Stability of the Peutz–Jeghers Syndrome Kinase LKB1 Requires Its Binding to the Molecular Chaperones Hsp90/Cdc37

Stability of the Peutz–Jeghers syndrome kinase LKB1 requires its binding to the molecular chaperones Hsp90/Cdc37

Pascale Nony1,He´ le` ne Gaude1, Mireille Rossel2, Laurence Fournier1, Jean-Pierre Rouault3 and Marc Billaud*,1

1Laboratoire Ge´ne´tique et Cancer, FRE 2692, Domaine Rockefeller, 8, avenue Rockefeller, 69373 Lyon Cedex 08, France; 2Laboratoire Biologie Cellulaire Quantitative, Ecole Pratique des Hautes Etudes, INSERM E0343, Place E Bataillon, 34095 Montpellier Cedex 05, France; 3Laboratoire de Communications Cellulaires et Diffe´renciation, Inserm U418, INRA UA 953, Hoˆpital Debrousse, 29 rue Sœur Bouvier, 69322 Lyon Cedex 05, France

Peutz–Jeghers syndrome (PJS) is an autosomal dominant particularly of the breast, intestine, testis, cervix and disorder characterized by the presence of multiple pancreas (Giardiello et al., 1987; Spigelman et al., 1989). gastrointestinal polyps and an increased risk for various Germline mutations of the LKB1 gene (also called types of cancers. Inactivating germline mutations of the STK11) located on chromosome 19pl3.3 and encoding a LKB1 gene, which encodes a serine/threonine kinase, are serine/threonine kinase are the underlying genetic cause responsible for the majority of PJS cases. Here, we show of the majority of PJS cases (Hemminki et al., 1998; that the heteromeric complex containing the molecular Jenne et al., 1998). These mutations which include small chaperones Hsp90 and Cdc37/p50 interacts with the deletions and point mutations, disrupt LKB1 function kinase domain of LKB1. Treatment of cells with either (Mehenni et al., 1998; Ylikorkala et al., 1999). The loss geldanamycin or novobiocin, two pharmacological inhibi- of heterozygosity (LOH) at the LKB1 locus found in a tors of Hsp90 causes the destabilization of LKB1. fraction of PJS polyps together with the inactivating Furthermore, geldanamycin treatment leads to the ubiqui- nature of the PJS germline mutations are consistent with tination and the rapid degradation of LKB1 by the a tumor suppressor role for LKB1 (Hemminki et al., proteasome-dependent pathway. In addition, we found 1998; Avizienyte et al., 1999). While rare somatic that a LKB1 point mutation identified in a sporadic mutations of LKB1 have been identified in sporadic testicular cancer, weakens the interaction of LKB1 with cancers, inactivation of both LKB1 alleles appears a both Hsp90 and Cdc37/p50 and enhances its sensitivity to relatively common event in adenocarcinoma of the lung the destabilizing effect of geldanamycin. Collectively, our (Sanchez-Cespedes et al., 2002). results demonstrate that the Hsp90/Cdc37 complex is a The LKB1 gene has been conserved across species and major regulator of the stability of the LKB1 tumor orthologs have been identified in mouse, Xenopus suppressor. Furthermore, these data draw attention to the (Xeek1), Drosophila and Caenorhabditis elegans ( par-4) possible adverse consequences of antitumor drugs that (Su et al., 1996; Smith et al., 1999; Watts et al., 2000; target Hsp90, such as antibiotics related to geldanamycin, Martin and St Johnston, 2003). LKB1 is widely which could disrupt LKB1 function and promote the expressed during embryonic development and mice development of polyps and carcinomatous lesions. homozygous for an inactivating LKB1 mutation die at Oncogene (2003) 22, 9165–9175. doi:10.1038/sj.onc.1207179 midgestation with multiple developmental defects including abnormal vasculogenesis, neural tube mal- Keywords: LKB1; Peutz-Jeghers; Tumor suppressor; formation and mesenchymal apoptosis (Ylikorkala Hsp90 et al., 2001). Immunostaining of LKB1 in mouse and human small intestine has revealed that LKB1 is expressed in two distinct topographic regions, the crypts which contain rapidly dividing stem cells and Introduction the tip of the villi where cells undergo apoptosis and are shed into the lumen (Karuman et al., 2001; Bardeesy The Peutz–Jeghers syndrome (PJS) is a dominantly et al., 2002). Furthermore, LKB1 is also detected inherited syndrome characterized by the presence of in the epithelium of the gastric mucosa. Consistent with hamartomatous polyps in the gastrointestinal tract and these data suggesting a role for LKB1 during mor- by melanin spots on the lips, buccal mucosa and digits phogenesis of the stomach and the intestine, several (Peutz, 1921; Jeghers et al., 1949). Patients with PJS groups have recently reported that heterozygous have an increased risk of developing various cancers, LKB1 knockout mice develop multiple gastrointestinal polyps which are histologically similar to the *Correspondence: M Billaud; E-mail: [email protected] hamartomas found in PJS patients (Bardeesy et al., Received 2 April 2003; revised 23 August 2003; accepted 9 September 2002; Jishage et al., 2002; Miyoshi et al., 2002; Rossi 2003 et al., 2002). Hsp90/Cdc37 regulate LKB1 stability P Nony et al 9166 Human LKB1 consists of 433 amino acids and of either formation or maintenance of epithelial contains both a nuclear localization signal in its polarity. N-terminal noncatalytic domain and a prenylation Here, we provide evidence that the binding of consensus sequence in its C-terminus (Smith et al., molecular chaperones Hsp90 and Cdc37 to LKB1 is 1999; Collins et al., 2000; Sapkota et al., 2001). LKB1 is critical for LKB1 stabilization. In addition, treatment located in the nucleus of cells, but is also detected in the with geldanamycin, an antibiotic that disrupts the cytoplasm and at cell membranes (Smith et al., 1999; complex formed between LKB1 and Hsp90, promotes Collins et al., 2000; Sapkota et al., 2001). The ubiquitination of LKB1 and its subsequent degradation subcellular localization of LKB1 is regulated via its by the proteasome. These data suggest that the interaction with specific proteins such as LIP-1 (Smith association of Hsp90/Cdc37 to LKB1 is a rate-limiting et al., 2001), a novel molecule displaying leucine-rich event that controls LKB1’s biological activity. repeats, and STRADa, a STE20-like pseudokinase (Baas et al., 2003). Both LIP-1 and STRADa anchor LKB1 in the cytoplasm. However, LIP-1 forms a Materials and methods ternary complex with LKB1 and Smad-4 to regulate Vectors the TGF-b signalling (Smith et al., 2001) while STRADa is phosphorylated by LKB1 and enhances the catalytic The pSG-Flag eukaryotic expression vector used in this study activity of LKB1 (Baas et al., 2003). Although the was derived from the pSG5 plasmid (Stratagene) and signalling pathways which regulate LKB1 activity are contained the sequence coding for the Flag epitope tag. The human LKB1 (hLKB1) cDNA was PCR amplified from yet poorly understood, it is known that in response to human intestinal cDNA using the following primers: ionizing radiation, the ataxia-telangectasia mutated Forward: 50-GATCAGATCTCAATTGGCCGCCATG- (ATM) kinase phosphorylates LKB1 at Thr366 (Sap- GAGGTGGTGGACCCG-30 Reverse: 50-GTCCAGATCTG kota et al., 2002), whereas upon stimulation with TCGACTCACTGCTGCTTGCAGGC-30. agonists of adenylate cyclase or growth factors, both The PCR fragment was digested with BglII and subcloned the cAMP-dependent protein kinase A and p90RSK into the BamHI restriction site of the pSG5/Flag plasmid. The phosphorylates LKB1 at Ser428 (Sapkota et al., 2001). hLKB1 insert was entirely sequenced and corresponded to the It has been further proposed that LKB1 is a mediator of wild-type sequence. The pSG/Flag-LKB1 vector encodes the p53-dependent apoptosis (Karuman et al., 2001), is human LKB1 fused N-terminally to the Flag epitope. The necessary for the brahma-related gene (BRG1)-induced pSG/Flag control plasmid contains the human LKB1 cDNA carrying a missense mutation that replaces the 9th codon by a cell cycle arrest (Marignani et al., 2001) and is involved stop codon. in the regulation of the NF-kB pathway via its To generate the pSG/Flag-D243 plasmid which encodes a interaction with a novel protein called FLIP-1 (Liu truncated mutant of LKB1 (1–243 amino acids), the pSG/ et al., 2003). Finally, previous studies have revealed Flag-LKB1 plasmid was digested with BstEII which cuts at that the reintroduction of LKB1 in tumor codon 241 of the LKB1 coding sequence. After treatment with cells which lack expression of LKB1 induces a G1 T4 DNA polymerase, an adapter containing a stop codon at cell cycle arrest mediated by the CDK inhibitor codon 243 was ligated into the BstEII digested vector. To p21WAFÀ1/CIP1, thus reinforcing the notion that LKB1 generate the pSG/Flag-D88 plasmid, the pSG/Flag-LKB1 acts as a tumor suppressor (Shen et al., 2002; Tiainen plasmid was digested with BamHI which cuts at codon 88 of et al., 1999, 2002). the LKB1 coding sequence, treated with the Klenow DNA polymerase (Biolabs) and then digested with BglII. This cDNA Mutations in par-4, the C. elegans orthologue of fragment which encodes a truncated form of LKB1 (88–433 LKB1, disrupt several aspects of embryonic asymmetry amino acids) was subcloned into the pSG/Flag plasmid. The that are required for the determination of cell fates in pSG/Flag-D146–186 was obtained by partial digestion with early blastomeres (Watts et al., 2000 and references XmnI of pSG/Flag-LKB1 followed by a SgraI digestion (which herein). par-4 mutation also interfere with the determi- released a cDNA fragment coding for the region comprised nation of the worm intestine, an intriguing result between amino acids 147–185 of LKB1), treated with the considering the fact that intestinal hamartomatous Klenow enzyme and then religated. The pSG/Flag-D317 polyps are one of the cardinal clinical features of PJS plasmid was generated by directed mutagenesis (U.S.E (Watts et al., 2000 and references herein). The past year Mutagenesis Kit from Amersham Pharmacia Biotec Inc.) revealed that Drosophila LKB1 is involved in the using an oligonucleotide which contains a stop codon at codon 317: establishment of polarity in the germ line and in epithelial follicular cells (Martin and St Johnston, 50ÀGAAACATCCTCCGGCTTAAGCACCAGTG 2003). Furthermore, Blumer and co-workers have 0 recently showed that LKB1 interacts with, and phos- CCCATCCCÀ3 : phorylates (Blumer et al., 2003), the activator of G- protein signalling 3 (AGS3), a receptor-independent The following missense mutations of LKB1 were obtained by directed mutagenesis as described above, using the following activator of G-protein signalling known to be involved oligonucleotides: in spindle positioning during asymmetric cell divisions (Blumer et al., 2003 and references herein). These results K78A: 50-GTGCAGGAGGGCCGTCGC indicate that LKB1 functions have been evolutionarily GATCCTCAAGAAGAAGAAG-30, conserved and that hamartomas and tumors which M136R: 50GAGTACTGCGTGTGTGGCAGGCA develop in PJS patients could result from a perturbation GAAATGCTGGACAGCGTG-30,

Oncogene Hsp90/Cdc37 regulate LKB1 stability P Nony et al 9167 G163D: 50-CTGTCAGCTGATTGACGACCTGGAG- immunoprecipitation with the anti-Flag antibody. 1 mg of TACCTGCATAGC-30. soluble proteins was incubated in 1 ml of lysis buffer in the presence of 1 mg of antibody. After 2 h of incubation at 41C, The vector expressing hemagglutinin-tagged ubiquitin immune complexes were collected on protein G coupled to (HA-Ub) was kindly provided by D Bohmann (University sepharose beads (Amersham Pharmacia Biotech) and washed of Rochester, Medical Center, Rochester, New York, three times in lysis buffer. The resulting complexes were NY, USA). The pSG5/Flag-FRS2 expresses the human resuspended in Laemmli buffer, boiled for 5 min, separated on docking protein FRS2 fused N-terminally with the Flag tag a SDS–PAGE polyacrylamide gel. The gel was either silver- epitope. stained (Bio-Rad reagents) or transferred on PVDF membranes for Western blotting analyses. For large-scale purifica- Antibodies tion of LKB1-associated proteins, the equivalent of 10 immunoprecipitations was loaded in one well of the poly- The Anti-Flag monoclonal antibody (clone M2) was pur- acrylamide gel and resolved by SDS–PAGE. The proteins were chased from Sigma. The anti-Hsp90 monoclonal (clone sc- then revealed by Coomassie staining (Coomassie stain solu- 13119) and polyclonal (clone sc-8262, from goat) antibodies, tion, Bio-Rad). Protein band was excised from the gel and kept the anti-Cdc37 polyclonal antibody (clone sc-6396, from goat), in 10% ethanol for mass spectrometry analysis. the anti-Raf-1 polyclonal antibody (clone sc-133, from rabbit) For coimmunoprecipitation experiments of endogenous and the anti-EB1 polyclonal antibody (clone sc-6412) were LKB1 and Hsp90 proteins, sodium molybdate (20 mM final obtained from Santa Cruz Biotechnology, Inc. The anti- concentration, purchased from Sigma) was added both in the ubiquitin monoclonal antibody was purchased from Zymed lysis and washing buffers. Laboratories Inc. To immunoprecipitate endogenous LKB1, the anti-LKB1 (N19, clone sc-8185, from goat) obtained from Santa Cruz Biotechnology, Inc. was used. In Western Blot LKB1 autokinase assay experiments, LKB1 protein was revealed using a rabbit Flag-LKB1 protein was immunoprecipitated with the anti- antibody raised against two peptides (VAEALHPFAADDTC Flag antibody and the immunocomplex was incubated in a and SRAEGRAPNPARKAC) and produced by the Covalab 30 ml mixture reaction containing 20 mM HEPES pH 7.4, company (Lyon, France). HRP-secondary antibodies were 10 mM MgCl2,10mM MnCl2,10mM b-glycerophosphate, 1 ml purchased from Amersham. All commercially available anti- [g-32P]ATP (10 mCi/ml–4500 Ci/mmol, ICN). After 30 min at bodies were used according to the manufacturers’ recommen- 301C, the reaction was stopped by addition of 6 ml of the SDS dations. sample buffer (120 mM Tris-HCl pH 6.8, 4% SDS, 2% glycerol, 0.01% bromophenol blue, 10% b-mercaptoethanol). Cell line and transfection assays Samples were electrophoresed on a SDS–PAGE polyacrylamide gel (10%), transferred to PVDF membrane and analysed HBL100 human breast epithelial cell line was cultured in by autoradiography and by Western blot. Dulbecco’s modified Eagle’s medium with 10% of heat- inactivated fetal bovine serum supplemented with 50 U/ml of penicillin and 50 U/ml of streptomycin and incubated at 371C Mass spectrometry analysis in 5% CO2. Plasmids encoding LKB1 cDNAs were transfected Mass spectrometry analyses and similarity search in SWISS- into cells at 80% of confluence using ExGen 500 (Euromedex) PROT/TrEMBL was performed by the Swiss-2D Service as a transfection reagent, according to the manufacturer’s company (Geneva). instructions. Treatment with Hps90 and protease inhibitors Protein analyses Geldanamycin (GA), novobiocin, the proteasome inhibitor Cells were lysed 48 h post-transfection in a lysis buffer benzyloxycarbonyl-Leu-Leu-Leu-aldehyde (MG132) and Pep- containing: 20 mM Tris-HCl (pH 7.5), 150 mM NaCl, 2 mM statin A were purchased from Sigma. Leupeptin was purchased EDTA and 1% Nonidet P-40, supplemented with a mixture of from Alexis Coger. Stock solutions were prepared as follows: proteases and phosphatases inhibitors. After 15 min on ice, GA was resuspended in DMSO at the final concentration of lysates were cleared by centrifugation at 14 000 r.p.m. for 400 mM and used at the final concentration of 2 mM; novobicin 15 min at 41C. Depending on the experiment, either the soluble was resuspended in water and used at a final concentration of cellular fraction or both the soluble and insoluble fractions 1mM; MG132 was dissolved in water at the final concentration (the insoluble fraction corresponds to the pellet) were analysed of 50 mM and used at the final one of 50 mM; Pepstatin A was by Western blotting. Protein concentration was determined by prepared in methanol at 1.45 mM, final concentration, and a modified Bradford assay (Bio-Rad). Proteins were resus- used at the final one of 100 mM; Leupeptin was dissolved in pended in the Laemmli buffer (20 mM Tris-HCl (pH 6.8), 2% water at 20 mM final concentration and used after a 1000-fold SDS, 20% glycerol, 20 mg of bromophenol blue per ml) in the dilution. Drugs or vehicles were added directly in the cell presence of 1.4 M of 2-mercaptoethanol and separated on a culture media for the times indicated in the figure legends. SDS–PAGE polyacrylamide gels, then transferred to poly- Cells were exposed to drugs 48 h post-transfection. vinylidenedifluoride (PVDF) membranes (Immobilon P Milli- pore). Membranes were incubated with the appropriate Pulse-chase analysis antibody as previously described. To solubilize the NP40 insoluble cellular fraction, cells were At 48 h post-transfection, transfected HBL100 cells were lysed in a buffer containing: 20 mM Tris-HCl (pH 7.5), 1% washed with methionine/cysteine-free DMEM (Sigma). Cul- SDS, 1% Nonidet P-40, 150 mM NaCl, 10 mM iodoacetamide ture medium was then replaced with methionine/cysteine-free and a mixture of protease and phosphatase inhibitors. After DMEM containing 10% SVF, penicillin/streptomycin anti- clearing by centrifugation, the supernatant was treated with biotics and 62.5 mCi/ml of [35S]methionine/[35S]cysteine bensonase (5 UI/ml) (Sigma) for 20 min on ice. After cell lysis, (TRAN35S-Label, ICN) and incubated 15 h at 371C. At this the buffer was diluted up to 0.3% of SDS and used for time point, cells were washed two times in methionine/cysteine-

Oncogene Hsp90/Cdc37 regulate LKB1 stability P Nony et al 9168 free DMEM containing 10% SVF, penicillin/streptomycin and protein kinases (Pearl and Prodromou, 2000; antibiotics and then either lysed immediately or incubated in Young et al., 2001). Hsp90 cellular functions require complete medium in the presence of either GA or DMSO for its association with a set of specific cofactors which the indicated times prior to cell lysis. Flag-LKB1 protein was ensures the binding selectivity of Hsp90 to its targets immunoprecipitated from the cleared lysates with the anti-Flag and assists its chaperone activity. The co-chaperone antibody as previously described and separated by SDS– PAGE. Gels were dried and exposed for autoradiography. The Cdc37/p50 interacts with Hsp90 and evidence has been relative amount of radiolabelled LKB1 protein was quantified provided that Cdc37/p50 acts as an adaptator subunit by densitometry using the fluorimeter FluorS and the Quantity One software (BioRad, Ivry, France).

Results

To characterize proteins that interact with LKB1, we transfected the mammary epithelium cell line HBL100 with an eukaryotic expression vector expressing the human LKB1 protein tagged with an N-terminal Flag epitope. As a control, we also expressed in the same cell line the N-terminally flagged docking protein FRS2 (Kouhara et al., 1997). LKB1 was immunoprecipitated from transfected cells and the immunoprecipitate was fractionated by SDS–PAGE and silver stained to visualize proteins which bind to LKB1. In these conditions, a band migrating at approximately 90 kDa was repeatedly detected in the lane corresponding to LKB1 but not in the FRS2 control (Figure 1a). Western blotting with the anti-Flag antibody confirmed that both LKB1 and FRS2 were correctly immunoprecipitated in these conditions and expressed at similar levels (Figure 1b). To determine the identity of the 90 kDa protein, we pooled several immunoprecipitates to enrich the 90 kDa protein and excised the band from the gel after Coomassie-staining. MALDI mass spectrometry of the purified material revealed that this band corresponds to the cytosolic molecular chaperone heat-shock protein 90 alpha (Hsp90a). Previous studies have shown that Hsp90 isoforms a and b display very similar functions (Pearl and Prodromou, 2000; Young et al., 2001) and in the rest of the study, we have used immunological reagents which recognize both isoforms. To confirm that Hsp90 forms a specific complex with LKB1, we immunoprecipitated LKB1 with the Flag antibody and probed the Western blot with an anti-Hsp90 serum. As shown in Figure 1c, a band corresponding to Hsp90 was clearly detected in the LKB1 lane and not in the FRS2 control. Hsp90 is an ubiquitous molecule that plays a key role in the stabilization and the conformational regulation of signalling effectors such as steroid hormone receptors

Figure 1 LKB1 forms a complex with both Hsp90 and Cdc37. (a) The human breast epithelial mammary cell line HBL100 was transfected with a vector expressing either the Flag-LKB1 or the Flag-FRS2. At 48 h post-transfection cells were lysed, LKB1 and FRS2 proteins were immunoprecipitated with an anti-Flag antibody, the immunocomplexes were separated by SDS–PAGE and the gel was silver stained to visualize proteins associated with LKB1 and FRS2. (b) Samples analysed in panel a were immunoblotted with the anti-Flag antibody. (c,d) Flag-LKB1 and Flag-FRS2 were immunoprecipitated from HBL100 whole cells lysates. Immunoprecipitates were subjected to Western blotting using an anti-Hsp90 antibody (c) and an anti-Cdc37/ p50 antibody (d)

Oncogene Hsp90/Cdc37 regulate LKB1 stability P Nony et al 9169 that binds to, and is needed, for the activity of an array of protein kinases (Hunter and Poon, 1997). To examine whether LKB1 is also associated with Cdc37/p50, LKB1 was immunoprecipitated and Western blotted with an anti-Cdc37 antibody. As shown in Figure 1d, Cdc37 was found in the LKB1 immunoprecipitates but not in the FRS2 control, thus confirming that LKB1 forms a complex with Cdc37/p50. To further investigate the relevance of the LKB1- Hsp90 interaction, we analysed whether endogenous LKB1 binds to Hsp90. For that purpose, we lysed HBL100 cells in a buffer containing sodium molybdate at a concentration of 20 mM to preserve the putative complex formed with Hsp90 (Stancato et al., 1997). Endogenous LKB1 was immunoprecipitated with an antibody raised against the N-terminus part of the protein (Figure 2a), and we confirmed that endogenous LKB1 coimmunoprecipitated with Hsp90 in these conditions (Figure 2b). It has been previously established that Hsp90 interacts with the Raf-1 kinase (Perdew et al., 1997; Stancato et al., 1997). In accordance with these results, we found that Hsp90 co-precipitates with Raf-1, but not with the microtubule-binding protein EB1, thus ensuring the specificity of our immunoprecipitation conditions. We next defined the region in LKB1 responsible for its interaction with Hsp90. A series of four LKB1 deletion mutants was generated and expressed in HBL100 cells (Figure 3a). Immunoblotting with the anti-Flag antibody confirmed that these LKB1 mutants were expressed after transfection, although the expression level of D243 and D317 mutants was lower than for the two other LKB1 mutants (Figure 3b). We also checked that these four truncation products were Figure 2 Endogenous LKB1 is associated with Hsp90. (a) Total immunoprecipitated with the anti-Flag antibody (data cell lysates prepared from HBL100 cells were subjected to immunoprecipitation with either an anti-LKB1 antibody or an not shown). To analyse their binding properties, LKB1 antibody that reacts with the microtubule-binding protein EB1. truncation products were immunoprecipitated and The immunoprecipitates were resolved by SDS–PAGE and Western blotted with an anti-Hsp90 antibody. As shown Western blotted with an anti-LKB1 polyclonal serum. (b) Whole in Figure 3c, deletion of the noncatalytic carboxy- cell lysates prepared from HBL100 cells were immunoprecipitated terminal region (D317) did not prevent the binding of with various antibodies and Western blotted with an anti-Hsp90 polyclonal serum Hsp90. However, the D243 mutant formed a complex with Hsp90, thus indicating that both the N-terminal region and a portion of the kinase domain of LKB1 are sufficient for the interaction with Hsp90. In addition, truncation of the N-terminal amino acids of LKB1 analysis an LKB1 mutant (K78A) devoid of catalytic (D88) significantly reduced its interaction with Hsp90 activity. Immunoprecipitation of LKB1 mutants with considering that the relative expression level of this the anti-Flag antibody followed by immunoblotting mutant is lower compared to LKB1 wild-type and to the with both the anti-Hsp90 and Cdc37/p50 antibodies D243 mutant. Finally, an internal deletion within the revealed that the mutation G163D reduced drastically catalytic domain (D146–186) abolished the association the association with both Hsp90 and Cdc37/p50 whereas with Hsp90. Taken together, these results indicate that the K78A and the PJS mutations did not affect the LKB1 kinase domain but not the carboxy-terminal significantly the interaction of LKB1 with the Hsp90/ region is required for the interaction with Hsp90. Cdc37 complex (Figure 4a). Although, Hsp90 was We next investigated whether LKB1 mutations hardly detected in association with Flag-LKB1 located within the catalytic domain and found either (G163D) using this procedure, metabolic labelling of in a PJS patient (M136R) (Olschwang et al., 2001) or in transfected HBL100 cells showed that LKB1-G163D a sporadic testicular tumor (G163D) (Ylikorkala et al., retained some capacity to interact with Hsp90, albeit 1999) interfere with the binding of Hsp90 and Cdc37/ with a reduced efficiency compared to LKB1 wild type p50. The G163D mutation is located within the region (data not shown). We further assessed the LKB1 of the LKB1 kinase domain that we found to be crucial catalytic activity of these mutants by measuring the for the binding to Hsp90. We also included in this ability of LKB1 to autophosphorylate. As already

Oncogene Hsp90/Cdc37 regulate LKB1 stability P Nony et al 9170 activity as previously reported (Ylikorkala et al., 1999) (Figure 4c). Thus, these results indicate that the binding of Hsp90/Cdc37 to LKB1 does not depend on LKB1 protein kinase activity and also strengthen the notion that the LKB1 kinase domain contains the major regions of contact necessary for the assembly of Hsp90/Cdc37 complex. The drug geldanamycin (GA) interacts with the NH2- terminal ATP/ADP binding pocket of Hsp90 and acts as a competitive inhibitor of its ATPase activity, thereby leading to the dissociation and eventual degradation of Hsp90 client proteins (Whitesell et al., 1994; Schulte et al., 1995; Prodromou et al., 1997; Stebbins et al., 1997). To examine whether LKB1 stability depends on Hsp90-binding, HBL100 cells were transfected either with a vector expressing the Flag-LKB1 or the Flag- FRS2 and then treated with 2 mM of GA. As shown in Figure 5a, 2 h of GA treatment caused a marked diminution of the level of LKB1 whereas it had no significant effect on the stability of FRS2 which does not interact with Hsp90. To validate these results, we used novobiocin, another known Hsp90 inhibitor which is structurally unrelated to GA and binds to a recently characterized ATP-binding domain located in the C- terminal part of Hsp90 (Marcu et al., 2000). As displayed in Figure 5b, novobiocin treatment of transfected HBL100 cells led to a significant decrease of LKB1, whereas novobiocin treatment did not substantially alter the level of the FRS2 protein. Quantification of the amount of LKB1 normalized with the amount of actin indicates that 8 and 24 h of novobiocin treatment induced a 60 and 50% decrease in the levels of LKB1, respectively. Next, we monitored the degradation of LKB1 upon GA treatment and determined its half-life. HBL100 cells expressing Flag- LKB1 were labelled with a mixture of [35S] methionine/ [35S] cysteine and then chased with a nonradioactive medium at various times. In these experiments, either GA or the diluent DMSO were added together with the nonradioactive medium at the time of the chase. As shown in Figure 5c, LKB1 was found to be relatively stable in cells treated with DMSO, whereas this half-life Figure 3 LKB1 kinase domain is required for specific was dramatically shortened in cells treated with GA. association with Hsp90. (a) A series of four LKB1 deletion Pulse-chase experiments also confirmed that GA led to a mutants was generated: D-243, D-317, D-88 and D-146- very rapid disassembly of the Hsp90-LKB1 complex, 186. The black bold line represents the regions of the detectable as soon as 5 min after the addition of the LKB1 protein retained in the deletion mutants. The kinase domain of LKB1 spans the region between amino acids 49 compound to the culture medium (Figure 5c). We and 310. The black box depicted in the N terminal part of confirmed by Western blotting that the band detected by LKB1 corresponds to the nuclear localization signal. autoradiography at 90 kDa and visualized in Figure 5c (b) Wild-type or mutant Flag-LKB1 proteins were corresponded to Hsp90 (data not shown). To compare expressed in HBL100 cells and the proteins were revealed by Western blot with the anti-Flag monoclonal antibody. the respective half-lives of both LKB1 and G163D (c) HBL100 cells expressing either LKB1 wild-type or the mutants, we carried out pulse-chase experiments with deletion mutants were lysed, immunoprecipitated with the HBL100 transfectants and then quantified the levels of anti-Flag antibody and Western blotted with the anti- radioactive material. As shown in Figure 5d, the half-life Hsp90 polyclonal serum of wild-type LKB1 was 2 h whereas the half-life of LKB1-G163D mutant was 30 min, a finding in agreement with the observed reduced affinity of Hsp90 to this described, the K78A mutation compromised the LKB1 mutant form of LKB1. Finally, we examined the effects autokinase activity (Figure 4c). Furthermore, the PJS of GA on endogenous LKB1 and found that treatment mutation M136R disrupted LKB1 autokinase activity, of HBL100 cells with 2 mM of GA resulted in a drastic while the G163D mutant retained a weak autocatalytic decrease of the amount of LKB1 (Figure 5e). During the

Oncogene Hsp90/Cdc37 regulate LKB1 stability P Nony et al 9171

Figure 4 Tumor mutation G163D impairs the interaction between LKB1 and Hsp90/Cdc37. (a, b) Wild-type LKB1, a kinase dead mutant (K78A), a Peutz–Jeghers mutant (M163R) and a testicular tumor mutant (G163D) that have been expressed in HBL100 cells were immunoprecipitated with the anti-Flag antibody and Western blotted with both the anti-Hsp90 and anti-Cdc37 antibodies (a)or with the anti-Flag monoclonal antibody (b). (c) Immunokinase assay. Wild-type LKB1 or mutant forms of LKB1 were immunoprecipitated with the anti-Flag antibody and the immunocomplexes were incubated for 30 min with [g32-p]ATP, separated by SDS–PAGE and the gel was autoradiographed

course of this study, we noticed that the diminution of 2002) treatment with both GA and MG132 led to a endogenous LKB1 required a longer duration of drastic increase of the level of LKB1 in the insoluble incubation with GA, an observation which suggests fraction (Figure 6b), whereas other protease inhibitors that endogenous LKB1 has a longer half-life than did not show any effects (Figures 6a,b). These data exogenous Flag-LKB1. Taken together, these results indicate that MG132 stabilizes LKB1 upon GA treat- indicate that Hsp90 binding is necessary for LKB1 ment, thus supporting the hypothesis that LKB1 stabilization. undergoes proteosomal degradation when the LKB1- A number of studies have now demonstrated that Hsp90 complex is dissociated. GA-induced degradation of Hsp90 target proteins is Targeting of proteins to the proteasome requires their preceded by their ubiquitination and subsequent target- tagging by the ubiquitin system which covalently ing to the proteasome (Mimnaugh et al., 1996; Whitesell conjugates polyubiquitin chains onto proteins destined et al., 1997). To investigate whether a similar mechanism to proteolysis (Ciechanover and Schwartz, 1998). Our could account for GA-mediated decay of LKB1, we results obtained with the proteasome inhibitor MG132 blocked the proteasome function with a speciﬁc prompted us to study whether LKB1 was ubiquitinated inhibitor, MG132. When HBL100 cells expressing when the complex formed with Hsp90 was disrupted. Flag-LKB1 were incubated with GA plus MG132 and For that purpose, HBL100 cells were either transfected lysed in a buffer containing 1% of NP-40, we did not with a vector expressing the Flag-LKB1 or cotransfected observe a signiﬁcant increase of LKB1 in the soluble with the Flag-LKB1 expressing vector and a plasmid fraction (Figure 6a). However, as previously described coding for ubiquitin. Western blotting with the anti-Flag for other protein kinases known to bind Hsp90 such as antibody revealed a series of bands forming a smear PDK-1 and MOK (Miyata et al., 2001; Fujita et al., above LKB1 in the lanes corresponding to the cells

Oncogene Hsp90/Cdc37 regulate LKB1 stability P Nony et al 9172

Figure 5 Hsp90 regulates LKB1 stability. (a) HBL100 cells were transiently transfected with plasmids expressing either the Flag- LKB1 protein or the Flag-FRS2. Cells were treated either for 2 or 8 h in the presence of 2 mM of Geldanamycin (GA) or with the vehicle DMSO. At the end of the indicated times or before the addition of GA (T0), cells were resuspended in lysis buffer and proteins were analysed by Western Blot using an anti-Flag M2 specific antibody. (b) HBL100 cells were transiently transfected with plasmids expressing either the Flag-LKB1 protein or the Flag-FRS2. Cells were treated either for either 3, 8 or 24 h in the presence of 1 mM of novobiocin (final concentration) or with the vehicle H2O. At the end of the indicated times, cells were resuspended in lysis buffer and proteins were analysed by Western blot using an anti-Flag M2 specific antibody or an anti-actin. (c) HBL100 cells expressing Flag- LKB1 protein were pulse labelled with [35S]methionine/[35S]cysteine and chased with nonradioactive medium for the indicated times in the presence of GA (lanes 16–19) or in absence (DMSO, lanes 12–14.) At the end of the indicated times (T0), cells were lysed and the Flag-LKB1 protein was immunoprecipitated using the anti-Flag M2 antibody and resolved on a polyacrylamide gel. The dried gel was autoradiographed. The Hsp90 protein which coimmunoprecipitates with LKB1 is also revealed as indicated, as confirmed by Western blot experiment (data not shown). (d) HBL100 cells expressing either LKB1 wild type or the G163D mutant were metabolically labelled, chased with nonradiocative medium and the radioactive material was then quantified. Similar results were obtained in two independent experiments. (e) Endogenous LKB1 protein is also destabilized by GA treatment. HBL100 cells were treated for 24 h in the presence of 2 mM of geldanamycin (GA) or its absence (DMSO), lysed and endogenous LKB1 was immunoprecipitated. Immunoprecipitation reactions were analysed by Western blot using an anti-LKB1 antibody

treated with both GA and MG132 (Figure 7a). When bands was clearly enriched in cells cotransfected with the cells were cotransfected with the vectors expressing both vectors (Flag-LKB1 and HA-ubiquitin) and LKB1 and ubiquitin, these bands were detected in the incubated with GA plus MG132 (Figure 7b). Collec- control lanes (DMSO and GA), but were markedly tively, these results provide evidence that LKB1 is increased in the lane corresponding to the GA plus ubiquitinated and degraded by the proteasome when MG132 treatment. These data are in agreement with the Hsp90 is released by the action of GA. idea that MG132 prevents the proteolysis of unstable multiubiquitinated forms of LKB1. To conﬁrm these results, we transfected HBL100 cells using the same conditions and performed an immunoprecipitation with Discussion the anti-Flag antibody followed by a Western blot with an antiubiquitin antibody. A smear consistent with the In this work, we have provided evidence that the covalent attachment of ubiquitin chains to LKB1 was molecular chaperones Hsp90 and Cdc37 bind to human detected in cells transfected with the Flag-LKB1 vector LKB1 and regulate its stability. Using LKB1 truncation and treated with GA plus MG132, and this smear of and deletion mutants, we found that the kinase domain

Oncogene Hsp90/Cdc37 regulate LKB1 stability P Nony et al 9173

Figure 6 Degradation of LKB1 is mediated by the proteasome. HBL100 cells transiently transfected with plasmid expressing Flag- LKB1 proteins were treated for 4 h in the presence of GA (lanes 3 and 10) or with the vehicle (DMSO, lanes 2 and 9). The proteasome inhibitor MG132 was added in the presence of either GA (lanes 4 and 11) or lysosomal protease inhibitors (Pepstatin, lanes 5 and 12 or its diluent methanol [MeOH], lanes 6 and 13, and Leupeptin, lanes 7 and 14). Both soluble and insoluble fractions were analysed by Western blot using the anti-Flag M2 antibody (see experimental procedures for further details). T0 corresponds to the time before drug addition

of LKB1 is required for the association with Hsp90, whereas the 116 amino acids of the C-terminus are dispensable. Our results also suggest that a portion of the LKB1 N-terminal part might be involved in the binding to Hsp90. However, we have not precisely Figure 7 Dissociation of the LKB1-Hsp90 complex promotes mapped the boundary of the N-terminal region involved LKB1 ubiquitination. (a) HBL100 cells were transiently transfected in this association and it is possible that only the 39 with a plasmid expressing Flag-LKB1 or cotransfected with amino acids of the LKB1 kinase domain which are plasmid expressing Flag-LKB1 and HA-ubiquitin (pHA-Ub). At lacking in the LKB1-D88 truncated mutant, are involved 48 h post-transfection cells were treated during 16 h either by DMSO, GA, GA and MG132. Cells treated with these compounds in this process. An LKB1 mutation identified in a were lysed and analysed by Western blot using the anti-Flag testicular tumor (G163D) and located within the antibody (b) HBL100 cells transfected as described in the panel a catalytic core domain (Ylikorkala et al., 1999) impairs were lysed, subjected to immunoprecipitation with the anti-Flag the formation of the LKB1/Hsp90-Cdc37 heterocom- antibody and Western blotted with either an anti-ubiquitin plex, thus reinforcing the idea that Hsp90/Cdc37 antibody or the anti-Flag monoclonal antibody interacts with the LKB1 kinase domain. Recently, Boudeau and colleagues reported that Hsp90/Cdc37 bind to mouse LKB1 and they further obtained evidence that this complex associates with the kinase domain of Pharmacological inhibition of Hsp90 with GA and LKB1 (Boudeau et al., 2002), a finding in accordance novobiocin was found to promote dissociation of the with our own results. Interestingly, we repeatedly LKB1/Hsp90-Cdc37 complex and the subsequent de- noticed that the amount of Hsp90 co-precipitated either gradation of LKB1. Metabolic labelling and pulse-chase with LKB1 deleted of the C-terminus (D317) or with experiments confirmed that GA induced a rapid LKB1 wild type was comparable, despite a consistent dissociation of the Hsp90-LKB1 complex. In these lower expression of the LKB1-D317 protein. This experiments, the half-life of wild-type LKB1 was 2 h observation may indicate that the C-terminal domain whereas the half-life of the LKB1-G163D mutant was of LKB1 is flexible and hinders the binding of 30 min, a result in agreement with the observed reduced Hsp90, thus providing a possible mechanism for affinity of Hsp90 to this mutant form of LKB1. The controlling the dynamics of the LKB1 and Hsp90/ fundamental role of the proteasome in the control of Cdc37 interactions. protein turnover has now been firmly established

Oncogene Hsp90/Cdc37 regulate LKB1 stability P Nony et al 9174 (Ciechanover and Schwartz, 1998), therefore suggesting regulatory complex that binds to human LKB1 and that the degradation of LKB1 might be regulated by this regulates its stability. Our results suggest that Cdc37/ proteolytic pathway. Consistent with this idea, we found p50 is the targeting subunit which directly interacts with that GA-induced degradation of LKB1 was blocked the LKB1 kinase domain as already demonstrated for with MG132, a specific inhibitor of the proteasome, several protein kinases (Hunter and Poon, 1997). whereas other protease inhibitors did not prevent LKB1 Furthermore, LKB1 catalytic activity was not required destabilization. Unexpectedly, MG132 induced an for its association with Hsp90/Cdc37. However, it is increase of the amount of LKB1 in the insoluble cellular possible that Hsp90 assists the folding and the matura- fraction and not in the soluble cellular extracts. Similar tion of the LKB1 kinase domain. A classical example of findings have been reported for certain protein kinases the role of Hsp90 relates to its association with the which associate with Hsp90, such as MOK (Miyata steroid hormone receptors which is thought to maintain et al., 2001) and PDK-1 (Fujita et al., 2002), and this the receptor in a conformation competent for the phenomenon is likely to reflect the fact that proteins binding of the hormone (Toft, 1999). Similarly, it is which fail to be correctly folded aggregate in the conceivable that Hsp90 interaction with LKB1 induces a insoluble fraction when both Hsp90 and the proteasome conformational change in LKB1 compatible with the are inhibited. In agreement with our data, Boudeau and loading of additional signalling partners such as p53, co-workers observed that either GA or another un- BRG-1 or LIP-1. Future experiments should examine related Hps90 inhibitor, radicicol, induce a decrease of the dynamics of the association and dissociation of the the mouse LKB1 level which could be prevented by the LKB1-Hsp90 complex, since it is plausible that the addition of proteasome inhibitors (Boudeau et al., disassembly of this multiprotein complex might be 2002). physiologically relevant during the cell cyle, thus leading Proteins fated to be degraded by the proteasome are to the destruction of LKB1 and elimination of its tagged with polyubiquitin chains through the sequential growth arrest function. action of E2 ubiquitin-conjugating enzymes and E3 One important consideration resulting from this ubiquitin ligases (Ciechanover and Schwartz, 1998). research relates to the pharmacological inactivation of Consistent with these biochemical processes, disruption Hsp90 with drugs structurally related to GA, such as 17- of the LKB1/Hsp90 complex induced LKB1 ubiquitina- allylaminogeldamycin (AAG) (Neckers, 2002). This is a tion and its subsequent degradation by the proteasome. promising strategy in cancer therapeutics with AAG It has been recently reported that the co-chaperone currently in phase I clinical trial (Neckers, 2002). Our CHIP (carboxyl terminus of Hsc70-interacting protein) results suggest however that Hsp90 inhibitors may also is an E3 ubiquitin ligase that promotes ubiquitination of elicit adverse clinical effects and favor the development Hsp90 client proteins such as the cystic fibrosis of tumors by disrupting LKB1 tumor suppressor transmembrane-conductance regulator, glucocorticoid function, a question which needs to be carefully studied receptor and the Erb2 tyrosine kinase receptor (Ballin- before wider clinical use of such compounds is ger et al., 1999; Connell et al., 2001; Jiang et al., 2001; implemented. Meacham et al., 2001). By remodelling Hsp90 complexes, CHIP is involved in the protein triage system which controls the equilibrium between chaperone- mediated protein refolding and degradation by the Acknowledgements proteasome. It is thus conceivable that CHIP might be We gratefully acknowledge Janet Hall for critical reading of one of the E3 ligases that interacts with LKB1 and the manuscript. We also thank Dirk Bohmann for the kind gift induces its ubiquitination when the heterocomplex of the plasmid expressing HA-tagged ubiquitin. This work was supported by the Ligue Nationale Contre le Cancer (group Hsp90/LKB1 is dissociated, a question which warrants Marc Billaud ‘labellise´ e’ by the Ligue Nationale Contre le further investigation. Taken together, our results pro- Cancer). During this work, PN, HG and MR have been vide the first evidence that LKB1 turnover is regulated the recipients of a fellowship from the Ligue Nationale by the ubiquitin–proteasome proteolytic system. Contre le Cancer, Comite´ de Haute-Savoie and Ligue In summary, the data described in this study identify Nationale and from the Fondation pour la Recherche the molecular chaperones Hsp90 and Cdc37 as a major Me´ dicale, respectively.

References

Avizienyte E, Loukola A, Roth S, Hemminki A, Tarkkanen Bardeesy N, Sinha M, Hezel AF, Signoretti S, Hathaway NA, M, Salovaara R, Arola J, Butzow R, Husgafvel-Pursiainen Sharpless NE, Loda M, Carrasco DR and DePinho RA. K, Kokkola A, Jarvinen H and Aaltonen LA. (1999). Am. J. (2002). Nature, 419, 162–167. Pathol., 154, 677–681. Blumer JB, Bernard ML, Peterson YK, Nezu J, Chung P, Baas AF, Boudeau J, Sapkota GP, Smit L, Medema R, Dunican DJ, Knoblich JA and Lanier SM. (2003). J. Biol. Morrice NA, Alessi DR and Clevers HC. (2003). EMBO J., Chem., 278, 23217–23220. 22, 3062–3072. Boudeau J, Deak M, Lawlor MA, Morrice NA and Alessi DR. Ballinger CA, Connell P, Wu Y, Hu Z, Thompson LJ, (2002). Biochem. J., 19, 849–857. Yin LY and Patterson C. (1999). Mol. Cell. Biol., 19, Ciechanover A and Schwartz AL. (1998). Proc. Natl. Acad. 4535–4545. Sci. USA, 95, 2727–2730.

Oncogene Hsp90/Cdc37 regulate LKB1 stability P Nony et al 9175 Collins SP, Reoma JL, Gamm DM and Uhler MD. (2000). Perdew GH, Wiegand H, Vanden Heuvel JP, Mitchell C and Biochem. J., 345 (Part 3), 673–680. Singh SS. (1997). Biochemistry, 36, 3600–3607. Connell P, Ballinger CA, Jiang J, Wu Y, Thompson LJ, Peutz JL. (1921). Ned. Tijdschr. Geneeska, 10, 134–146. Hohfeld J and Patterson C. (2001). Nat. Cell. Biol., 3, 93–96. Prodromou C, Roe SM, O’Brien R, Ladbury JE, Piper PW Fujita N, Sato S, Ishida A and Tsuruo T. (2002). J. Biol. and Pearl LH. (1997). Cell, 90, 65–75. Chem., 277, 10346–10353. Rossi DJ, Ylikorkala A, Korsisaari N, Salovaara R, Luukko Giardiello FM, Welsh SB, Hamilton SR, Offerhaus GJ, K, Launonen V, Henkemeyer M, Ristimaki A, Aaltonen LA Gittelsohn AM, Booker SV, Krush AJ, Yardley JH and and Makela TP. (2002). Proc. Natl. Acad. Sci. USA, 99, Luk GD. (1987). N. Engl. J. Med., 316, 1511–1514. 12327–12332. Hemminki A, Markie D, Tomlinson I, Avizienyte E, Roth S, Sanchez-Cespedes M, Parrella P, Esteller M, Nomoto S, Trink Loukola A, Bignell G, Warren W, Aminoff M, Hoglund P, B, Engles JM, Westra WH, Herman JG and Sidransky D. Jarvinen H, Kristo P, Pelin K, Ridanpaa M, Salovaara R, (2002). Cancer Res., 62, 3659–3662. Toro T, Bodmer W, Olschwang S, Olsen AS, Stratton MR, Sapkota GP, Boudeau J, Deak M, Kieloch A, Morrice N and de la Chapelle A and Aaltonen LA. (1998). Nature, 391, Alessi DR. (2002). Biochem. J., 362, 481–490. 184–187. Sapkota GP, Kieloch A, Lizcano JM, Lain S, Arthur JS, Hunter T and Poon RYC. (1997). Trends Cell. Biol., 7, 157– Williams MR, Morrice N, Deak M and Alessi DR. (2001). J. 161. Biol. Chem., 276, 19469–19482. Jeghers H, McKusick VA and Katz KH. (1949). N. Engl. J. Schulte TW, Blagosklonny MV, Ingui C and Neckers L. Med., 241, 992–1005. (1995). J. Biol. Chem., 270, 24585–24588. Jenne DE, Reimann H, Nezu J, Friedel W, Loff S, Jeschke R, Shen Z, Wen XF, Lan F, Shen ZZ and Shao ZM. (2002). Clin. Muller O, Back W and Zimmer M. (1998). Nat. Genet., 18, Cancer Res., 8, 2085–2090. 38–43. Smith DP, Spicer J, Smith A, Swift S and Ashworth A. (1999). Jiang J, Ballinger CA, Wu Y, Dai Q, Cyr DM, Hohfeld J and Hum. Mol. Genet., 8, 1479–1485. Patterson C. (2001). J. Biol. Chem., 276, 42938–42944. Smith DP, Rayter SI, Niederlander C, Spicer J, Jones CM and Jishage K, Nezu J, Kawase Y, Iwata T, Watanabe M, Miyoshi Ashworth A. (2001). Hum. Mol. Genet., 10, 2869–2877. A, Ose A, Habu K, Kake T, Kamada N, Ueda O, Kinoshita Spigelman AD, Murday V and Phillips RK. (1989). Gut, 30, M, Jenne DE, Shimane M and Suzuki H. (2002). Proc. Natl. 1588–1590. Acad. Sci. USA, 99, 8903–8908. Stancato LF, Silverstein AM, Owens-Grillo JK, Chow YH, Karuman P, Gozani O, Odze RD, Zhou XC, Zhu H, Shaw R, Jove R and Pratt WB. (1997). J. Biol. Chem., 272, 4013– Brien TP, Bozzuto CD, Ooi D, Cantley LC and Yuan J. 4020. (2001). Mol. Cell., 7, 1307–1319. Stebbins CE, Russo AA, Schneider C, Rosen N, Hartl FU and Kouhara H, Hadari YR, Spivak-Kroizman T, Schilling J, Pavletich NP. (1997). Cell, 89, 239–250. Bar-Sagi D, Lax I and Schlessinger J. (1997). Cell, 89, Su J-Y, Erikson E and Mailer JL. (1996). J. Biol. Chem., 271, 693–702. 14430–14437. Liu WK, Chien CY, Chou CK and Su JY. (2003). J. Biomed. Tiainen M, Vaahtomeri K, Ylikorkala A and Makela TP. Sci., 10, 242–252. (2002). Hum. Mol. Genet., 11, 1497–1504. Marcu MG, Chadli A, Bouhouche I, Catelli M and Neckers Tiainen M, Ylikorkala A and Makela TP. (1999). Proc. Natl. LM. (2000). J. Biol. Chem., 275, 37181–37186. Acad. Sci USA, 96, 9248–9251. Marignani PA, Kanai F and Carpenter CL. (2001). J. Biol. Toft DO. (1999). Control of Hormone Receptor Function by Chem., 276, 32415–32418. Molecular Chaperones and Folding Catalysts. Academic Martin SG and St Johnston D. (2003). Nature, 421, Publishers, Harwood, Amsterdam. 379–384. Watts JL, Morton DG, Bestman J and Kemphues KJ. (2000). Meacham GC, Patterson C, Zhang W, Younger JM and Cyr Development, 127, 1467–1475. DM. (2001). Nat. Cell. Biol., 3, 100–105. Whitesell L, Mimnaugh EG, De Costa B, Myers CE and Mehenni H, Gehrig C, Nezu J, Oku A, Shimane M, Rossier C, Neckers LM. (1994). Proc. Natl. Acad. Sci. USA, 91, 8324– Guex N, Blouin JL, Scott HS and Antonarakis SE. (1998). 8328. Am. J. Hum. Genet., 63, 1641–1650. Whitesell L, Sutphin P, An WG, Schulte T, Blagosklonny MV Mimnaugh EG, Chavany C and Neckers L. (1996). J. Biol. and Neckers L. (1997). Oncogene, 14, 2809–2816. Chem., 271, 22796–22801. Ylikorkala A, Avizienyte E, Tomlinson IP, Tiainen M, Roth S, Miyata Y, Ikawa Y, Shibuya M and Nishida E. (2001). J.Biol. Loukola A, Hemminki A, Johansson M, Sistonen P, Markie Chem., 276, 21841–21848. D, Neale K, Phillips R, Zauber P, Twama T, Sampson J, Miyoshi H, Nakau M, Ishikawa TO, Seldin MF, Oshima M Jarvinen H, Makela TP and Aaltonen LA. (1999). Hum. and Taketo MM. (2002). Cancer Res., 62, 2261–2266. Mol. Genet., 8, 45–51. Neckers L. (2002). Trends Mol. Med., 8, S55–S61. Ylikorkala A, Rossi DJ, Korsisaari N, Luukko K, Alitalo K, Olschwang S, Boisson C and Thomas G. (2001). J. Med. Henkemeyer M and Makela TP. (2001). Science, 293, 1323– Genet., 38, 356–360. 1326. Pearl LH and Prodromou C. (2000). Curr. Opin. Struct. Biol., Young JC, Moareﬁ I and Hartl FU. (2001). J. Cell Biol., 154, 10, 46–51. 267–273.

Oncogene