Geldanamycin induces ErbB-2 degradation by proteolytic fragmentation.

Exposure of carcinoma cell lines to the antibiotic geldanamycin induces the degradation of ErbB-2, a co-receptor tyrosine kinase that is frequently overexpressed in certain tumors. Using ErbB-2 mutants expressed as chimeric receptors or green fluorescent protein fusion proteins, we report that the kinase domain of ErbB-2 is essential for geldanamycin-induced degradation. The kinase domain of the related epidermal growth factor receptor was not sensitive to this drug. The data further indicate mechanistic aspects of ErbB-2 degradation by geldanamycin. The data show that exposure to the drug induces at least one cleavage within the cytoplasmic domain of ErbB-2 producing a 135-kDa fragment and a 23-kDa fragment. The latter represents the carboxyl-terminal domain of ErbB-2, whereas the former represents the ectodomain and part of the cytoplasmic domain. Degradation of the carboxyl-terminal fragment is prevented by proteasome inhibitors, whereas degradation of the membrane-anchored 135-kDa ErbB-2 fragment is blocked by inhibitors of the endocytosis-dependent degradation pathway. Confocal microscopy studies confirm a geldanamycin-induced localization of ErbB-2 on intracellular vesicles.

ErbB-2 is a Type 1 transmembrane tyrosine kinase that functions as a co-receptor by forming dimers with other members of the ErbB receptor family (ErbB-1 (EGF 1 receptor), ErbB-3, and ErbB-4; Refs. 1 and 2). Although ErbB-2 has a potential ligand-binding ectodomain, no direct ligand has yet been identified. In its role as a co-receptor, ErbB-2 enhances the signaling capacity of its dimerization partners. The association of ErbB-2 with these various receptors is, however, entirely ligand-dependent. In the absence of growth factor ErbB-2 is reported to interact with CD44, an adhesion receptor, in ovarian carcinoma cell lines (3) and with a large plasma membrane glycoprotein complex in microvilli of a mammary adenocarcinoma cell line (4). ErbB-2 has also been demonstrated to form ligand-dependent complexes with the IL-6 receptor component gp130 (5) and Trk A (6), the nerve growth factor receptor.
ErbB-2 was originally identified as the transforming oncogene neu in which a point mutation in the transmembrane domain is responsible for its oncogenic potential (7,8). ErbB-2 also functions as an oncogene when overexpressed (9,10) and in humans is frequently overexpressed in breast and ovarian tumors (11). ErbB-2 overexpression in breast cancer is associated with a poor prognosis (12), and hence it is a target for therapeutic reagents, including monoclonal antibodies and drugs (13). Frequently, antibodies that decrease the growth of ErbB-2-expressing tumors also reduce the level of ErbB-2 by a mechanism that is unclear. Hence, the transforming activity of ErbB-2 is related to structural changes or changes in its level of expression.
The benzoquinoid ansamycin antibiotics geldanamycin and herbimycin were first isolated from the culture broths of several actinomycete species (14,15) and described as inhibitors of tyrosine kinase-dependent growth (16,17). These compounds, particularly geldanamycin, have tumorical activity toward numerous tumor cell lines (18), including those that overexpress ErbB-2 (19). This action toward tumor cell lines is attributed to the capacity of geldanamycin to induce the degradation of several important signal transducers important in mitogenic pathways. These targets include protein kinases, such as Src, Raf, FAK, and ErbB-2, and other growth regulating proteins, such as p53 (20). The mechanism for the geldanamycin-induced degradation of these various molecules is centered on the Hsp90 family of chaperones, because Hsp90 is the major intracellular protein that binds geldanamycin (20,21). Geldanamycin has been shown to dissociate Hsp90 from various proteins and thereby inhibit their function, such as the nuclear translocation of glucocorticoid receptors, or to induce their metabolic degradation, such as Src, Raf, and p53.
In the case of ErbB-2, association with Hsp90 has not been reported. However, it has been reported that the glucose-regulated chaperone GRP94, an Hsp90 family member that is localized to the lumen of the endoplasmic reticulum, does associate with ErbB-2 in a geldanamycin-sensitive manner (22). Geldanamycin-induced degradation of ErbB-2 is reported to involve, presumably as a consequence of the dissociation of GRP94, the polyubiquitination of ErbB-2 and its proteosomal degradation (23). On the basis of GRP94 localization, these studies would suggest an interaction with the ErbB-2 ectodomain in the lumen of the endoplasmic reticulum and druginduced degradation during receptor biosynthesis. Others, however, have suggested that this interpretation does not account for the quantitative aspects of ErbB-2 degradation induced by geldanamycin (24).
We have explored the question of how geldanamycin induces ErbB-2 degradation and show that the ErbB-2 kinase domain is essential for sensitivity of geldanamycin. Also we show that geldanamycin induces fragmentation of ErbB-2 within the carboxyl-terminal region of the cytoplasmic domain and that the resulting transmembrane fragment is degraded by a mechanisms that involves the formation of intracellular vesicles.
Cell Culture and Transfection-Human mammary tumor-derived SKBr3 cells were grown in 5% CO 2 at 37°C in McCoy medium with 10% fetal bovine serum, COS 7 cells were grown in Dulbecco's modified Eagle's medium with 10% fetal bovine serum, and all fibroblast cell lines were grown in Dulbecco's modified Eagle's medium containing 10% calf serum. Cells were grown to about 80% confluency then washed two times, and treated with indicated drugs in serum-free Dulbecco's modified Eagle's medium. For transfection and expression of GFP fusion protein, COS 7 cells were grown to ϳ70% confluency overnight and transfected with LipofectAMINE (Life Technologies, Inc.) according to manufacture's recommendations (10 g of plasmid DNA mixed with 16 l of LipofectAMINE were used per 60-mm tissue culture dish). The cells were grown for 48 h before assays.
Construction of GFP Fusion Proteins-The ErbB-2 kinase domain and cytoplasmic region cDNA fragments were generated by PCR with high fidelity VENT R DNA polymerase (New England BioLabs). Following the numbering of Yamamoto et al. (26), the kinase domain fragment and cytoplasmic domain fragment correspond, respectively, to residues 715-990 and 676 -1255 of ErbB-2. To prepare these two fragments the following primers were synthesized: upstream primer with SacI restriction site 5Ј-GGG ATC CTC ATC AAA CGA GCT CAG AAG ATC-3Ј (primer 1), downstream primers with XbaI restriction sites 5Ј-ACT ACG TCC AGT TCT AGA TCA CAC TGG CAC GTC CAG ACC-3Ј (primer 2), 5Ј-GTA GAA GGT GCT GTC TAG AGG ACT GGC TGG-3Ј (primer 3). PCR products were cloned into pEGFP-C1 vector by using SacI and XbaI sites.
Constructs with ErbB-2 carboxyl terminus and its truncation fragments fused to GFP were prepared by megaprimer approach. The carboxyl-terminal domain fragment corresponds to residues 991-1255 or ErbB-2 (26). In the first round of PCR the following primers that contain stop codons were used as 5Ј primers: 5Ј-GTA CCC CTG CCC TAA GAG ACT GAT GGG-3Ј, 5Ј-CAG CCC CCT TAA CCC CGA GAG GGC-3Ј, 5Ј-CCC CAG TAC TAA ACA CCC CAG GGA-3Ј, 5Ј-CCC AGC ACC TAA AAA GGG ACA CCT-3Ј and primer 2 as the 3Ј primer. One strand of this product was used as the 3Ј primer in a second round of PCR with primer 1.
The following primers were used to prepare construct of GFP fusion protein with EGF receptor kinase domain (residues 683-698, according to the numbering of Ullrich et al. (27)): 5Ј-CGA AGG CGC CAC AGA GCT CGG AAG CGC ACG-3Ј (upstream primer with SacI restriction site) and 5Ј-GTA GAA GTT GGA GTC TAG AGG ACT TGG-3Ј (downstream primer with XbaI restriction site). The cDNA of ErbB-2 carboxyl terminus (residues 991-1255, according to Yamamoto et al. (26)) was prepared with 5Ј-TCC CGC ATG GCC AGA GCT CCC CAG-3Ј upstream primer with SacI restriction site and primer 2 as downstream primer.
To prepare construct with ErbB-2 kinase domain and EGF receptor carboxyl terminus (residues 959 -1186, according to Ullrich et al. (27)), SacI restriction site in EGF receptor carboxyl terminus was mutated in 1 round PCR with primers 5Ј-CTC CTA AGT TCT CTG AGT GCA ACC-3Ј (upstream primer) and 5Ј-TCA TAC TAT GGT GTC GAC TCA TGC TCC AAT AAA TTC ACT GCT TTG-3Ј (downstream primer). PCR product was used in second round of PCR to generate EGF receptor carboxyl terminus with primer 5Ј-CAT TTG CCA AGT CCT CTA GAC TCC AAC TTC-3Ј (upstream primer). The cDNA fragment for ErbB-2 kinase domain was prepared as described above with primers 1 and 3. Both products were cut with XbaI and ligated. Then cDNA fragment was cloned into pEGFP vector through SacI and SalI restriction sites. All the constructions described above were verified by sequencing in the regions that underwent genetic manipulations.
Immunoprecipitation and Immunoblotting-After indicated treatments, the cells were solubilized by scraping with rubber policeman into cold lysis buffer (10 mM Tris-HCl, pH 7.5, 150 mM NaCl, 2 mM EDTA, 1% Nonidet P-40, 1 mM phenylmethylsulfonyl fluoride, 10 g/ml aprotinin, 10 g/ml leupeptin, 1 mM Na 3 VO 4 ). The lysates were then clarified by centrifugation (14,000 ϫ g, 10 min). Receptors were immunoprecipitated with 1 g of the indicated antibody immobilized on protein G or protein A by incubation for 1 h at 4°C. Subsequently, the complexes were washed with lysis buffer three times and resuspended in Laemmli sample buffer for 7.5% SDS-PAGE. After electrophoresis, proteins were transferred to nitrocellulose membranes, and the membranes were blocked by incubation with 5% bovine serum albumin in PBS for 1 h at room temperature. The membranes were then incubated 1 h at room temperature with the indicated blotting antibody in TBSTw buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 0.05% Tween 20, 0.5% nonfat milk), washed three times in the same buffer, and incubated 1 h with horseradish peroxidase-conjugated mouse antibody or protein A. The membranes were then washed five times in TBSTw and visualized by ECL.
Confocal Microscopy-Cells were grown 1-2 days on Lab-Tek Chamber slides to 50 -70% confluency, treated for indicated times with geldanamycin, and fixed by adding freshly prepared 4% paraformaldehyde in PBS and incubated for 1 h or overnight. Fixed SKBr3 and 3T3 cells were washed three times in PBS, permeabilized by incubation in 0.5% Triton X-100 in PBS for 10 min, and blocked by incubation with 5% bovine serum albumin in PBS for 1 h. Subsequently, the cells were incubated with fluorescene dye-conjugated secondary antibody for 1 h, washed five times, and dried. Fixed COS 7 cells expressing GFP fusion protein were observed directly.

Influence of ErbB-2 Kinase Domain on Geldanamycin-induced
Degradation-To measure the extent of ErbB-2 degradation in cells treated with geldanamycin, we have used, in parallel, antibodies that react with ectodomain or carboxylterminal domain epitopes of ErbB-2. The results, as shown in Fig. 1A, demonstrate that immunoreactivity to both ectodomain and carboxyl-terminal domain antibodies is rapidly lost following exposure to geldanamycin. The cells employed in this experiment, SKBr3, overexpress ErbB-2, and quantitation of the data indicates that in these cells the half-life of ErbB-2 is approximately 2 h in the presence of geldanamycin. Others have reported a similar half-life for ErbB-2 under these conditions (19) and contrasts with the reported half-life of about 7-9 h for ErbB-2 under normal conditions (28,29). This influence of geldanamycin on ErbB-2 metabolic stability is reasonably specific because no significant decrease in the structurally related receptors ErbB-1 or ErbB-4 was detected following geldanamycin treatment of A-431 cells or T47-17 cells, respectively (Fig.  1B).
To map the region of ErbB-2 that mediates geldanamycin sensitivity, we initially employed chimeric receptors in which the cytoplasmic domain of the EGF receptor is replaced by that of ErbB-2 (EGFR/ErbB-2 CD ) or in which the carboxyl-terminal domain of the EGF receptor is replaced by the corresponding region of ErbB-2 (EGFR/ErbB-2 CT ). These chimeric receptors are expressed in NIH 3T3 cells and have been described previously (25). The data in Fig. 2 show that following the addition of geldanamycin the chimeric receptor having the entire ErbB-2 cytoplasmic domain is sensitive to drug-induced degradation, whereas the chimeric receptor that contains only the carboxyl-terminal domain of ErbB-2 is not influenced by the presence of geldanamycin. In this experiment the EGFR/ErbB-2 CD receptor has a half-life of approximately 3.9 h in the pres-ence of geldanamycin. In the same background the ErbB-2 receptor has a half-life of about 3.3 h in the presence of this drug (data not shown). Therefore, the ectodomain of ErbB-2 does not have a significant role in determining the sensitivity of this receptor to geldanamycin-induced degradation. Given the data in Fig. 1B showing that the EGF receptor is not sensitive to geldanamycin, these results suggest that the kinase domain and/or juxtamembrane region of the ErbB-2 cytoplasmic domain mediate sensitivity to geldanamycin-induced degrada-tion. This conclusion is consistent with the results of Miller et al. (19), who reported that an ErbB-2 internal deletion mutant lacking the kinase domain was not degraded in the presence of geldanamycin.
To directly test the possibility that the kinase domain determined geldanamycin sensitivity, we constructed GFP fusion proteins with the ErbB-2 cytoplasmic domain (GFP-ErbB-2 CD ), the ErbB-2 kinase domain (GFP-ErbB-2 KD ), the ErbB-2 carboxyl-terminal domain (GFP-ErbB-2 CT ), or the EGF receptor kinase domain (GFP-EGFR KD ). These fusion proteins were expressed in COS 7 cells and tested for sensitivity to geldanamycin by blotting lysates with antibody to GFP after incubating the cells in the absence or presence of the drug. The results are shown in Fig. 3. As a control, we tested the sensitivity of GFP to geldanamycin, and, as shown in Fig. 3A (lanes 1 and 2), treatment with geldanamycin for 6 h induced no decrease in the cellular level of GFP. Also, there was no geldanamycininduced degradation of the GFP fusion proteins containing the carboxyl terminus of ErbB-2 (Fig. 3A, lanes 3 and 4) or the entire cytoplasmic domain of ErbB-2 (Fig. 3A, lanes 5 and 6). However, the GFP fusion protein containing the ErbB-2 kinase domain (Fig. 3B, lanes 1-4) was rapidly degraded in the presence of geldanamycin, whereas there was no degradation of a fusion protein containing the EGF receptor kinase domain (Fig.  3B, lanes 5-8). The results in Fig. 3B are quantitated in Fig. 3C and show the increased sensitivity of the ErbB-2 kinase domain to geldanamycin compared with the EGF receptor kinase domain.
In this system, the lack of sensitivity to geldanamycin is, on the basis of previous data as expected for fusion proteins containing the ErbB-2 carboxyl-terminal domain or the EGF receptor kinase domain. That the fusion protein containing only the ErbB-2 kinase domain is rapidly degraded in the presence of geldanamycin indicates that the kinase domain is sufficient to mediate its degradation. However, the lack of degradation of the fusion protein containing the entire ErbB-2 cytoplasmic domain was unexpected and is in apparent discordance with the geldanamycin-induced degradation of the EGF receptor/ ErbB-2 chimeric receptor, which contains the entire ErbB-2 cytoplasmic domain (Fig. 2, lanes 1-4).
These latter results suggested that perhaps within the context of cytosolic GFP fusion proteins, but not in the transmembrane ErbB-2 molecule, the ErbB-2 carboxyl terminus had a protective effect on the sensitivity of the kinase domain to degradation induced by geldanamycin. To test this possibility we constructed a series of GFP fusion proteins containing the ErbB-2 kinase and carboxyl-terminal domains with progressive deletions of the ErbB-2 carboxyl-terminal domain. The geldanamycin sensitivity of these constructs, when expressed in COS 7 cells, is shown in Fig. 4. Deletion of the carboxylterminal 19 residues (⌬1236 -1255, lanes 3 and 4) or 59 residues (⌬1196 -1255, lanes 5, 6) did not increase sensitivity of the fusion proteins to degradation in the presence of geldanamycin. However, increased sensitivity to geldanamycin-induced degradation was observed when deletions of 105 or more residues were made in the ErbB-2 carboxyl-terminal domain (lanes 7 -14). Hence, loss of the amino-terminal half of this carboxylterminal domain significantly increases the sensitivity of the ErbB-2 kinase domain to geldanamycin.
In this series of fusion proteins we also determined whether the carboxyl-terminal domains of the EGF receptor would abrogate sensitivity of the ErbB-2 kinase domain to geldanamycin. Hence, we prepared a fusion protein construct to encode the kinase domain of ErbB-2 and carboxyl-terminal domain of the EGF receptor. When this molecule was expressed in COS 7 cells, the results (Fig. 4B) showed complete sensitivity to geldanamycin-induced degradation in contrast to the fusion protein containing the carboxyl-terminal sequences of ErbB-2 (lanes 1 and 2).

Detection of Geldanamycin-induced ErbB-2 Fragments-The
above results, which indicate a role of the ErbB-2 carboxylterminal domain in determining the sensitivity of ErbB-2 kinase domain to geldanamycin-induced degradation, suggested that perhaps the carboxyl-terminal region of the transmembrane receptor might be cleaved following the addition of geldanamycin and that this event may be necessary for subsequent degradation of the ErbB-2 molecule, perhaps in a manner that required localization at the plasma membrane and not in the cytosol. Therefore, we re-examined the geldanamycininduced degradation of the chimeric EGF receptor/ErbB-2 CD molecule, whose degradation in the presence of geldanamycin was shown in Fig. 2 (lanes 1-4). If the carboxyl terminus is, in fact, cleaved prior to degradation, then the antibodies used in the experiment shown in Fig. 2 would not detect the remaining membrane-localized fragment because they are to an epitope in the ErbB-2 carboxyl terminus. Hence, we tested the geldanamycin-induced degradation of this molecule using an antibody to the EGF receptor ectodomain. As shown in Fig. 5, this antibody detected both the native 185-kDa form of the chimeric receptor plus a geldanamycin-induced fragment of approximately 135 kDa (lanes 2 and 3). As shown in lane 5, the formation of this fragment in cells exposed to geldanamycin was blocked by the presence of the protease inhibitor ALLN and to a lesser extent by lactacystin (data not shown).
We next attempted to confirm that geldanamycin induced a carboxyl-terminal cleavage in the native ErbB-2 molecule as well as the chimeric receptor. To test this we used SKBr3 cells and an antibody to an epitope in the ErbB-2 ectodomain. As shown in Fig. 6A (lane 2) incubation of these cells with geldanamycin for 6 h resulted in the loss of the native ErbB-2 molecule, and no fragment was detected. This result is similar to that reported in Fig. 1A (lanes 5-8). We reasoned that perhaps in these cells the fragment was metabolically unstable and might be detectable if the geldanamycin treatment were performed at a low temperature to reduce metabolic degradation. In this part of the experiment (lanes 3 and 4), geldanamycin was added to the cells for 1 h at 37°C, and then the cells were cooled to 4°C and incubated for an additional 5 h. Under these conditions an  1 and 2) or were truncated at the indicated residues in the carboxyl-terminal domain (lanes 3-14). The cells were then treated with (ϩ) or without (Ϫ) geldanamycin (3 M) for 6 h. Lysates were prepared, and equal aliquots (30 g) were electrophoresed on 10% SDS-PAGE gels. Subsequently, Western blotting with anti-GFP was used to detect the fusion protein bands, which were visualized by ECL. B, a construct encoding a fusion protein of GFP with the ErbB-2 kinase domain and the EGF receptor carboxyl terminus (GFP-ErbB2 KD EGFR CT ) was expressed in COS 7 cells. The cells were then treated with (ϩ) and without (Ϫ) geldanamycin (3 M) for 6 h, and lysates were prepared. Equal aliquots (30 g) of each lysate were electrophoresed and blotted with antibody to GFP. Bands were visualized by ECL. ErbB-2 fragment of 135 kDa was detectable, and in the presence of ALLN this fragment was not detected. These data indicated that endogenous ErbB-2 is cleaved within the cytoplasmic domain in geldanamycin-treated cells.
A number of protease inhibitors were tested without success for their capacity to prevent degradation of the 135-kDa ErbB-2 fragment produced in response to geldanamycin. However, the cathepsin B inhibitor CA074-Me (30) did stabilize the level of the ErbB-2 fragment in geldanamycin-treated cells (Fig. 6B). Because cathepsin B is mainly localized in late endosomes (31), we tested compounds that interfere in the acidification and/or processing of endosomes. As shown in Fig. 6C (lanes 1-5), chloroquine, monensin, and folimycin each significantly increased accumulation of the 135-kDa ErbB-2 fragment in geldanamycin-treated cells. In the absence of geldanamycin, none of these compounds revealed major immunoreactive bands other than the native ErbB-2. Detection of this accumulated 135-kDa fragment was not possible when an antibody to the ErbB-2 carboxyl-terminal domain was employed (lanes 6 -10), supporting the conclusion that this fragment is produced by a cleavage at the carboxyl terminus of ErbB-2.
If ErbB-2 is cleaved at the carboxyl terminus such that an antibody epitope is lost from the native molecule, an antibody to a carboxyl-terminal epitope may be able to detect the released fragment if it is sufficiently metabolically stable. That such a carboxyl-terminal ErbB-2 fragment can be detected in lysates from geldanamycin-treated cells blotted with an antibody to the ErbB-2 carboxyl terminus is shown in Fig. 7. As demonstrated in Fig. 7A, the presence of protease inhibitors ALLN or proteasome inhibitor I reveals the presence of a 23-kDa ErbB-2 carboxyl-terminal fragment produced during geldanamycin exposure of SKBr3 cells. The data in Fig. 7B show the influence of incubation time in geldanamycin on the accumulation of this fragment in SKBr3 cells. The fragment is readily detected within 1 h following the addition of geldanamycin. The previously described 135-kDa ErbB-2 fragment is also readily detected in the same period of time (data not shown).
Geldanamycin-induced Intracellular ErbB-2 Containing Vesicles-The capacity of folimycin, chloroquine, or monensin to increase accumulation of the 135-kDa ErbB-2 fragment suggests that this fragment is normally degraded by a mechanism that involves endocytic vesicles. Therefore, we used confocal microscopy to determine whether ErbB-2 is internalized during geldanamycin treatment. Following drug or vehicle exposure for 6 h, the cells were fixed and permeabilized prior to incubation with antibody to ErbB-2 and fluorescene-conjugated second antibody. Shown in Fig. 8 are SKBr3 cells with endogenous ErbB-2 (Fig. 8, A and B), NIH 3T3 cells stably expressing transfected ErbB-2 (Fig. 8, C and D), and COS 7 cells transiently expressing a GFP fusion protein with the ErbB-2 kinase domain (Fig. 8, E and F). In the absence of geldanamycin, the transmembrane form of ErbB-2 is clearly expressed at the cell surface in SKBr3 (Fig. 8A) and NIH 3T3 cells (Fig. 8C), whereas the GFP fusion protein is located in the cytosol of COS 7 cells (Fig. 8E). In each case the distribution is changed dramatically following geldanamycin incubation, such that immunoreactivity is concentrated in intracellular vesicles, which  3 and 4). Cell lysates were then prepared, and equal (30 g) aliquots of each lysate were analyzed by electrophoresis and Western blotting, using antibody to an epitope in the ErbB-2 ectodomain (anti-ErbB-2 (ED)) and ECL. B, SKBr3 cells were preincubated for 1 h with the cathepsin B inhibitor CA074-Me (100 M) as indicated, and then geldanamycin (3 M) was added for 6 h. Following cell lysis, aliquots (30 g) of each lysate were electrophoresed and blotted with antibody to the ErbB-2 ectodomain (lanes 1-3) or antibody to the ErbB-2 carboxyl-terminal domain (lanes 4 -6). Bound antibody was then visualized by ECL. C, SKBr3 cells were preincubated for 1 h as indicated with chloroquine (100 M), monensin (50 M), or folimycin (1.0 g/ml). Geldanamycin (3 M) was then added, and the incubation was continued an additional 6 h. Cell lysates were then prepared, and equal aliquots (30 g) of each lysate were electrophoresed and Western blotted (WB) with antibody to the ectodomain of ErbB-2 (lanes 1-6) or antibody to the carboxyl terminus of ErbB-2 (lanes 7-11).Bound antibody was visualized by ECL. mimic the appearance of lysosomes (Fig. 8, B, D, and F). The vesicles could be detected within 2 h of geldanamycin consistent with the time course of ErbB-2 degradation. The presence of either ALLN or folimycin decreased the redistribution of ErbB-2 immunoreactivity observed in the presence of geldanamycin (data not shown). Also, the GFP fusion protein with the entire ErbB-2 cytoplasmic domain, which we have previously shown is metabolically stable in the presence of geldanamycin (Fig. 3A, lanes 5 and 6), was not redistributed in the presence of the drug (data not shown).

DISCUSSION
In this manuscript we report several novel aspects of the mechanism by which geldanamycin induces the degradation of ErbB-2. Using antibodies to both ectodomain and carboxylterminal epitopes, we detect two fragments of ErbB-2 produced following geldanamycin incubation. One fragment of 135 kDa represents the ErbB-2 ectodomain plus the transmembrane domain and part of the kinase cytoplasmic domain. A second fragment of approximately 23 kDa is also detected under these conditions and represents the carboxyl-terminal domain of ErbB-2. These fragments are detectable within 30 -60 min of geldanamycin addition to cells and continue to be detected for several hours. These two fragments do not, however, account for the entire mass of the native ErbB-2 molecule. In previous studies of geldanamycin-induced ErbB-2 degradation, fragments of this molecule have not been detected. This is likely due to two factors. The first is the absence of appropriate inhibitors and the second is the use of different antibodies for sequential precipitation (cytoplasmic domain epitope) and blotting (ectodomain epitope).
Based on the characteristics of these fragments, at least two mechanisms can be proposed to account for the generation of these fragments (Fig. 9). In model I, an endoproteolytic cleavage within the carboxyl-terminal domain could directly generate the observed 23-kDa fragment plus a 160-kDa fragment representing the rest of the ErbB-2 molecule. Subsequent cleavage of this latter fragment within the kinase domain could generate the observed 135-kDa fragment plus a small fragment of approximately 27 kDa that is undetectable with available immunologic reagents. Alternatively, it can be proposed that fragments are produced in the order predicted in model II. In this case an endoproteolytic event within the kinase domain would directly generate the observed 135-kDa transmembrane fragment plus a fragment of about 50 kDa. Subsequent cleavage of this latter fragment would produce the observed 23-kDa carboxyl-terminal domain fragment.
In either model intermediate fragments have not been detected (the 160-kDa fragment in model I or the 50-kDa fragment in model II), and this may be due to the rapidity with which the second cleavage occurs. We prefer the scheme depicted in model I for the following reasons. Our data show that expression of the ErbB-2 cytoplasmic domain as a cytosolic GFP fusion protein is not sensitive to geldanamycin-induced degradation, whereas a GFP fusion with the ErbB-2 kinase domain without the carboxyl-terminal domain is sensitive to geldanamycin-induced degradation. Also, the transmembrane ErbB-2 receptor is sensitive to degradation induced by geldanamycin. This suggests that proteolytic cleavage of the carboxylterminal domain may be restricted topologically within the cell to an area near the cytoplasmic face of the plasma membrane and not available to mediate cleavage of the cytosolic GFP fusion protein containing the entire ErbB-2 cytoplasmic domain. The exact protease(s) that generate these ErbB-2 fragments in response to geldanamycin have not been identified.
The metabolic stability of ErbB-2 in cells is likely complex, and ectodomain cleavage by metalloprotease activity has been reported (32)(33)(34)(35). In these experiments we have not observed ectodomain fragmentation, which produces fragments of different sizes than those we have detected in geldanamycin-treated cells.
The data in this manuscript indicate that complex series of proteolytic events are involved in the degradation of ErbB-2. This would include the proteosome as well as cathepsin B, an endosomal protease. Mimnaugh et al. (23) reported that lactacystin (10 M), a proteosome inhibitor, blocked the geldanamycin-induced loss of the 185-kDa native form of ErbB-2. We find that lactacystin (10 -40 M) only partially prevents cleavage of the native ErbB-2 molecule and that ALLN is more effective in this regard. ALLN, like lactacystin, is a proteosome inhibitor, but it also inhibits other proteases such as calpain and cathepsin B and L. However, various calpain inhibitors (calpain inhibitor III, calpain inhibitor V, and calpastatin) do not block geldanamycin-induced fragmentation of ErbB-2 in our system.
In our experiments geldanamycin induces the formation of intracellular vesicles containing ErbB-2. Previously the accumulation of such vesicles was only reported after a prolonged 22-h incubation in the presence of geldanamycin and was attributed to the relocalization of newly synthesized ErbB-2 molecules (22). We observed the much more rapid formation of intracellular ErbB-2-containing vesicles, within 2 h of geldanamycin exposure. These vesicles also form when cells expressing the cytosolic GFP fusion protein with the ErbB-2 kinase domain are treated with geldanamycin. Also, the formation of these vesicles containing ErbB-2 is blocked by agents that interfere in the processing and acidification of endosomes, such as chloroquine, folimycin, and monensin. Hence, we conclude that these vesicles containing ErbB-2 are derived from the plasma membrane and that endosomal proteases, such as cathepsin B, participate in the degradation of internalized ErbB-2. The manner in which these vesicles are actually formed is, however, not clear. Inhibitors of cathepsin B, such as CA074-Me, have been shown to inhibited degradation of EGF and the EGF receptor within endosomes (30).
Lastly, our results indicate that the kinase domain of ErbB-2 mediates sensitivity to geldanamycin. Previously, Miller et al. (19) have shown that the ErbB-2 mutants lacking the entire cytoplasmic domain or the kinase domain are not degraded in cells treated with geldanamycin. Our analysis of chimeric receptors and GFP fusion protein agrees with that data.
Geldanamycin induces the metabolic degradation of other protein kinases, such as Raf and Src. In the case of Raf, in particular, the geldanamycin-binding protein Hsp90 along with another protein p50 cdc37 interact with the raf kinase domain (36,37). It seems likely that a geldanamycin-binding protein, perhaps Hsp90, also interacts with the ErbB-2 kinase domain and, in the absence of geldanamycin, assists in the maintenance of this kinase domain in an active conformation. Given the intracellular abundance of proteins such as Hsp90, this interaction could be of low affinity and not readily detected by assays such as co-immunoprecipitation. Additional experi-ments will explore this issue.