Cytoskeletal Modulation of Lipid Interactions Regulates Lck Kinase Activity*

Background: Rafts are important for phosphoregulation of Lck, but how they are formed and maintained in cell membranes is poorly understood. Results: Disrupting the actomyosin cytoskeleton declusters raft lipophilic reporters and deregulates Lck. Conclusion: The actomyosin cytoskeleton maintains lipid interactions that sustain rafts for Lck regulation. Significance: These data provide new information regarding how rafts are maintained for Lck regulation. The actin cytoskeleton promotes clustering of proteins associated with cholesterol-dependent rafts, but its effect on lipid interactions that form and maintain rafts is not understood. We addressed this question by determining the effect of disrupting the cytoskeleton on co-clustering of dihexadecyl-(C16)-anchored DiO and DiI, which co-enrich in ordered lipid environments such as rafts. Co-clustering was assayed by fluorescence resonance energy transfer (FRET) in labeled T cells, where rafts function in the phosphoregulation of the Src family kinase Lck. Our results show that probe co-clustering was sensitive to depolymerization of actin filaments with latrunculin B (Lat B), inhibition of myosin II with blebbistatin, and treatment with neomycin to sequester phosphatidylinositol 4,5-bisphosphate. Cytoskeletal effects on lipid interactions were not restricted to order-preferring label because co-clustering of C16-anchored DiO with didodecyl (C12)-anchored DiI, which favors disordered lipids, was also reduced by Lat B and blebbistatin. Furthermore, conditions that disrupted probe co-clustering resulted in activation of Lck. These data show that the cytoskeleton globally modulates lipid interactions in the plasma membrane, and this property maintains rafts that function in Lck regulation.

Structurally, rafts are posited to consist of lipids that are ordered through interactions with cholesterol. Cholesterol-dependent lipid ordering to form discrete domains is represented in model membranes by combinations of cholesterol and fluid phase lipids, notably sphingolipids, that form a cholesterol-dependent liquid-ordered (L o ) phase (6 -8). Some bilayer compositions contain coexisting L o and liquid-disordered (L d ) phases (8,9), and this is suggested to be representative of coexisting raft and nonraft environments in cell membranes (10). Coexisting ordered and disordered phases can be produced in blebs that are generated by chemical treatment of cells (11), showing that the plasma membrane contains lipid mixtures that will undergo phase separations in certain conditions. Consistent with the membrane raft model, proteins and lipids that favor L o lipid environments exhibit a cholesterol-dependent clustering in the plasma membrane (12,13). Furthermore, clustering of raft markers in the plasma membrane often correlate with the F-actin content of the cell (14,15), suggesting that the cytoskeleton can promote, by poorly understood mechanisms, lipid interactions that transition them to a more ordered state (16). This interpretation is supported by data showing that lipids in model membranes undergo ordering as a result of attachment of actin filaments to the bilayer (17). However, cytoskeletal ordering of lipids in minimalist systems such as lipid vesicles may not be representative of cell membranes. Emission from the lipophilic probe Laurdan, which is sensitive to solvent polarity, is shifted by disrupting the cell cytoskeleton in a manner that is consistent with disordering of plasma membrane lipids (16), yet this may reflect changes in probe fluorescence by parameters unrelated to lipid ordering. Also unknown is whether the effect of the cytoskeleton on lipid interactions is specific to lipids that favor ordered lipid environments such as rafts.
Most rafts have a diameter less than ϳ20 nm (18), thus challenging their characterization in intact cells because they are too small to visually resolve by light microscopy. Nonetheless, separate approaches exist for resolving nanoscale complexes, one being fluorescence resonance energy transfer (FRET) (13, 19 -21). Measuring FRET between dialkyl forms of the carbocyanine dyes DiO and DiI, Baird and co-workers (21) showed evidence of cholesterol-dependent lipid heterogeneities in the plasma membrane of RBL cells. However, the role of the cytoskeleton in forming these lipid complexes was not reported.
Herein, we report experiments that measured the effect of altering the cytoskeleton on FRET between membrane-anchored forms of DiO and DiI. Our data show that conditions that disrupt either the structural or functional integrity of the cytoskeleton resulted in a decrease in probe clustering associated with lipid ordering. These conditions also altered phosphorylation of Lck on its regulatory tyrosines to increase the amount of active protein. Altogether, these data favor a model where the cytoskeleton regulates formation of lipid domains that are necessary for efficient regulation of Lck.

EXPERIMENTAL PROCEDURES
Cell Culture and Sample Preparation-Jurkat T cells (clone E6-1) were prepared for microscopy by seeding ϳ10 6 cells onto a poly-L-lysine-coated (Sigma) coverslip, followed by washing with RPMI containing 50 mM HEPES (pH 7.4) (RPMI-HEPES), and then adding RMPI-HEPES containing either 1% dimethyl sulfoxide (Sigma) alone, or drug diluted from a 100ϫ stock solution in dimethyl sulfoxide. Final drug concentrations were 5 M latrunculin B (Lat B) (Calbiochem, La Jolla, CA), 50 M blebbistatin ((Ϫ) stereoisomer) (Calbiochem), and 5 g/ml filipin (Cayman Chemicals, Ann Arbor, MI). Incubations with either drug or vehicle alone were for 30 min at 37°C. Alternatively, cells were treated with 20 mM neomycin overnight at 37°C in RPMI-HEPES containing 2% FBS. Following treatment, the samples were labeled with C 16 -DiO and either C 16 -DiI or C 12 -DiI by incubating for 10 min at 4°C in RPMI-HEPES containing 1 M label. Samples were maintained at 4°C until imaging, after which they were equilibrated to and maintained at 37°C.
In experiments employing immunoblotting, cells were treated in suspension at a density of 10 6 /ml using the media and drug concentrations described above. Following incubation, the cells were sedimented, quickly resuspended in Laemmeli sample buffer containing 1% 2-mercaptoethanol, and then incubated at 100°C for 5 min. Proteins were separated by gel electrophoresis (10% acrylamide) and then immunoblotted using monoclonal antibody to Lck (clone 2B, BD Biosciences), rabbit antibody to pTyr 505 of Lck (Cell Signaling Technology; Danvers, MA), or rabbit antibody to pSrc 416 (Cell Signaling Technology). Immunoblots were developed by ECL (Amersham Biosciences, Rockford, IL), and detected using a G:Box imaging system (Syngene, Frederick, MD).
Ca 2ϩ Flux Measurements-Ca 2ϩ flux from T cell stimulation was measured using the Ca 2ϩ indicator Indo-1 (Invitrogen) as described previously (22). Indo-1 fluorescence was measured by flow cytometry (MOFLO; Dako Cytomation; Fort Collins, CO) based on the fluorescence emission at 475 and 400 nm. Excitation for Indo-1 was at 390 nm. A base line was achieved by passing the cells for approximately 1 min prior to addition of ␣CD3 monoclonal antibody (OKT3) for cell stimulation. The temperature of the instrument was maintained at 37°C, and the flow rate was maintained by a base sheath pressure of 60 psi Ϯ 1 psi.
Preparation of GUVs-0.05 mg of lipid containing 0.1 mole % C 16 -DiO and C 12 -DiI was dried on to a indium tin oxide-coated slide (Delta Technologies, Stillwater, MN), first using a gentle stream of Ar gas, followed by incubation under vacuum for 1 h at room temperature. Giant unilamellar vesicles (GUVs) were generated by electroformation as described (23), using a sandwich composed of two indium tin oxide-coated slides and containing ϳ0.5 ml of 300 mM sucrose in Milli-Q water (Millipore, Billerica, MA). GUVs were grown using a sinusoidal wave current (1 V, 10 Hz) for not Ͻ1 h at 60°C. The GUV fraction was collected and diluted with an equal volume of 150 mM NaCl in Millipore water for imaging.
Flow Cytometry-Jurkat T cells were fixed in 2% (w/v) paraformaldehyde for 20 min at 37°C and permeabilized in 0.1% (v/v) Triton X-100 in PBS-glycine for 10 min at room temperature. Staining with Texas Red-conjugated phalloidin was done as we have described (24). Cells were analyzed on an LSR II Flow Cytometer (BD Biosciences). 10,000 events were counted from each sample. Cells were gated on Forward Scatter and Side Scatter using untreated cells, and this gate was applied to the remaining samples for measurement of Texas Red fluorescence.
Cell Imaging and Analysis-Unless otherwise noted, imaging was performed using a Zeiss LSM 510 META confocal microscope (Oklahoma Medical Research Foundation imaging core facility) equipped with a 63ϫ water objective (NA, 1.2) and a thermo-controlled chamber for maintaining the samples at 37°C (19). Image processing and quantitation were performed using iVision imaging software (version 4.0, BioVision Technologies, Exton, PA).
FRET was measured by detecting sensitized emission of acceptor following donor excitation. Accordingly, images were acquired in three separate channels: donor channel (DiO, 488 nm excitation/505 to 550 nm emission), acceptor channel (DiI, 543 nm excitation/560 nm Ͻ emission), and FRET channel (488 nm excitation/560 nm Ͻ emission). FRET efficiency was calculated using images collected in the FRET channel. Images of cells labeled with either C 16 -DiO or C 16 -DiI alone were collected in the DiO and DiI channels to determine correction factors necessary to eliminate contributions from donor and acceptor bleed-through to the FRET channel (25)(26)(27). The parameter K A for conversion of fluorescence ratios to FRET efficiency values (see Equation 1) (27) was determined by measuring quenching of donor fluorescence following addition of acceptor. Specifically, cells labeled with C 16 -DiO alone were imaged in the donor channel (F D ) and then re-imaged in the donor channel following addition of an equal amount of C 16 -DiO (F DA ). The decrease in donor fluorescence by labeling with acceptor represents the FRET efficiency, E, calculated using the equation: K A was calculated from E quenching as described (see Equation 5 in Ref. 27).

Cell Labeling and Detection of FRET between Lipophilic
Carbocyanine Dyes-We used dihexadecyl (C 16 )-and didodecyl (C 12 )-anchored forms of DiO and DiI to measure by FRET microscopy membrane heterogeneities associated with lipid ordering. The principal behind this approach is illustrated in Fig. 1A. Specifically, the C 16 tail targets the probes to domains composed of ordered lipids, whereas C 12 -anchored probes favor less-ordered L d phase lipid environments (28,29). Accordingly, rafts composed of ordered lipids co-label with C 16 -anchored DiO (C 16 -DiO) and C 16 -anchored DiI (C 16 -DiI), minimizing their intermolecular distance to increase the FRET efficiency relative to that between C 16 -DiO and C 12 -anchored DiI (C 12 -DiI) (21).
We show examples of the distinct affinities of the C 16 -and C 12 -anchored probes for L o phase lipids in Fig. 1B. Each row consists of images of a GUV double-labeled with C 16 -DiO and C 12 -DiI. In the top row is a GUV containing co-existing L o and L d phases; C 16 -DiO and C 12 -DiI segregate due to enrichment in the separate lipid phases. Conversely, in the bottom row is a GUV containing only L d phase lipids, and C 16 -DiO and C 12 -DiI colocalize throughout the vesicle. Measurement of multiple (n ϭ 31) mixed L o /L d phase GUVs showed that 20% of the C 16 -DiO signal colocalized with regions of C 12 -DiI enrichment. Thus, 80% of the C 16 -DiO was restricted to the L o phase.
We controlled for probe co-clustering absent lipid ordering by measuring FRET in GUVs composed of L d phase lipids. Specifically, demixing of probes due to unexpected interactions between C 16 -DiO and either C 16 -DiI or C 12 -DiI will show as an elevated FRET efficiency. However, we observed no significant difference in FRET efficiency between the separate donor-acceptor pairs (Fig. 1C), showing that unequal mixing of C 16 -DiO with either C 16 -DiI versus C 12 -DiI was not detected even at nanoscale dimensions.
Lipid Ordering in T Cell Plasma Membrane by Cytoskeleton-We measured the plasma membrane of Jurkat T cells, where rafts are critical for cell regulation and stimulation through the T cell receptor (4,30). The cells were double-labeled using conditions that minimized internalization of the probes (see "Experimental Procedures"), while also producing efficient labeling of the plasma membrane and a FRET signal that was specific to cells that contained both DiO and DiI (Fig. 1D). To control for perturbation of the plasma membrane by the labeling, we measured the increase in intracellular Ca 2ϩ , or Ca 2ϩ flux, which occurs following cross-linking the T cell receptor as this is sensitive to changes in plasma membrane permeability (31). We observed that double-labeling cells with C 16 -DiO and C 16 -DiI had no affect on either the magnitude or duration of the Ca 2ϩ flux (Fig. 1E), thus indicating no significant perturbation of the outer membrane by the probes.  16 -anchored dialkyl carbocyanine dyes incorporate into ordered phase lipids, such as the L o phase rafts, whereas the C 12 -anchored probes prefer L d phase lipids (21). In membranes double-labeled with C 16 -DiO and C 16 -DiI, the probes will co-localize in ordered lipid microenvironments, thus elevating FRET efficiency relative to that of C 16 -DiO and C 12 -DiI. B, epifluorescence images of GUVs double-labeled with C 16 -DiO and C 12 -DiI. In the top row, the vesicle is composed of DPPC:DOPC:cholesterol (2:2:1), which forms co-existing L d and L o lipid phases at room temperature (8). The vesicle in the bottom row is composed of DOPC alone, which occurs as a single L d phase at room temperature. C, FRET efficiency values measured between C 16 -DiO and either C 16 -DiI (C 16 /C 16 ) or C 12 -DiI (C 16 Fig. 2A are results from measuring FRET between C 16 -DiO and either C 16 -DiI or C 12 -DiI in double-labeled Jurkat cells containing the indicated ratios of donor-toacceptor (D:A). These data show co-clustering of C 16 -DiO and C 16 -DiI was significantly greater than that of C 16 -DiO and C 12 -DiI. For example, FRET efficiency for C 16 -anchored DiO and DiI was ϳ2-fold or greater than that of C 16 -DiO and C 12 -DiI at each D:A ratio. Furthermore, FRET between C 16 -anchored DiO and DiI was sensitive to the D:A ratio, which is another signature of co-clustering of the donor and acceptor (32). The elevated co-clustering of C 16 -DiO and C 16 -DiI was cholesterol-dependent because sequestering cholesterol with filipin decreased their FRET efficiency to values similar to that of C 16 -DiO and C 12 -DiI, and the FRET became independent of the D:A ratio (Fig. 2B). Importantly, the decrease in FRET efficiency for the C 16 -anchored probes was not due to anomalous quenching of the DiI by filipin because fluorescence intensities of cells labeled with C 16 -DiI were unchanged upon addition of the drug (supplemental Fig. S1). Altogether, these data are consistent with the notion of cholesterol-dependent rafts in the plasma membrane that are composed of L o phase lipids.

Summarized in
To determine whether co-clustering of C 16 -anchored DiO and DiI was sensitive to disruption of the actomyosin cytoskeleton, we measured cells treated with either latrunculin B (Lat B) to disrupt actin filaments, or blebbistatin to inhibit nonmuscle myosin II (NM II). Each of these conditions was as efficient as filipin in reducing FRET between C 16 -DiO and C 16 -DiI, and each resulted in the FRET efficiency becoming independent of the D:A ratio (Fig. 2B). Importantly, the decrease in probe coclustering by filipin and blebbistatin was not due to a loss of F-actin, as treated and untreated cells showed similar amounts of staining by phalloidin (Fig. 3). Altogether, these data show that ordered lipid domains detected by FRET between C 16 -anchored probes are sensitive to conditions that either disrupt F-actin or inhibit NM II activity.
Cytoskeleton Elevates Interactions between L o -and L d -preferring Probes-In contrast to our results with C 16 -DiO and C 16 -DiI, FRET between C 16 -DiO and C 12 -DiI was not affected by filipin, Lat B, or blebbistatin (supplemental Fig. S2). This suggests that co-clustering of C 16 -DiO and C 12 -DiI is minimal and therefore not further reduced by the respective treatments. Alternatively, lipid heterogeneities co-labeled with C 16 -DiO and C 12 -DiI may exist but were not distinguished by the FRET analysis in Fig. 2. To discriminate these separate interpretations, we employed an alternative approach to assess probe coclustering using FRET efficiency values. This consisted of measuring the effect of acceptor concentration (F) on FRET efficiency, which can distinguish clustering events not resolved by measuring FRET over a narrow range of acceptor concentrations (32). Co-clustering of each donor-acceptor pair was quantitated as described (13), which consisted of fitting FRET efficiency (E) values to the isothermal binding equation, where K in Equation 2 is analogous to a disassociation constant for the donor and acceptor (13), decreasing in value as their co-clustering increases.
Fitted curves generated from FRET between C 16 -DiO and either C 16 -DiI or C 12 -DiI are shown in Fig. 4A, and K determined for each experiment is plotted in Fig. 4B. Variance in the E values that is inherent in this approach was addressed by using large data sets for the curve fitting, which minimized the 95% confidence intervals in the fitted curve (blue dashed lines). Consistent with our findings in Fig. 2 showing an elevated coclustering of C 16 -anchored probes, K from FRET between C 16 -DiO and C 16 -DiI was 5-fold less than K for C 16 -DiO and C 12 -DiI. To test the specificity of the curve fitting, FRET efficiency values measured for C 16 -DiO and C 12 -DiI were fit to Equation 2 using K determined for C 16 -DiO to C 16 -DiI FRET. This resulted in a fitted curve that deviated from the experimental values and produced large and nonrandom residuals (supplemental Fig.  S3), thus showing that resolved K values were specific to the respective data sets. In summary, the separate FRET donoracceptor pairs produced distinct sets of FRET efficiency values over a range of a acceptor values, and this resulted in unique values for K. Also plotted in Fig. 4B are K values determined for cells treated with filipin, Lat B, or blebbistatin, each resolved from FRET between C 16 -DiO and either C 16 -DiI or C 12 -DiI. Summarizing our results, filipin increased K for the C 16 -DiO and C 16 -DiI FRET pair alone, whereas K trended toward even larger values for both sets of FRET pairs when cells were treated with either Lat B or blebbistatin. Furthermore, the fitted curves for the Lat B-and blebbistatin-treated samples were linear over much of the range of acceptor values, making K approximate to the largest acceptor intensity value ( Fig. 4B and supplemental  Fig. S4). This property is indicative of a random or nonclustered distribution of the donor and acceptor (13), showing efficient demixing of both L d and L o phase probes by Lat B and blebbistatin. Thus, the cytoskeleton affected co-clustering of probes independent of their affinity for ordered lipid environments, indicating that cytoskeletal modulation of lipid interactions was not restricted to the raft L o phase.
Clustering of Lipophilic Probes Requires PIP 2 -To identify membrane signals that activate lipid clustering, we measured FRET in cells where the lipid co-factor phosphatidylinositol 4,5-bisphosphate (PIP 2 ) was sequestered using neomycin (33,34). FRET efficiency and K values determined in this experiment are plotted in Fig. 5. These data show that treatment with neomycin decreased co-clustering of C 16 -DiO and C 16 -DiI, evidenced by a decrease in FRET efficiency (Fig. 5A) and an increase in K ( Fig. 5B and supplemental Fig. S5). Furthermore, this effect was specific because FRET efficiency (Fig. S2) and K (Fig. 5B) in cells double-labeled with C 16 -DiO and C 12 -DiI were not changed significantly by the neomycin. In a separate experiment, we observed that sequestering PIP 2 by overexpressing the pleckstrin homology domain of phospholipase C-␦ caused a measurable and significant decrease in the generalized polarization of Laurdan (supplemental Fig. S6), which is a signature of decondensation of membrane lipids associated with disruption of rafts (35,54).
Because PIP 2 signals that structure the cytoskeleton are cholesterol-dependent (34), we also measured the effect of co-treating cells with neomycin and filipin on probe co-clustering. As we show in Fig. 5, these conditions disrupted co-clustering of C 16 -DiO with both C 16 -DiI and C 12 -DiI, indicated by K values that trended toward the maximum acceptor intensity for each donor-acceptor pair ( Fig. 5B and supplemental Fig. S5). Similarly, FRET efficiency for C 16 -DiO and C 16 -DiI became minimal and independent of the D:A ratio (Fig. 5A). Importantly, neither neomycin alone, nor neomycin plus filipin,  affected phalloidin staining (Fig. 3), again showing that the changes in probe co-clustering were not due to a decrease in the amount of F-actin. Altogether, these data are evidence that cytoskeletal effects on lipid ordering require cholesterol-dependent PIP 2 signals.

SFK Phosphoregulation by Signals That Modulate Lipid
Interactions-SFKs undergo regulation through alternate phosphorylation and dephosphorylation of separate regulatory tyrosines. For Lck, this consists of phosphorylation of its regulatory C-terminal Tyr 505 by Csk to down-regulate activity (36) and dephosphorylation of phospho-Tyr 505 (pTyr 505 ) by CD45 for activation (37).
Previous findings suggest membrane rafts are necessary to maintain negative regulation of Lck. For example, detergent fractionation studies show Csk co-associates with Lck in a detergent-resistant membrane fraction, which is posited to be representative of the composition of rafts (5,38). Conversely, CD45 is excluded from detergent-resistant membranes (4). Similarly, treating T cells with either filipin or Lat B using conditions that disrupt rafts results in dephosphorylation of pTyr 505 of Lck (19). To determine whether either neomycin or blebbistatin also affected Lck regulation, we measured phosphorylation of Tyr 505 (pTyr 505 ) in cells treated with neomycin alone, neomycin plus filipin, or blebbistatin alone. Whole cell lysates prepared from treated and untreated control Jurkat cells were immunoblotted with separate antibodies that recognized either pTyr 505 or the Lck N terminus to report total Lck. Detection of Lck and pTyr 505 was specific because signal associated with Lck was absent in immunoblots of Lck-deficient JCaM1.6 cells (Fig. 6A). Representative data from separate immunoblotting measurements are shown in Fig. 6B, showing that neomycin, neomycin plus filipin, and blebbistatin were each as effective as Lat B in reducing the pTyr 505 content of Lck.
Another site of phosphoregulation in SFKs is a conserved tyrosine present in the activation loop of the kinase domain, which is phosphorylated to up-regulate SFK activity (39). In Lck, this residue is Tyr 394 , and we therefore measured the effect of the separate conditions on the phospho-Tyr 394 (pTyr 394 ) content of Jurkat cells. We detected the pTyr 394 by immunoblotting with an antibody made to the phosphorylated form of the equivalent site in Src, Tyr 416 . Immunoblotting lysate from JCaM1.6 cells showed signal from the anti-pTyr 416 antibody was specific to Lck expression (Fig. 6A), thus representing pTyr 394 . In Fig. 6, C and D, are representative immunoblots with the anti-pTyr 394 antibody, again measuring lysates of cells that were treated with the indicated conditions. These data show that treatment with either neomycin alone, or neomycin plus filipin, increased the pTyr 394 content of Lck by 2-fold and 30%, respectively. Conversely, blebbistatin produced only a nominal increase in pTyr 394 , and Lat B had no effect (Fig. 6D).
pTyr 394 is a substrate for CD45 (40), and CD45 may therefore quench increases in pTyr 394 that result from treatment with either Lat B or blebbistatin. We therefore also measured pTyr 394 levels in Lat B-and blebbistatin-treated J45.01 cells, which are a CD45-deficient clone of Jurkat cells (41). This showed that blebbistatin caused a robust increase in the pTyr 394 (Fig. 6D), whereas Lat B caused a modest decrease in pTyr 394 . In summary, Fig. 6 shows that conditions that disrupt lipid interactions that we detected by FRET results in activation of Lck, indicating that interactions between Lck and its regulators are modulated by lipid domains that are maintained by the cytoskeleton. The results with Lat B-treated cells, however, indicate that other, lipid-independent mechanisms are also important for regulating Lck during homeostasis.

DISCUSSION
We report here findings showing that nanoscopic lipid domains composed of ordered lipids are maintained by the actomyosin cytoskeleton, evidenced by declustering of C 16 -anchored DiO and DiI by either depolymerizing actin filaments with Lat B, or by inhibiting NM II activity with blebbistatin. Probe co-clustering was sensitive to sequestration of cholesterol with filipin, evidence that lipid domains detected by C 16 -DiO to C 16 -DiI FRET fall within the broad generalization of rafts, namely, lipid domains that form through interactions between cholesterol, other membrane lipids, and membrane proteins (42).
The phosphoinositide PIP 2 and its PI3K produce phosphatidylinositol 3,4,5-trisphosphate are critical co-factors in actin polymerization and attachment of actin filaments to cell membranes. Furthermore, we observed that co-clustering of the C 16 -anchored dyes was reduced in cells treated with neomycin to sequester PIP 2 . Similarly, sequestering PIP 2 by overexpressing pleckstrin homology domain of phospholipase C-␦ produced a decondensation of plasma membrane lipids as reported by the generalized polarization of Laurdan. The augmented effect in disrupting probe clustering from co-treating cells with neomycin and filipin is consistent with findings that show cytoskeletal architecture is regulated by both PIP 2 and cholesterol (34) and suggests PIP 2 signals that modulate the cytoskeleton are compartmentalized to rafts. Similarly, we previously showed that selectively elevating or depleting raft pools of PIP 2 by expressing separate forms of the polyphosphate 5Ј-phosphatase Inp54p resulted in robust and distinct changes in cell mor- phology and adhesion (22), which are products of cytoskeletal architecture regulated by PIP 2 .
The plasma membrane is suggested to exist at a critical point that is near the transition between separate fluid phases with different degrees of lipid ordering, such that small changes in composition, temperature, or other physical properties of the membrane favor formation of ordered lipid domains (42,43). Furthermore, protein binding to the PIP 2 headgroup imparts an ordering effect in its lipid hydrocarbon chains (44), which may, in turn, impact lipid interactions proximal to PIP 2 . Accordingly, we posit that interactions between the cytoskeleton and plasma membrane via PIP 2 favor formation of ordered lipid domains of various sizes and compositions, producing the elevated FRET between the lipophilic probes that was reported here (Fig. 7). Proteins that bind both phosphoinositides and the cytoskeleton and are therefore predicted to be important in the cytoskeletal modulation of lipid ordering include ezrin-radixinmoesin, Rho GTPases, annexins, and filamin A (16,45,46). Finally, the lipid ordering by the cytoskeleton to stabilize rafts contrasts with the notion that cytoskeleton facilitates raft formation through caging or direct binding of proteins to the actin network that underlies the plasma membrane (47).
Disruption of probe co-clustering by inhibiting NM II without effecting the F-actin content suggests that mechanical properties of the cytoskeleton such as tension generated by myosin activity is important for lipid ordering effects by the cytoskeleton. It is interesting to note that actin filaments can exert considerable compressive force upon lipid bilayers, evidenced by robust changes in the shape of liposomes and lipid droplets that contain attached F-actin (48,49). Furthermore, compression applied to membrane surfaces can order underlying fluid phase lipids (50). However, further studies are needed to show whether compression applied by a combination of crosslinked actin filaments and NM II activity contribute to lipid ordering and formation rafts in cell membranes.
Previous interpretations have stressed the similarities between L o phase lipids in model membranes and that deduced for lipids in rafts. Namely, rafts, similar to L o phase lipids, form by interactions between fluid lipids and cholesterol to form ordered lipid domains (9,51). However, recent studies of membrane blebs show that the notion that rafts are represented by L o phase lipids alone is likely an oversimplification. For example, blebs with different compositions and degrees of lipid ordering can be produced by relatively modest changes in preparation (52), and some species of blebs produce ordered domains that do not exclude L d markers (11). These properties suggest that the plasma membrane is composed of a continuum of domains with separate degrees of lipid ordering, each established by a combination of their respective protein and lipid composition. Similarly, we showed here a cytoskeleton-dependent co-clustering of the L o /L d FRET pair C 16 -anchored DiO and C 12 -anchored DiI that was not affected by filipin. This suggests co-labeling of cholesterol-independent lipid domains that are also influenced by the cytoskeleton (Fig. 7). Alternatively, the cytoskeleton may produce a global effect that increases lipid ordering regardless of lipid phase.
Conditions that decreased co-clustering of the raft markers also decreased pTyr 505 levels in Lck. Furthermore, we showed for the first time that, in some instances, conditions that disrupted rafts also caused activation of Lck as reported by an elevation of pTyr 394 . We interpret these findings as evidence that cyoskeletal rafts are important for maintaining signal quiescence in Lck. Detergent fractionation studies show evidence that Csk co-associates with Lck in rafts and that rafts sequester Lck from CD45. Both of these properties would underlie inhibition of raft pools of Lck, first by phosphorlyation of Tyr 505 by Csk and then sequestration of the resulting pTyr 505 from CD45. Accordingly, disruption of rafts is predicted to elevate interactions between Lck and CD45 while decreasing its interactions with Csk. This would account for the observed decrease in Lck pTyr 505 content in cells treated with Lat B, neomycin, and blebbistatin. The corresponding increase in pTyr 394 that was observed in most conditions could occur by activation of Lck from release of its autoinhibition by pTyr 505 dephosphorylation. Other proteins likely participate in the down-regulation of raft Lck because blebbistatin increased pTyr 394 in CD45deficient J45.01 cells. One candidate is PEP (PEST domain-enriched tyrosine phosphatase), which associates with Csk to inhibit SFK signaling in T cells (53). Interactions between Lck and PEP via Csk may therefore be disrupted by blebbistatin. Altogether, further study is necessary to identify these candidates and the role of rafts in regulating their interactions with Lck.
In summary, we measured FRET between lipophilic carbocyanine dyes to assess lipid heterogeneities occurring in the plasma membrane of T cells. Our data show a cholesterol-dependent lipid ordering that is sensitive to disruption of F-actin and inhibition of either NM II or PIP 2 . Measuring tyrosine phosphorylation of Lck, we showed that conditions that disrupt lipid domains formed by the cytoskeleton also cause deregulation of Lck. Altogether, these data show cholesterol-dependent rafts that occur as a result of interactions between the cytoskel-FIGURE 7. Membrane lipid ordering by the cytoskeleton. Association of the cytoskeleton with the plasma membrane activates lipid interactions in the underlying membranes to produce ordered lipid domains that include L o phase rafts and cholesterol-independent L d * domains. Arrows indicate sites of PIP 2 -dependent interactions between the cytoskeleton and plasma membrane that are necessary for cytoskeletal effects on lipid ordering. PIP 2 -dependent lipid ordering is predicted to be mediated by cytoskeletal proteins that bind simultaneously to F-actin and PIP 2 , such as ezrin-radixin-moesin proteins, filamin A, and some species of annexins. These are indicated in the schematic as the "PIP 2 -binding protein," the different colors indicating separate species of these factors that may reside in different membrane compartments. NM II-dependent increases in lipid ordering may occur through tension that is imparted in the cytoskeleton and transduced to the membrane surface.