Palmitoylation and Intracellular Domain Interactions Both Contribute to Raft Targeting of Linker for Activation of T Cells*

Some transmembrane proteins must associate with lipid rafts to function. However, even if acylated, transmembrane proteins should not pack well with ordered raft lipids, and raft targeting is puzzling. Acylation is necessary for raft targeting of linker for activation of T cells (LAT). To determine whether an acylated transmembrane domain is sufficient, we examined raft association of palmitoylated and nonpalmitoylated LAT transmembrane peptides in lipid vesicles by a fluorescence quenching assay, by microscopic examination, and by association with detergent-resistant membranes (DRMs). All three assays detected very low raft association of the nonacylated LAT peptide. DRM association was the same as a control random transmembrane peptide. Acylation did not measurably enhance raft association by the first two assays but slightly enhanced DRM association. The palmitoylated LAT peptide and a FLAG-tagged LAT transmembrane domain construct expressed in cells showed similar DRM association when both were reconstituted into mixed vesicles (containing cell-derived proteins and lipids and excess artificial raft-forming lipids) before detergent extraction. We conclude that the acylated LAT transmembrane domain has low inherent raft affinity. Full-length LAT in mixed vesicles associated better with DRMs than the peptide. However, cells appeared to contain two pools of LAT, with very different raft affinities. Since some LAT (but not the transmembrane domain construct) was isolated in a protein complex, and the Myc- and FLAG-tagged forms of LAT could be mutually co-immunoprecipitated, oligomerization or interactions with other proteins may enhance raft affinity of one pool of LAT. We conclude that both acylation and other factors, possibly protein-protein interactions, target LAT to rafts.

Some transmembrane proteins must associate with lipid rafts to function. However, even if acylated, transmembrane proteins should not pack well with ordered raft lipids, and raft targeting is puzzling. Acylation is necessary for raft targeting of linker for activation of T cells (LAT). To determine whether an acylated transmembrane domain is sufficient, we examined raft association of palmitoylated and nonpalmitoylated LAT transmembrane peptides in lipid vesicles by a fluorescence quenching assay, by microscopic examination, and by association with detergent-resistant membranes (DRMs). All three assays detected very low raft association of the nonacylated LAT peptide. DRM association was the same as a control random transmembrane peptide. Acylation did not measurably enhance raft association by the first two assays but slightly enhanced DRM association. The palmitoylated LAT peptide and a FLAG-tagged LAT transmembrane domain construct expressed in cells showed similar DRM association when both were reconstituted into mixed vesicles (containing cell-derived proteins and lipids and excess artificial raft-forming lipids) before detergent extraction. We conclude that the acylated LAT transmembrane domain has low inherent raft affinity. Full-length LAT in mixed vesicles associated better with DRMs than the peptide. However, cells appeared to contain two pools of LAT, with very different raft affinities. Since some LAT (but not the transmembrane domain construct) was isolated in a protein complex, and the Myc-and FLAG-tagged forms of LAT could be mutually co-immunoprecipitated, oligomerization or interactions with other proteins may enhance raft affinity of one pool of LAT. We conclude that both acylation and other factors, possibly proteinprotein interactions, target LAT to rafts.
Recent years have seen an explosion of interest in membrane microdomains called lipid rafts (1)(2)(3). Rafts have been implicated in processes as diverse as signal transduction (1,4), membrane trafficking (5,6), and apoptosis (7). In addition, many pathogenic viruses and bacteria hijack host cell rafts during infection (8 -10). In all of these cases, function depends on selective enrichment of a subset of membrane proteins in rafts. For this reason, it is important to determine how proteins are targeted to rafts.
A key feature of raft structure is the tight packing of lipid acyl chains. Raft lipids are probably in the liquid-ordered (l o ) 1 phase, in which lipid acyl chains are extended and ordered (11,12). Many proteins are targeted to rafts by their favorable association with these ordered lipids. For example, raft proteins such as glycosylphosphatidylinositol-anchored proteins, Src family kinases, and heterotrimeric G protein ␣ subunits are linked to saturated acyl chains, which partition well into rafts. Because of this tight acyl chain packing, raft lipids and proteins are insoluble in nonionic detergents and can be isolated from cell lysates as DRMs. Although DRM association of a protein may not provide a quantitative measure of its association with rafts in cells (13)(14)(15), detergent insolubility is a valuable and widely used tool for identifying proteins that have a high affinity for rafts and are likely to function in rafts. We will use "rafts" to describe l o phase domains in model membranes and similar domains thought to exist in cells and "DRMs" to refer specifically to detergent-resistant membranes isolated from either source.
A major unanswered question in the raft field is how transmembrane proteins are targeted to rafts. Because of the difficulty in packing membrane-spanning ␣ helices into the ordered raft environment, transmembrane proteins are not expected to associate well with rafts. In support of this idea, model membrane studies have shown that transmembrane peptides and proteins are excluded from l o phase domains and also from cholesterol-free gel phase domains, in which lipids are even more tightly packed (16 -21) (reviewed in Ref. 22). As expected, most transmembrane proteins appear to be excluded from DRMs isolated from cells (23,24). Nevertheless, some transmembrane proteins, including important signaling receptors, are present in rafts (1,4). Little is known about how these proteins are targeted to rafts. Although raft-targeting domains or motifs have been identified in some transmembrane proteins (25)(26)(27)(28)(29)(30)(31), no consensus targeting motifs have been defined. Protein-protein interactions can enhance raft affinity of some proteins. For instance, raft targeting of the T cell co-receptor protein CD4 is enhanced by binding to the raft-associated tyrosine kinase Lck (32), and co-ligation to a protein complex including the tetraspanin CD81 prolongs the association of the B cell receptor with rafts during signaling (33). Palmitoylation is also required for DRM-association of some transmembrane proteins (24, 27, 32, 34 -37). Palmitoylation is required for function of some of these proteins, suggesting that raft targeting is important in function (32, 34 -36, 38). However, not all acylated transmembrane proteins are enriched in DRMs. For instance, the transferrin receptor (TfR) is excluded from rafts (39,40) despite being palmitoylated (41). Furthermore, acylation of integral membrane proteins is not always required for DRM association (42,43). It is not known whether acylation is sufficient for raft targeting of any transmembrane protein. It seems unlikely that acylation could overcome the large energetic cost of packing transmembrane helices into rafts. In agreement with this expectation, an artificial acylated transmembraneous peptide was excluded from DRMs after detergent extraction of raft-containing model membranes (20).
We wanted to examine raft targeting of one transmembrane protein in detail. Although packing of most transmembrane proteins into rafts is likely to be similarly difficult, we selected LAT, which plays an essential signaling role in T cells and some other hematopoietic cells (44 -46), as a model. LAT contains a short 3-residue extracellular domain, a single membrane-spanning domain, and an intracellular domain and is found in the plasma membrane. Upon activation of the T cell receptor, LAT is phosphorylated on multiple sites by the nonreceptor tyrosine kinase ZAP-70. Phosphorylated LAT is a docking site for signaling proteins that include Grb-2, phospholipase C␥, and phosphatidylinositol 3-kinase. Recruitment of these proteins to the membrane, through binding to LAT, is required for their activation (46).
Several studies have provided evidence that LAT must associate with rafts in order to function. LAT is palmitoylated on two Cys residues close to the cytoplasmic membrane interface (34). A nonpalmitoylated mutant LAT was found to be correctly targeted to the plasma membrane of a LAT-negative line of Jurkat cells, which are derived from T cells. However, the mutant did not associate with DRMs and thus had a reduced affinity for rafts. Furthermore, ligation of the T cell receptor failed to stimulate tyrosine phosphorylation of this LAT mutant, and the protein was defective in signal transduction (34,38). Nevertheless, function could be restored when LAT, lacking its own palmitoylation sites, was targeted to rafts through linkage to the dually acylated amino-terminal domain of Lck (38).
The goal of our study was to determine whether the acylated transmembrane domain was sufficient for raft targeting of LAT. We took several complementary in vivo and in vitro approaches to this question. First, we used three different assays to compare the raft partitioning of artificial acylated and nonacylated LAT transmembrane domain peptides in model membranes. Next, we compared DRM association of the acylated LAT peptide, full-length LAT, and a LAT transmembrane domain construct, expressing the last two in cells. In one approach, we reconstituted all three species into lipid vesicles of the same raft-forming lipid composition. Comparing the acy-lated peptide and the transmembrane domain construct this way allowed us to determine whether synthesis of the transmembrane domain in cells affected raft affinity. In a complementary approach, we compared DRM association of LAT and the transmembrane domain construct in DRMs isolated from detergent-extracted cells. This allowed us to determine whether cell-specific factors that were not recapitulated in model membranes affected differences in DRM association. These complementary approaches gave a detailed picture of the role of the acylated transmembrane domain in targeting LAT to rafts. Our main conclusion, that the acylated transmembrane domain is not sufficient for full raft targeting, is likely to apply to a wide variety of transmembrane raft proteins.
Peptides-Acetyl-K 2 W 2 L 8 AL 8 W 2 K 2 -amide (LW peptide), LAT peptide (EADWLSPVGLGLLLLPFLVTLLAALCVRCRE; corresponding to residues 2-32 of murine LAT (44), including the transmembrane domain and both acylation sites, with substitution of Trp for Ala 4 ), and a control LAT peptide (differing from LAT peptide in containing Ala at position 4 and Trp replacing Leu at position 18 and used to provide evidence that it and, by extension, probably LAT peptide as well were transmembraneous) were purchased from Research Genetics (Carlsbad, CA). In DOPC vesicles, the control LAT peptide showed a highly blue-shifted Trp fluorescence, with max ϭ 322-323 nm, showing that the Trp was near the center of the bilayer and thus that the peptide was transmembraneous (47). Peptides were purified via reverse phase high pressure liquid chromatography using a C18 column with a 2-propanol/ water gradient containing 0.5% (v/v) trifluoroacetic acid (48). Peptide purity was confirmed by matrix-assisted laser desorption ionization time-of-flight (MALDI-TOF) mass spectrometry. LW peptide dissolved in ethanol and LAT peptide dissolved in 0.1% trifluoroacetic acid, 75% ethanol (all liquid mixtures v/v), were stored at Ϫ20°C until use. Peptide concentrations were determined by spectrophotometry using a molar absorptivity value at 280 nm of 5560 M Ϫ1 cm Ϫ1 per Trp residue.
Antibodies-Mouse anti-TfR antibodies were from Zymed Laboratories (San Francisco, CA); rabbit anti-LAT antibodies were from Upstate Biotechnology, Inc. (Lake Placid, NY); rabbit anti-FLAG antibodies were from Affinity BioReagents (Golden, CO), mouse anti-FLAG antibodies were from Sigma, mouse anti-Myc antibodies were from Invitrogen, and rabbit anti-placental alkaline phosphatase antibodies from Dako (Carpinteria, CA). Rabbit anti-mouse antibodies, horseradish peroxidase-conjugated goat anti-mouse IgG ϩ IgM (H ϩ L chains) and donkey anti-rabbit IgG (H ϩ L chains), and fluorescein isothiocyanateconjugated goat anti-rabbit IgG (H ϩ L chains) were from Jackson ImmunoResearch Laboratories (West Grove, PA).
Plasmids-Plasmids encoding C-terminally FLAG-or Myc-tagged human LAT in pCEFL or pcDNA3 (34,44) were generous gifts of Dr. L. E. Samelson (National Institutes of Health, Bethesda, MD). Plasmids encoding LAT(TMD)WT (residues 1-36 of murine LAT, followed by a 6-residue linker sequence and a FLAG tag) and LAT(TMD)CA (a double Cys to Ala mutant of LAT(TMD)WT) were made by inserting stop codons immediately after the FLAG tags in plasmids encoding LAT/ SHP-1 and LAT(CA)/SHP-1 (generous gifts of Dr A. Kosugi, Osaka University, Japan; described in Ref. 49) using the QuikChange kit (Stratagene, La Jolla, CA). Schematic diagrams of the LAT peptide and mature LAT(TMD)WT and LAT(TMD)CA are shown in Fig. 1. A plasmid encoding human placental alkaline phosphatase has been described (23).
Cell Lines and Transfection-COS-1 and Jurkat cells (American Type Culture Collection, Manassas, VA) were maintained in Dulbecco's modified Eagle's medium with 5% iron-supplemented calf serum or in RPMI 1640 with 10% fetal calf serum, respectively. COS cells were transfected with 1-2 g of DNA per dish using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions and harvested 1-2 days after transfection.
Other Reagents-MitoTracker Deep Red 633 was from Molecular Probes. PVDF membrane was from Millipore Corp. (Bedford, MA). X-omat AR film was from Eastman Kodak Co. (Rochester, NY), and chemiluminescence reagent was from PerkinElmer Life Sciences. Serva Blue G was from SERVA Electrophoresis GmbH. Latrunculin A was from BioMol (Plymouth Meeting, PA). Other reagents were from Sigma.

Methods
Peptide Palmitoylation-1 mg of LAT peptide was incubated with 1.2 mg of palmitoyl-CoA in 0.1% NH 4 OH, 1 mM dithiothreitol, 5% 2-propanol at 37°C overnight (final volume 333 l). The reaction mixture was dried under N 2 gas, lyophilized, suspended in 300 l of water, and subjected to centrifugation for 10 min at top speed in a microcentrifuge. Water-soluble materials in the supernatant were removed, and insoluble pellets containing peptide were dissolved in 100 l of 0.1% trifluoroacetic acid, 50% 2-propanol. Palmitoylated LAT peptide was purified either by high performance thin layer chromatography using n-butyl alcohol/acetic acid/water (5:2:3, v/v/v) as the development solvent or by high pressure liquid chromatography as described above. When using high performance thin layer chromatography, the acylated peptide band was identified on a parallel analytical plate by ninhydrin reactivity and detected on the preparative plate using a UV lamp. The palmitoylated LAT peptide was scraped from the plate and extracted from the silica gel with 0.5 ml of 0.1% trifluoroacetic acid, 80% 2-propanol. Palmitoylation and peptide purity were confirmed by MALDI-TOF mass spectrometry.
Rhodamine Labeling of Peptides-Peptide (400 g) was mixed with 1 mg of Lissamine rhodamine B sulfonyl chloride (Molecular Probes) in a final volume of 1 ml of 50% ethanol, 50% 50 mM sodium bicarbonate buffer (pH 9.0) and incubated overnight at room temperature. Rhodamine-labeled peptide (rhodamine/peptide ratio about 1:3) was separated from free rhodamine by gel filtration column chromatography using Sephadex LH-20 (Amersham Biosciences) equilibrated with 0.1% trifluoroacetic acid, 75% ethanol. Fractions containing rhodamine-labeled peptide were pooled, and the labeled peptide was further purified by high performance thin layer chromatography as described above. When preparing dually modified LAT peptide, rhodamine labeling was performed before palmitoylation.
Fluorescence Quenching Assay-Synthetic peptides were incorporated into ethanol dilution vesicles prepared as described (21) except that peptides and lipids were initially dissolved in 0.02% phosphoric acid, 75% ethanol, and the composition of the phosphate-buffered saline (PBS) used in this study was 150 mM NaCl, 20 mM phosphate buffer, pH 7.4. Quenching of peptide Trp fluorescence was measured as previously described (21). For thermal scanning experiments, the fluorescence in samples was measured at a series of increasing temperatures. For experiments in which lipid concentration was varied, the fluorescence in each sample was measured at 23°C.
Giant Unilamellar Vesicle (GUV) Preparation and Analysis-GUVs were prepared as described previously (50,51). In brief, lipid mixtures (compositions given throughout) together with 0.5 mol % rhodaminelabeled peptide (except as noted) and trace amounts of fluorescent lipid analogs (0.1 mol % BODIPY-PC and 0.2 mol % perylene) were dissolved in chloroform/methanol (2:1). The mixtures were dried into a film on the walls of glass tubes at high vacuum. The films were slowly hydrated with water-saturated N 2 gas to form bilayers. Buffer (10 mM KCl, 1 mM EDTA, 2 mM HEPES, pH 7.0) was added to the tube, and the membranes were allowed to slough off and form vesicles overnight. Mixtures were maintained above the main phase transition temperature of the lipids to avoid demixing before the vesicles formed. The vesicles were allowed to slowly cool to room temperature over 6 h. Confocal images were obtained with a laser-scanning confocal microscope (Leica TCS SP2) at 488-nm excitation, with a 522/535-nm band pass emission filter for BODIPY-PC and a 585-nm long pass emission filter for rhodaminelabeled peptides. For the experiment in Fig. 8A, COS cells in one confluent 35-mm dish (estimated to contain 180 nmol of total membrane lipid, assuming similar yield to cells examined previously (23)) were lysed by incubation in 400 l of TNE/OG and a mixture of protease inhibitors (PIM; 0.5 g/ml leupeptin, 0.7 g/ml pepstatin, and 0.2 mM phenylmethylsulfonyl fluoride) at 4°C for 1 h. Insoluble material was removed by centrifugation for 5 min at top speed in a microcentrifuge. 2 mol of purified lipids (composition given in figure legends; all lipid ratios mol/mol) dissolved in 500 l of TNE/OG were mixed with clarified cell lysates, incubated at 4°C for 1 h, and then dialyzed against PBS at 4°C for 20 h to form mixed vesicles. These were mixed with an equal volume of TNE plus 80% sucrose, placed in an ultracentrifuge tube, overlaid with 1.7 ml of TNE plus 35% sucrose and 1 ml of TNE plus 5% sucrose, and subjected to ultracentrifugation at 200,000 ϫ g for 3 h. Vesicles were harvested from the 5%/35% sucrose interface, diluted with TNE, and collected as a pellet by ultracentrifugation at 200,000 ϫ g for 30 min.

Reconstitution of Peptides and Cell Lysates in Vesicles for Detergent
For the experiment described in Fig. 9B, 5 ϫ 10 7 Jurkat cells were extracted with 0.5 ml of TNE containing 20 mM 3-(3-cholamidopropyl)dimethylammonio-1-propanesulfonate (CHAPS) and PIM for 30 min at 4°C. Lysates were mixed with an equal volume of TNE plus 80% sucrose in an ultracentrifuge tube and overlaid with 2.5 ml of TNE plus 38% sucrose and then 1 ml of TNE plus 5% sucrose. Gradients were subjected to centrifugation at 200,000 ϫ g for 3 h. The top 1.5 ml (containing DRMs) and the bottom 1.5 ml (containing solubilized material) were harvested separately. Each was adjusted to 60 mM OG by the addition of 0.5 ml of TNE plus 240 mM OG, incubated 1 h at 4°C, mixed with 2 mol of lipid dissolved in 0.5 ml of TNE plus 60 mM OG, and dialyzed against PBS for 20 h at 4°C. The resulting vesicles were collected as a pellet by ultracentrifugation at 200,000 ϫ g for 30 min and resuspended in 450 l of TNE. The same method was used for the experiment shown in Fig. 9C, except that COS cells in one confluent 35-mm dish (with or without transfection of LAT) were substituted for Jurkat cells, and appropriate DRM and CHAPS-soluble fractions from the first sucrose gradient (described in the figure legend) were mixed together before the addition of OG (60 mM final concentration) and 2 mol of OG-solubilized lipid and dialysis.
Detergent Insolubility Experiments-For determining DRM association of proteins or peptides in vesicles, vesicles prepared as described above were suspended in 500 l of TNE plus 1% Triton X-100 and incubated at 4°C for 30 min. Extracts were mixed with an equal volume of TNE plus 80% sucrose (all sucrose solutions contained PIM), placed in an ultracentrifuge tube, and overlaid with 2.5 ml of TNE plus 38% sucrose and then 1 ml of TNE plus 5% sucrose. Gradients were subjected to centrifugation for 3 h at 200,000 ϫ g at 4°C using an SW60Ti rotor. 12 ϫ 360-l fractions were harvested from the top, and the pellet was collected. For samples containing peptides, each fraction was brought to 800 l with TNE, and rhodamine fluorescence was measured (excitation, 565 nm; emission, 585 nm). For mixed vesicles containing proteins from cell lysates, aliquots of each gradient fraction were subjected to SDS-PAGE as described (24). Proteins were transferred to PVDF membranes and subjected to Western blotting, detecting proteins by chemiluminescence after incubation of membranes with appropriate primary antibodies and horseradish peroxidase-conjugated secondary antibodies, as described (24). Bands were quantified using NIH Image. For determining DRM association of proteins in transfected COS cell lysates (Fig. 6), cells in one confluent 35-mm dish were extracted with 0.5 ml of TNE plus 1% Triton X-100 and PIM for 30 min at 4°C. Extracts were then subjected to sucrose gradient centrifugation, SDS-PAGE, and Western blotting as described above for vesicle-derived extracts. For the latrunculin experiment (Fig. 10D), aliquots of 5 ϫ 10 7 cells were adjusted to 2 ϫ 10 7 cells/ml and incubated for 30 min with or without 5 M latrunculin A at 37°C, collected by centrifugation, and lysed in 1 ml of TNE plus 1% Triton X-100 and PIM for 20 min on ice. Nuclei and cell debris were removed by centrifugation for 10 min at 900 ϫ g at 4°C. Supernatants were mixed with an equal volume of TNE plus 80% sucrose and overlaid in SW41 tubes with 6 ml of TNE plus 38% sucrose and 3 ml of TNE plus 5% sucrose. After overnight centrifugation at 28,000 rpm in an SW41 rotor, 1-ml aliquots were harvested from the top. Aliquots of each fraction were analyzed by SDS-PAGE and Western blotting. All blots were probed for endogenous TfR as well as LAT, and data from any gradients in which TfR was not essentially confined to the lowest load fractions were discarded.
Immunofluorescence Microscopy-Immunofluorescence microscopy was as reported (52). Briefly, paraformaldehyde-fixed, permeabilized, coverslip-grown COS cells, expressing the proteins listed in the legend to Fig. 7, were incubated with rabbit anti-FLAG and then fluorescein isothiocyanate-conjugated goat anti-rabbit antibodies. In some cases, before fixation and permeabilization, cells were removed from polylysine-coated coverslips by sonication for 4 -6 s with a probe sonicator placed about 5 mm above the coverslip, leaving only coverslip-adherent plasma membrane sheets. Preliminary experiments using anti-Myc antibodies verified that Myc-and FLAG-tagged LAT had the same localization pattern when expressed in COS cells and that the localization of FLAG-tagged LAT appeared the same when detected with either anti-FLAG or anti-LAT antibodies. Cells expressing LAT(TMD)CA were fixed and permeabilized and then treated with 600 nM Mito-Tracker Deep Red 633 for 30 min at room temperature to label mitochondria, before incubation with antibodies.
Blue Native (BN)-PAGE-BN-PAGE was performed according to Ref. 53 with changes as noted below. Cells were lysed in lysis buffer (500 mM 6-amino caproic acid, 2 mM EDTA, 25 mM Bistris, pH 7.0) containing 60 mM OG or 1% digitonin for 30 min at 4°C. Lysates were clarified by centrifugation for 5 min at top speed in a microcentrifuge and then by ultracentrifugation at 100,000 ϫ g for 20 min in a Beckman TLA100 ultracentrifuge. Supernatants were mixed with one-tenth volume sample buffer (5% Serva Blue G, 500 mM 6-aminocaproic acid, 100 mM Bistris, pH 7.0) and one-tenth volume glycerol. Proteins were separated on linear 5.5-16% acrylamide gradient gels run at 100 V at 4°C for 3-4 h. Blue cathode buffer (50 mM Tricine, 15 mM Bistris, 0.02% Serva Blue G, pH 7.0) was replaced by colorless cathode buffer (lacking Serva Blue G) after 1 h. Proteins were transferred to PVDF membranes and detected by Western blotting. Sizes were roughly estimated by comparison with standards, visualized by staining gels with Coomassie Blue: ␤-amylase (200 kDa), alcohol dehydrogenase (150 kDa), bovine serum albumin (66 kDa), and carbonic anhydrase (29 kDa).
Co-immunoprecipitation-COS cells in one confluent 35-mm dish expressing FLAG-and/or Myc-tagged LAT were lysed in 0.5 ml of TNE plus 1% digitonin for 30 min at 4°C. After centrifugation at top speed in a microcentrifuge for 5 min, supernatants were incubated with either rabbit anti-FLAG or mouse anti-Myc antibodies at 4°C for 18 h. Primary antibodies were precipitated by incubation with Pansorbin (Calbiochem) alone or bound to rabbit anti-mouse antibodies for 1 h at 4°C. Pansorbin-bound complexes were washed with TNE and boiled in SDS sample buffer. Eluted proteins were resolved on SDS-PAGE, transferred to PVDF membranes, and detected by Western blotting.

Detecting Raft Association of LAT Peptide by Fluorescence
Quenching-We first measured partitioning of palmitoylated and nonpalmitoylated LAT peptide (schematized in Fig. 1) between co-existing ordered and disordered domains in model membranes. We used a fluorescence quenching assay, detecting intrinsic Trp fluorescence of the peptide in vesicles that contained 12SLPC, a spin-labeled lipid that quenches Trp fluorescence, to assay partitioning behavior (16,21,54). We compared peptide fluorescence in matched pairs of vesicles, each containing the same amount of 12SLPC but containing different unlabeled lipids. The unlabeled lipids were chosen so that one set of vesicles contained a uniform disordered phase, whereas the other might contain co-existing ordered and disordered domains (21). In two-phase vesicles, 12SLPC is highly concentrated in disordered domains and correspondingly depleted from ordered domains (21). Thus, if a peptide is also concentrated in disordered domains, it will encounter a higher concentration of 12SLPC than it would in the control one-phase vesicles, and its fluorescence will be more highly quenched (21). The degree to which quenching in two-phase vesicles is enhanced over that in uniform vesicles can be used to determine the partition constant (K p ) of the peptide between the two phases (16,21).
We first compared fluorescence quenching of the peptides in control and experimental vesicles, maintaining the lipid composition of each constant and varying the temperature. In this experiment, the control vesicles contained a 3:1 mixture of DOPC and SLPC, which mix uniformly to form the l c phase under all conditions examined here (54). The experimental vesicles contained a 3:1 mixture of DPPC/SLPC. At 23°C, the lowest temperature examined, the DPPC/SLPC vesicles contain co-existing gel and l c phase domains (54). As the temperature is increased, gel phase domains melt and mix with the l c phase domains. At high temperatures, the vesicles exist in a uniform l c phase.
At low temperatures, nonpalmitoylated LAT peptide was quenched more strongly in two-phase vesicles than in control one-phase vesicles ( Fig. 2A, closed squares, experimental DPPC-containing vesicles; open squares, control single-phase DOPC-containing vesicles), showing that the peptide was concentrated in the disordered domains in the two-phase vesicles. By contrast, quenching was similar in both vesicles at or above 40°C, when both vesicles existed in a single disordered phase. This behavior supported the conclusion that the increased quenching in the DPPC/SLPC vesicles at lower temperatures resulted from the presence of two phases in those vesicles and preferential partitioning of the peptide into the quencher-rich domains. We repeated the experiment on vesicles containing 25% cholesterol, along with either DOPC/12SLPC or DPPC/ 12SLPC in a 3:1 ratio. The DOPC/12SLPC/cholesterol vesicles are present in a uniform liquid-disordered (l d ) phase under all conditions examined, whereas DPPC/12SLPC/cholesterol vesi- cles form co-existing l o phase and l d phase domains at the lowest temperatures examined (21). (We will refer to the cholesterol-containing disordered phase as l d , to distinguish it from the cholesterol-free disordered phase, referred to as l c . Ordered and disordered domains can co-exist in both the cholesterol-containing and cholesterol-free vesicles that we examined. However, without cholesterol, the ordered domains are in the gel state, whereas in the presence of cholesterol, the ordered domains are in the l o state). Results (shown in Fig. 2B) were similar to those in the cholesterol-free vesicles, except that enhanced quenching in DPPC-containing vesicles persisted to higher temperatures, consistent with the ability of cholesterol to promote and stabilize phase separation (54 -56). Quenching of the palmitoylated LAT peptide was similar to that of the nonpalmitoylated peptide, within experimental error, in both cholesterol-containing and cholesterol-free vesicles (Fig. 2, C and D). We conclude that both acylated and nonacylated LAT peptide are more concentrated in disordered than ordered domains when the two co-exist in the same bilayer.
We next examined quenching of LAT peptide fluorescence in a series of vesicles of varying lipid composition at room temperature (Fig. 3). In this case, control vesicles, present in a single disordered phase under all conditions examined, contained graded mixtures of DOPC and 12SLPC. The experimental vesicles contained graded mixtures of DPPC and 12SLPC. We also examined vesicles that contained 25 mol % cholesterol in addition to the various mixtures of 12SLPC with either DPPC or DOPC. At lower concentrations of 12SLPC, DPPC-containing vesicles contain co-existing ordered and disordered domains (54). At high 12SLPC concentrations, the vesicles are present in a uniform disordered phase. When two phases co-exist, 12SLPC is more highly concentrated in disordered domains than in ordered domains. A peptide that is also more highly concentrated in these domains will be more highly quenched than in control single phase vesicles. Both palmitoylated and nonpalmitoylated LAT peptide, with and without cholesterol, showed this behavior (Fig.  3) (i.e. F/F 0 was always lower in the two-phase vesicles than in corresponding single phase vesicles). At high concentrations of 12SLPC, DPPC-containing vesicles were present in a single disordered phase. In these cases, fluorescence quenching of the peptides was similar to that in uniform DOPC-containing vesicles. These results confirmed that both acylated and nonacylated LAT peptides were more concentrated in the disordered domains and showed that this behavior was largely independent of the amount of lipid present in ordered domains. Comparison with theoretical curves of quenching versus 12SLPC concentration suggested a 5-10-fold higher partitioning into the disordered phase over the ordered phase, similar to that previously observed for LW peptide, a simple nonacylated model transmembrane peptide (21).
Partitioning of LAT Peptide between l o and l d Domains in GUVs-As an another way of examining raft targeting in model membranes, we incorporated rhodamine-labeled palmitoylated and nonpalmitoylated LAT peptides into GUVs that contained co-existing l o and l d domains, visualized using fluorescent lipid analogs that partitioned preferentially into one domain or the other. We first examined two-phase vesicles containing SM/DOPC/DOPG/cholesterol (100:95:5:100) and trace amounts of BODIPY-PC (green in Fig. 4) and perylene (blue) as markers of the disordered and ordered phases, respectively (51,57). Vesicles also contained 0.5 mol % rhodaminelabeled palmitoylated or nonpalmitoylated LAT peptides (red). Equatorial confocal sections of GUVs containing unpalmitoylated (Fig. 4A) or palmitoylated (Fig. 4B) LAT peptides are shown. A tangential confocal slice near the top of a GUV containing palmitoylated LAT peptides is shown in Fig. 4C. In each of panels A-C, BODIPY-PC fluorescence is shown in the top image, peptide fluorescence is shown in the middle image, and a merged image is shown at the bottom. In each case, LAT peptide co-localized with the l d phase marker BODIPY-PC and was excluded from the co-existing l o phase domains. The l o phase marker perylene is not shown in Fig. 4, A-C. A lower resolution three-color image of a GUV containing palmitoylated LAT peptide, taken with a wide field microscope, is shown in Fig. 4D. Perylene (Fig. 4D, bottom left image) is clearly concentrated in the lower region of the GUV that excludes both BODIPY-PC (Fig. 4D, upper left image) and peptide (Fig. 4D,  upper right image). In many cases, as is apparent in Fig. 4D, the line tension between the coexistent phases deformed the GUVs (57) . We conclude that both palmitoylated and nonpalmitoylated LAT peptide partitioned strongly into the l d phase domains in two-phase GUVs and that the peptides were distributed fairly uniformly in these domains. The exact lipid composition of the ordered and disordered phases had little effect on peptide partitioning. We observed no peptide-induced perturbation of the phase behavior of the vesicles when compared with peptide-free control vesicles. Such perturbation would be expected if the peptide associated preferentially with domain boundaries (data not shown).
Detergent Insolubility of Peptides in Artificial Vesicles-We next compared the raft affinity of different peptides by measuring their association with DRMs present after Triton X-100 extraction of raft-containing vesicles that contained the peptides. Peptides were incorporated into vesicles of a variety of lipid compositions by OG co-dialysis. Peptides associated efficiently with all vesicles examined (data not shown). Vesicles were incubated with TNE plus 1% Triton X-100 at 4°C for 30 min, and extracts were subjected to sucrose density gradient centrifugation. Gradients were fractionated, and rhodamine fluorescence in each fraction was measured. (Rhodamine-labeled peptides were used for these experiments, since Triton X-100 fluorescence makes Trp fluorescence difficult to detect.) The percentage of total peptide fluorescence present in DRMs (shown in Fig. 5) was calculated by dividing the fluorescence in the DRM-containing gradient fraction 3 by the total rhodamine fluorescence in the gradient. (The distribution of palmitoylated LAT peptide across one representative gradient is shown in Fig. 8B).
We first measured peptide DRM association in vesicles containing 33 mol % cholesterol, along with various mixtures of DPPC and DOPC, in which DPPC ranged from 25 to 75 mol % of total phospholipid. Previous fluorescence-quenching results suggest that l d and l o phase domains co-exist in these vesicles (54). We examined DRM association of acylated and nonacylated LAT peptides and also of the control transmembrane peptide, LW peptide (21), incorporated into these vesicles. LW peptide has a simple sequence not based on that of any natural protein. As expected, LW peptide and nonpalmitoylated LAT peptide showed low DRM association in all vesicles examined, and the behavior of the two peptides was indistinguishable (Fig. 5A, diamonds and open squares). DRM association of both these peptides increased as the amount of DPPC, and thus the amount of the bilayer present in the l o phase, was increased, consistent with a low but detectable partitioning of the peptide into the l o phase domains. Since LW peptide and nonacylated LAT peptide showed the same behavior, the amino acid sequence of the LAT peptide did not measurably enhance raft affinity in the absence of peptide acylation. The palmitoylated LAT peptide (Fig. 5A, closed squares) showed slightly but significantly higher DRM association than the other two peptides except at the lowest concentration of DPPC, when very little l o phase membrane was present. DRM association of the palmitoylated LAT peptide, like that of the other peptides, increased with increasing amounts of DPPC.
We next determined whether altering lipid composition of raftcontaining vesicles could enhance the difference between DRM association of acylated and nonacylated LAT peptides. We examined the effect of increasing the amount of cholesterol to 50 mol %, substituting a mixture of sphingolipids for DPPC, and substituting either POPC or a mixture of POPC, 1-palmitoyl-2-oleoyl phosphatidylethanolamine, and 1-stearoyl-2-oleoyl phosphatidylserine for DOPC. None of these changes greatly affected the degree to which acylation of the LAT peptide enhanced DRM association (Fig. 5B). We conclude that the effect of palmitoylation on detergent insolubility of the LAT peptides depended mostly on the presence of l o phase domains rather than on the specific lipid composition of the co-existing domains. Raft Association of Full-length LAT Versus the LAT Transmembrane Domain-Since the acylated LAT peptide associated only modestly with DRMs, it seemed possible that features of LAT other than the acylated transmembrane domain contributed to its raft affinity. We considered two possibilities: first, that sequences present in LAT but not in the acylated peptide might be required, and second, that expression in cells might contribute to raft affinity. This might occur through interaction with cellular proteins or lipids. To test these possibilities, in the next series of experiments, we examined fulllength LAT, the synthetic acylated LAT peptide, and a LAT transmembrane-domain construct expressed in cells. This construct, LAT(TMD)WT, contained the first 35 amino acids of mature murine LAT, including the transmembrane domain and the two palmitoylation sites and a C-terminal FLAG tag (Fig. 1). We also examined a control construct, LAT(TMD)CA, that could not be palmitoylated, since both Cys were replaced by Ala.
DRM Association of LAT and LAT(TMD)WT in COS Cells-We first examined association of LAT and the LAT(TMD) constructs with DRMs isolated from Triton X-100extracted transfected COS cells. We separated DRMs and Triton X-100-soluble materials in cell lysates by sucrose gradient centrifugation. We then fractionated the gradients and performed SDS-PAGE and Western blotting to detect proteins of interest and quantitated the results by densitometry. A small amount of LAT(TMD)WT (8.2%) and somewhat more LAT (15.0%) were present in DRMs (Fig. 6). The difference between these two values was statistically significant (p Ͻ 0.05). The fraction of endogenous Jurkat cell LAT associated with DRMs was similar to that of transfected LAT expressed in COS cells (see Fig. 10D). By contrast, LAT(TMD)CA (not shown) and endogenous COS cell TfR, a control nonraft protein (39,40), were not detected in DRMs. A glycosylphosphatidylinositolanchored protein, placental alkaline phosphatase, associated substantially more efficiently with DRMs than LAT.
LAT and the LAT(TMD) Constructs Have Different Patterns of Cellular Localization-Even proteins with a high affinity for rafts are solubilized by Triton X-100 if they are present in raft-poor membranes at the time of Triton X-100 extraction (23,58). Thus, the difference between DRM association of LAT and LAT(TMD)WT might reflect their residence in different cellular membranes. For this reason, we examined the cellular distribution of the two proteins in COS cells by immunofluorescence microscopy. LAT had a relatively diffuse plasma membrane distribution (Fig. 7A), as verified by the presence of the protein in plasma membrane sheets derived from transfected cells by sonication (Fig. 7B). By contrast, LAT(TMD)WT was present in puncta (Fig. 7C) that did not co-localize with caveo-lin-1 (not shown). At least some LAT(TMD)WT was present on the plasma membrane, since it was detected in plasma membrane sheets (Fig. 7D). LAT(TMD)WT often had a more punctate distribution in the sheets than did full-length LAT, mirroring the differences seen in whole cells. LAT(TMD)CA (Fig.  7E) was concentrated in mitochondria, identified by Mito-Tracker Deep Red 633 (Fig. 7F), although faint staining of the endoplasmic reticulum was also detected. We presume that this protein was aberrantly recognized by the mitochondrial import machinery (59). This targeting may have followed deinsertion of the protein from the ER membrane. Such deinsertion could have been facilitated by the short length of the extracellular domain (which consisted of the sequence EAD) and the absence of positively charged residues from this domain. By contrast, LAT(TMD)WT, which was never seen in mitochondria, might have been stably anchored in the ER membrane by acylation, allowing it to be delivered to the plasma membrane via the secretory pathway. In any event, the constructs clearly had different cellular distribution patterns, which might have contributed to differential association with DRMs after Triton X-100 extraction of cells.

DRM Association of LAT and LAT(TMD)WT after Reconstitution into Raft-containing Mixed
Vesicles-To compare the DRM association of LAT and LAT(TMD)WT in membranes of the same lipid composition, circumventing any problem resulting from differential cellular localization, we reconstituted both proteins into mixed vesicles, which contained both cell-derived proteins and lipids and an excess of artificial lipids of a raftforming lipid composition. To do this, COS cells expressing the proteins were lysed with OG, to solubilize rafts (60). Clarified lysates were mixed with excess OG-solubilized artificial lipids of a raft-forming composition (DPPC/DOPC/cholesterol, 1:1:1), and vesicles were formed by dialysis. Although the mixed ves- icles were expected to contain a variety of cellular proteins and lipids, the artificial lipids, present in at least a 10-fold excess over cellular lipids, should largely determine their phase behavior. LAT and LAT(TMD)WT associated efficiently with these vesicles and floated with them to a low density position on sucrose density gradients (data not shown). LAT(TMD)CA could not be detected in the vesicles (not shown) and was not examined further. DRM association of proteins in the vesicles was determined by Triton X-100 extraction and analysis by sucrose gradient centrifugation, SDS-PAGE, and Western blotting. Before performing the experiment, we verified that LAT in lysates of cells extracted with OG did not float in sucrose gradients after OG extraction of cells but remained in the bottom load fractions, suggesting that it was fully solubilized (data not shown).
31.0% of LAT and 16.3% of LAT(TMD) in these vesicles associated with DRMs (Fig. 8A). The difference between the two values was statistically significant (p Ͻ 0.05). By contrast, TfR was largely Triton X-100-soluble. DRM association of LAT(TMD)WT was very similar to that of the palmitoylated LAT peptide when both were reconstituted into vesicles of the same lipid composition (Fig. 8B). This similarity suggested that LAT(TMD)WT was likely to be palmitoylated in vivo. Assuming that it was, the similarity also suggested that features unique to expression of LAT(TMD)WT in cells, but not present in the artificial peptide, were not important for DRM association. Reconstitution of LAT(TMD)WT or of palmitoylated LAT peptide, together with full-length LAT in the same mixed vesicles did not affect detergent insolubility of either the peptide or of LAT(TMD)WT (data not shown). Thus, although palmitoylation increased the DRM association of the transmembrane domain of LAT, a palmitoylated transmembrane span was not sufficient for full DRM association of LAT.
Some LAT Has Very High and Stable DRM Association-We imagined two alternate ways in which the fraction of LAT in DRMs might be determined. First, all LAT molecules might be equivalent, and interact in the same way with lipids. In this case, the degree of DRM association would simply result from the inherent affinity of the protein for rafts or its partition constant (K p ). Alternatively, two pools of LAT, with different affinities for rafts, might be present in cells. We took the following approach (diagrammed schematically in Fig. 9A) to distinguish between these possibilities. We separated DRM and detergent-soluble fractions from Jurkat cell lysates on sucrose density gradients and then reconstituted them separately into vesicles. For this experiment, we lysed cells with CHAPS, which is useful for preparing DRMs (61) but can be removed by dialysis more easily than Triton X-100. LAT associated about equally well with Triton X-100-DRMs and CHAPS-DRMs (not shown). After separation on sucrose flotation gradients, CHAPS-resistant membranes (R) and CHAPSsolubilized material (S) were each incubated with sufficient OG to solubilize DRMs. In support of the idea that this treatment fully solubilized all membranes, we found that no LAT floated to a low density position when lysates were subjected to sucrose density gradient centrifugation (not shown). Clarified lysates were mixed with excess OG-solubilized lipids (DPPC/DOPC/ cholesterol, 1:1:1). Mixed vesicles were formed by dialysis and then extracted with Triton X-100. DRMs and Triton X-100- soluble materials were separated on sucrose density gradients. Gradients were fractionated, and the distributions of LAT and TfR between floating DRMs and Triton X-100-soluble material were determined by SDS-PAGE and Western blotting. For comparison, whole Jurkat cell OG lysates, not separated into DRM and detergent-soluble fractions, were reconstituted into vesicles of the same lipid composition and subjected to the same procedure. As shown in Fig. 9B, LAT from cell-derived DRMs (R) associated much better with vesicle-derived DRMs than did LAT that was solubilized during extraction of cells (S). DRM association of this reconstituted S pool of LAT appeared similar to that of the acylated LAT peptide in reconstituted in vesicles of the same lipid composition. LAT from whole cell lysates (W) appeared to show intermediate behavior. In contrast, TfR was always efficiently solubilized after reconstitution (Fig. 9B).
(The small amount of TfR in the R fraction from cells was only detected in gradient fractions after Triton X-100 extraction of mixed vesicles by long exposure of the blot.) Although artificial lipids were present in excess over cellular lipids, it was possible that the presence of different cellular lipids in the R and S fractions led to the differential DRM association of LAT after reconstitution. To control for this possibility, we repeated the experiment using transfected COS cells, under conditions where vesicles had identical lipid compositions. Parallel dishes of COS cells were either transfected with LAT or left untransfected. Cells in both dishes were extracted with CHAPS, and DRM and detergent-soluble fractions were separated on sucrose gradients as for Jurkat cells. DRMcontaining fractions from transfected cells were named R, and DRM-containing fractions from untransfected cells were named R. Similarly, detergent-soluble fractions from transfected and untransfected cells were named S and S, respectively. R and S fractions were mixed with each other, as were R and S fractions. Both pools were then mixed with OG to solubilize DRMs and mixed with OG-solubilized artificial lipids (DPPC/DOPC/cholesterol, 1:1:1). Mixed vesicles were formed by dialysis, extracted with Triton X-100, and subjected to sucrose density gradient centrifugation to separate DRMs and solubilized material. Gradient fractions were analyzed for the presence of LAT by SDS-PAGE and Western blotting. (The procedure was identical to that shown in Fig. 9A except that DRM and detergent-soluble fractions, one from transfected cells and the other from untransfected cells, were mixed together with artificial lipids to form vesicles). As shown in Fig.  9C, LAT derived from cellular DRMs (R) associated better with vesicle-derived DRMs than did LAT that was solubilized from cells by CHAPS (S), ruling out artifacts due to nonidentical lipid compositions.
Oligomerization of LAT-We do not know why some LAT had such high and stable DRM association. However, clustering of proteins can increase their DRM association (62,63), and proteins can be recruited to DRMs by binding to other proteins (32,33,64). Thus, oligomerization of LAT or association with other proteins could enhance its raft affinity. As a first step toward addressing this possibility, we used BN-PAGE to determine the apparent size of LAT isolated from transfected COS cells. In BN-PAGE, proteins are treated with Coomassie Blue, to impart a negative charge, and 6-amino-n-hexanoic acid, to maintain solubility (53). However, proteins are not denatured, and protein complexes often remain intact. This method provides only a rough estimate of size but can detect the difference between large and small complexes.
When we examined OG lysates of LAT-transfected COS cells by BN-PAGE and Western blotting, we detected several LATcontaining bands (Fig. 10A). The most intense band, which migrated near the 66-kDa standard, could have been either a small oligomer or a monomer. (The predicted molecular mass of LAT is 24,985 daltons, but it migrates as a 36 -38-kDa protein on SDS-PAGE (44).) We also saw a band that migrated near the 150-kDa standard. Finally, we often also saw one or two closely spaced bands that migrated much more slowly than the 200-kDa marker. When we repeated the experiment, lysing cells with digitonin, which often preserves protein-protein interactions that are disrupted by other detergents, we also saw several LAT-containing bands. Sometimes, as shown in (Fig. 10A), most of the protein migrated very slowly. In this case, a slowly migrating smear partially obscured one or two distinct bands that co-migrated with the slowly migrating bands in the OG lysates. We also saw bands that migrated slightly more slowly than the most prominent bands in the OG lysate. In the experiment shown here, these bands were very faint. On other gels, Lysates were adjusted to 40% sucrose, overlaid with a sucrose step gradient, and subjected to ultracentrifugation (100,000 ϫ g for 3 h). LAT in gradient fractions was detected by SDS-PAGE and Western blotting. C, COS cells were transfected with LAT-Myc, LAT-FLAG, or both, as indicated (m, LAT-Myc; f, LAT-FLAG). Cells were lysed with TNE plus 1% digitonin, and clarified lysates were subjected to immunoprecipitation using either mouse anti-Myc (m) or rabbit anti-FLAG (f) antibodies as indicated. Immune complexes were subjected to SDS-PAGE and Western blotting, probing blots with either mouse anti-Myc or mouse anti-FLAG antibodies and appropriate secondary antibodies as indicated. HC, mouse antibody heavy chain. D, Jurkat cells were treated (Latrun.) or not (Con.) with latrunculin A before preparation of DRMs on sucrose gradients and analysis of gradient fractions by Western blotting, probing with anti-LAT and anti-TfR as in Fig. 6. The percentage of total LAT in DRMs is indicated. these bands were substantially more prominent (not shown). Endogenous Jurkat cell LAT, solubilized by digitonin or OG, showed similar behavior to that presented here for LAT expressed in COS cells (data not shown). To ensure that the slowly migrating LAT seen after digitonin lysis was not contained in partially solubilized membrane fragments or other protein-lipid complexes, we subjected digitonin lysates of LATtransfected cells to sucrose gradient centrifugation, under conditions where any lipid-rich complexes would float to a low density position, but soluble protein complexes would move only a small distance. LAT was only detected in the bottom gradient fractions in which it was loaded, suggesting that it was fully solubilized (Fig. 10B). LAT(TMD)WT migrated at the dye front on BN gels, regardless of the detergent used to lyse cells (Fig. 10A). These results suggested that LAT oligomerized or associated with other proteins, whereas LAT(TMD)WT did not. Further work will be required to characterize these interactions more fully, to determine the basis of the differential effects of the two detergents, and to determine whether protein-protein interactions enhance raft affinity of LAT.
To determine whether individual LAT molecules interacted with each other, we co-expressed FLAG-and Myc-tagged LAT in COS cells. Cells were lysed with digitonin, and either protein was recovered by immunoprecipitation with appropriate antibodies. Immune complexes were subjected to SDS-PAGE and Western blotting, probing separately with antibodies to each tag. Each form of LAT could be co-immunoprecipitated with antibodies to the other form (Fig. 10C). Control experiments on singly transfected cells confirmed the identity of the bands (Fig. 10C). This result showed that complexes containing more than one LAT monomer are present in digitonin extracts. Co-immunoprecipitation of the two forms of LAT after OG extraction of co-transfected cells was weaker and variable (not shown).
To determine whether association with the actin cytoskeleton enhanced the DRM association of LAT, we treated Jurkat cells with or without the actin filament-disrupting agent latrunculin A before isolating DRMs. Latrunculin A treatment had no major effect on DRM association of LAT (Fig. 10D). DISCUSSION Biophysical considerations suggest that transmembrane helices, even if they are acylated, should be very difficult to pack into the ordered lipid environment found in rafts. However, some transmembrane proteins have been found in rafts. For some, like LAT, good evidence supports the idea that raft association is required for function (34,38). The mechanism of raft association of these proteins has been a puzzle. We found that at least two mechanisms contributed to efficient raft targeting of LAT. The first was acylation, and the second depended on some feature of LAT that was not present in its acylated transmembrane domain alone.
Quenching, Microscopy, and DRM Association All Show Low Raft Association of the Acylated LAT Peptide-An important finding of this work, supported by results from all three assays, was that the acylated transmembrane domain of LAT associated poorly with rafts. We did not detect a difference in raft association between acylated and nonacylated LAT peptides by fluorescence quenching or microscopy and observed only a modest enhancement by the detergent insolubility assay. Similarly, Killian and co-workers (20) found that an acylated transmembrane peptide of random sequence associated poorly with rafts. The slight discrepancy that we observed between the three assays probably resulted from differences in sensitivity. Although the shapes of the fluorescence quenching assay curves depend on K p , the difference in curve shape for fluorophores with similar K p values is small (54). Similarly, detection of partitioning of the peptides between domains in GUVs is not expected to be sensitive to small changes in K p . By contrast, acylation slightly enhanced peptide DRM association.
Thus, acylation had only a slight effect on raft association of the LAT peptide. Nevertheless, acylation, most likely via its effect on raft association, is clearly important for function of LAT and some other transmembrane proteins (32, 34 -36, 38). This suggests that cells may be highly sensitive to very small changes in raft partitioning of proteins, such that even the relatively modest inherent raft association of LAT can support raft-dependent functions. Furthermore, in addition to acylation, transmembrane proteins may generally require second raft-targeting mechanisms, such as protein-protein interactions (32,33), for full raft association.
Concentration of Acylated Peptides at Domain Boundaries-The simplest theoretical considerations of partition behavior would predict that a palmitoylated TM polypeptide would concentrate to some degree at the boundary between ordered and disordered domains, such that the palmitoyl groups were immersed in the ordered domains and the polypeptide immersed in the disordered domains. However, we saw no such edge concentration by microscopic examination of GUVs. Moreover, we previously calculated the theoretical effect of preferential localization of fluorophores at domain boundaries on the shape of the fluorescence quenching curve (see Fig. 5D in Ref. 21). This analysis predicted that in vesicles containing high concentrations of quencher lipid, quenching of a fluorophore with a high preference for domain boundaries would be lower in twophase vesicles than in uniform mixtures containing the same amount of quencher. However, even in two-phase vesicles containing 60 -70% 12SLPC, F/F 0 for the acylated peptide was higher in one-phase than in two-phase vesicles ( Fig. 3 and data not shown), consistent with preferential partitioning of the acylated peptide into the bulk l d phase domains. The fact that we did not detect a higher concentration of palmitoylated LAT peptide at domain boundaries may reflect a relatively modest affinity of the LAT palmitoyl groups for ordered domains. Consistent with this possibility, Wang et al. (65) found that dipalmitoylation of a nontransmembraneous peptide sometimes induced only a moderate degree of raft association in model membranes. It is also possible that the partition behavior of peptide-linked palmitoyl groups is affected by their interaction with the TM polypeptide.
Although we did not detect the palmitoylated LAT peptide at domain boundaries, it is an intriguing possibility that some acylated peptides or proteins may concentrate there. The assays we used might underestimate or fail to detect such a localization. Microscopic detection of domains in GUVs could underestimate edge concentration if the pixel size were sufficiently larger than the width of the edge domain. DRM association could also underestimate edge concentration if molecules at the edge of ordered domains were more soluble in detergent than those in the core of the domains. Furthermore, the edges of ordered domains may comprise a small fraction of the total amount of membrane within ordered domains. This would make it more difficult to detect enrichment of a molecule at edges by either DRM association or fluorescence quenching.
Nonacylated Peptides Show Some DRM Association-Even the nonacylated transmembrane peptides we examined were not completely excluded from DRMs. DRM association of these peptides increased with increasing amounts of ordered lipid domains in the bilayer (Fig. 5A). This behavior suggested that these peptides partitioned into ordered domains to a low but real degree. Acylation of the LAT peptide modestly enhanced DRM association (Fig. 5A). This result suggests that the distinction between "raft-associated" and "non-raft-associated" transmembrane proteins may often be more quantitative than qualitative.
Effect of Transmembrane Domain Sequence on Raft Targeting-We found no difference between DRM association of the nonpalmitoylated LAT peptide and the control peptide LW peptide. Furthermore, fluorescence quenching of the LAT peptide was very similar to that of the LW peptide reported previously (21). Thus, the amino acid sequence of the LAT peptide did not contribute measurably to raft affinity in the absence of acylation. However, proper amino acid sequence may enhance raft affinity of some acylated transmembrane proteins. For instance, transmembrane domain mutations, especially those in residues predicted to contact the outer leaflet of the bilayer, can reduce DRM association of influenza hemagglutinin (HA), even when palmitoylation is not affected (66,67). Nevertheless, no conserved "raft-targeting" motifs have been identified in the transmembrane domain of any protein. The transmembrane domains of two different HA subtypes show little sequence similarity, although mutation of each can affect DRM association (66,67). The fact that HA is a trimer further complicates analysis. If interactions between monomers in the trimer enhance raft association, then mutations might affect raft affinity indirectly by changing these interactions. Although these and other studies have defined regions of some transmembrane proteins that are important in raft targeting (25)(26)(27)(28)(29)(30)(31), no model of how certain residues could favor raft association has emerged.
Enhanced DRM Association of Full-length LAT-Our results suggested that cells contain two pools of LAT, with very different raft affinities. When LAT that was CHAPS-soluble in cells was incorporated into artificial raft-containing vesicles and then extracted with Triton X-100, the protein was recovered in both DRM and Triton X-100-soluble fractions, although only a small amount was in DRMs. We speculate that this LAT partitioned between rafts and nonrafts in the vesicles according to its inherent raft affinity. This inherent affinity was probably determined largely by the acylated transmembrane domain, because about the same amount of the acylated LAT peptide was recovered in DRMs from vesicles of the same lipid composition.
However, a second pool of LAT seemed to have a much higher raft affinity. This conclusion was based on the finding that when LAT isolated from cell-derived DRMs was incorporated into raft-containing vesicles and then subjected to Triton X-100 extraction, almost all of the LAT was recovered in DRMs. Thus, a second factor, in addition to the acylated transmembrane domain, strongly enhanced raft affinity of this pool of LAT. The total raft association of LAT in cells is probably a weighted average of the raft association of the two pools.
Oligomerization, or binding to other protein(s) with higher raft affinity, can enhance DRM association of some proteins. For instance, antibody-mediated cross-linking of glycosylphosphatidylinositol-anchored proteins increases their DRM association (62). Furthermore, DRM association of CD4 was found to depend both on its own acylation and on binding to Lck (32). Likewise, the association of the B cell receptor with rafts, which is important for signaling, is prolonged by co-ligation of the receptor with a cell surface protein complex (33). The tetraspanin CD81 plays a key role in recruiting this complex to rafts (33). Protein-protein interactions might enhance raft association of LAT as well. Consistent with this possibility, we detected at least some LAT, but no LAT(TMD)WT, in protein complexes by BN-PAGE. Alternatively, it is possible that another region of LAT, distinct from the transmembrane domain, can interact directly with the bilayer to promote raft association.
In conclusion, we found that although acylation is known to be required for raft targeting of LAT, raft targeting of the acylated LAT transmembrane domain alone was very inefficient. As might have been expected, then, acylation was not sufficient to completely overcome the difficulty of packing a hydrophobic helix into an ordered raft domain. The acylated and nonacylated LAT peptides differed only slightly in their DRM association. Further work will more fully elucidate how cells harness the relatively modest inherent raft affinity of LAT and probably of other transmembrane proteins for use in raftdependent functions.