Localization of epidermal growth factor-stimulated Ras/Raf-1 interaction to caveolae membrane.

An essential step in the epidermal growth factor (EGF)-dependent activation of MAP kinase is the recruitment of Raf-1 to the plasma membrane. Here we present evidence that caveolae are the membrane site where Raf-1 is recruited. Caveolae fractions prepared from normal Rat-1 cells grown in the absence of serum were highly enriched in both EGF receptors and Ras. Thirty seconds after EGF was added to these cells Raf-1 began to appear in caveolae but not in non-caveolae membrane fractions. The maximum concentration was reached at 3 min followed by a decline over the next 60 min. During this time EGF receptors disappeared from the caveolae fraction while the concentration of Ras remained constant. The Raf-1 in this fraction was able to phosphorylate MAP kinase kinase, whereas cytoplasmic Raf-1 in the same cell was inactive. Elevation of cellular cAMP blocked the recruitment of Raf-1 to caveolae. Overexpression of Ha-Ras caused the recruitment of Raf-1 to caveolae independently of EGF stimulation, and this was blocked by the farnesyltransferase inhibitor BZA-5B. Finally, prenylation appeared to be required for localization of Ras to caveolae.

The binding of EGF 1 to its receptor initiates a kinase cascade that results in the activation of MAP kinase (1)(2)(3). A key intermediate in this cascade is the GTP binding protein Ras. Ras appears to be a molecular switch (4) that controls the recruitment of Raf-1 to the plasma membrane after EGF binding. At the membrane, Raf-1 is activated and becomes available to phosphorylate MAP kinase kinase, the next kinase in the cascade. A direct interaction between Ras and Raf-1 appears to be required for recruitment to the plasma membrane (5)(6)(7).
The Raf-1 that comes to the cell surface after EGF binding may be randomly distributed on the membrane or localized to a specific membrane domain. Recently we reported that EGF receptors are highly enriched in caveolae-rich fractions from human fibroblasts (8). Previous work (9) suggests that ϳ5% of the EGF receptors in A431 cells, which are high affinity receptors, are in a Triton X-100 insoluble complex. Moreover, the binding of EGF stimulates the appearance of Raf-1 in a Triton X-100 insoluble fraction of plasma membrane (10). Triton X-100 insolubility is a well established biochemical characteristic of caveolae (11,12). These observations raise the possibility that EGF bound to receptors in caveolae stimulates the recruitment of Raf-1 to this domain.
Caveolae were first identified in thin section, electron microscopic images of gallbladder epithelial cells (13) and endothelial cells (14). They have a characteristic membrane coat that can only be seen with the scanning electron microscope (15) or by rapid-freeze, deep-etch electron microscopy (16). A striking feature of the coat is that it decorates flat segments of membrane as well as membrane in different stages of invagination. Only the highly invaginated portions of membrane can be recognized as caveolae in thin section images of the cell. The variable morphology of this coated membrane suggests that caveolae, like clathrin-coated pits (17), are able to internalize molecules by forming buds that eventually seal off from the plasma membrane. Indeed, biochemical (18 -21), morphological (22,23), pharmacological (24), and molecular biology experiments (25) have documented the delivery of 5-methyltetrahydrofolate to the cytoplasm of MA 104 cells by a caveolae-dependent, receptor internalization cycle.
A different perspective of caveolae has emerged from studying membrane fractions enriched in caveolae membrane (26). Relative to the plasma membrane these fractions contain a high concentration of signal transducing molecules (8,12,27), receptors that use these transducers during cell signaling (28 -31), and lipids known to be intermediates in cell signaling (32). In the current study we use cell fractionation to gather evidence that the initial events in EGF-stimulated Ras/Raf-1 interaction also occur in this compartment. The presence of all these molecules at a specific location on the cell surface suggests that multiple signaling pathways originate from this membrane domain.
Antibodies were obtained from the following sources. Anti-caveolin IgG (mAb) was a gift from Dr. J. Glenney (Glentech, Inc., Lexington, KY); anti-Mek-1 IgG (mAb), anti-Raf-1 IgG (mAb), anti-Ha-Ras IgG (mAb), anti-Sos-1 IgG (mAb), and anti-Grb2 IgG (monoclonal IgG) were from Transduction Laboratories (Lexington, KY). Anti-14 -3-3 polyclonal IgG was a gift from Dr. Pallas (Harvard Medical School, Boston). Anti-Raf-1 rabbit polyclonal antibody was from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA). Anti-pan Ras IgG (mAb) was from Oncogene Science (Uniondale, NY), and anti-EGF receptor IgG (mAb) was from Calbiochem. * This work was supported by National Institutes of Health Grants HL 20948, GM 43169, and GM 15631 and the Perot Family Foundation. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
¶ Supported by a postdoctoral fellowship from the Helen Hay Whitney Foundation.

Methods
Cell Culture-Rat-1 cells (6 ϫ 10 5 ) were seeded in T-75 culture flasks and grown in 10 ml of DMEM supplemented with 10% (v/v) fetal calf serum for 4 days. Cells were then incubated for 24 -48 h in DMEM that did not contain serum before EGF stimulation. Rat-1 cells overexpressing oncogenic Ha-Ras V12 were described previously (33). These cells were seeded on day 0 at a density of 2 ϫ 10 5 cells per 100-mm dish in 10 ml of DMEM supplemented with 10% (v/v) fetal calf serum and 500 g/ml G418. Cells were fed again on days 1 and 3 with 7.5 ml/dish of medium that contained 100 M BZA-5B in 1% DMSO plus 100 M DTT or DMSO plus DTT alone (33).
Preparation of Caveolae Fraction-Detergent-free caveolae fractions were prepared by the method of Smart et al. (8). All steps were carried out at 4°C. Plasma membranes were isolated by scraping cells from the dish in 5 ml of ice-cold buffer A (0.25 M sucrose, 1 mM EDTA, 20 mM Tricine, pH 7.8) and pelleting them at 1,400 ϫ g for 5 min. Cells were resuspended in 1.0 ml of buffer A and homogenized 15 times with a tight Dounce homogenizer. The postnuclear supernatant fraction obtained by centrifugation at 1,000 ϫ g for 10 min was layered on top of 23 ml of a 30% Percoll solution prepared in buffer A and centrifuged at 84,000 ϫ g for 30 min in a Beckman Ti60 rotor. The plasma membrane band was collected and the volume adjusted to 2.0 ml with buffer A. The volume of the plasma membrane fraction was adjusted to 2.0 ml with buffer A and sonicated 6 times with a Vibra Cell sonicator (Sonics & Materials, Danbury, CN). The sonicate was mixed with 1.84 ml of 50% OptiPrep prepared in buffer B (0.25 M sucrose, 6 mM EDTA, 120 mM Tricine, pH 7.8) plus 0.16 ml of buffer A to make a 23% OptiPrep solution. This was placed on the bottom of a Beckman SW41 centrifuge tube, and a linear 20 to 10% OptiPrep gradient (prepared by diluting 50% OptiPrep in buffer B with buffer A) was layered on the top. The sample was then centrifuged at 52,000 ϫ g for 90 min in a SW41 Beckman swinging bucket rotor. A sample of the bottom fraction (fractions 12 and 13) was collected and designated non-caveolae membrane. The top 5 ml of the gradient (fractions 1-6) was collected and mixed with 50% OptiPrep in buffer B. This was overlaid with 2 ml of 5% OptiPrep in buffer A and centrifuged at 52,000 ϫ g for 90 min. An opaque band located just above the 5% interface was collected and designated caveolae fraction.
Protein Determination-Protein concentrations were determined using Bio-Rad DC assay for the samples that contain detergent and a Bio-Rad Bradford assay for the samples that contain OptiPrep.
Electrophoresis and Immunoblots-Each sample was concentrated by trichloroacetic acid precipitation and washed in acetone. Pellets were suspended in Laemmli sample buffer (34), heated at 95°C for 3 min, and loaded onto 12.5% SDS-polyacrylamide gel using the method of Laemmli (34). The separated proteins were transferred to nylon by the method of Towbin et al. (35). The nylon was blocked in buffer C (20 mM Tris, pH 7.6, 137 mM NaCl, 0.5% Tween 20) containing 5% dry milk for 1 h at room temperature. Primary antibodies were diluted in buffer C containing 1% dry milk and incubated with the nylon samples for 1 h at room temperature. The nylon was washed 4 times for 10 min each in buffer C plus 1% dry milk and incubated with the appropriate horseradish peroxidase-labeled anti-IgG for 1 h at room temperature. The nylon was then washed and the bands visualized using the enhanced chemiluminescence method. Nylon is the only transfer material that will bind caveolin for immunoblotting.
Raf-1 Kinase Assay-Raf-1 kinase activity was measured as described previously (33). Raf-1 was immunoprecipitated as described above. The protein A-agarose beads were first washed with buffer F (20 mM Tris-HCl, pH 7.5, 0.14 M NaCl, 25 mM ␤-glycerophosphate, 1% v/v Triton X-100, 2 mM EDTA, 2 mM sodium pyrophosphate, 1 mM sodium orthovanadate, 1 mM phenylmethylsulfonyl fluoride, and 10% glycerol) followed by two washes with 50 mM NaHepes, pH 8.0. The washed beads were resuspended in a final volume of 45 l in buffer G (25 mM NaHepes, pH 7.4, 25 mM ␤-glycerophosphate, 10 mM MgCl 2 , 1 mM DTT) containing 1.5 g of purified recombinant kinase-inactive MAP kinase kinase-1. Five l of buffer H (50 mM NaHepes, pH 8.0, 1 mM ATP, and 5 mCi/ml [␥-32 P]ATP) was added and the mixture incubated at room temperature for 30 min. After centrifugation for 15 s at 12,000 ϫ g at 4°C, the supernatant fraction was immediately mixed with the sample buffer and separated by polyacrylamide gel electrophoresis using 12.5% gels. The gels were fixed and dried before the radioactive intensity of the MAP kinase kinase band (46 kDa) was measured using a Molecular Dynamics PhosphorImager (Sunnyvale, CA).

RESULTS
To determine if Ras was in caveolae we used a detergent-free method for preparing caveolae-rich fractions (8) from cells that had been incubated in the presence (ϩEGF) or absence (ϪEGF) of EGF (Fig. 1). We set up a standard immunoblot assay to measure the relative concentration of different proteins. Samples of whole cell lysate (lanes 1 and 7, 75 g/lane), postnuclear supernatant (lanes 2 and 8, 5 g/lane), plasma membrane (lanes 3 and 9, 5 g/lane), cytosol (lanes 4 and 10, 5 g/lane), non-caveolae plasma membrane fraction (lanes 5 and 11, 5 g/lane), and caveolae fraction (lanes 6 and 12, 5 g/lane) were separated by electrophoresis and immunoblotted with either anti-caveolin IgG (Caveolin), anti-Raf-1 IgG (Raf-1), or anti-Ha-Ras IgG (Ras). All three proteins could be detected in 75 g of whole cell lysate from both control and EGF-treated cells (lanes 1 and 7). Caveolin was highly enriched in the caveolae fraction compared with the plasma membrane (compare lanes 6 with lanes 3 and 5), and the intensity of the band was about the same in EGF-treated cells (compare lanes 6 with 12). Exactly the same pattern was seen for Ha-Ras. Relative to the plasma membrane, we only detected Ha-Ras in the caveolae fraction (compare lanes 6 with lanes 3 and 5). We also found that K-Ras was concentrated in the caveolae fraction (data not shown). When we blotted with an anti-Raf-1 IgG, very little Raf-1 was detected in the caveolae fraction from control cells (lane 6). EGF-treated cells, by contrast, had a strong Raf-1-specific band in this fraction (lane 12). EGF did not stimulate the appearance of Raf-1 in either the non-caveolae fraction (compare lanes 5 and 11) or the cytosol (compare lanes 4 and 10). These results indicate that Raf-1 is recruited to the caveolae fraction and show Ras to be highly enriched in this fraction as well. Similar results were obtained using a standard Triton X-100 insolubility protocol for purifying caveolae (12,36) with the exception that Ras was not retained in these caveolae (data not shown).
GTP-Ras is the active form that recruits Raf-1 to the plasma membrane (5,37,38). Nucleotide binding to Ras, in turn, is controlled by the nucleotide exchange factor SOS-1 (39). SOS-1 also appears to be recruited to the membrane in response to EGF, in a complex with Grb2 (1, 40 -42). We found both Grb2 and SOS-1 to be present in the caveolae fraction of unstimulated cells (Fig. 2, Grb2 and SOS-1, lane 6). Grb2 appeared to be enriched in this fraction relative to the plasma membrane fraction (Grb2, compare lane 6 with lane 4) whereas the concentration of SOS-1 was about the same in both fractions (SOS-1, compare lane 6 with lane 4). After 3 min of EGF treatment, however, the concentration of both proteins was increased specifically in the caveolae fraction. The amount of Raf-1 in caveolae fraction was also increased (Raf-1, compare lane 6 and 12), but Ras was unchanged (Ras, compare lanes 6 and 12).
As we showed previously (8), EGF receptors were highly enriched in the caveolae fraction of unstimulated cells (Fig. 2, EGF, lane 6). Receptors could not even be detected in 75 g of whole cell lysate (EGF, lane 1). Surprisingly, EGF receptors were no longer detectable in the caveolae fraction after cells were exposed to EGF for 3 min (EGF, compare lane 6 with 12).
Another group of proteins that have been implicated in Raf-1 activation are the 14 -3-3 proteins (43,44). These proteins were not present in the caveolae fractions from unstimulated cells (Fig. 2, 14-3-3, lane 6), although they were detected in the non-caveolae membrane and the cytosol fractions (14 -3-3,  lanes 5 and 3). EGF caused the appearance of 14 -3-3 proteins in the caveolae fraction (14 -3-3, lane 6) but the intensity of the band was the same as the band in the plasma membrane fraction (14 -3-3, lane 4). Therefore, 14 -3-3 proteins may not be enriched in the caveolae fraction relative to the plasma membrane after exposure to EGF but are, nevertheless, available in caveolae for interaction with Raf-1.
The caveolae purification procedure begins with the separation of light membrane components of the plasma membrane from the bulk of the plasma membrane protein by density gradient centrifugation. These light membranes are then further purified and concentrated on a second gradient to obtain the final caveolae fraction (8). Any caveolae-associated protein that is exclusively in this domain should only be present in the light membrane fractions from the first gradient. To determine the membrane distribution of Ras, Raf-1, and the EGF receptor, plasma membranes from Rat-1 cells were purified on Percoll gradients, sonicated, and separated on the first OptiPrep gradient (see "Methods"). The entire protein content of each fraction (Fig. 3B) was loaded on gels, separated by electrophoresis, and transferred to nylon for immunoblotting with the indicated antibody (Fig. 3A). Fig. 3 shows that Ras (Ras) and caveolin (Caveolin) could only be detected in the light membrane fractions (lanes 1-9). Before EGF stimulation, a small amount of Raf-1 was detected in the bottom fractions but none was seen in the light fractions (Raf Ϫ EGF). Exposure to EGF dramatically increased the concentration of Raf-1 in the light plasma membrane fractions (Raf ϩ EGF, lanes 1-9) whereas barely changing the amount in the bottom fractions (lanes 10 -14). The light membrane fractions also contained all of the detectable EGF receptors in cells grown in the absence of EGF for 24 h (EGF Receptor).
In contrast to these proteins neither integrin nor receptorbound transferrin was enriched in caveolae. Virtually all of the detectable integrin in the plasma membrane was in the heavy  9 -14). Furthermore, Rat-1 cells incubated in the presence of 125 I-transferrin for 1 h at 4°C, washed, and processed to isolate caveolae had 103,333 dpm of radioactivity in the starting plasma membrane, 73,860 dpm (71%) in the non-caveolae membrane fraction, and 2236 dpm (2%) in the caveolae fraction from the second gradient. These results are consistent with our previous finding that transferrin receptors are not concentrated in the caveolae fraction (8).
If caveolae are a specific site for Raf-1 recruitment, then kinase activation should take place at this location (45,46). An immune complex kinase assay (33) was used to measure the effects of EGF on Raf-1 kinase activity (Fig. 4, Kinase Activity) in the cell lysate (lanes 1 and 4), non-caveolae membrane (lanes 2 and 5), and caveolae membrane (lanes 3, 6, and 7) fractions from Rat-1 cells. The relative quantity of Raf-1 in each fraction was determined by immunoblotting (Fig. 4, Immunoblot). In the absence of EGF (lanes 1-3), all of the immune complexes had a basal kinase activity that was equal to the activity observed when a nonimmune antibody was used (compare with lane 7). EGF stimulated the kinase activity 4.8-fold in the caveolae fraction (compare lanes 3 with 6) but did not stimulate activity in the non-caveolae fraction (lane 5) nor in the cytoplasmic fraction (data not shown). At the concentrations of protein used for the immunoprecipitation, we also could not detect an increase in activity in the cell lysate (lane 4). Immunoblotting showed that the amount of Raf-1 in immunoprecipitates of the cell lysate changed little after EGF treatment although there was a slight increase in the amount of Raf-1 in the non-caveolae fraction. By contrast, EGF stimulated a dra-matic increase in the amount of Raf-1 in the caveolae fraction (compare lane 3 with lane 6). These results indicate that the Raf-1 retrieved from the caveolae fraction by immunoprecipitation was active. Fig. 5A shows the kinetics of Raf-1 recruitment to the caveolae fraction. Cells were incubated in the presence of EGF for the indicated time before the fractions were isolated. The caveolae fraction was separated by gel electrophoresis and immunoblotted with anti-Raf-1 IgG (Raf-1) and anti-Ha-Ras IgG (Ras). Initially we could not detect any Raf-1 in the fraction (lane 1). Within 30 s after the addition of EGF (lane 2), Raf-1 began to appear. The concentration of Raf-1 peaked at 3 min (lane 3) and then gradually declined over the next 57 min (lanes 4 -6). Very little Raf-1 was in the fraction after 60 min (lane 6). Throughout this time period, the amount of Ras in the caveolae fraction remained unchanged (Ras, lanes 1-6).
At the same time as the Raf-1 was recruited to the caveolae fraction, the EGF receptor disappeared (Fig. 5B). Cells were incubated for the indicated time in the presence of EGF before caveolae fractions were prepared and immunoblotted with anti-EGF receptor IgG. Within 30 s after the addition of EGF to the medium, the concentration of receptor in the fraction began to decline (compare lane 0 with lane 0.5). Receptor was no longer detected after 60 min of incubation (lane 60).
Elevation of cAMP has recently been shown to inhibit the transmission of growth stimulating signals through the Ras pathway (47,48). Fig. 6 shows that both 8-bromo-cAMP and forskolin blocked the recruitment of Raf-1 to the caveolae fraction. Cells that were not incubated in the presence of EGF had a basal amount of Raf-1 in the caveolae fraction (lane 1), but after EGF treatment for 3 min, the concentration of Raf-1 increased (lane 2). This increase in Raf-1 was completely prevented either by raising the endogenous level of cAMP (lane 3) or by adding an exogenous source of cAMP (lane 4). Neither of these treatments had any effect on the concentration of Ras in the fraction.
Since Raf-1 binds to GTP-Ras, cells expressing an oncogenic Ha-Ras that cannot hydrolyze GTP should constitutively have Raf-1 present in the caveolae fraction. Cells overexpressing Ha-Ras V12 , which cannot hydrolyze GTP (49), were grown in the presence or absence of the farnesyltransferase inhibitor BZA-5B for 3 days (Fig. 7A). BZA-5B reverses the transformed phenotype of the cell by preventing the farnesylation of onco-

FIG. 4. EGF stimulates Raf-1 activation in caveolae fraction.
Rat-1 cells were grown for 24 h in the absence of serum and incubated in the presence (lanes 4-7) or absence (lanes 1-3) of 50 ng/ml EGF for 3 min. Samples of whole cell lysate (100 g), noncaveolae membrane fraction (10 g), and caveolae membrane fraction (10 g) were prepared from Rat-1 cells as described in Fig. 1 and immunoprecipitated on agarose beads as described previously (33) using either an anti-Raf-1 IgG (lanes 1-6) or a nonimmune IgG (lane 7). One set of immunoprecipitates was assayed for Raf-1 kinase activity (Kinase activity) as described previously (33), and the density of the radioactive bands quantified on a PhosphorImager (Molecular Dynamics). Another sample of immunoprecipitates was separated by gel electrophoresis and immunoblotted (Immunoblot) with mAb anti-Raf-1 IgG (lanes 1-7). genic Ha-Ras (50). We then prepared the indicated fractions (lanes 1-12) from both sets of cells and separated the proteins by gel electrophoresis. Samples were immunoblotted with anti-Raf-1 IgG and anti-Ha-Ras IgG using either a 3-min or a 5-s exposure time. Large amounts of Ras were in the caveolae fraction (Ras, lanes 1-6) of control cells, although it was also present in the non-caveolae membrane fraction (Ras, lane 4). By contrast, Raf-1 was not detected in either the plasma membrane (Raf-1, lane 4) or the non-caveolae membrane fractions (Raf-1, lane 5) but was present in the caveolae fraction (Raf-1, lane 6). When cells were grown in the presence of BZA-5B (ϩBZA-5B, lanes [7][8][9][10][11][12], two Ras bands were observed in the cell lysate and the postnuclear supernatant fractions (Ras, lanes 7 and 8, 5-s exposure). These correspond to the nonprenylated Ha-Ras (upper band) and the farnesylated Ha-Ras (lower band) (33,50). Only the farnesylated Ras was present in the caveolae fraction (5-s exposure, lane 12), indicating prenylation is required for localization to caveolae. Moreover, the concentration of Ras in the caveolae fraction was reduced compared with control cells (compare lanes 6 and 12). There also was significantly less Raf-1 in the caveolae fraction (compare Raf-1 in lanes 6 and 12) of BZA-5B-treated cells. As an additional control (Fig. 7B), we compared the concentration of Raf-1 in the caveolae fraction of cells expressing either wild type Ha-Ras (lane 1), Ha-Ras V12 (lane 2), or pp60 src kinase (lane 3). Even though the amount of Ras in the caveolae fractions from both wild type and Ha-Ras V12 expressing cells was the same (Ras), Raf-1 (Raf-1) was only detected in the caveolae from cells expressing Ha-Ras V12 . Raf-1 was not present in the caveolae fraction of cells overexpressing pp60 src kinase. These results suggest that Raf-1 binds to GTP-Ras in caveolae and that BZA-5B reverses the effects of Ha-Ras V12 -induced transformation by reducing the amount of Raf-1 in this membrane domain. DISCUSSION We have detected a highly organized population of EGF receptors in normal Rat-1 cells. Rather than being randomly distributed on the surface, most of these receptors appear to be concentrated, together with Ras, in a highly select piece of membrane that has the biochemical properties of caveolae. The available antibodies could not detect these molecules by immunocytochemistry, although previously we localized the Rasrelated protein, Rap1A/B, in isolated caveolae (27). Nevertheless, each of these proteins was found primarily in the light membrane fraction from the first gradient (Fig. 2), indicating that caveolae are most likely the preferred location on the surface of unstimulated cells. The binding of EGF leads to the recruitment of Raf-1 to this membrane where the kinase becomes active. Kinase activation may require a molecular modification or a co-factor that can only be acquired at this location. Termination of Raf-1 recruitment may occur in part by the migration of receptors out of this membrane domain after ligand binding. These migrant receptors would no longer be associated with Ras and, therefore, unable to continue activating the GTPase.
Some cells express both high affinity and low affinity EGF receptors (51). Caveolae may be the preferred location for high affinity EGF receptors. Inactivation of the high affinity receptors with a specific monoclonal antibody prevents early events in EGF signal transduction (52). The high affinity population appears to be stabilized by oligomerization (53) and sphingolipids (54). EGF can stimulate oligomerization in vivo (55). Time-resolved fluorescence imaging microscopy has recently found, however, that up to 12% of the EGF receptors in A431 cells are oligomerized in the absence of EGF (56). These preoligomerized receptors appear to be the high affinity population in these cells. Our results suggest that caveolae are able to concentrate EGF receptors. The clustering of receptors most likely would maximize receptor-receptor interactions that favor oligomer formation. The high concentration of sphingomyelin in caveolae (32) should stabilize these oligomers and thereby maintain them in the high affinity state.
One way to view the results of this study is that caveolae FIG. 6. cAMP blocks the EGF-stimulated recruitment of Raf-1 to the caveolae fraction. Rat-1 cells were grown for 24 h in the absence of serum. Cells were either not treated (lanes 1 and 2), incubated for 15 min in the presence of 50 M forskolin (lane 3), or incubated in the presence of 500 M 8-bromo-cAMP for 10 min (lane 4). The indicated dishes then received 50 ng/ml EGF and were further incubated for 3 min. At the end of the incubations, caveolae fractions were prepared as described in Fig. 1, and 5 g of protein was separated by gel electrophoresis before immunoblotting with either mAb anti-Raf-1 IgG or mAb anti-Ras IgG.  lanes 1 and 7), postnuclear supernatant (5 g, lanes 2 and 8), cytosol (5 g, lanes 3 and 9), plasma membrane (5 g, lanes 4 and 10), noncaveolae plasma membrane fraction (5 g, lanes 5 and 11), and caveolae fraction (5 g; lanes 6 and 12) were separated by gel electrophoresis and immunoblotted with either anti-Raf-1 IgG or anti-Ras IgG. Immunoblots were exposed for either 3 min or 5 s. B, caveolae fractions from Rat-1 cells expressing either Ha-Ras (lane 1), Ha-Ras V12 (lane 2), or pp60 src kinase (lane 3) were prepared and immunoblotted with antibodies against either Raf-1 (Raf-1) or Ras (Ras). function like a signaling depot. Signal transducing molecules such as Ras, PKC␣ (36,57), nonreceptor tyrosine kinases (12,58), and heterotrimeric GTP binding proteins (8,12,27) reside in this location waiting for input from migrating receptors or other stimulants. When a receptor such as the EGF receptor moves into a caveola, Ras-mediated signal transduction can occur if EGF is present. From this location the receptor can also interact with other resident transducing molecules to generate complex signaling patterns that affect cell behavior. Superimposed on these patterns are the modulatory influences of locally produced lipid signaling molecules such as ceramide (32). In addition, the caveolae internalization cycle may influence the signal output from these receptors. For example, initial signaling by the EGF receptors in the current study probably came from caveolae that were open at the cell surface because the receptors rapidly migrated away after ligand binding. If the receptors had instead become sequestered in plasmalemmal vesicles after ligand binding, as occurs with the cholecystokinin receptor (59), then activation of Raf-1 might have continued. Nutrients such as folate (22) that are concentrated and internalized by caveolae may also influence the signal from these receptors, providing the cell with important information about the nutritional environment. Caveolae appear to be a versatile site where signal transduction is organized at the cell surface.