Tyrosine kinase receptors concentrated in caveolae-like domains from neuronal plasma membrane.

Recent evidence suggests that tyrosine kinases are highly organized in caveolae of tissue culture cells. We now report the isolation of a membrane domain from neuronal plasma membranes that has the biochemical characteristics of caveolae. A low density membrane (LDM) fraction with the same density as caveolae was highly enriched in tyrosine kinases such as insulin receptors, neurotrophin receptors, Eph family receptors, and Fyn. Grb2, Ras, heterotrimeric GTP-binding proteins, and Erk2 were also concentrated in the LDM. Incubation of the LDM fraction at 37°C stimulated the phosphorylation on tyrosine of multiple, resident proteins, whereas the bulk membrane fraction was devoid of tyrosine kinase activity. The LDM, which makes up ∼5-10% of the plasma membrane protein, appears to be organized for signal transduction.

Cell fractionation has dramatically advanced our understanding of how caveolae function in cells. Originally, they were thought only to be involved in transcytosis across endothelial cells (1) because morphologic evidence indicated caveolae can form plasmalemmal vesicles capable of ferrying cargo molecules from the blood to tissue spaces (2). An endocytic role for caveolae has been confirmed with the discovery that the cytoplasmic delivery of folate by potocytosis (3) is mediated by a hormonally regulated, caveolae internalization cycle (4). The localization of multiple signal transducing molecules to caveolae (5,6), however, was unexpected. Caveolae appear to be involved in processing and integrating cellular information at the cell surface (7). The concentration of active receptor tyrosine kinases in caveolae (8,9), for example, raises the possibility that growth factor receptors may use caveolae to carry out specific signaling events in all cells. Exactly how caveolae coordinate signal transduction with the internalization of molecules and ions remains to be determined.
The principle method used to identify caveolae has been the characteristic flask shape these membranes display in thinsection electron microscopy images. Membranes with this appearance have been seen on the surface of most cells but appear to be particularly abundant in endothelial (10,11) and muscle cells (12). Unlike the clathrin-coated pit, which can be recognized regardless of membrane shape, misshapen caveolae are not easily detected by this method. Furthermore, ultrastructural criteria are difficult to apply in tissue cells such as neu-rons, which contain many vesicular and invaginated membrane profiles that cannot be distinguished from each other without specific markers. Several markers for invaginated caveolae have been identified, including glycosylphosphatidylinositol-anchored membrane proteins (13,14), G M1 gangliosides (15), and caveolin (16). These markers may in the future allow one to define this membrane domain in complex tissues using ultrastructure.
In addition to morphology, other criteria now exist for determining if a cell or tissue of interest has a caveolae-like membrane domain. Receptor tyrosine kinases (8), tyrosine-phosphorylated proteins (8), and Erk 1 have been found to be enriched in isolated caveolae and mapped by immunocytochemistry to invaginated, fibroblast caveolae. These and other new marker proteins help to further define and identify caveolae membrane during purification. Cell fractionation has also provided valuable information about the physical characteristics of this membrane, such as partial insolubility in Triton X-100 (5) and a low buoyant density (17). Thus far, these physical properties seem to be common to all caveolae membranes.
Growth factors and their receptors play an important role in brain development and function. Caveolae-like structures have been seen in thin-section images of neuronal N2A cells (18) and nerve endings of the hypothalamus (19). In addition, Triton X-100-insoluble membrane fractions have been isolated from cerebellum (20) and cortical neurons (21). This suggests that a similar membrane domain exists in neurons. We have used a detergent-free method of isolating caveolae from tissue cultured cells (17) to see if neuronal plasma membranes contain a domain with the characteristics of caveolae. Because caveolin is difficult to detect in nerve cells, we have used several different markers to create a physical map between invaginated caveolae, which are easily recognized with the electron microscope in fibroblast (18) or epithelial cells (13), and a low density membrane (LDM) 2 fraction prepared from a synaptic membrane preparation. Our results suggest that neuronal plasma membrane does contain a caveolae-like domain highly enriched in functional tyrosine kinases, as well as other signal-transducing molecules.
ies were a gift from Dr. Luis Parada (The University of Texas Southwestern Medical Center at Dallas). Anti-Hsp70 and anti-BiP monoclonal antibodies were purchased from StressGen (Victoria BC Canada). Anti-PrP polyclonal antibodies were a gift from Dr. Stanley Prusiner (University of California, San Francisco, CA). Anti-P75 NGFR polyclonal antibody was a gift from Dr. Moses V. Chao (Cornell University, NY). Anti-fodrin polyclonal antibodies were a gift from Dr. Peter Michaely (The University of Texas Southwestern Medical Center at Dallas). Anti-insulin receptor ␤ subunit, anti-Lerk2, and anti-Hek7 (used to blot rat Ehk1) polyclonal antibodies were purchased from Santa Cruz Biotechnologies (Santa Cruz, CA). Anti-NR1A and antiphosphotyrosine (4G10) were from UBI (Lake Placid, NY). Cholera toxin subunit ␤ conjugated to horseradish peroxidase was purchased from Sigma. OptiPrep was from Life Technologies, Inc. Chemicals for the 5Ј-nucleotidase and succinate dehydrogenase activity assays were purchased from Sigma and were performed according to the protocols described in Centrifugation Techniques V (NYCOMED PHARMA, Oslo, Norway).
Isolation of Synaptic Plasma Membrane (SPM)-An established protocol was followed to prepare membranes enriched in synaptic plasma membranes from rat forebrains (22). Briefly, 4 -6-week-old Sprague-Dawley rats were decapitated, and the forebrains were removed quickly. Pooled forebrains of six rats were homogenized in ice-cold 0.3 M sucrose using a motor-driven, Teflon glass homogenizer. The homogenate (H) was first centrifuged at 800 ϫ g for 20 min to remove debris, and the resulting supernatant fraction was centrifuged at 9000 ϫ g for 20 min. The crude synaptosome pellet (P1) was washed once in 0.3 M sucrose (P2) and hypotonically lysed in 0.03 M sucrose containing 0.2 mM phenylmethylsulfonyl fluoride, 0.5 g/ml leupeptin, and 1 M pepstatin for 20 min at 4°C. Large pieces of membrane were collected by centrifuging at 25,000 ϫ g for 20 min (P3). The pellet was resuspended in 5 ml dH 2 O and manually homogenized 5-6 times. The sample was then adjusted to 1.1 M sucrose and 5 mM Tris-HCl (pH 7.5) with a total volume of 15 ml before being placed at the bottom of a Beckman SW28 tube. A 15-ml sample of 0.8 M sucrose was overlaid onto the top of the sample, followed by 5 ml of 0.3 M sucrose, all in 5 mM Tris-HCl (pH 7.5). The sample was centrifuged at 19,000 rpm for 2.5 h in a Beckman SW28 rotor. The synaptic plasma membranes were recovered from the 0.8 -1.1 M sucrose interface, washed with dH 2 O, and stored at Ϫ70°C until use. The myelin fraction (MYL) was recovered from the 0.3-0.8 M sucrose interface, and the pellet corresponds to the mitochondria fraction (MIT).
Isolation of Low Density Membrane Domains from SPM-The method used to isolate low density membrane domains was adapted from the method used to isolate caveolae membrane from tissue-cultured cells (17). Approximately 500 g of SPM membrane protein were placed in a TH641 tube and sonicated in 2 ml of buffer A (0.25 M sucrose, 1 mM EDTA, and 20 mM Tricine-NaOH, pH 7.8). The sample was adjusted to 4 ml with 50% OptiPrep in buffer A (final concentration, 23%) and overlaid with 6 ml of a 10 -20% OptiPrep gradient as described previously (14). The sample was centrifuged using a TH641 rotor at 17,500 rpm for 2.5 h. The bottom 5 ml of the gradient were pooled and designated the high density membrane (HDM) fraction. The top 4 ml of the sample were pooled and mixed with 1 ml of buffer A plus 4 ml of 50% OptiPrep and placed in a fresh TH641 tube. A 1-ml sample of 15% OptiPrep was overlaid onto the top of the mixture, followed by 0.5 ml of 5% OptiPrep, all in buffer A. The sample was further centrifuged at 17,500 rpm for 1.5 h using a TH641 rotor. The resulting membrane fraction recovered from the 5-15% OptiPrep was designated the LDM fraction.
Electrophoresis and Western Blotting Analysis-Various samples were analyzed using SDS-PAGE (23). For Western blotting, the Immobilon-P membrane (Millipore) was blocked in 5% nonfat milk in TBS (20 mM Tris-HCl, pH 7.5, 150 mM NaCl). The membrane was first probed with primary antibodies at the dilution suggested by the suppliers, except for anti-P75 NGFR , which was used at 1:5000 dilution. Usually the membrane was incubated with primary antibodies overnight at 4°C and washed with 50 ml of TBS for 5 min, 50 ml of TBST (0.05% Tween 20 in TBS) for 5 min, and 50 ml of TBS for 5 min. Secondary goat anti-rabbit or goat anti-mouse IgGs conjugated to horseradish peroxidase (Cappel, Durham, NC) were diluted in TBS plus 2% nonfat milk and used at room temperature for 1-2 h. Depending on the primary antibodies, the secondary antibodies were diluted from 1:2,000 to 1:30,000. The membrane was washed three times with TBS, TBST, and TBS as described above. The enhanced chemiluminescence (ECL; Amersham Corp.) or SuperSignal (Pierce) was used to visualize the reactive proteins.
Lipid Extraction, Separation, and Detection-The lipids were extracted according to the method of Bligh and Dyer (24). Briefly, various samples were adjusted with dH 2 O to 1 ml and mixed with 1.2 ml chloroform (HPLC grade) and 1.2 ml methanol containing 2% acetic acid (HPLC grade). The samples were vortexed three times (20 s each time) and centrifuged in a table top centrifuge. The organic phase was collected and dried using liquid nitrogen. The sample was redissolved in 40 l of chloroform and spotted onto a TLC plate along with various lipid standards. To visualize sphingomyelin, the lipid samples were separated in a solvent system of chloroform:acetone:methonal:acetic acid:H 2 O (10:4:3:2:1, v/v/v/v) and developed in vaporized I 2 . For cholesterol visualization, the extracted lipid samples were redissolved in 50 l of petroleum ether:ethyl ether:acetic acid (80:20:1, v/v/v) and spotted onto a TLC plate alone with cholesterol standard. The TLC plate was developed in the same solvent system. The plate was then sprayed with sulfuric acid-dichromate. Cholesterol was detected by heating the plate for 5-15 min at 180°C. For visualizing the ganglioside G M1 , the samples were first separated by electrophoresis using a 4 -15% SDS-PAGE gradient gel and transferred to the Immobilon-P membrane. The membrane was incubated in the presence of cholera toxin b subunit conjugated to horseradish peroxidase at 1:30,000 dilution in TBS plus 0.1% bovine serum albumin at room temperature for 1 h. ECL was used to visualize the conjugate.
Electron Microscopy-Samples from purified LDM fractions were fixed with 2% glutaraldehyde in 0.1 M phosphate buffer, pH 7.5, for 10 min. The fixed samples (2-3 l) were air-dried on carbon-coated grids. The grids were washed in the phosphate buffer, postfixed with 1% osmium tetroxide in the phosphate buffer, for 10 min, and stained sequentially for 10 min each with 1% tannic acid, 4% uranyl acetate, and 2% lead citrate.

RESULTS
The starting material for the membrane fractionation procedure was a SPM fraction prepared from rat forebrain by the method of Jones and Matus (22). This method yields seven fractions: brain homogenate (H), first pellet (P1), second pellet (P2), third pellet (P3), myelin (MYL), mitochondria (MIT), and SPM. Fig. 1 shows an analysis of membrane and cytoplasmic markers. The SPM fraction was highly enriched in neuroligins compared to the initial homogenate. This fraction also con- , synaptic plasma membrane (SPM) were prepared as described under "Experimental Procedures." Each fraction was assayed for the plasma membrane marker, 5Ј nucleotidase, and the mitochondria marker, succinate dehydrogenase, using standard enzymatic assays. Other indicated markers were detected by immunoblotting, using 10 g/lane of protein. The various markers used are: neuroligins for the plasma membrane, synaptophysin for the synaptic vesicle, HSP70 for the cytosol, and BiP for the endoplasmic reticulum. Specific activity for 5Ј-nucleotidase is nM P i released/g protein, and for succinate dehydrogenase, relative absorbance (490 nm)/g protein.
tained synaptophysin, but both a cytoplasmic (HSP70) and an endoplasmic reticulum (BiP) chaperon protein were barely detectable. We recovered 78% of the 5Ј-nucleotidase activity in the SPM but only 11% of the mitochondria marker, succinate dehydrogenase. Therefore, the SPM fraction was specifically enriched in plasma membrane markers and relatively devoid of markers for other cellular compartments.
Application of Caveolae Isolation Procedure-We described previously a three-step purification method for preparing a caveolae-rich fraction from tissue culture cells (17). This method relies upon the light buoyant density of caveolae membrane relative to the remainder of the plasma membrane. Although caveolin was the initial marker used to establish that these fractions contained invaginated caveolae seen with the electron microscope (16), a number of additional markers have since been identified (4,8,15,25). We applied this technique to the SPM fraction.
We first sonicated the SPM and loaded the sonicate on the bottom of an OptiPrep gradient ( Fig. 2A, Opti-1, E) before centrifugation for 150 min. Fourteen fractions were collected from the gradient and analyzed for protein content. The protein profile was identical to that obtained with sonicated plasma membranes from fibroblasts (17). The lower seven fractions, which contained the bulk of the protein, were saved and designated HDM, whereas the upper seven fractions were pooled and applied to a second OptiPrep gradient ( Fig. 2A, Opti-2, Ⅺ).
All of the protein in the second gradient was in fraction 2, which is the same fraction in the gradient where fibroblast caveolae collect (17). Approximately 5-10% of the SPM membrane protein was in this LDM fraction. Samples of initial homogenate H (20 g), SPM (10 g), HDM (10 g), and LDM (10 g) were separated by PAGE and stained (Fig. 2B). We identified (arrows) at least 15 bands in the LDM that were either enriched or uniquely present compared to the HDM. A number of bands in the HDM were not present in the LDM, but there also were bands common to the two fractions. This suggests the LDM is a specific membrane domain with the physical characteristics of caveolae membrane.
We used blotting to determine if any of the caveolae protein and lipid markers were enriched in the LDM fraction. None of the anti-caveolin 1 IgGs we tested detected caveolin in the LDM or any other fraction (data not shown). A well established morphologic (15,26) and cell fractionation (17) marker for fibroblast caveolae is G M1 ganglioside, which binds cholera toxin. The entire protein content of each Opti-1 gradient fraction was separated by gel electrophoresis and blotted with cholera toxin (Fig. 3A, Opti-1). The majority of the detectable G M1 was in fractions 1-7. When we combined these fractions and loaded them on an Opti-2 gradient (Fig. 3A, Opti-2), GM1 was highly concentrated in the LDM fraction (fractions 1 and 2). Clusters of prion protein have also been localized by morphology to fibroblast caveolae (18). We found rat brain prion to be present in all of the Opti-1 gradient fractions (Fig. 3B,  Opti-1). Nevertheless, we pooled fractions 1-7, ran them on the Opti-2 gradient, and found prion to be enriched in the LDM fraction (Fig. 3B, Opti-2). In contrast to these markers, nearly all of the detectable PSD 95 (Fig. 3C, Opti-1), NR1A glutamate receptor subunit (Fig. 3D, Opti-1) and fodrin (Fig. 3E, Opti-1) were in the bottom seven fractions of Opti-1 that make up the HDM fraction.
Another characteristic of fibroblast caveolae is that they are enriched in cholesterol (27) and sphingomyelin (28). We used thin layer chromatography to analyze the sphingomyelin (Fig.  4A) and cholesterol (Fig. 4B) content of the various brain fractions. Lipids were extracted from brain homogenate (H), SPM, HDM, and LDM fractions. The extracts were separated and visualized directly. Compared to the SPM, the LDM was clearly enriched in both cholesterol and sphingomyelin. In addition, phosphatidylcholine and phosphatidylethanolamine were en-

FIG. 3. Distribution of marker proteins and glycolipids in Opti-Prep gradients.
Opti-1 and Opti-2 gradients were prepared and fractionated as described. Total proteins in each fraction were precipitated with 10% trichloroacetic acid, washed three times in acetone, air-dried, redissolved in SDS-loading buffer, and separated using a 4 -15% SDS-PAGE. The proteins were transferred onto the Immobilon-P membrane and either blotted with the indicated antibody or with the ␤ subunit of cholera toxin conjugated to horseradish peroxidase (G M1 ). riched in the LDM.
Heterotrimeric G ␣ and G ␤␥ have been localized by immunogold to isolated caveolae prepared from smooth muscle cells (6). In addition, these proteins are clearly enriched in caveolae fractions from human fibroblast plasma membrane (17). Immunoblotting of the Opti-1 fractions from rat brain SPM with specific antibodies against these two proteins (Fig. 5) showed that, like fibroblast plasma membrane, most of the G ␣ (A, Opti-1) and G ␤ (B, Opti-1) subunits were in the top seven fractions. When these fractions were pooled and separated on the Opti-2 gradient, both G protein subunits were concentrated in the LDM fractions (Fig. 5, A and B, Opti-2). We also saw a similar fractionation pattern for Ras (Fig. 5C, Opti-1, -2), which has been found to be involved in recruiting Raf kinase to Rat-1 cell caveolae after epidermal growth factor binding (9).
We analyzed the morphology of these membrane preparations using whole mount electron microscopy (Fig. 6). Numerous vesicle-like structures were seen scattered across the grid. The size of these structures was variable. Of 264 vesicles, ϳ75% had a diameter between 40 and 120 nm, with an average of ϳ109 nm for all vesicles analyzed. We were unable to detect any coat material on these membrane structures.
Tyrosine Kinases Concentrated in LDM Fraction-The epidermal growth factor receptor (9), platelet-derived growth factor receptor (8), and Src kinase (5,8) are tyrosine kinases enriched in fibroblast caveolae. We used immunoblotting to determine if similar kinases were concentrated in the LDM fraction (Fig. 7). Equal protein samples (10 g) of SPM, HDM, and LDM fractions were separated by gel electrophoresis and immunoblotted with anti-insulin receptor ␤ subunit IgG (Fig.  7A). Based on the intensity of the band, the receptor was depleted from the HDM and enriched in the LDM fraction. When fractions were blotted with an antibody directed against the TrkB neurotrophin receptor (Fig. 7B), two intensely reactive bands corresponding in molecular weight to the full-length (TrkB) and truncated (TrkBЈ) TrkB molecules were present in the LDM but not the HDM fraction. Along with TrkB, the low-affinity NGF receptor, p75 NGFR (Fig. 7C), was highly enriched in the LDM fraction. Finally, the Src-related tyrosine kinase, Fyn (Fig. 7D) was concentrated in the LDM fraction.
The Eph-related family of tyrosine kinase receptors are important for neural development (29,30). An unusual feature of the ligands for these receptors is that they are membrane bound (31), in some cases by a glycosylphosphatidylinositol anchor (32). Two members of the family are the receptor Ehk1 and the ligand Lerk2. Immunoblotting showed that both proteins were highly enriched in the LDM fraction (Fig. 7, E and F).
Tyrosine kinases use multiple intermediate molecules to transmit signals between cellular compartments (33). We have already shown that Ras is highly enriched in the LDM (Fig.  5C). Immunoblotting showed that Grb2 (Fig. 7G) and Erk2 (Fig. 7H) were also concentrated in the LDM. In experiments not shown, we found protein kinase C␣, calmodulin, and phosphatidylinositol 3Ј-kinase to be enriched in LDM, although these intermediates were in the HDM as well. Each of these molecules have previously been found to be concentrated in preparations of fibroblast caveolae (8).
To confirm the localization of tyrosine kinases in the LDM, we looked for the presence of kinase substrates using an antibody that recognizes phosphotyrosine residues (Fig. 8). Fractions were prepared in the presence of phosphatase inhibitors to preserve any endogenously phosphorylated substrates that might be present (Fig. 8). As long as the fractions were maintained at 4°C, however, we did not detect any phosphorylated proteins (0 min). After as little as 5 min at 37°C, we detected numerous phosphotyrosine-containing proteins in the SPM and the LDM (5 min). No phosphorylated proteins could be detected in the HDM. The staining intensity increased in both SPM and LDM with time at 37°C (30 min), but still no bands appeared in the HDM fraction (Fig. 8). At all times, the intensity of the bands was greater in the LDM than the SPM. The appearance of these substrates did not require the addition of receptor ligands to the fractions (data not shown). This suggests that active tyrosine kinases together with multiple substrates are organized in the LDM.  5. Distribution of heterotrimeric G ␣ (A) and G ␤ (B) and Ras (C) in fractions from OptiPrep gradients. Opti-1 and Opti-2 were prepared, fractionated, and immunoblotted as described in Fig. 3.

FIG. 6. Whole-mount electron microscopic analysis of LDM.
LDM was fixed, applied to a carbon-coated electron microscopy grid, stained, air-dried, and viewed with a JOEL 100-CX electron microscope. Bar, 0.3 m.

DISCUSSION
We have prepared a membrane fraction from synaptic plasma membrane that shares many properties with fibroblast caveolae. The main difference between the two is the lack of detectable caveolin 1. Several other isoforms of caveolin have already been identified (34,35), but none of them seem to be present in the brain. The absence of caveolin does not necessarily mean that the LDM is not structurally and functionally related to caveolae. Caveolin appears to be involved in shut-tling cholesterol between the endoplasmic reticulum and the plasma membrane (36) and in organizing caveolae proteins such as heterotrimeric GTP-binding proteins (37). Other proteins might perform these functions in nerve cells. Alternatively, neurons may express an isoform of caveolin that is not detected by our antibodies.
Only recently have methods for isolating caveolae membrane been available; therefore, little is known about how the chemistry, structure, and function of this membrane differs among cell types. We cannot be sure, therefore, that the LDM fraction rightfully deserves the designation "caveolae." The current results are the starting point for determining what similarities exist between the LDM and the caveolae membrane found in tissues and tissue culture cells. Thus far, we have obtained a similar membrane fraction from every type of cell we have tested (data not shown). Combined with structural methods, it should now be possible to elucidate the range of functions carried out by membranes with a common set of physical and molecular characteristics.
The first class of signaling molecules found to be shared by both LDMs and fibroblast caveolae are the tyrosine kinases. Platelet-derived growth factor receptors have been mapped by immunocytochemistry to caveolae (8), and cell fractionation has identified ligand-specific signaling activities for both platelet-derived growth factor (8) and epidermal growth factor (9) in caveolae membranes. Furthermore, insulin receptors are concentrated in caveolae-like invaginated membrane (38) as well as caveolae fractions (data not shown) of fibroblasts. Caveolinrich, Triton X-100-insoluble complexes (39) and isolated, detergent-free, caveolae membrane (8) are enriched in Src family kinases. We found both nonreceptor and receptor kinases to be highly enriched in LDMs. These kinases also appear to be active in isolated LDMs because only this fraction was able to phosphorylate resident substrates on tyrosine. Finally, both LDMs and fibroblast caveolae are enriched in adapter molecules known to function as intermediates in tyrosine kinase signal transmission (33). The LDMs should be a useful starting point for further studying how tyrosine kinases function in adult and developing nervous tissue.
Implications of Tyrosine Kinase Localization-An emerging area of cell regulation research is the study of combinatorial interactions between multiple ligands and receptors during kinase activation. Trk/p75/neurotrophin (40,41), CNTFR␣/ CNTF/LIFR␤/gp130 (42,43), GDNF/GDNFR␣/Ret (44), and Eph/Lerk (30,31) are four different molecular cassettes that encode information for intracellular kinase intermediates. The code is generated by the pattern of interactions among the molecular units of each cassette. Three types of molecules are in each cassette; soluble, transmembrane, and glycosylphosphatidylinositol-anchored. The kinase activating molecule is always transmembrane, whereas the ligands can either be soluble or membrane-anchored. Many times the latter are anchored by glycosylphosphatidylinositol. If LDMs are caveolae, they can do two things that might critically influence the interactions among members of the cassette: concentrate lipidanchored membrane proteins (13,14) as well as glycolipids (15); and sequester molecules away from the extracellular space (3). The presence of a lipid anchor (either glycosylphosphatidylinositol (45) or fatty acid (39,46)) automatically targets a ligand or ligand adapter to caveolae, placing the molecule in a membrane environment favorable for integrating information from other regions of the extracellular and intracellular space (7). The clustering will also help modulate interactions among members of a cassette. For example, Eph ligands do not stimulate their respective receptor unless they are membrane-bound or clustered (47); therefore, caveolae could The samples were first incubated on ice for 30 min and then shifted to 37°C for the indicated times. The reaction was terminated by adding ice-cold TCA to a final concentration of 10%. Total proteins were recovered, separated in a 4 -15% SDS-PAGE, and immunoblotted with anti-phosphotyrosine IgG. ECL was used to visualize the results. regulate kinase activation by controlling ligand clustering. Molecular sequestration, on the other hand, has several possible functions: (a) create an environment (e.g. low pH (48)) that modulates the outcome of ligand/receptor interaction; (b) hold the ligand/receptor complex in either an active or inactive state for prolonged periods of time; (c) promote the covalent modification of cassette molecules by local enzymes.
Finding both p75 and Trk to be enriched in LDM suggests how this domain might function in signal integration. In this location, these two receptor types are favorably positioned to interact with each other as well as lipid intermediates during signal transduction. Both receptors are known to interact immediately after exposure to neurotrophins, while mediating high affinity ligand binding (41). These interactions will occur more efficiently as a consequence of being in a common location at the cell surface. Just as importantly, this organization can coordinate interactions among receptors with the release of lipid second messengers such as ceramide (40). Like caveolae, LDM appear to be enriched in sphingomyelin. Fibroblast caveolae are known to be a site where sphingomyelin is converted to ceramide (28), and the release of ceramide regulates plateletderived growth factor receptor function. NGF binding to p75 also stimulates ceramide formation (49,50), and this may occur in LDMs. Ceramide is an active intermediate in apoptosis (51). Interestingly, neural cells transfected with p75 constitutively undergo cell death (52).
The LDM is a likely target for certain diseases. There are two of particular interest: Alzheimer's and prion diseases. A pool of amyloid precursor protein exists in a caveolae-like, detergentresistant membrane fraction with many of the same properties as the LDM (21). This raises the possibility that the release of amyloidogenic ␤A4 peptide occurs in the LDM, thus contributing to the formation of amyloid. Amyloid might have a direct effect on the molecular organization of the LDM, leading to aberrant signal transduction from this site. A similar phenomenon may be involved in prion disease. We found prions to be enriched in the LDM. They have also been localized to fibroblast caveolae membrane (18), Triton X-100-insoluble membrane (53), and a caveolae-like membrane domain from N2A cells (18,54). If infectious prion formation occurs in the LDM (53), then the presence of malfolded prions could easily disrupt signal transduction.
The membrane fraction we have prepared is highly enriched in tyrosine kinases and associated molecules. Regardless of whether the LDM corresponds to caveolae, the ability to retrieve a membrane domain containing an organized array of these molecules offers the opportunity to study tyrosine kinase signal transduction in a format that more closely mimics how they are found in the cell. Tyrosine kinases have an important role to play during neuronal development (41,55) and in nerve plasticity (56); therefore, the LDM fractions may yield new insights into how kinases control these activities. The existence of LDMs indicates signal transduction by tyrosine kinases is just as highly organized as chemical transmission by neurotransmitters.