase in both p38 phosphorylation and Cdc42 activation (Fig. 4C) with sequential repletion over that observed with initial cholesterol depletion. As shown in Fig. 6A, the raft specificity of m␤CD-mediated actin polymerization was instead demonstrated by effective blockade of m␤CD-induced actin polymerization by the inclusion of cholesterol in an iso-repletion protocol. Confocal microscopy of PMNs stained with rhodamine-phalloidin indicates that LPS and m␤CD induce markedly different morphologies of F-actin remodeling, the former a pattern of cytoskeletal polarization, as reported previously (46), and the latter a cortical rim of F-actin (Fig. 6B). DISCUSSION Lipid rafts are cholesterol-rich membrane microdomains that are thought to organize agonist-induced interactions among signaling proteins (1,2). They have been demonstrated to play an important regulatory role in multiple signaling pathways, including B and T cell receptor signaling, PDGF signaling, and most recently, LPS signaling. As a physicochemically isolable signaling compartment, rafts have provided a new window upon novel and sometimes unexpected signaling interactions.
By embarking upon previous reports of dynamic translocation of Rho family GTPases to rafts (11), and of a role for these proteins in the LPS signaling cascade in cell lines (13,14), we report that LPS activates Cdc42 by a p38-dependent mechanism in the human PMN and induces translocation of Cdc42 to PMN rafts. Raft-associated Cdc42 and p38 are found to be activated after LPS exposure. We confirm the presence of the LPS receptor component, CD14, within human PMN rafts. Moreover, selective perturbation of raft function with m␤CD suggests a role for rafts in both 1) the maintenance of p38 and Cdc42 in their inactive form in the resting human PMN, and 2) the transmission of the LPS signal to both p38 and Cdc42. Regarding the former, a previous report (47) describing no activation of p38 with m␤CD treatment of NIH 3T3 cells suggests cell type specificity for this phenomenon. A report of ligand-independent activation of the epidermal growth factor receptor by m␤CD (48) suggests a possible mechanism, by analogy, for the recapitulation by m␤CD of multiple effects of LPS noted in the present study (i.e. activation of p38 and Cdc42 but not Rac; O 2 . priming; and actin polymerization). However, this phenomenon is unlikely to apply to the LPS receptor because m␤CD activates extracellular signal-regulated kinase in the human PMN (data not shown) but LPS does not (33). Furthermore, m␤CD reversibly primes for O 2 . release (Fig. 5) and remodels actin in a clearly distinct manner from LPS (Fig.  6B). Regarding a role for rafts in transmission of the LPS signal to p38, it has been reported previously (9) that tumor necrosis factor-␣ secretion by LPS-stimulated human monocytes is blocked by preincubation with m␤CD. Most interestingly, raft integrity appears to be somewhat specific for transmission of the LPS signal to p38, because m␤CD pretreatment does not inhibit hyperosmotic shock-induced p38 phosphorylation (47). Cdc42 is activated selectively in PMN rafts, whereas p38 is activated both within and outside PMN rafts. Although we are not aware of previous reports of p38 within lipid rafts, a precedent for raft compartmentalization of p38 activity is provided by one report that p38 inhibition decreases stress-induced phosphorylation of the raft-localized protein caveolin-1 in NIH 3T3 cells (47). LPS has also been reported to activate c-Jun  (9), and m␤CD has been reported to activate extracellular signal-regulated kinase within COS-1 cell rafts and to prime for epidermal growth factor-induced activation of extracellular signal-regulated kinase in COS-1 lysates (4). Activation and translocation of Rac1 and RhoA to fibroblast rafts have been reported to be induced by PDGF (12). Consistent with these reports on other Rho GTPases, the present work indicates that PMN rafts play an important role in Cdc42 regulation. This is evinced both by the selective, raft-localized activation of Cdc42 induced by LPS and by detectable increases in GTP-Cdc42 from lysates of m␤CD-treated cells compared with untreated cells (Fig. 4A). We found that m␤CD activates p38 and Cdc42 in resting PMNs and yet inhibits LPS-induced activation of p38 and Cdc42 (Fig. 4). We speculate that these findings can be reconciled by the hypothesis that resting PMN rafts exert a negative regulatory influence upon Cdc42, and yet also contain feed-forward positive signals important to LPSinduced activation of Cdc42. The similar temporal and quantitative increases of total Cdc42 and GTP-Cdc42 in rafts induced by LPS (Fig. 2) suggest that most, if not all, raft-localized Cdc42 is activated, as has been suggested for other Rho GTPases (12). This, in turn, is consistent with either one of two mechanistic possibilities: 1) activation of Cdc42 induces its translocation to lipid rafts or 2) translocation of Cdc42 to lipid rafts induces its activation in that location. These two mechanisms are not distinguished by the present study. The first is supported by a previous report (12) that GTP␥S treatment of fibroblast lysates induces translocation of RhoA to caveolae. The second is supported by our observation of the co-localization within rafts of two positive Cdc42 regulators, p38 and moesin, the latter reported to activate Cdc42 by exclusion of Rho GDP-dissociation inhibitor (38). Of interest, moesin has been reported previously to link lipid rafts with the actin cytoskeleton in T lymphocytes (49), to be a potential receptor for LPS on monocytes (50), and consistent with our findings (Fig. 2B), to further translocate to the PMN plasma membrane upon activation (51).
In addition to demonstrating a role for rafts in regulation of Cdc42 activity, we also demonstrate raft regulation of two Rho GTPase-dependent functions, O 2 . production and actin polymerization. Most interestingly, cholesterol depletion was a potent, yet reversible priming mechanism for fMLP-triggered O 2 . production (Fig. 5). Although the mechanism underlying this phenomenon remains unclear, it cannot be explained solely by m␤CD-induced activation of p38 or Cdc42. fMLPinduced O 2 . release is p38-dependent in the PMN (18). However, sequential cholesterol repletion of depleted cells increases p38 phosphorylation, yet reverses m␤CD-induced O 2 . priming (Figs. 4C and 5). Similarly, whereas Cdc42, like Rac, has been linked to the oxidative burst of the PMN (52), sequential repletion also increases its activation over that observed with initial cholesterol depletion (Fig. 4C). Like m␤CD, LPS primes the human PMN for fMLP-triggered O 2 . production without activating Rac (Figs. 1, 4, and 5). Whereas the mechanism of both m␤CD-and LPS-induced O 2 . priming remains somewhat unclear in the literature, it is tempting to speculate that they both may operate through relocalization of Rac away from the raft compartment ( Fig. 2B) or, potentially, through relocalization of other proteins, such as the Phox components. M␤CD induces actin polymerization in a subcortical distribution markedly distinct from that induced by LPS (Fig. 6B). Of interest, this pattern of actin remodeling is very similar to that noted in the PMN within the first 30 s of fMLP exposure (53), suggesting that M␤CD and fMLP may operate through like mechanisms. The comparison is, however, complicated by the fact that fMLP is reported to activate RhoA, Cdc42, and Rac (the last through a Cdc42-dependent mechanism) in the human PMN (54,55). Unlike O 2 . production and most other physiologically relevant functions of the LPS-stimulated human PMN (e.g. tumor necrosis factor-␣ production, adhesion, and chemotaxis), actin polymerization in the human PMN is p38-independent (33). This suggests that LPS-induced activation of Cdc42 is not central to, or at least not a rate-limiting step in, the remodeling of actin produced by LPS. Future studies will need to test whether LPS and/or m␤CD activate RhoA.
FIG. 6. Effect of cholesterol depletion upon actin polymerization. A, PMNs were either left untreated (buffer), incubated with 100 ng/ml E. coli LPS (45 min), incubated with 5 mM m␤CD, or cholesterol-iso-repleted with m␤CD/10 M cholesterol (chol). F-actin was then assayed by cell fixation/permeabilization followed by staining with NBD-phallacidin and flow cytometry (31). Actin assembly was quantified by the relative fluorescence index (RFI) or mean fluorescence of the sample divided by that of nonstimulated cells. The values (ϮS.E.) shown were generated from three separate experiments. B, PMNs were either left untreated, incubated with LPS, or incubated with m␤CD, as described above in A, and then fixed, permeabilized, and stained with rhodamine-phalloidin (Molecular Probes, Eugene, OR). Photographs were taken using an Olympus Vanox Microscope with 60ϫ oil objective using Slidebook software (Intelligent Imaging, Denver, CO).
Limitations of the present study should be noted. Although we did not directly measure intrinsic GTPase activity, quantitation of GTP-Cdc42 with PBD affinity precipitation is a well accepted and reliable technique (19). The terminally differentiated human PMN cannot be transfected, thus limiting studies such as the present one to the use of chemical inhibitors. Although some controversy exists regarding the relationship between "detergent-resistant membranes," Triton-insoluble low density membrane fractions, as isolated in the present study, and "lipid rafts," this nevertheless represents a well accepted methodology for raft isolation (20). The demonstration of two proteins in raft isolates does not prove their physical association. In fact, rafts are found not only in the plasma membrane but also within intracellular membranes (1,2), and rafts from both locations are expected to be collected together by our methodology. The significance of a broader distribution of rafts within the sucrose density gradient identified by flotillin-1 as compared with alkaline phosphatase activity ( Fig. 2A) remains somewhat uncertain; nevertheless, flotillin-1 is an integral membrane protein that has been confirmed to be raftspecific in multiple cells, including PMNs (22,23). Moreover, we confirmed both translocation of total Cdc42 to, and activation of Cdc42 in, raft fractions identified by the narrower range of alkaline phosphatase activity (Fig. 2C). In any event, several groups have reported the existence of heterogeneous subtypes of lipid rafts within single cell types, and in fact, one group, like us, has reported previously (56) that raft-localized Cdc42 has a slightly denser distribution compared with other raft markers such as CD55 and CD14.
Regarding the results derived from cholesterol depletion, we cannot exclude the possibility that we may have obtained different results with a different depletion protocol, such as one using a different cyclodextrin compound or extracting a higher percentage of cellular cholesterol. Comparable or less aggressive cholesterol depletion protocols have been validated previously for the neutrophil (57,58). Although some groups have determined that m␤CD is not specific for raft cholesterol (27,59), the depletion protocols that have typically been used (e.g. 10 mM m␤CD for 60 min (59) or 20 mM m␤CD for 30 min (27)) would be expected to remove substantially more cholesterol than our own, and therefore may well be less selective. Groups like ours that have normalized cholesterol measurements to protein or phospholipid in an attempt to quantify specific cholesterol content have also found that cyclodextrins are specific for raft cholesterol extraction (4). Our depletion protocol did not disperse the distribution within sucrose density gradients of flotillin, CD55, or alkaline phosphatase activity (data not shown). Others have similarly noted that cholesterol depletion does not invariably decompartmentalize raft-localized proteins, such as linker for activation of T cells in human peripheral blood T cell lymphoblasts (61), or GM1, Thy1, and Lck in murine splenic T cells (62). These observations likely indicate the following: 1) the existence of raft subtypes of varying cholesterol composition and sensitivity (62); 2) the potential for varying thresholds of cholesterol sensitivity among different proteins for their raft compartmentalization (63); and 3) the possibility that sucrose density gradient separation may be too insensitive to detect all aspects of raft organization (61). In this light, the phrase "raft perturbation" may be both more appropriate and more to the point than "raft disruption," as a descriptor for the m␤CD effect. In the present study, the specific effect of m␤CD on raft structure/function was indicated not only by selective cholesterol depletion from rafts ( Fig. 3A) but also by the multiple m␤CD-modulated signaling and functional phenomena (i.e. activation of p38 and Cdc42, O 2 . release, and actin polymerization) that were shown to be cholesterol-de-pendent by their blockade/reversal with concomitant/sequential cholesterol treatment. Moreover, some of these phenomena (e.g. activation of p38 and Cdc42) were localized to lipid rafts (Fig. 2). In summary, LPS activates Cdc42 by a p38-dependent mechanism in the human PMN, inducing its recruitment to lipid rafts, where it and p38 are both found to be activated. Lipid rafts exert an important regulatory influence upon the activation state of p38 and Cdc42 and upon transmission of the LPS signal to both p38 and Cdc42. In parallel, lipid rafts also regulate the Rho GTPase-dependent functions, O 2 . production, and actin polymerization, albeit in a distinct manner. Raft perturbation via cholesterol depletion primes for fMLP-triggered O 2 .
production in a reversible manner but directly induces actin polymerization in an apparently irreversible manner. Our demonstration of the important regulatory influence of rafts upon the inflammatory phenotype of the human PMN raises several interesting questions. Because serum and cell membrane cholesterol have been reported to be correlated (64), we speculate that our findings may underlie both leukocyte-mediated inflammation associated with acute hypercholesterolemia in animal models (65), and poorer clinical outcomes in critically ill patients who sustain acute declines in serum cholesterol (66). Recent reports (60) of a decline in raft cholesterol with aging also raise the possibility that this phenomenon may be causally linked to age-related diseases. Thus, manipulation of cellular cholesterol content may perhaps represent another approach to modification of the innate immune response.