Nucleophosmin-anaplastic lymphoma kinase (NPM-ALK) is a constitutively active fusion tyrosine kinase involved in lymphomagenesis of human anaplastic large cell lymphomas (ALCL), the maturation and activity of which depend on the association with the heat shock protein (hsp) 90 protein chaperone. Targeting hsp90 by the ansamycins geldanamycin and 17-allyl-amino-demethoxygeldanamycin (17-AAG) promotes degradation of several proteins through the ubiquitin-proteasome pathway, including oncogenic Raf, v-Src, erbB2, and BCR-ABL. We have previously shown that 17-AAG prevents hsp90/NPM-ALK complex formation and fosters NPM-ALK turnover, perhaps through its association with the hsp70 chaperone. Here, we show that inhibition of the proteasome activity by the potent and specific compound pyrazylcarbonyl-Phe-Leu-boronate (PS-341) blocks 17-AAG-induced down-regulation of NPM-ALK, which becomes detergent-insoluble and relocates into ubiquitin-rich perinuclear vesicles that represent aggregated polyubiquitinated forms of the protein. Kinase activity was not mandatory for proteasomal degradation of NPM-ALK, because kinase-defective NPM-ALK was even more rapidly degraded upon 17-AAG treatment. Prolonged exposure to the proteasome inhibitor was shown to trigger caspase-3-mediated apoptosis in proliferating ALCL cells at nanomolar concentrations. However, we verified that the accumulation of detergent-insoluble NPM-ALK in ALCL cells was not a spurious consequence of PS341-committed apoptosis, because caspase inhibitors prevented poly(ADP-ribose) polymerase cleavage whereas they did not affect par-titioning of aggregated NPM-ALK. In line with these observations, the carboxyl hsp70-interacting ubiquitin ligase (CHIP), was shown to increase basal ubiquitination and turnover of NPM-ALK kinase, supporting a mechanism whereby NPM-ALK proceeds rapidly toward hsp70-assisted ubiquitin-dependent proteasomal degradation, when chaperoning activity of hsp90 is prohibited by 17-AAG.
Anaplastic large cell lymphomas (ALCLs) show recurring chromosomal translocations between chromosome 2 and chromosomes 1, 3, or 5 and in a few cases, chromosomes 17 and 19 (1) . The majority of the anaplastic lymphoma kinase (ALK)-positive ALCLs carries a t(2;5)(p23;q35) reciprocal translocation that originates a Mr 80,000 constitutively active tyrosine kinase protein, termed nucleophosmin (NPM)-ALK or p80, where the oligomerization motif within the NH2-terminal portion of the NPM protein is juxtaposed to the catalytic domain of the insulin-like ALK tyrosine kinase receptor (2) . The oncogenic potential of NPM-ALK has been demonstrated both in vivo and in vitro, because mice that receive injections of NPM-ALK-transformed murine BaF3 or Fr3T3 cells develop hematological malignancies, and ectopic expression of NPM-ALK accelerates cell cycle entry of rat fibroblasts, which also acquire the capability to grow on semi-solid support (3) . Moreover, transgenic mice in which the full-length cDNA of the npm-alk gene is under control of the murine CD4 promoter spontaneously develop T-cell lymphomas or plasma cell neoplasms (4) and lethally irradiated interleukin 9 transgenic mice reconstituted with NPM-ALK-transduced mouse bone marrow progenitors develop either B-cell lymphomas or T-cell lymphoblastic lymphomas (5) . Cytoplasm-localized NPM-ALK homodimers phosphorylate canonical receptor tyrosine kinase adapter proteins such as Shc, IRS-1, Grb2, or PLCγ1, leading to activation of signaling pathways involved in both cell proliferation and survival (6 , 7) . Indeed, whereas NPM-ALK association with and phosphorylation of p85 regulatory subunit of phosphatidylinositol 3′-kinase activates Akt and reduces Bad-dependent apoptosis (8 , 9) , phosphorylation of STAT3 transcription factor results in the up-regulation of antiapoptotic protein Bcl-xL, as well as in transcription of cell-cycle-related genes (10 , 11) . We have recently shown that expression of mature NPM-ALK kinase requires the interaction with the molecular chaperone heat shock protein (hsp) 90, and inhibition of the chaperoning activity by the benzoquinone 17-allyl-amino-demethoxygeldanamycin (17-AAG) reduces NPM-ALK content and negatively regulates its signaling pathway in ALCL cells (12) . The exact mechanism of NPM-ALK degradation is still unknown, but our previous findings that destabilization of NPM-ALK kinase is preceded by binding hsp70 (a chaperone involved both in folding and proteasomal degradation of certain hsp90-client proteins) led us to speculate in favor of the 26S proteasome as degradation pathway of misfolded NPM-ALK. Several other hsp90 chaperone substrates, such as Akt (13) , Raf-1 (14) , BCR-ABL (15) , and CFTR (16) , have been reported to bind hsp70, before degradation, in cells treated with hsp90 inhibitors. The ability of hsp70 to drive protein degradation, rather than folding, depends on the interaction with the E3 ubiquitin ligase carboxyl terminus hsp70-interacting protein (CHIP). CHIP binds to hsp70 through NH2-terminal tetratricopeptide (TPR) motifs (17) while mediating ubiquitination of chaperone-bound polypeptides through a COOH-terminal U-box domain (18) . Here, we demonstrate that the inhibition of proteasome activity prevents 17-AAG-dependent degradation of NPM-ALK, resulting in the accumulation of NPM-ALK kinase in the detergent-insoluble fraction of treated cells. Furthermore, we provide evidence that CHIP ubiquitin ligase associates with NPM-ALK and negatively regulates NPM-ALK steady state in transfected cells by promoting its ubiquitination.
MATERIALS AND METHODS
The human ALK-positive ALCL cell lines Karpas299 and SR786 and the ALK-negative FE-PD (19) were maintained in RPMI 1640 containing 15% heat-inactivated FCS, 2 mmol/liter glutamine, 100 units/ml penicillin, and 100 μg/ml streptomycin under standard tissue-culture conditions. COS-7 monkey renal epithelial cells were cultivated in Dulbecco’s modified medium, supplemented with 10% heat-inactivated FCS, glutamine, penicillin, and streptomycin, as described above.
Reagents and Antibodies.
17-AAG and pyrazylcarbonyl-Phe-Leu-boronate (PS-341) were obtained from L. Neckers and Edward Mimnaugh (National Cancer Institute, Bethesda, MD). Anti-poly(ADP-ribose) polymerase (PARP) and anti-CHIP antibodies were purchased from Calbiochem-Novabiochem Corp. (San Diego, CA). Monoclonal antiactin IgM ascites and horseradish peroxidase goat antimouse IgM were bought from Oncogene Research Products (Boston, MA). Rabbit polyclonal antihistone deacetylase 1 and monoclonal antiubiquitin (P4D1) antibodies and antiphosphotyrosine (PY99) and anti-hsp70 monoclonal antibodies were from Santa Cruz Biotechnology (Santa Cruz, CA). Anti-ALK (ALKc) and anti-NPM (NPM-376) monoclonal antibodies were a generous gift from Prof. B. Falini (University of Perugia, Perugia, Italy). Caspase inhibitors Z-VAD-FMK, Z-AEVD-FMK, DEVD-CHO, and Z-LEHD-FMK and protease inhibitors N-acetyl-l-leucyl-l-leucyl-norleucinal (ALLnL), N-acetyl-Leu-Leu-Met-CHO (LLM), pepstatin, [4-(2-aminoethyl) benzenesulfonylfluoride, HCl] (AEBSF), calpeptin, and chathepsin inhibitor I were purchased from Calbiochem. Phenylmethylsulfonyl fluoride was purchased from Sigma-Aldrich Co. (St. Louis, MO). Leupeptin, aprotinin, and rabbit antimouse IgG1 were obtained from CAPPEL (ICN Biomedicals Inc., Costa Mesa, CA). 4′,6-Diamidino-2-phenylindole nucleic acid stain, fluorophore-conjugated goat antimouse (Alexa546), and goat antirabbit (Alexa488) antibodies were bought from Molecular Probes (Eugene, OR). Horseradish peroxidase-conjugated sheep antimouse and donkey antirabbit antibodies were purchased from Amersham Biosciences (Uppsala, Sweden), as well as protein A-Sepharose beads and protein G-Sepharose Fast-Flow beads. BCA protein assay and Western blot chemiluminescence reagents were all from Pierce Chemical Co. (Rockford, IL). Protran nitrocellulose membranes were from Schleicher & Schuell. All of the other chemicals used in this study were purchased from Sigma-Aldrich.
Immunoprecipitation and Immunoblotting.
ALCL and COS-7 cells were lysed as described previously (12) . In brief, cells were washed twice in ice-cold 1× PBS and incubated on ice for 20 min in 1% Triton X-100 lysis buffer. After high-speed clarification at 4°C, supernatants were recovered, and pellets were resuspended in the same volume of lysis buffer and sonicated on ice. Thirty to 60 μg of lysates from both fractions were loaded onto 10% SDS-PAGE and electrotransferred to nitrocellulose membranes. Alternatively, when immunoprecipitation was performed, 500 μg of lysates were incubated for 2–12 h at 4°C with anti-ALK, anti-CHIP, or anti-hsp70 antibody, where indicated. Immunocomplexes were adsorbed to 30 μl of protein G-Sepharose beads or to 150 μl of protein A-Sepharose beads previously conjugated to rabbit antimouse IgG1 antibody and incubated 2 h at 4°C. The immunoadsorbed pellets were washed in 1% Triton X-100 lysis buffer and heated at 95°C in 1× reducing Laemmli loading buffer, before loading onto SDS-PAGE. To assess NPM-ALK ubiquitination, cell lysates were prepared in radioimmunoprecipitation assay buffer (Triton lysis buffer supplemented with 0.1% SDS, 0.5% sodium deoxycholate, 1 mm phenylmethylsulfonyl fluoride (PMSF), 1 mm sodium orthovanadate, and 10 mm sodium fluoride), and SDS-PAGE resolved samples were transferred onto polyvinylidene difluoride membranes. Proteins were visualized by chemiluminescence using a commercial kit (Pierce Chemical Co.). Films (Hyperfilm; Amersham Biosciences) were scanned, and results were analyzed by using image analysis software (NIH Image).
Preparation of Cytoplasmic and Nuclear Fractions.
To obtain purified cytoplasmic and nuclear fractions, cells were processed according to the protocol described by Rosette and Karin (20) . After washing in PBS, cells were frozen in dry ice and immediately thawed on ice in pH 7.9 hypotonic buffer [10 mm 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid, 10 mm KCl, 0.2 mm EDTA, 0.1 mm EGTA, 1 mm DTT, 1 mm phenylmethylsulfonyl fluoride, 20 μg/ml leupeptin, and 20 μg/ml aprotinin]. NP40 (2%) was then added, and cells were further lysed by passing through a 26-gauge needle. Soluble cytoplasmic proteins were recovered by clarification at 10,000 × g, whereas intact nuclei were resuspended in pH 7.9 hypertonic buffer [20 mm 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid, 0.4 m NaCl, 1 mm EDTA, and 1 mm DTT] containing protease inhibitors. Nuclear extracts were recovered from pestle-broken nuclei (Kimble/Kontes Scientific Glassware Instruments, Vineland, NJ), rocked on ice for 30 min, and clarified by high-speed centrifugation (10,000 × g). Supernatant represents soluble nuclear extracts.
Preparation of Expression Constructs and Transient Transfection.
The whole npm-alk coding region, originally cloned into the pSRαMSVtkneo retroviral vector (3) , was subcloned into the pcDNA3 vector (Invitrogen). Point mutant NPM-ALKK210A (K210A) was created with QuikChange Site-Directed Mutagenesis Kit system (Stratagene, La Jolla, CA). Wild-type CHIP and the two point mutants K30A and H260Q, cloned into the pcDNA3.1 vector, are kind gifts from L. Neckers. To transiently express the constructs described above, COS-7 cells (0.2 × 106) were plated in 100-mm dishes; after 72 h, log-phase growing cells were transfected with 35 μg each of NPM-ALK or NPM-ALKK210 or with 20 μg when either construct was cotransfected with equal amounts of β-galactosidase (β-gal) or CHIP expression vectors, by calcium-phosphate DNA coprecipitation. Thirty-six h after transfection, COS-7 cells were lysed in ice-cold lysis buffer or additionally treated with 17-AAG and/or PS-341, where indicated. Lysates were clarified, immunoprecipitated, and resolved by SDS-PAGE as described above.
Karpas299 cells (3 × 103) were plated onto 8 well/chamber slides (Becton Dickinson, Franklin Lakes, NJ) and grown for 72 h. Cells were then incubated for 24 h with 17-AAG, PS-341, or both compounds, as reported in Fig. 3 ⇓ . Exponentially growing cells were fixed in 3.7% formaldehyde, permeabilized in 0.2% Triton X-100, and blocked in 10% FCS-PBS, followed by incubation in 100 mm glycin. After permeabilization, cells were incubated 1 h at 37°C with ALK-11 rabbit anti-ALK antibody (a kind gift of S. W. Morris, St. Jude Children’s Research Hospital, Memphis, TN) and with a mouse antiubiquitin antibody. After extensive washing in PBS, cells were incubated with fluorophore-conjugated secondary antibodies goat antirabbit (Alexa488) and goat antimouse (Alexa546) at a 1:500 dilution of 2 mg/ml stock. Cells were then washed in PBS and mounted onto slides with 1:1 PBS/glycerol, with addition of 4′,6-diamidino-2-phenylindole nucleic acid stain at a 1:1000 dilution. Cells were observed at ×63/0.75 numerical aperature in a Leica DMBL microscope. Images were obtained with a Leica DC 300F digital camera and prepared for reproduction with Leica IM1000 imaging software (Leica Microsystems Ltd., Wetzlar, Germany).
NPM-ALK Accumulates in the Detergent-Insoluble Cell Fraction after Proteasome Inhibition.
Degradation of receptor tyrosine kinases still remains an intriguing issue, because immature receptor proteins are usually degraded through the 26S proteasome, whereas proteolysis of mature receptors upon ligand-dependent endocytosis and ubiquitination may be both lysosome and proteasome dependent (21, 22, 23, 24, 25, 26) . Because we have previously shown that the ALK receptor protein moiety is responsible for the accelerated degradation of NPM-ALK kinase when hsp90-folding activity is inhibited by 17-AAG (12) , we screened a panel of selective protease inhibitors. We included inhibitors of both proteasomal and lysosomal function to characterize the degradation pathway of NPM-ALK in 17-AAG-treated ALCL cells Karpas299 and SR786 (Fig. 1, A and B ⇓ , respectively). Exponentially growing cells were treated for 8 h with inhibitors of lysosomal (leupeptin and pepstatin), calpain (cathepsin inhibitor I and calpeptin), proteasomal (PS-341, ALLnL and LLM), and serine/threonine (AEBSF, PMSF) proteases in the presence of 10 μm 17-AAG. Surprisingly, none of the inhibitors used had an effect on NPM-ALK expression in the clarified lysates of both cell lines. However, in ALCL cells incubated with proteasome inhibitors PS-341, ALLnL, and LLM, degradation of NPM-ALK caused by 17-AAG was somehow prevented, because NPM-ALK accumulated consistently in detergent-insoluble pellet fractions (Fig. 1, A and B ⇓ , graph). Given that boronic acid dipeptides (PS-341) and peptide aldehydes (ALLnL and LLM) are promiscuous inhibitors, because they also target certain serine and cysteine proteases besides the proteasome (27) , we analyzed more specific inhibitors of such enzymes AEBSF, cathepsin inhibitor I, and calpeptin. Yet, because no similar accumulation of NPM-ALK protein occurred with such more specific compounds, it is likely that NPM-ALK is subjected to proteasomal degradation when its hsp90-assisted maturation is impeded by 17-AAG.
To investigate the kinetics of NPM-ALK degradation, Karpas299 cells were exposed to 0.5 μm 17-AAG for 24 h, and proteasome activity was inhibited for 16 or 24 h (Fig. 1C ⇓ , Lanes 1–7). Unlike the Mr 38,000 protein NPM (Fig. 1C ⇓ , Lanes 8–14), NPM-ALK protein level was reduced both in clarified lysates and detergent-insoluble pellet fractions after treatment with 17-AAG (Fig. 1C ⇓ , Lane 3). When added to 17-AAG, PS-341 inhibitor did not prevent the down-regulation of detergent-soluble NPM-ALK but caused time-dependent partitioning of NPM-ALK in the insoluble pool (Fig. 1C ⇓ , Lanes 6 and 7). When Karpas299 cells were grown in the presence of both 17-AAG and PS-341 for increasing time intervals, time-dependent down-regulation of soluble NPM-ALK caused by 17-AAG paralleled exactly the up-regulation of insoluble NPM-ALK mediated by PS-341 (Fig. 1D) ⇓ , in accordance with what was shown previously for other hsp90-client proteins (28, 29, 30) .
PS-341-Induced Apoptosis Does Not Affect Proteasomal Degradation of NPM-ALK.
PS-341 was the first proteasome inhibitor to enter Phase I clinical trial for its antitumor activity, because tumor cells are more sensitive than normal cells to PS-341-dependent impairment of cell growth and induction apoptosis (31 , 32) , as recently shown in multiple myeloma cells both in vivo and in vitro (33 , 34) . Pro-apoptotic activity of PS-341 was assessed both in t(2;5)-positive (Karpas299, SUDHL1, and SR786) and t(2;5)-negative (FE-PD) ALCL cell lines, by exposing cells for 48 h to nanomolar concentrations of inhibitor. As shown in Fig. 2A ⇓ , in vitro cell proliferation was inhibited by more than 90% at 10 nm in all of the cell lines and regardless of NPM-ALK expression (Fig. 2A ⇓ , graph). Indeed, exposure to PS-341 caused time-dependent cleavage of the Mr 116,000 apoptotic hallmark PARP (Fig. 2A ⇓ , panels) into its caspase-cleaved fragment of Mr 89,000 (Fig. 2A ⇓ , arrowheads) in both ALK-positive (SUDHL1 and SR786) and ALK-negative (FE-PD) cells.
Thus, to rule out that the accumulation of insoluble NPM-ALK might be an artifact of PS-341-driven apoptosis rather than inhibition of its proteasomal degradation, Karpas299 and SR786 cells (Fig. 2B ⇓ , Lanes 1–6 and 7–12, respectively) were treated with two pan-caspase inhibitors before being exposed for as long as 8 h to higher doses of 17-AAG and PS-341. Consistently with PARP cleavage shown in Fig. 2A ⇓ , the two cell lines responded differently to PS-341 treatment, because down-regulation of mature PARP protein was strongly induced in SR786 but not in Karpas299. In SR786 cells pretreated with Z-VAD-FMK and Z-AEVD-FMK inhibitors, caspase-dependent PARP cleavage was completely blocked (Fig. 2B ⇓ , graph), and PARP protein level rose even higher than in untreated cells, which might possess a constitutive caspase activity, as appears from PARP immunoblots of Fig. 2A ⇓ . In contrast, after incubation with both 17-AAG and PS-341, aggregated NPM-ALK was found in the pellet fraction both in the absence of apoptosis (Fig. 2B ⇓ , Lanes 4–6) and when PS-341-induced apoptosis (SR786) was prevented by the two caspase inhibitors (Fig. 2B ⇓ , Lanes 10–12). When Karpas299 cells were treated longer (24 h) with PS-341, extensive PARP cleavage occurred (Fig. 2C ⇓ , graph), and it was totally prevented by the two specific caspase-3 and caspase-9 inhibitors, DEVD-CHO and Z-LEHD-FMK, respectively. However, redistribution of NPM-ALK upon proteasome inhibition observed in the detergent-insoluble fraction of 17-AAG-treated cells (Fig. 2C ⇓ , Lane 4) was not prevented when both caspase-3 and caspase-9 enzymes were inhibited (Fig. 2C ⇓ , Lanes 5 and 6), hence proving that increase of insoluble NPM-ALK and apoptosis are two independent phenomena in PS-341-treated ALCL cells.
NPM-ALK Relocates in Ubiquitin-Rich Perinuclear Structures after Proteasome Inhibition.
Covalent attachment of multiple molecules of ubiquitin to specific lysine residues of target proteins is a strict prerequisite for recognition and subsequent degradation by the 26S proteasome, and inhibition of the proteasome activity normally leads to accumulation of polyubiquitinated proteins into cells (35) . Thus, we performed fluorescent microscopy analysis of ALCL cells stained for NPM-ALK and ubiquitin (Fig. 3A ⇓ , panels A–D and E–H, respectively), to verify whether detergent-insoluble NPM-ALK corresponds to deposition of ubiquitinated forms of the protein in cells treated with both 17-AAG and PS-341. As shown in Fig. 3A ⇓ , exposing Karpas299 to PS-341 alone caused accumulation of ubiquitinated proteins into the cytoplasm (Fig. 3A ⇓ , panel G), which had no correspondence with the wide distribution of NPM-ALK (Fig. 3A ⇓ , panel C). On the other hand, when the cells were exposed to both 17-AAG and PS-341, NPM-ALK fluorescence showed a perinuclear distribution that perfectly merged with the ubiquitin staining (Fig. 3A ⇓ , panels D, H, and N, arrowheads). Besides, when part of the cells were lysed and cytosolic and nuclear soluble fractions were purified and analyzed together with sonicated detergent-insoluble pellets, accumulation of NPM-ALK was found only in the pellet fraction of cells exposed to both 17-AAG and PS-341 (Fig. 3 ⇓ B, Lane 12). This indicates that juxtanuclear-localized NPM-ALK observed in fixed cells represents insoluble ubiquitinated NPM-ALK. Furthermore, we observed that chaperones hsp70 and hsp40, but not hsp90, were increased consistently in the detergent-insoluble fraction of NPM-ALK-transfected cells treated with 17-AAG and PS-341 (data not shown). Because this is consistent with the evidence that, upon prolonged inhibition of proteasome activity, molecular chaperones relocate in large perinuclear structures with polyubiquitinated misfolded client proteins (36 , 37) , it is likely that the deposition of detergent-insoluble NPM-ALK may result from the impairment of its chaperone-assisted proteasomal degradation.
17-AAG-Induced Degradation of NPM-ALK Is Independent of Its Tyrosine Kinase Activity.
Ubiquitin-dependent proteasomal degradation of several hsp90-client proteins has been shown to rely upon interaction with the hsp70 chaperone, independently from their biological functions (38, 39, 40) . To test whether kinase activity was mandatory for NPM-ALK proteasomal degradation, we generated a kinase-defective mutant of NPM-ALK by replacing lysine 210 of the nucleotide-binding domain with alanine (K210A). Wild-type NPM-ALK and kinase-defective K210A mutant, when transiently expressed in COS-7 cells, displayed comparable steady-state levels, although the K210A mutant failed to react with the antiphosphotyrosine antibody that marked NPM-ALK immunoprecipitates (Fig. 4A) ⇓ . As expected, 17-AAG inhibited hsp90-driven maturation of both wild-type NPM-ALK and NPM-ALK K210A, causing their rapid depletion in COS-7 cells (Fig. 4B ⇓ , Lane 2). Nevertheless, when added to 17-AAG, PS-341 prevented the loss of the two misfolded proteins and enhanced the extent of their aggregation in the detergent-insoluble pellet fraction (Fig. 4B ⇓ , Lane 8). Interestingly, the K210A mutant showed enhanced sensitivity to hsp90 inhibition (Fig. 4B ⇓ , Lane 2), similarly to what was found recently for kinase-dead EGFR (41) . When the kinetics of the down-regulation of the two proteins were compared, 10 μm 17-AAG led to almost complete depletion of the K210A mutant after just 2 h of treatment, whereas wild-type NPM-ALK was still detectable even after 8 h (Fig. 4C ⇓ , arrowhead).
The Ubiquitin Ligase CHIP Down-Regulates NPM-ALK Kinase Expression.
Chaperone-assisted folding requires the cooperation of several scaffold proteins and cofactors of the hsp70 and hsp90 multimolecular chaperone complexes (42) . The TPR-containing protein CHIP has been shown to suppress the folding and catalyze the ubiquitination of certain hsp90-client proteins when bound to hsp70, by virtue of a COOH-terminal U-box domain (43 , 44) . Point mutations at lysine 30 within the TPR domain or at histidine 260 of the U-box domain abrogate respectively CHIP ability to complex with hsp70 and its ubiquitin ligase activity, both mandatory for ubiquitin-dependent proteasomal degradation of chaperone-bound substrates (45) . Because we have recently shown that after 17-AAG treatment, NPM-ALK/hsp70 association increases and precedes NPM-ALK down-regulation in ALCL cells, we wondered whether CHIP could regulate NPM-ALK turnover, perhaps throughout hsp70. Therefore, we examined NPM-ALK steady state in COS-7 cells expressing wild-type CHIP or each of the two point mutants described above (CHIP K30A and CHIP H260Q) or β-gal protein. In accordance with our hypothesis, wild-type CHIP protein markedly diminished NPM-ALK expression when compared with β-gal or the two point mutants (Fig. 5A ⇓ , Lanes 1–4), and it stimulated kinase-defective NPM-ALK K210A degradation (Fig. 5A ⇓ , Lanes 5–9). When an anti-CHIP antibody was used for immunoprecipitation, NPM-ALK precipitated from clarified lysates of CHIP-transfected and CHIP H260Q-transfected cells (Fig. 5B ⇓ , Lanes 4 and 6) but not from β-gal or K30A cell lysates (Fig. 5B ⇓ , Lanes 2 and 5). Control immunoprecipitation, performed by immunoadsorbing NPM-ALK with ALK-C antibody, ruled out any spurious effect by detecting NPM-ALK in all of the above described cell lysates, except in lysates of untransfected COS-7 (Fig. 5B ⇓ , Lane 8). In addition, we carried out reverse co-immunoprecipitation and confirmed that, unlike CHIP K30A, wild-type CHIP and CHIP H260Q associated with hsp70 and NPM-ALK (Fig. 5B ⇓ , Lanes 13 and 14 and 19 and 20, respectively) as well as with kinase-defective NPM-ALK K210A (data not shown). Taken together, these data indicate that recruitment of CHIP and subsequent degradation of NPM-ALK kinase depends on NPM-ALK binding to hsp70 rather than on its kinase activity.
Additionally, we investigated the combinatorial effect of CHIP expression and 17-AAG treatment, because we noticed a similar inhibitory effect on NPM-ALK steady state. Indeed, CHIP and 17-AAG seemed to cooperate in this activity, because 17-AAG-induced depletion of soluble NPM-ALK further increased in the presence of CHIP (Fig. 5C) ⇓ . In contrast, CHIP H260Q mutant, which lacks enzymatic activity, neither affected soluble NPM-ALK level in untreated cells nor prevented its down-regulation in the presence of 17-AAG. Yet, it is worth pointing out that CHIP may cause apparent down-regulation of hsp90 substrates independently of its ubiquitin ligase activity, by partitioning misfolded hsp90 substrates in the insoluble pool (46) . In fact, cotransfection of NPM-ALK with both wild-type CHIP and CHIP H260Q resulted in the redistribution of NPM-ALK from the soluble to the insoluble pool even in the absence of 17-AAG (Fig. 5C) ⇓ , whereas exposure to 17-AAG treatment was additive to such an effect.
CHIP Enhances NPM-ALK Ubiquitination in COS-7 Cells.
To assess whether CHIP ubiquitin ligase activity stimulates NPM-ALK ubiquitination, COS-7 cells untransfected or transfected with plasmids expressing NPM-ALK, β-gal, or CHIP were incubated with 10 μm 17-AAG, 3 μm PS-341, and the combination thereof for 8 h. As shown in Fig. 6A ⇓ , treating β-gal/NPM-ALK-expressing cells with 17-AAG was enough to elicit NPM-ALK ubiquitination, thus reinforcing the concept that NPM-ALK is efficiently ubiquitinated before proteasome degradation when hsp90-folding activity is inhibited. Blockade of proteasome activity by PS-341 had only a slight effect on NPM-ALK ubiquitination, whereas it enhanced the level stabilized by 17-AAG (Fig. 6A ⇓ , Lanes 1–4). In the presence of CHIP, however, polyubiquitinated NPM-ALK was precipitated even from the clarified lysates of untreated cells, and the extent of NPM-ALK ubiquitination was augmented dramatically after treatment with 17-AAG and PS-341 (Fig. 6A ⇓ , Lanes 9–12). As expected, immunoadsorbed lysates from untransfected (Fig. 6A ⇓ , Lanes 5–8) or β-gal/CHIP-transfected cells (Fig. 6A ⇓ , Lanes 13 and 14) did not contain any higher molecular-weight form of ubiquitinated NPM-ALK, which indicates the relative specificity of CHIP-induced NPM-ALK ubiquitination. Remarkably, we found that the association between NPM-ALK and CHIP was stabilized by the proteasome inhibitor PS-341 in the presence of 17-AAG (Fig. 6A ⇓ , bottom), although CHIP steady state remained unaltered (Fig. 6B) ⇓ . Thus, we can conclude that CHIP supports the ubiquitination of NPM-ALK per se and in the presence of 17-AAG, and blockade of the proteasome catalytic activity results in the accumulation of CHIP-bound polyubiquitinated NPM-ALK.
NPM-ALK is a Mr 80,000 fusion tyrosine kinase originating from a reciprocal translocation between chromosomes 2 and 5, which regulates cell-cycle progression of ALK-positive ALCL lymphomas (47) and antiapoptotic behavior of transformed cells (3) upon self-association and phosphorylation. Ansamycin drugs such as herbimycin, geldanamycin, and its derivative 17-AAG have been found to diminish NPM-ALK expression and tyrosine phosphorylation in ALCL cells, leading to G1-phase arrest and apoptosis (48 , 49) . Despite their initial use as tyrosine kinase inhibitors, ansamycins are a class of small molecules with no inhibitory kinase activity in vitro (50) but are capable of binding to and competing with ATP for the ATP-binding site of the molecular chaperone hsp90, which regulates late-stage folding and maturation of several signaling proteins. Drug-induced impairment of hsp90-chaperoning activity has been shown to promote ubiquitin-dependent proteasomal degradation of non-native substrates (51 , 52) , and we have previously shown that inhibition of hsp90 activity by 17-AAG prevents NPM-ALK/hsp90 complex formation and accelerates NPM-ALK turnover (12) .
Covalent attachment of multiple ubiquitin molecules to selected lysine residues targets for ATP-dependent proteolysis by the 26S proteasome both normal native and unfolded proteins, as well as damaged polypeptides (35) . However, ubiquitin-dependent degradation of growth factor receptors is still an intriguing process, because poly-ubiquitination of ligand-activated receptors at the plasma membrane has been shown to accelerate endocytosis (22 , 53, 54, 55) , whereas subsequent proteolysis of endosome-sequestered receptors may be hindered either by proteasomal or lysosomal inhibitors (24 , 25) . To investigate the degradation pathway of NPM-ALK, we screened a panel of lysosomal, endosomal, and proteasomal inhibitors, because the COOH-terminal portion of NPM-ALK corresponds to the entire cytoplasmic domain of the transmembrane receptor ALK, possibly including sequences relevant for both proteasomal and/or lysosomal proteolysis. When the ALCL cell lines Karpas299 and SR786 were maintained in the presence of both 17-AAG and the different protease inhibitors described in Fig. 1 ⇓ , 17-AAG-induced degradation was prevented and NPM-ALK accumulated in the detergent-insoluble fraction only in the presence of proteasome inhibitors. The observation that aggregated NPM-ALK did not accumulate when selective inhibitors of serine, cysteine, and lysosomal proteases were used suggests that the proteasome may represent the preferred degradation pathway of NPM-ALK in 17-AAG-treated cells.
Abrogation of proteasomal activity by PS-341 inhibitor has been found to block cell cycle progression of tumor cells (31 , 32) and to induce apoptosis through nuclear factor κB (NFκB) antiapoptotic activity overcome (56) , cytochrome c release (57) , Bcl-2 cleavage (58) , and c-Jun NH2-terminal kinase-dependent caspase-3 and caspase-8 activation (59) . The effect, which increases tumor cell sensitivity to DNA-damaging chemotherapy both in vitro and in vivo (60 , 61) , can be prevented by pan-caspase or caspase-3 inhibitors (58) . In line with these observations, we found that PS-341 did block in vitro growth of all of the ALCL cell lines tested, regardless of NPM-ALK kinase expression, and caused apoptosis in a time- and dose-dependent manner, as assessed by caspase-dependent cleavage of PARP protein (Fig. 2A) ⇓ . However, even though PS-341-induced fragmentation of PARP was prevented by inactivation of both caspase-3 and caspase-9 (Fig. 2C) ⇓ or by using pan-caspase inhibitors (Fig. 2B) ⇓ , aggregated NPM-ALK was still found in the detergent-insoluble pellet fraction of 17-AAG/PS-341-treated cells, proving that proteasomal degradation of NPM-ALK and apoptosis occur independently of each other.
Down-regulation of detergent-soluble NPM-ALK and increase of its detergent-insoluble pool after exposing 17-AAG-treated cells to PS-341 occurred simultaneously in a time-dependent manner (Fig. 1, C and D) ⇓ , similarly to what was described recently for the hsp90 client protein Akt (28) . The loss of solubility seems to be featuring many other misfolded chaperone substrates when their binding to hsp90 is prevented and proteasome activity is sequentially inhibited (15 , 29 , 30 , 62 , 63) , and this correlates to intracellular deposition of polyubiquitinated protein aggregates with a preferred juxtanuclear distribution (64 , 65) into vimentin-surrounded inclusions known as aggresomes (37 , 65) . Indeed, PS-341 caused accumulation of ubiquitinated proteins in Karpas299 cells, and the extent further increased in the presence of 17-AAG (Fig. 3A) ⇓ . PS-341 treatment did not significantly alter steady-state and cellular localization of NPM-ALK fusion kinase in the absence of 17-AAG, but it caused NPM-ALK to colocalize with ubiquitin, when combined to the hsp90 inhibitor, into perinuclear vesicles resembling the polyubiquitinated protein aggregates described (Fig. 3) ⇓ . The data presented above support a mechanism whereby 17-AAG-induced misfolding and degradation of NPM-ALK occurs rapidly and efficiently unless proteasome activity is perturbed, likely indicating a relationship between chaperone-assisted folding of NPM-ALK and the ubiquitin/proteasome pathway.
As recently shown, the balance between protein folding and destruction is reached through the activity of chaperone-bound cofactors, such as CHIP and BAG-1, that diverge misfolded or damaged chaperone-client proteins to proteasomal degradation rather than allowing their folding. CHIP, in particular, acts as an E3 ubiquitin ligase and promotes ubiquitination of hsp90-client proteins by virtue of a COOH-terminal U-box domain (43) that recruits the E2-conjugating enzyme UBCH5a (18) . Through multiple TPR repeats, CHIP binds to hsp70 and hsp90 in a mutually exclusive manner, preventing hsp70-folding activity and active hsp90/p23 complex formation, respectively (17) . Point mutation analysis has shown that both the TPR repeats and U-box domain are mandatory for CHIP ubiquitin ligase activity, because rapid degradation of misfolded proteins in the presence of wild-type CHIP cannot be elicited by either TPR (CHIP K30A) or U-box (CHIP H260Q) point mutants (44 , 45 , 64 , 66) . Consistent with this mechanism, we found that CHIP associated with and increased protein turnover of NPM-ALK when expressed in COS-7 cells (Fig. 5) ⇓ . In contrast, in the presence of CHIP H260Q, which binds stronger to hsp70 and NPM-ALK but shows impaired ubiquitin-ligase activity, or CHIP K30A, which does not bind hsp70, NPM-ALK steady state was unaffected, hence proving that ubiquitin-dependent degradation of NPM-ALK takes place when the mutual interaction between CHIP and hsp70 exists. Interestingly, CHIP and 17-AAG exert a similar effect on NPM-ALK stability, which indicates that although some redundancy is likely to occur, a cooperation between the two mechanisms, as previously found for erbB2 (45 , 67) , may be feasible. Indeed, when COS-7 cells were exposed to 17-AAG, drug-induced degradation of exogenous NPM-ALK was enhanced in the presence of wild-type CHIP, whereas U-box mutant CHIP H260Q did not counteract 17-AAG-induced depletion of soluble NPM-ALK but rather stabilized its insoluble pool (Fig. 5C) ⇓ . By acting as an hsp70-folding inhibitor, CHIP may indeed decrease soluble protein levels without stimulating the ubiquitin-dependent degradation pathway but by causing aggregation of misfolded proteins (46) . Accordingly, the partition of aggregated NPM-ALK in the detergent-insoluble pellet fraction was observed in cells expressing wild-type or U-box mutant CHIP (Fig. 5C) ⇓ but not in those cells transfected with CHIP K30A (data not shown). However, when the ubiquitination status was assessed, we verified that CHIP dramatically raised the content of ubiquitinated NPM-ALK when proteasome activity was inhibited in the presence of 17-AAG (Fig. 6) ⇓ , thus indicating that CHIP represents the E3 ubiquitin ligase responsible for the ubiquitination of misfolded NPM-ALK in 17-AAG-treated cells.
Ubiquitin-dependent down-regulation of activated receptor tyrosine kinases at the plasma membrane has been shown to occur upon recruitment of another E3 ubiquitin ligase, Cbl, although that drives endocytosis and lysosomal degradation of monoubiquitinated receptor proteins (68) . Unlike CHIP, Cbl needs to be tyrosine phosphorylated before activation and catalytically inactive tyrosine kinases are not affected by Cbl (24 , 69 , 70) . In contrast, in transiently transfected COS-7 cells, kinase-dead NPM-ALK K210A and CHIP associated (data not shown) and CHIP dictated the down-regulation of kinase-dead NPM-ALK (Fig. 5A) ⇓ . Besides, kinase-defective NPM-ALK K210A appeared to be more dependent on hsp90-chaperoning activity than wild-type NPM-ALK, as confirmed by the higher responsiveness to and the enhanced decay rate of K210A mutant in the presence of 17-AAG (Fig. 4) ⇓ . In agreement with our data, kinase-dead EGFR was recently shown to bind to hsp90 with higher affinity than wild-type EGFR and to be more rapidly degraded by the hsp90 inhibitor geldanamycin (41) . Similarly, folding and stabilization of newly synthesized kinase-inactive heme-regulated eukaryotic initiation factor kinase has been found to depend on hsp90 and to be sensitive to the inhibitory effect of geldanamycin until its kinase domain is activated (71) .
In conclusion, the present study strongly supports the hypothesis that 17-AAG-induced down-regulation of NPM-ALK kinase occurs through the ubiquitin/proteasome degradation pathway, because inhibition of proteasome activity by PS-341 prevents the depletion of both endogenous and exogenous NPM-ALK induced by 17-AAG. We have hereby shown that combined inhibition of hsp90 and proteasome activity causes NPM-ALK to accumulate insoluble and to relocate in ubiquitin-rich perinuclear vesicles that, resembling aggresome formations, couple NPM-ALK dislocation to degradation. Furthermore, we found a correlation between chaperone-assisted folding and proteasomal degradation of NPM-ALK by showing that CHIP ubiquitin ligase, when bound to hsp70, associates with and enhances the turnover of ubiquitinated NPM-ALK. Finally, what we find particularly interesting from a pharmacological viewpoint, is that inactivating mutations of kinase activity were found to increase NPM-ALK responsiveness toward inhibitors of the hsp90-chaperoning machinery, which makes the development of NPM-ALK kinase inhibitors a promising anticancer strategy when used in combination with hsp90 inhibitors.
We thank Dr. B. Falini for anti-ALK (ALK-C) and anti-NPM (NPM-376) monoclonal antibodies; Dr. S. W. Morris (St. Jude Children’s Research Hospital) for ALK-11 polyclonal antibody; Dr. Martina Boscaro (Education and Communication Office, Istituto Oncologico Veneto) for help in editing the manuscript; and E. Tosato for technical support.
Grant support: Consiglio Nazionale delle Ricerche-Ministero Istruzione Università e Ricerca, Settore Oncologia, Legge 449/97, and Fondazione Città della Speranza. P. Bonvini was a recipient of a fellowship from Istituto Oncologico Veneto.
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.
Requests for reprints: Paolo Bonvini, Clinica di Oncoematologia Pediatrica, Azienda Ospedaliera-Università di Padova, via Giustiniani 3, 35128 Padova, Italy. Phone: 39-049-8215678; Fax: 39-049-8215679; E-mail:
- Received November 11, 2003.
- Revision received February 10, 2004.
- Accepted February 26, 2004.
- ©2004 American Association for Cancer Research.