Disruption of the Raf-1-Hsp90 molecular complex results in destabilization of Raf-1 and loss of Raf-1-Ras association.

Cytosolic Raf-1 exists in a high molecular weight complex with the heat shock protein Hsp90, the purpose of which is unknown. The benzoquinone ansamycin, geldanamycin, specifically binds to Hsp90 and disrupts certain multimolecular complexes containing this protein. Using this drug, we are able to demonstrate rapid dissociation of both Raf-1-Hsp90 and Raf-1-Ras multimolecular complexes, concomitant with a markedly decreased half-life of the Raf-1 protein. Continued disruption of the Raf-1-Hsp90 complex results in apparent loss of Raf-1 protein from the cell, although Raf-1 synthesis is actually increased. Prevention of Raf-1-Hsp90 complex formation interferes with trafficking of newly synthesized Raf-1 from cytosol to plasma membrane. These data indicate that association with Hsp90 is essential for both Raf-1 protein stability and its proper localization in the cell.

Raf-1, a serine/threonine kinase, is part of a highly conserved kinase cascade that mediates signaling by extracellular growth factors and leads to the stimulation of mitogen-activated protein kinases (1,2). Raf-1 functions downstream of Ras, which in its active, GTP-bound state binds directly to the amino-terminal regulatory domain of Raf-1 (3). This interaction is transient and apparently serves to recruit Raf-1 to the cell membrane (4,5), a step that is necessary for Raf-1 activation. The requirement for Ras can be bypassed by coupling a plasma membrane targeting signal to Raf-1 (6). Following its recruitment by Ras, Raf-1 associates with cytoskeletal components via an unknown mechanism (7).
Although activated Raf-1 is plasma membrane-associated, this kinase is primarily cytosolic in location and exists in a native heterocomplex with the heat shock proteins Hsp90 and p50 (8). Hsp90 is an ubiquitously expressed molecular chaperone that has been found in complexes with a variety of proteins including steroid hormone receptors, dioxin receptor, actin, v-src, and other kinases (9 -13). Raf-1 binds to Hsp90 via its COOH-terminal catalytic domain (8) and remains complexed to Hsp90 and p50 even when bound to Ras at the plasma membrane (14). It is not clear why native Raf-1 associates with Hsp90, although it has been proposed that this heat shock protein is involved in Raf-1 transport to the cell membrane (15).
The benzoquinone ansamycin, geldanamycin (GA), 1 has been shown to bind specifically and directly to Hsp90 and to disrupt the Hsp90-pp60 src molecular complex, leading to destabiliza-tion of pp60 src (16). Although originally described as tyrosine kinase inhibitors, benzoquinone ansamycins have been shown to be inactive when added directly to purified tyrosine kinases at concentrations Ͼ1500 times their effective in vivo dose (17,18). Additionally, several attempts to demonstrate direct association of ansamycins with tyrosine kinases in vivo and in vitro have been unsuccessful (16,19). This class of drug is now thought to exert kinase inhibitory activity indirectly by somehow destabilizing these proteins (18,20,21). Consistent with this hypothesis, binding to Hsp90 and destabilization of the Hsp90-pp60 src complex occurs both in vivo and in vitro at nanomolar concentrations of ansamycin (16). These drug levels are very similar to the concentration of ansamycin previously reported to produce decrements in cellular lck, v-src, and epidermal growth factor receptor protein level and activity (18,20,21), leading to the hypothesis that benzoquinone ansamycins are tyrosine kinase inhibitors because they disrupt Hsp90kinase heterocomplexes (16).
We now report that GA also disrupts the association between Hsp90 and the serine/threonine kinase Raf-1. The purpose of this study was to use GA to analyze the function of the Raf-1-Hsp90 complex.

EXPERIMENTAL PROCEDURES
Materials-GA was obtained from the Developmental Therapeutics Program, NCI. Culture media were purchased from Biofluids, Inc., [ 35 S]methionine was obtained from ICN Biomedicals, Inc., and protein A-Sepharose beads were purchased from Pharmacia Biotech Inc. Raf-1 antibody (clone C-12) was purchased from Santa Cruz Biotechnology, Hsp90 antibody (SPA-830) was purchased from Stressgen, and Ras antibody (pan-ras Ab-3) was purchased from Oncogene Science. A horseradish peroxidase-conjugated secondary antibody to rabbit (Raf-1) or mouse (Hsp90, Ras) IgG was purchased from Amersham Corp. and used in conjunction with Western blot chemiluminescence reagent (Renaissance, Du Pont). Nitrocellulose membrane was obtained from Schleicher & Schuell. All other chemicals were of highest available commercial grade.
Tissue Culture-MCF7 cells were obtained from the American Type Culture Collection and maintained in Dulbecco's modified Eagle medium containing 10% bovine calf serum and 10 mM HEPES. CHP100 cells were obtained from Dr. A. Evans (Children's Hospital of Philadelphia) and grown in RPMI 1640 medium with 10% bovine calf serum and 10 mM HEPES. For labeling proteins with [ 35 S]methionine, cells were washed with phosphate-buffered saline and kept in methionine-free media for 30 min. Then, 100 Ci/ml [ 35 S]methionine were added for 2 h. If the experiment required a chase, the cells were subsequently washed with phosphate-buffered saline and kept in medium containing nonradioactive methionine for 3-24 h.
Gel Electrophoresis and Western Blotting-Cell lysates or immuno-* 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.
§ To whom correspondence should be addressed: Clinical Pharmacology Branch, NCI, NIH, Bldg. 10, Rm. 12N226, Bethesda, MD 20892. 1 The abbreviations used are: GA, geldanamycin; PAGE, polyacrylamide gel electrophoresis. adsorbed protein A-Sepharose pellets were heated at 90°C in SDS sample buffer for 5 min, chilled on ice, and electrophoresed through 8 or 10% SDS-polyacrylamide gels. Proteins were electrophoretically transferred to nitrocellulose membranes. The membranes were blocked for 2 h with a solution containing 5% nonfat dry milk, 10 mM Tris-HCl, pH 7.5, 2.5 mM EDTA, pH 8, 50 mM NaCl, and 0.05% Tween 20 and then probed with the primary antibodies mentioned above diluted in blocking solution. After six washes with washing buffer (10 mM Tris-HCl, 2.5 mM EDTA, pH 8, 50 mM NaCl, 0.05% Tween 20), the membranes were exposed to horseradish peroxidase-labeled secondary antibody diluted in blocking solution. After an additional six washes, the proteins were visualized with Western blot chemiluminescence reagent following manufacturer's instructions.
[ 35 S]Methionine-labeled proteins were immunoprecipitated and electrophoresed as described above. The SDS-PAGE gel was fixed with a solution of 10% acetic acid and 50% methanol, washed copiously in water, and enhanced with Enlightning solution (DuPont NEN) prior to gel drying and autoradiography. Films of either chemiluminescent or radioactive blots were scanned into a Macintosh computer using a Foto/Eclipse Gel Analysis system (Fotodyne), and band intensities were quantified using Collage Analysis software (Fotodyne). Raf-1 half-life was determined by regression analysis of log transformed Raf-1-specific band intensities.
Fractionation of Cytoplasmic and Membrane Components-Cells were homogenized with a Dounce homogenizer in TESV buffer (TNESV buffer without detergent) with protease inhibitors and further disrupted by sonication. After ultracentrifugation at 100,000 ϫ g for 60 min at 4°C, the supernatant fraction, representing cytosolic components, was set aside, and the pellet was homogenized, sonicated, and extracted with TENSV buffer (containing 1% Nonidet P-40) with protease inhibitors. After centrifugation at 16,000 ϫ g for 15 min at 4°C, the supernatant, representing the Nonidet P-40-soluble membrane fraction, was set aside, and the remaining pellet, representing the Nonidet P-40-insoluble membrane fraction, was solubilized with 6% SDS. Protein concentrations of each fraction were determined using the BCA reagent (Pierce). The percentage of total cellular Raf-1 present in the various fractions was estimated after quantification of Raf-1 Western blots, by taking into account the total protein content of each fraction.

RESULTS AND DISCUSSION
GA Disrupts the Raf-1-Hsp90 Heterocomplex as Well as Complexes between Raf-1 and Ras-We chose to study in detail two cell lines in which Raf-1 is abundantly expressed: the breast cancer cell line MCF7 and the neuroepithelioma cell line CHP100. MCF7 cells were grown in log phase and lysed with TENSV buffer after treatment with and without GA for 4 h. 1800 g of total protein were immunoprecipitated with Raf-1 antibody, separated by SDS-PAGE, and immunoblotted. We were able to demonstrate coprecipitation of both Hsp90 and Ras with Raf-1 (Fig. 1, lane 1). Coprecipitation of Hsp90 and Ras with Raf-1 disappeared after brief treatment with GA ( Fig.  1, lane 2). Disruption of these heterocomplexes occurred in the absence of detectable changes in either cellular Hsp90 or Ras protein levels as assayed by Western blotting of 50 g of total protein (Fig. 1, lanes 3 and 4).
Although cellular Raf-1 protein was reduced by 55% in GAtreated cells (compare Raf-1 signal in lanes 3 and 4, Fig. 1), Raf-1-specific immunoprecipitation from 1.8 mg of total protein resulted in apparent antibody saturation, because drug treatment only minimally reduced the amount of Raf-1 recoverable by immunoprecipitation (compare Raf-1 signal in lanes 1 and 2,  Fig. 1). Because native Raf-1-Hsp90 heterocomplexes are unstable compared with the pp60 v-src -Hsp90 complex (8), the data shown in Fig. 1 do not represent a stoichiometric coprecipitation. However, this technique yields qualitative data demonstrating the disruption of existing Raf-1-Hsp90 and Raf-1-Ras complexes. Because Hsp90 is a very abundant protein and serves as a chaperone for a variety of other proteins, only a minor fraction of it is associated with Raf-1, although at least all cytosolic Raf-1 appears to occur in a complex with heat shock proteins (14). The amount of GTP-Ras bound to Raf-1 depends on the activation status of the cell.

The Stability of Raf-1 Is Significantly Decreased after the Disruption of the Raf-1-Hsp90
Complex-Because brief exposure to GA results in a significant decrement in total cellular Raf-1 protein (compare Raf signal in Fig. 1, lanes 3 and 4), we next asked the question whether Raf-1 is destabilized after treatment with GA. We compared steady-state levels of total cellular Raf-1 with newly synthesized Raf-1 in cells that had been treated with GA for 16 h. Total cellular Raf-1 was assayed by Western blotting of 10 g of total cellular protein from CHP100 cells ( Fig. 2A, lanes 1 and 2). Newly synthesized Raf-1 was detected by labeling with a 2-h pulse of [ 35 S]methionine, followed by Raf-1 immunoprecipitation, SDS-PAGE, and autoradiography ( Fig. 2A, lanes 3 and 4). Even though GA increases Raf-1 synthesis, steady-state levels of the protein were markedly reduced. These results are consistent with a marked reduction in Raf-1 stability following GA treatment.
In order to test this hypothesis directly, a [ 35 S]methionine pulse-chase protocol was used to determine Raf-1 half-life in both CHP100 and MCF7 cells exposed to GA for 16 h. GA exposure reduced Raf-1 half-life from 17.5 to 4 h in CHP100 cells and from 11 to 4 h in MCF7 cells (Fig. 2B). The halfmaximal GA concentration necessary to produce these effects was determined to be 25 nM (data not shown).

Treatment with GA Destabilizes Both Cytosolic and Rasassociated Raf-1 and Prevents Newly Synthesized Raf-1 from
Trafficking to the Membrane-Raf-1 can exist in any of three subcellular compartments. Newly synthesized and enzymatically inactive protein occurs in the cytosol and is recruited to the plasma membrane by Ras subsequent to extracellular signals (4,5). Ras recruitment is apparently necessary for Raf-1 activation, but Raf-1-Ras association is transient and is followed by Ras-independent association of Raf-1 with undefined cytoskeletal elements (7). We next studied whether Raf-1 destabilization following GA treatment occurred in more than one subcellular compartment or was limited to the cytosolic fraction of the protein. We chose MCF7 cells for this study because upon cellular fractionation, we found that 44% of Raf-1 protein was cytolsolic, 45% was extractable from membrane pellets with 1% Nonidet P-40 (the Ras-associated fraction), and 11% was found in the Nonidet P-40-insoluble membrane fraction (Raf-1 associated with cytoskeletal elements). SDS-PAGE and  1 and 2). As well, 50 g of total protein from cell lysates were used for Western blotting (lanes 3 and 4).
immunoblotting of extracts prepared from the three subcellular fractions revealed that Raf-1 disappeared equally from the three fractions upon treatment with GA (t1 ⁄2 of 4 -6 h; Fig. 3, A and D). Ras protein was localized to the detergent-soluble membrane fraction and remained unaltered, even after prolonged exposure to GA (Fig. 3B). These data suggest that disruption of the Raf-1-Hsp90 complex does not preferentially affect either the cytosolic or membrane components of Raf-1 protein.
Given the long half-life of Raf-1 in these tumor cells (see Fig.  2B), the data are consistent with several events. First, disruption of preformed cytosolic Raf-1-Hsp90 complexes by GA leads to rapid destabilization of cytosolic Raf-1 protein. Continued exposure of cells to GA also prevented newly synthesized Raf-1 from associating with Hsp90 (data not shown). Second, Raf-1-Hsp90 complex disruption affects the detergent-soluble membrane component of Raf-1 as rapidly as it does the cytosol component (Fig. 3, A and D). Loss of Raf-1 from this fraction occurred too quickly to be due solely to inability to recruit new Raf-1 from the cytosol. Immunoprecipitation of Raf-1 no longer co-precipitated Ras at a time when significant amounts of Raf-1 were still found in the detergent-soluble membrane fraction (compare Fig. 3A, lane 6 with Fig. 1, lane 2), suggesting that Ras association with Raf-1 requires the continued participation of Hsp90. Furthermore, Raf-1 stability, even when the protein is associated with Ras, depends on the presence of Hsp90.
Finally, the kinetics of Raf-1 turnover in the detergent-insoluble membrane fraction are also affected by GA (Fig. 3, A and  D). Whether this means that anchorage of Raf-1 to cytoskeletal elements requires participation of Hsp90 remains to be determined, because the harsh conditions necessary to solubilize cytoskeleton-bound Raf-1 preclude recovery of protein heterocomplexes. Although a model in which enzymatically active Raf-1 is no longer associated with Hsp90 would be appealing in that such a model would be strikingly similar to that proposed for association of Hsp90 with pp60 v-src (22), such a model is not consistent with our current data.
Together with destabilization, Raf-1-Hsp90 complex disruption might also affect the ability of cytosolic Raf-1 to be recruited by Ras. To address this question, we pulsed MCF7 cells, which had been exposed to GA for 16 h, with [ 35 S]methionine and followed the labeling with a 4 h chase (Fig. 3C). As described above (see Fig. 2A), Raf-1 synthesis in drug-treated cells was elevated (approximately 3-fold). However, in two separate experiments, the chase period was sufficient to allow 68% of newly synthesized Raf-1 (61 and 76%, respectively) to appear in the Nonidet P-40-soluble membrane fraction of untreated cells, whereas 32% of the labeled protein (24 and 39%, respectively) was recovered from the cytosol. In contrast, in GAtreated cells only 30% of radiolabeled Raf-1 (25 and 35%, respectively) was recovered from the Nonidet P-40-soluble membrane fraction, whereas 70% (65 and 75%, respectively) remained associated with the cytosol. These data are consistent with a model in which disruption of Raf-1-Hsp90 cytosolic complexes not only destabilizes Raf-1 but also interferes with its proper intracellular trafficking and recruitment to the membrane by Ras.
Destabilization of Raf-1 by GA Occurs in Various Cell Lines-Because Raf-1 is a ubiquitously expressed and highly conserved protein (2,23,24), its destabilization by GA should be a general phenomenon. To test this hypothesis, we chose a diverse panel of cell lines representing nonmalignant mouse fibroblasts (NIH 3T3) as well as sarcomatous (CHP100), carcinomatous (HeLa, MCF7, and DU145), lymphomatous (Raji), and leukemic (CEM) human cell lines. Although these cell lines differed in terms of Raf-1 levels, they all showed a marked reduction in the steadystate level of Raf-1 protein after GA treatment (Fig. 4).

FIG. 2. GA increases Raf-1 synthesis while decreasing Raf-1 half-life.
A, CHP100 cells were pretreated with or without GA (1 M) and exposed to a 2-h [ 35 S]methionine pulse (100 Ci/ml of methionine-free medium) before lysing the cells in TENSV buffer. Lanes 1 and 2 show total Raf-1 determined by Western blotting. Additionally, 200 g of total protein was immunoprecipitated with Raf-1 antibody, separated on an 8% SDS-PAGE gel, and visualized by autoradiography (lanes 3 and 4). Data shown were obtained by immunoprecipitating equal amounts of protein, but similar results were obtained if equivalent acid-precipitable radioactivity was immunoprecipitated. B, determination of Raf-1 half-life in CHP100 and MCF7 cells by [ 35 S]methionine pulse-chase. Cells were preincubated with or without GA (2 M) for 16 h. They were then pulsed for 2 h with [ 35 S]methionine (100 Ci/ml) and chased with nonradioactive methionine-containing medium for 3-24 h. The cells were lysed in TENSV buffer, and Raf-1 was immunoprecipitated and analyzed as above.
Taken together, these data point to a role for Hsp90 in Raf-1-mediated signal transduction. Although the half-life of Raf-1 is approximately 10 -20 h in the cell lines examined, such prolonged stability requires association of Hsp90 with both cytosolic Raf-1 as well as the Raf-1 found in the cell membrane fraction. Disruption of this complex is followed by a marked reduction in Raf-1 stability, even though synthesis of the protein is actually elevated. However, newly synthesized Raf-1, to which Hsp90 has been prevented from binding, cannot translocate efficiently to the Nonidet P-40-soluble membrane fraction. Additionally, Raf-1, already associated with this fraction at the time of Hsp90 dissociation, rapidly dissociates from Ras. Therefore, Hsp90 association also appears necessary for recruitment of cytosolic Raf-1 to the plasma membrane, as well as for its maintenance there in a Ras-bound state.
Because binding to Ras is part of the mechanism that leads to activation of Raf-1 and because the most highly enzymatically active Raf-1 is found in the Nonidet P-40-insoluble cellular membrane fraction (5), prevention of Raf-1 binding to Ras and depletion of Raf-1 from this and other cellular compartments are likely to have a negative impact on Raf-1 function.  3, 7, and 11), or 40 h (lanes 4, 8, and 12). The cells were lysed, and preparations of cytosol (lanes 1-4) and Nonidet P-40-soluble (lanes 5-8) and -insoluble (lanes 9 -12) membrane fractions were obtained. 15 g of total protein/lane were electrophoresed through 10% SDS-PAGE minigels. Western blotting was performed for Raf-1. B, Western blotting for Ras was performed using the fractions obtained from untreated and 16-h GA-treated samples. Lanes 1 and 2 represent cytosol from untreated and GA-treated samples, respectively; lanes 3 and 4 represent Nonidet P-40-soluble membrane preparations from untreated and GA-treated samples, respectively; and lanes 5 and 6 represent Nonidet P-40-insoluble membrane preparations from untreated and GA-treated samples, respectively. C, MCF7 cells were pulsed with [ 35 S]methionine after pretreatment without (lanes 1 and 2) or with (lanes 3 and 4) GA (2 M) and then chased for 4 h with nonradioactive media. Cytosol and Nonidet P-40-soluble membrane preparations were obtained, and equal amounts of total protein were immunoprecipitated with Raf-1 antibody, electrophoresed on an 8% SDS-PAGE gel, and visualized by autoradiography. Cytosolic fractions are represented in lanes 1 and 3; Nonidet P-40-soluble membrane fractions are represented in lanes 2 and 4. This experiment was performed twice. Densitometric analysis of the Raf-1 bands in untreated cells (in the two experiments) revealed 24 and 39% of the radioactivity in cytosol and 76% and 61% of the radioactivity in membrane. In GA-treated cells, 65 and 75% of the radioactivity was found in cytosol, compared with 25 and 35% in the membrane fraction. D, the experiment described in A was performed four times. Band densities were FIG. 4. GA depletes Raf-1 from both transformed and untransformed cell lines. Cells were grown in log phase culture for 16 h without (lanes 1, 3, 5, 7, 9, 11, and 13) or with GA (2 M) (lanes 2, 4, 6, 8, 10, 12, and 14). TENSV lysates were analyzed for Raf-1 content by Western blotting with Raf-1 antibody. The cells used were NIH 3T3 (lanes 1 and 2), CHP100 (lanes 3 and 4), HeLa (lanes 5 and 6), MCF7 (lanes 7 and 8), DU145 (lanes 9 and 10), Raji (lanes 11 and 12), and CEM (lanes 13 and 14).