A novel biotinylated lipid raft reporter for electron microscopic imaging of plasma membrane microdomains.

The submicroscopic spatial organization of cell surface receptors and plasma membrane signaling molecules is readily characterized by electron microscopy (EM) via immunogold labeling of plasma membrane sheets. Although various signaling molecules have been seen to segregate within plasma membrane microdomains, the biochemical identity of these microdomains and the factors affecting their formation are largely unknown. Lipid rafts are envisioned as submicron membrane subdomains of liquid ordered structure with differing lipid and protein constituents that define their specific varieties. To facilitate EM investigation of inner leaflet lipid rafts and the localization of membrane proteins therein, a unique genetically encoded reporter with the dually acylated raft-targeting motif of the Lck kinase was developed. This reporter, designated Lck-BAP-GFP, incorporates green fluorescent protein (GFP) and biotin acceptor peptide (BAP) modules, with the latter allowing its single-step labeling with streptavidin-gold. Lck-BAP-GFP was metabolically biotinylated in mammalian cells, distributed into low-density detergent-resistant membrane fractions, and was readily detected with avidin-based reagents. In EM images of plasma membrane sheets, the streptavidin-gold-labeled reporter was clustered in 20-50 nm microdomains, presumably representative of inner leaflet lipid rafts. The utility of the reporter was demonstrated in an investigation of the potential lipid raft localization of the epidermal growth factor receptor.

Piscataway, NJ) were provided by the indicated suppliers. Transferrin receptor-specifi c mouse monoclonal antibody H68.4 was kindly provided by Dr. David R. Sheff. Streptavidin-HRP was purchased from Perkin Elmer (Waltham, MA). Aurion BSA-C, cold water fi sh skin gelatin, normal goat serum, 6 nm gold particleconjugated streptavidin, and 10 nm gold particle-conjugated F(ab')2 fragment of goat anti-rabbit IgG were obtained from Electron Microscopy Sciences (Hatfi eld, PA). MDA-MB-231, MDA-MB-468, MCF7, and COS1 cells were purchased from the American Type Culture Collection (Manassas, VA) and cultured according to their recommendations unless otherwise noted. Transient transfection of cells with plasmid cDNAs was performed using Lipofectamine 2000 (Invitrogen, Grand Island, NY) according to the manufacturer's protocol.

Design and construction of the lipid raft reporter Lck-BAP-GFP
A plasmid vector expressing the tripartite Lck-BAP-GFP reporter ( Fig. 1 ) was constructed as follows. A cDNA encoding the 129-amino acid BAP module from the 1.3S transcarboxylase of Propionibacterium shermanii was amplifi ed by PCR with the PinPoint Xa-1 plasmid (Promega, Madison, WI) as template and cloned between the Bgl II and Sac II restriction sites of pDisplay (Invitrogen) exactly as described by Tannous et al. ( 15 ). The BAP module in this construct (kindly provided by Dr. David Motto) was subcloned via Bgl II and Sac II restriction enzymes into pLck-GFP, an expression vector based on pEGFP-N1 (Clontech, Mountain View, CA) that encodes a GFP protein with a leader peptide consisting of the 26-amino acid N-terminal membrane-targeting motif of the human Lck kinase ( 16 ) (kindly provided by Dr. Steven H. Green). The resulting construct, pLck-BAP-GFP, encodes a reporter comprising the N-terminal Lck membrane-targeting motif, an internal BAP module, and a C-terminal GFP module. The Lck-BAP-GFP encoding sequence was directly validated by DNA sequencing.

Western immunoblotting and streptavidin-HRP detection of biotinylated proteins after SDS-PAGE
Proteins resolved by SDS-PAGE were detected by standard Western immunoblotting procedures or streptavidin-HRP blotting in the case of biotinylated proteins. After proteins were transferred from SDS-PAGE gels to nitrocellulose membranes, the membranes were blocked with 3% BSA in TBS. Membranes were then probed with specifi c primary antibodies and either HRP-conjugated secondary antibodies detected by luminography with ECL Plus reagents (GE Healthcare, Pittsburgh, PA) or IRDye secondary antibodies detected with an Odyssey imaging system (LI-COR Biosciences, Lincoln, NE). Alternatively, biotinylated proteins were visualized by probing membranes with a streptavidin-HRP conjugate and ECL Plus reagents.

Gradient fractionation of low-density, detergent-resistant membranes
Fractionation of low-density, detergent-resistant membranes by sucrose density gradient fl otation centrifugation was performed much as described by Rodgers et al. ( 17 ). Briefl y, two ‫ف‬ 80% confl uent 150 mm dishes of MDA-MB-231 cells transiently expressing the Lck-BAP-GFP reporter were suspended in 0.75 ml of an ice-cold 1% Triton X-100 solution in TNE buffer ( MO]). This and all subsequent steps were performed at 4°C. The suspension was clarifi ed by centrifugation at 1,300 g for 5 min, then mixed and specifi cally label the antigens of choice and the optimization of immunolabeling conditions. To simplify the application of the immuno-EM membrane sheet method in visualizing lipid raft microdomains, we sought to identify a suitable raft marker that i ) was localized with high specifi city in lipid raft entities, ii ) was exposed on the inner surface of the cytoplasmic membrane (to facilitate labeling of inside-up membrane sheets), and iii ) could be effi ciently labeled with gold-conjugated reagents. The list of lipid raft markers that have been used in various studies includes i ) membrane outer leafl et markers, such as ganglioside GM1 (labeled with cholera toxin-based reagents) and various glycosylphosphatidylinositol (GPI)anchored proteins (e.g., Thy-1 and placental alkaline phosphatase); ii ) intrinsic membrane proteins, such as caveolin-1 (a marker for caveolae) and fl otillins (in our experience distributed in lipid rafts and the bulk membrane); and iii ) membrane inner leafl et markers, such as the multiply acylated peripheral membrane proteins G q ␣ (dually palmitoylated) and Lck (myristoylated and dually palmitoylated). Considering our criteria for a raft marker, the extent to which various raft constituents have been characterized, and our interest in noncaveolar lipid rafts, we identifi ed the Src family protein tyrosine kinase Lck as an ideal candidate and generated an Lck-derived reporter that could be readily labeled with avidin-biotin technology.
We present here a simplifi ed and robust method for EM imaging of lipid rafts in cellular cytoplasmic membranes. 2 This method uses a unique genetically encoded lipid raft reporter molecule that greatly facilitates the gold labeling of raft microdomains in immuno-EM membrane sheet experiments. This reporter comprises the N-terminal membrane-and lipid raft-targeting sequence of Lck, an enhanced green fl uorescent protein (GFP) module for the ready assessment of reporter expression, and a metabolically biotinylated biotin acceptor peptide (BAP) module, allowing the marker to be conveniently labeled with avidin reagents, such as streptavidin-gold conjugates in the context of EM imaging. We demonstrate the utility of this reporter in EM imaging of lipid rafts in plasma membrane sheets.

Reagents, cell culture, and cell transfection
Antibodies recognizing caveolin-1 (mouse monoclonal #610406; BD Transduction Labs, San Jose, CA), epidermal growth factor receptor (EGFR) (rabbit monoclonal SP9, Thermo Scientifi c, Rockford, IL), fl otillin-2 (rabbit polyclonal H-90, Santa Cruz Biotechnology, Santa Cruz, CA), and GFP (mouse monoclonal #N86/8, NeuroMab, Davis, CA; rabbit polyclonal A01704, GenScript, EM imaging of gold-labeled proteins in cellular plasma membrane sheets MDA-MB-468 breast cancer cells were plated on circular glass coverslips in 35 mm culture dishes at approximately 70% confl uency and cultured overnight in Leibovitz's L-15 medium with 10% FBS. The next day, cells were transfected with the pLck-BAP-GFP or pLck-GFP vector for 3 h in serum-free medium and incubated overnight in complete medium at 37°C. The following day, membrane sheets were isolated and fi xed on EM grids by a procedure adapted from those of Wilson et al. ( 19 ) and Prior et al. ( 9 ). Briefl y, cultures were washed twice with PBS, and individual coverslips were removed from the PBS, inverted, and lowered upon a group of formvar/carbon-coated EM grids that were precoated with poly-L-lysine. Gentle pressure was applied to the coverslip with the thumb for 10 s. The grids were then fl oated off the coverslip with PBS, fi xed with 2% paraformaldehyde, washed with PBS, and placed in blocking buffer (5% Aurion BSA-C [acetylated BSA], 0.1% cold water fi sh skin gelatin, and 5% normal goat serum in PBS) for 30 min. The grids were incubated overnight with 6 nm gold-conjugated streptavidin or, in the case of double-labeling experiments, in the presence of 6 nm gold-conjugated streptavidin and the antibody of choice (rabbit anti-EGFR or anti-GFP) diluted in incubation buffer (0.2% Aurion BSA-C in PBS). The following day, grids were washed six times with incubation buffer. Doublelabeled grids were further incubated with a 10 nm gold-conjugated F(ab')2 fragment of anti-rabbit-IgG followed by six washes with incubation buffer. Grids were postfi xed with 2.5% glutaraldehyde, washed once with PBS, washed fi ve times with water, stained with uranyl acetate, dried, and imaged with a JEOL JEM 1230 transmission electron microscope (12,000×).

Spatial pattern analysis of particles in EM images
The spatial distribution of gold-labeled particles in EM images was analyzed by use of the Ripley's K function statistic (9,20). This analysis is based on Ripley's function K ( r ), which for a distribution of one particle type is essentially the number of particles within a distance r of another divided by the average number of particles per unit area. For an ideally random particle distribution, K ( r ) = r 2 for all values of r . The normalized statistic SQRT[ K ( r )/ ] -r (as plotted in Fig. 6-8) then approximates a constant value of zero for a random particle distribution, and its deviation above zero indicates a tendency for the particles in the distribution to cluster and the size scale of this clustering ( 21 ). A similar bivariate K function was used to assess the tendency of particles of two different types to co-cluster. Before applying the K function statistic, the x -y coordinates of the centers of all goldlabeled particles in digital images were determined, and 6 nm and 10 nm particles were distinguished via the application of the automatic particle analysis and manual point selection tools in NIH ImageJ software application. Manual assessment was necessary when low contrast in images made automatic particle analysis error prone. Control experiments indicated that the 6 nm and 10 nm gold particles were readily distinguishable (see supplementary Fig. V). The x -y coordinates of the particle centers were then used to evaluate SQRT[ K ( r )/ ] -r over a range of r less than 1/5 of the image dimension via the SPPA software package generously provided by Dr. Peter Haase ( 20 ). This software uses the weighted edge correction method as originally described by Getis and Franklin ( 22 ) and modifi ed by Haase ( 23 ). This software also allows deviations from randomness in the spatial distribution to be assessed by performing 99 simulations of random particle distributions of an equivalent density, evaluating SQRT[K( r )/ ] -r for each distribution, and graphing the envelope of the maximum and minimum deviations of the SQRT[K( r )/ ] -r values from zero.
with an equal volume of 85% (w/v) sucrose in TNE buffer. The mixture was pipetted into a centrifuge tube and overlaid sequentially with 6 ml of 30% (w/v) sucrose and 3.5 ml of 5% (w/v) sucrose in TNE buffer. Gradients were centrifuged at 200,000 g for 14 h. After centrifugation, fractions were sequentially removed from the top of the gradient, and the presence of lipid raft and bulk membrane markers in the fractions was assessed by SDS-PAGE and immunoblotting with specifi c antibodies. Alternatively, the presence of the biotinylated Lck-BAP-GFP reporter was assessed by blotting with a streptavidin-HRP conjugate.

Fluorescence microscopy
For fl uorescence microscopy experiments, human MCF7 or simian COS1 cells were maintained in MEM with Earle's salts, 10% FBS, 1× GlutaMAX-I (Gibco, Grand Island, NY) or DMEM with 10% FBS (Hyclone, Rockford, IL), 1× GlutaMAX-I, respectively, in 5% CO 2 at 37°C. Before the experiments, cells were seeded in 35 mm dishes with cover glass bottoms (MatTek, Ashland, MA). The following day, cells at 40% or higher confl uency were transfected with 0.3-0.5 g of the pLck-GFP or pLck-BAP-GFP plasmid using 2-4 l of Lipofectamine 2000 transfection reagent (Invitrogen) in OptiMEM-I medium (Gibco). After 5 h of transfection, cells were placed in normal maintenance medium and allowed to recover overnight. The next day, the medium was replaced with low-serum medium (1.5% FBS), and the cells were examined by microscopy live or after fi xation in 4% paraformaldehyde. Confocal images of live MCF7 cells expressing the Lck-BAP-GFP reporter were acquired with a Zeiss 510 laser scanning confocal inverted microscope with a Plan-Apochromat 63× oil immersion objective. Total internal refl ection fl uorescence (TIRF) images of fi xed COS1 cells were acquired with a Leica AF 6000LX TIRF microscope with a 100× oil immersion objective.

Fluorescence recovery after photobleaching analysis of diffusional mobility
MCF7 cells grown in 35 mm culture dishes with cover glass bottoms were transfected with the pLck-GFP or pLck-BAP-GFP plasmid expression vector as described above. During fl uorescence recovery after photobleaching (FRAP) measurements, cells were bathed in HBSS to maintain a steady pH. A Zeiss 510 microscope confi gured with a 30°C heated stage, 63× objective, and 2× digital zoom was used for fl uorescence imaging. A circular region of interest (ROI) of 5 m diameter located in the plane of the cytoplasmic membrane was photobleached with a 488 nm argon laser (10 scans at 100% laser intensity). Using the 488 nm laser line at low power and a 505-550 nm band pass fi lter, preand postbleach images were captured as a time series with intervals of 2 s. The Zeiss LSM Image v4.2 microscope software was used to select a control ROI, which encompassed the bleached ROI, and to obtain the integrated fl uorescence intensity values for both ROIs as a function of time. To correct for the effect of gradual sample bleaching during laser scanning and instrument-related artifacts, a corrected fl uorescence intensity value for the bleached ROI at each time point was calculated as the ratio of bleached ROI and control ROI intensities, which was then normalized to an initial prebleach intensity of 100. Using the GraphPad Prism program, the corrected data were fi t by nonlinear regression with an analytical equation describing the FRAP recovery time course for the "bleached disk" experiment as derived by Soumpasis ( 18 ). The best-fi t equation yielded estimates for the characteristic diffusion time ( D ), the particle diffusion constant ( D = r 2 /4 D , where r is the radius of the bleached disk), and the fraction of reporter molecules that are mobile (mobile fraction). proteins endogenous to mammalian cells, carboxylases of 73, 75, 125, and 270 kDa that are mitochondrial or cytosolic and thus not present in the plasma membrane (26)(27)(28). The expression of the Lck-BAP-GFP reporter in cultured cells was also examined by TIRF and confocal fl uorescence microscopies ( Fig. 3 ). TIRF microscopic images clearly showed localized expression of the GFP-tagged protein in the cellular membrane proximal to the cover glass substrate. Transverse optical sections of cultured cells obtained by confocal fl uorescence microscopy showed expression of the reporter predominantly in the peripheral plasma membrane.

Diffusional mobility of the Lck-BAP-GFP reporter
Fluorescent reporters with multiply acylated N-termini similar to that of Lck (e.g., MyrPalm-YFP) ( 29 ) have been used in biophysical investigations of the molecular diffusion of lipid raft-associated proteins. Fluorescence recovery after photobleaching (FRAP) studies have shown that, although such reporters are lipid raft associated, they freely diffuse in the plasma membrane on the micron scale. Alternatively, fl uorescence microscopic, single-particle tracking experiments have shown that such reporters on the submicron scale show anomalous diffusion characterized by periods of unrestricted diffusion and episodes of restricted diffusion within submicron "transient confi nement zones" (30)(31)(32). If such transient confi nement zones are related to lipid rafts, these observations suggest that lipid raft microdomains are not stable structures but are dynamic in nature, with constituent raft proteins free to hop between lipid raft and bulk membrane environments ( 30,32 ) and rafts themselves possibly representing a transient coalescence of a liquid-ordered membrane phase ( 33 ).
The presence of the GFP module in the Lck-BAP-GFP reporter makes it applicable in such biophysical approaches.

A metabolically biotinylated and genetically encoded lipid raft reporter
To facilitate the gold labeling of lipid raft membrane microdomains in EM imaging studies, we developed an Lck-based lipid raft reporter construct designated Lck-BAP-GFP ( Fig. 1 ). Like the previously characterized fl uorescent reporter Lck-GFP ( 16 ), Lck-BAP-GFP incorporates the fi rst 26 N-terminal amino acids of the human Lck protein, which includes its post-translationally modifi ed membrane-targeting motif. This short N-terminal sequence has been shown in the context of chimeric proteins to be myristoylated, palmitoylated, and lipid-raft targeted ( 6,24,25 ). The Lck-BAP-GFP raft reporter developed here incorporates BAP and GFP modules ( Fig. 1 ). The GFP module enables the ready observation by fl uorescence microscopy of the expression and subcellular localization of the genetically encoded reporter. The BAP module is efficiently biotinylated by endogenous biotin ligases upon its expression in mammalian cells ( 15,26 ) and thus confers to proteins an ability to be detected with avidin-based reagents, such as streptavidin-HRP or streptavidin-gold particle conjugates.
In our characterization of the Lck-BAP-GFP reporter, we fi rst verifi ed its expression and metabolic biotinylation ( Fig. 2 ). Expression of Lck-BAP-GFP and the parental Lck-GFP reporter in transiently transfected cultured cells was demonstrated by immunoblotting of cell lysates with a GFP antibody. Immunoblotting also showed that the migration of each protein in SDS-PAGE was consistent with its predicted molecular mass. As anticipated, Lck-BAP-GFP was readily detected upon blotting of membrane transfers of SDS-PAGE gels with a streptavidin-HRP conjugate ( Fig. 2 ). Thus, the Lck-BAP-GFP protein appeared to be effi ciently biotinylated. Although the Lck-BAP-GFP reporter was the major band detected in streptavidin-HRP blots, longer-exposed blots of human cell lysates showed additional biotinylated proteins of approximately 75 and 125 kDa (see Fig. 5 ). The molecular masses of these proteins were consistent with three of the four known biotinylated Fig. 1. Schematic structure of a metabolically biotinylated lipid raft reporter, Lck-BAP-GFP. For the purpose of imaging lipid raft microdomains in cellular plasma membranes, a genetically encoded lipid raft-targeted reporter construct was developed that is metabolically biotinylated and thus detectible with avidin reagents. Schematic structure of the Lck-BAP-GFP reporter shows: Lck-N, the N-terminal lipid raft-targeting sequence of the Lck kinase with sites of myristoylation and dual palmitoylation indicated; BAP, the biotin acceptor peptide sequence that is metabolically biotinylated by endogenous mammalian biotin ligases; and GFP, enhanced green fl uorescent protein module for detection by fl uorescence microscopy. cancer cells expressing the Lck-BAP-GFP reporter after treatment with cold Triton X-100. In such gradient fl otation experiments, detergent-resistant membranes and lipid raft-associated proteins are expected to migrate to fractions of low buoyant density, whereas cytoplasmic proteins and detergent-solubilized nonraft membrane proteins will remain in the densest gradient fractions. Consistent with this expectation, immunoblotting showed the concentration of the lipid raft proteins caveolin-1 and fl otillin-2 in low-density gradient fractions (fractions 5 and 6), whereas the nonraft membrane protein transferrin receptor was found exclusively in the bottom gradient fractions (fractions [14][15][16]. Although some of the biotinylated Lck-BAP-GFP reporter was found in the cytosolic/solubilized membrane fractions, it was present at highest concentrations in the low-density, detergent-resistant membrane fractions. Because the association of Lck and Lck-derived reporters with detergent-resistant membranes is known to be dependent upon myristoylation and palmitoylation ( 17,24 ), these results indicate that the Lck-BAP-GFP reporter is subject to both translational modifi cations. In the cytosolic/solubilized membrane fractions (fractions [14][15][16], the Lck-BAP-GFP protein migrated as a doublet, with a higher mobility species evident therein. This second species might represent an Lck-BAP-GFP protein with incomplete post-translational modifi cation and thus may be devoid of a fully functional membrane-targeting motif. Also detected in the cytosolic/solubilized membrane fractions were biotinylated proteins of approximately 75 and 125 kDa, which likely represent endogenously biotinylated mitochondrial carboxylases (see above).
In agreement with earlier reports that the EGFR is solubilized in cold Triton X-100 ( 37,38 ), this putative lipid raft resident protein was also found in the solubilized gradient fractions. This suggested that if the EGFR is indeed As an example, to determine whether the Lck-BAP-GFP reporter behaved as previously characterized Lck-like reporters, we examined the macroscopic diffusion of Lck-GFP and Lck-BAP-GFP in the cellular membrane by FRAP ( Fig. 4 ). Lck-GFP and Lck-BAP-GFP showed unrestricted diffusion on the micron scale, with apparent diffusion constants of 0.44 and 0.32 m 2 /s, respectively, and mobile fractions close to 1.0. These diffusional characteristics closely agree with the results of previous FRAP studies of reporters with Lck-like membrane anchors ( 29,34 ). The high mobile fractions observed suggest that the diffusion of both reporters is unrestricted on the time-scale of these measurements and, importantly, that the incorporation of the BAP module into Lck-BAP-GFP does not cause the interaction of Lck-BAP-GFP with plasma membrane or cytoskeletal matrix proteins in a manner that signifi cantly impedes its diffusion.

Distribution of Lck-BAP-GFP into low-density, detergentresistant membrane fractions
Appending the N-terminal membrane-targeting motif of Lck to a reporter molecule of interest can confer a lipid raft localization to that reporter. In biochemical fractionation approaches, various Lck-based reporters have been found to distribute into detergent-resistant membrane fractions of low density. Although the relationship of such biochemically defi ned detergent-resistant membranes to lipid raft microdomains existing in intact cell membranes is unclear ( 35 ), detergent-resistant membranes are enriched in various lipid raft constituents (36,37). We sought to verify that the metabolically biotinylated Lck-BAP-GFP reporter would, like other characterized Lck-based probes and the Lck protein itself, distribute into detergent-resistant membranes ( 17,24 ). Fig. 5 shows a density gradient fractionation of membrane proteins from MDA-MB-231 breast streptavidin reagent ( Fig. 6 ). A representative image of gold-labeled Lck-BAP-GFP reporters in MDA-MB-468 breast cancer cell membranes is shown in Fig. 6A . A fairly dense labeling of the plasma membrane surface was generally seen, with an apparent clustering of the gold labels in microdomains of less than ‫ف‬ 100 nm. The companion Fig. 6B is a control image representative of membranes from cells expressing a nonbiotinylated raft reporter Lck-GFP and subjected to the same streptavidin-labeling procedure. In this case and as quantifi ed in Fig. 6C , a dramatically reduced labeling was evident, which indicated that the Lck-BAP-GFP reporter can be labeled in plasma membrane sheets with good specifi city. This was anticipated given the high specifi city of avidin-labeling technologies and that biotinylated proteins endogenous to mammalian cells are mitochondrial or cytosolic.
The spatial distribution of Lck-BAP-GFP reporters on the membrane surface and their apparent tendency to cluster were analyzed by use of the Ripley's K function statistic ( Fig. 6D ). By this analysis of the center-to-center interparticle distances, the distribution of gold-labeled Lck-BAP-GFP reporters in EM images was seen to be nonrandom, with a clustering of the reporters prominent on the size scale of 20-50 nm. This is similar to the size of Ras a lipid raft resident, it likely resides in rafts of different biochemical composition than of those into which the Lck-BAP-GFP reporter partitions (see Discussion).

EM imaging of the gold-labeled Lck-BAP-GFP reporter in plasma membrane sheets
Given the ease of its detection with avidin-based technologies, the Lck-BAP-GFP reporter described herein should have diverse applications in the investigation of lipid raft entities, for example its use as a marker of lipid rafts in biochemical fraction experiments ( Fig. 5 ). We envisioned it would be especially useful in EM characterization of protein microlocalization in plasma membrane sheets ( 3,8,9 ), wherein proteins of interest would be labeled with immunogold reagents and the biotinylated Lck-BAP-GFP lipid raft reporter with a streptavidin-gold particle conjugate. Thus, we next sought to demonstrate that the Lck-BAP-GFP reporter in cellular membranes could be effectively labeled with streptavidin-gold and its spatial distribution imaged by EM. To this end, membrane sheets derived from cultured cells ectopically expressing the Lck-BAP-GFP reporter were immobilized on polylysinecoated EM grids, and the membranes were fi xed and then labeled in a single step with a 6 nm gold particle-conjugated alternative gold-conjugated reagent that showed no tendency to self-cluster ( Fig. 7

Application of the Lck-BAP-GFP reporter as a lipid raft marker in EM imaging
The localization of the expressed Lck-BAP-GFP reporter in low-density, detergent-resistant membrane fractions and its observed clustering in plasma membrane domains of 20-50 nm are consistent with the reporter being targeted to lipid rafts (i.e., membrane microdomains with lipid composition distinct from the bulk membrane and with a liquid ordered structure; see Discussion). Given its ease of detection with streptavidin reagents, we anticipate that this reporter will be a useful inner leafl et lipid raft marker for biochemical experiments and various imaging applications. To demonstrate the utility of Lck-BAP-GFP as a raft marker in EM imaging, we applied it in assessing the potential lipid raft localization of EGFR molecules in breast cancer cell plasma membranes. Here, membrane sheets derived from cultured MDA-MB-468 breast cancer cells were simultaneously subjected to conventional immunogold labeling of the EGFR (with 10 nm gold particles) and streptavidin-gold labeling of the ectopically expressed Lck-BAP-GFP reporter (with 6 nm gold particles). EGFR and Lck-BAP-GFP were effectively gold labeled, and the two labeled species were readily distinguished (Fig. 8A).
When the spatial distributions of the labeled Lck-BAP-GFP and EGFR molecules were independently analyzed nanodomains identifi ed by Prior et al. in cellular plasma membranes ( 10 ) and is consistent with the proposed size of lipid raft structures ( 36 ).
To further validate the streptavidin-gold labeling strategy, we performed double-labeling experiments in which membrane sheets from Lck-BAP-GFP-expressing cells were labeled with streptavidin-gold and a GFPspecifi c antibody detected with a conventional immunogold reagent. The patterns of labeling by streptavidin-gold and the immunogold conjugate independently showed clustering on the scale of 20-50 nm ( Fig. 7 ). Importantly, there was signifi cant co-clustering of the two labels, as shown by a bivariate Ripley's K function analysis ( Fig. 7D ). Depending upon the effi ciencies of labeling of the Lck-BAP-GFP reporter with each of the two detection reagents and the number of reporter molecules present in individual rafts, either a perfect or partial overlap of streptavidin-gold and immunogold-labeled clusters would be predicted. Thus, we interpret the co-clustering of streptavidin-gold and anti-GFP/immunogold labeling ( Fig. 7D ) as evidence that both detection reagents effectively label the Lck-BAP-GFP reporter.
The 6 nm gold-conjugated streptavidin reagent displayed some tendency to self-cluster in control experiments in which the reagent was directly deposited on an EM grid (supplementary Figs. I and II). The observed clustering of Lck-BAP-GFP was much more extensive and was also observed by conventional immunogold labeling with an lipid raft species, at least not of the type into which a multiply acylated inner leafl et raft protein such as Lck would partition.

DISCUSSION
Lck is among those Src family protein tyrosine kinases having an N-terminal membrane-targeting motif subject to both myristoylation and palmitoylation. Thus, the Nterminal sequence of the nascent Lck peptide (MGCGCS in human Lck) is post-translationally modifi ed by proteolytic removal of the initiating methionine, followed by myristoylation of Gly-2 and palmitoylation of at least one of the cysteine residues Cys-3 and Cys-5 ( 17,42 ). Posttranslational modifi cation with multiple saturated fatty acid motifs confers to some proteins an affi nity for lipid raft domains, as demonstrated by a number of biochemical, biophysical, and imaging approaches (43). In the case of Lck, a variety of studies indicate that its N-terminal targeting motif is suffi cient to effect its localization in lipid rafts ( 6,17,24,25,42 ). Indeed, biophysical studies using fl uorescent ( Fig. 8B, C ), each showed signifi cant deviation from randomness and clustering on the submicron scale. Although the Lck-BAP-GFP reporter again showed clustering most predominant in the size range of 20-50 nm (compare Fig. 6D and 7B ), the EGFR showed clustering over a broader size range (25-150 nm). Such submicron-scale clustering of the EGFR in cellular plasma membranes has been previously characterized by immuno-EM ( 11,39 ) and single molecule-resolution fl uorescence microscopy (40), although the factors responsible for the establishment of these microdomains are the subject of continuing investigation ( 7,(39)(40)(41). By analyzing the co-clustering of the EGFR and Lck-BAP-GFP reporter with the bivariate Ripley's K function statistic ( Fig. 8D ), we addressed the issue of whether these observed EGFR receptor microdomains represent lipid raft entities. In this and four repeated experiments (data not shown), there was no signifi cant co-clustering of the EGFR and the lipid raft marker beyond what might be explained by a random coincidence of localization. We conclude that EGFR microdomains in MDA-MB-468 breast cancer cell membranes do not represent of plasma membrane-associated signaling molecules (many of which are peripheral membrane proteins or transiently associated with the plasma membrane inner surface) highlight the need for well characterized probes of membrane inner leafl et microdomains. The Lck-based lipid raft reporter described herein is thus expected to be particularly useful in view of its targeting to inner leafl et membrane microdomains (domains that are possibly distinct from those interrogated with commonly used outer leafl et reporters such as GPI-anchored proteins or ganglioside GM1-binding cholera toxin) and because of its amenability to the immuno-EM membrane sheet imaging methodology. The ease of its detection with avidin-based technology should make the Lck-BAP-GFP reporter also useful in the context of biochemical approaches for the study of lipid rafts.
The present work demonstrates the particular utility of the genetically encoded Lck-BAP-GFP reporter in the characterization of lipid raft entities by EM. The reporter was readily expressed in cultured cells and metabolically biotinylated therein, as indicated by its detection with streptavidin-HRP in blots of SDS-PAGE gels ( Fig. 2 ) and streptavidin-gold conjugates in EM imaging ( Figs. [6][7][8]. In membrane fractionation experiments, the reporter, as reporters with Lck-like targeting motifs have provided crucial evidence for the existence of lipid rafts in intact cells ( 6,25,44 ).
For some time the subject of controversy, lipid rafts are generally envisaged as small (<200 nm) membrane microdomains enriched with sphingolipids and cholesterol and possessing a more ordered structure than the bulk plasma membrane (which apparently underlies the enhanced affi nity of proteins with saturated fatty acid modifi cations for lipid rafts). The characterization of lipid raft entities has been confounded by the limited resolution of classical light microscopy, their intractability to biochemical fractionation, and the potentially transient nature of their formation (35,45). There is also signifi cant diversity among lipid raft species, with differing raft markers having been shown to occupy distinct domains ( 4,14 ). Thus, although EM examination of the spatial distributions of the plasma membrane outer leafl et lipid raft constituents sphingomyelin and ganglioside GM1 showed them each clustering within submicron domains, the observed sphingomyelin-and ganglioside GM1-rich microdomains were nonidentical (14). The paucity of well characterized lipid raft reporters, the recognized diversity of lipid raft species, and the signifi cant interest in characterizing the spatial organization investigations of membrane protein microlocalization and lipid raft association. We made a fi rst application of this reporter in an EM investigation of the proposed lipid raft localization of the EGFR ( Fig. 8 ). Although a number of biochemical fractionation studies have shown that the EGFR resides to a signifi cant extent in low-density membrane fractions believed to represent lipid rafts ( 37,38,(47)(48)(49)(50), our results indicated that the microdomains of EGFR localization observed by immuno-EM imaging of plasma membrane sheets ( 11,39 ) are not identical to those in which the inner leafl et raft marker Lck-BAP-GFP resides. Two previous EM studies, which used cholera toxin B-HRP staining of ganglioside-GM1 (41) or immunogold labeling of placental alkaline phosphatase (50) in ultrathin cryosections, showed a limited colocalization of EGFR molecules with plasma membrane regions containing outer leafl et raft markers. Also, multiple studies using biochemical techniques and EM imaging have shown that the EGFR, while present in low-density membrane fractions also rich in caveolin-1, does not reside to a signifi cant extent within caveolae ( 38,49,50 ). The present study suggests that EGFR microdomains observed by immuno-EM are also not identical to lipid rafts that are labeled with an Lck-based reporter. expected, was found in low-density detergent-resistant membrane fractions along with the traditional lipid raft markers caveolin-1 and fl otillin-2 ( Fig. 5 ). When the spatial distribution of the Lck-BAP-GFP reporter in the plasma membrane was examined by EM imaging of streptavidin-gold-labeled membrane sheets, the reporter showed a nonrandom localization with obvious clustering on the scale of 20-50 nm ( Figs. 6-8 ). The submicron-scale clustering of the Lck-BAP-GFP reporter, along with its observed fractionation in low-density detergentresistant membranes, is consistent with its anti cipated lipid raft localization. Nonetheless, in agreement with earlier studies of raft-associated proteins (34), the Lck-BAP-GFP reporter showed free diffusion on the micron scale ( Fig. 4 ), suggesting that its association with lipid rafts is transient in nature or that lipid rafts themselves exist only transiently. Interestingly, related single-particle tracking studies of Lck-based fl uorescent reporters have shown their diffusion on the scale of one micron, with periodic restrictions of diffusion within submicron zones of confi nement possibly representative of lipid raft entities ( 30,31,46 ).
The primary goal of this study was to create a genetically encoded lipid raft marker that would facilitate EM Fig. 8. Application of the Lck-BAP-GFP reporter in assessing the potential lipid raft localization of the EGFR. Membrane sheets from MDA-MB-468 cells expressing a high level of endogenous EGFR and the ectopic Lck-BAP-GFP lipid raft marker were analyzed by EM after gold labeling of the EGFR and Lck-BAP-GFP proteins. The EGFR was immunolabeled with a rabbit monoclonal antibody and 10 nm gold-conjugated secondary antibody, and the biotinylated raft marker was labeled with a 6 nm gold-conjugated streptavidin reagent. A: Representative transmission EM image (cropped version of the larger analyzed image) of 10 nm gold-labeled EGFR (arrowhead) and 6 nm gold-labeled Lck-BAP-GFP (arrow) proteins in a plasma membrane sheet (12,000×). B and C: Ripley's K function analysis of the clustering of the labeled Lck-BAP-GFP reporter (B) and EGFR (C). D: The evaluated bivariate K function (solid line) for the Lck-BAP-GFP and EGFR particle distributions in the representative image fell largely within the envelope of 99 simulated random distributions (dotted lines), which indicated little tendency for the EGFR and raft reporter to cocluster.
Our observation that the EGFR was apparently solubilized in cold Triton X-100 and was not present in the Triton X-100-resistant, caveolin-1-and Lck-BAP-GFP-enriched lipid raft fractions (see Fig. 5 ) is consistent with earlier investigations of the detergent-dependent solubility of EGFR-containing lipid raft species ( 37,38 ). In particular, Pike et al. ( 37 ) found that lipid rafts resistant to cold Triton X-100 (in which the Lck-BAP-GFP reporter was here found to be enriched) have a lipid composition distinct from that of EGFR-containing lipid rafts isolated using alternative detergents or a detergent-free method. This bolsters our conclusion that EGFR-containing membrane microdomains are distinct from those into which the Lck-BAP-GFP reporters partitions. Nonetheless, the biochemical characteristics of EGFR microdomains and the factors responsible for their formation remain open questions.
The novel Lck-BAP-GFP reporter described herein should be a valuable tool for the investigation of plasma membrane lipid rafts. This reporter is readily detected in biochemical fractionation and EM imaging applications by virtue of its constitutive biotinylation when expressed in mammalian cells. Our observation of Lck-BAP-GFP clustering within submicron domains suggests it might indeed reside in cholesterol-rich liquid ordered microdomains, as have been observed in model membrane systems. However, the lipid raft species in which this inner leafl et marker resides are likely distinct from those labeled by other raft reporters. Thus, the Lck-BAP-GFP reporter should not be considered a marker of lipid rafts in general but a specifi c marker of rafts for which a multiply acylated inner leafl et protein would have affi nity. We envision that additional lipid raft reporters exploiting the same avidin-based detection technology but derived from distinct raft resident proteins (e.g., caveolin or fl otillin proteins) would be of much utility in characterizing the diversity of lipid raft species.