Signal Responsiveness of IκB Kinases Is Determined by Cdc37-assisted Transient Interaction with Hsp90*

The IκB kinase (IKK) holocomplex, containing the kinases IKKα, IKKβ, and the scaffold NEMO (NF-κB essential modifier), mediates activation of NF-κB by numerous physiological stimuli. Heat shock protein 90 (Hsp90) and the co-chaperone Cdc37 have been indicated as additional subunits, but their specific functions in signal transduction are indistinct. Using an RNA interference approach, we demonstrate that Cdc37 recruits Hsp90 to the IKK complex in a transitory manner, preferentially via IKKα. Binding is conferred by N-terminal as well as C-terminal residues of Cdc37. Cdc37 is essential for the maturation of de novo synthesized IKKs into enzymatically competent kinases but not for assembly of an IKK holocomplex. Mature IKKs, T-loop-phosphorylated after stimulation either by receptor-mediated signaling or upon DNA damage, further require Hsp90-Cdc37 to generate an activated state. Thus, the present data denote Hsp90-Cdc37 as a transiently acting essential regulatory component of IKK signaling.

The IKK complex has a large apparent molecular mass of 700 -900 kDa and contains the kinases IKK␣ and IKK␤ as well as the regulatory nonenzymatic scaffold protein IKK␥, also known as NEMO (NF-B essential modifier) (1,7). Stoichiometric analyses indicated equimolar content of IKK␥ and kinase molecules (8,9) in the IKK complex, where a tetramer of IKK␥ is thought to bind to two kinase dimers (10). The scaffold protein ELKS has been proposed as a further regulatory component of cellular IKK complexes (11). IKK activation is dependent on phosphorylation at activation loop (T-loop) serines, either by upstream IKK kinases or by autophosphorylation. The activation process involves catalytic, nondestructive Lys-63-linked polyubiquitination (Ub K63 ) of IKK␥ as well as Ub K63 binding by IKK␥ (12,13). Moreover, conformational changes by induced protein interactions may also be a mechanism to stimulate IKK activity. A number of regulatory proteins have been suggested to interact with IKK components (reviewed in Refs. 1 and 7). Among them, the chaperones Hsp90 and Cdc37 were proposed as stoichiometric subunits of an IKK holocomplex (14).
Heat shock protein 90 (Hsp90), which represents 1-2% of cytosolic protein, acts as molecular chaperone and requires ATP binding and hydrolysis to maintain its function (15)(16)(17). Hsp90 client proteins are preferentially cellular signal transducers, such as protein kinases and transcription factors. Hsp90 promotes their proper folding, assembly, and transportation across different cellular compartments (18 -21). Hsp90 is regulated through sequential and cooperative binding of a set of co-chaperones that link Hsp90 to distinct classes of client proteins. The mammalian homologue of the budding yeast cell cycle control protein Cdc37 is one such co-chaperone (22,23). The immunophilin FK506-binding protein 51 (FKBP51), also suggested as a regulator of NF-B activity (24), is another co-chaperone that binds to Hsp90 via a tetratricopeptide repeat domain (18,25). Due to the complex biological functions of Hsp90 and its co-chaperones, knock-out studies may not be readily informative and in fact result in lethal phenotypes (26,27).
Studies using the Hsp90-specific ATPase inhibitor geldanamycin (GA) initially proposed a requirement of Hsp90 for NF-B activation (14, 28 -30). Hsp90 interacts with IB kinases and signaling proteins of the NF-B pathway, including MEKK3, NIK, RIP1, TAK1, and TBK1 (14, 24, 31-33). Pro-longed inhibition of Hsp90 by GA causes proteasomal degradation of IKK components and RIP1 (14,24,29,31,34). Hence, it is unclear to what extent Hsp90 inhibition affects NF-B pathways at the level of IKK versus that of upstream kinases and if chaperones control the signaling process per se or have homeostatic functions.
In the present study, we investigated how Hsp90 and Cdc37 interact with the IKK signalosome, utilizing RNA interference. With a compositional analysis, we demonstrate that the IKK␣, -␤, and -␥ core complex is not associated with additional stoichiometric subunits. Cdc37 recruits Hsp90 to the IKK complex in a transitory manner, preferentially via IKK␣, and is phospho-rylated by both IKK␣ and IKK␤. Interaction with the chaperone complex is essential to promote functional maturation of de novo synthesized IB kinases. Moreover, Hsp90-Cdc37 activity is required to trigger catalytic activation of T-loop phosphorylated IKKs following stimulation with TNF-␣ or DNA damage. These results provide evidence that the chaperone complex acts on the IKK complex in the process of its activation.

EXPERIMENTAL PROCEDURES
Cell Culture and Transfection-HeLa and 293 cells were grown in Dulbecco's modified Eagle's medium (PAA Laboratories), supplemented with 10% fetal calf serum, 1 mM sodium pyruvate, and 100 units/ml penicillin/streptomycin. 17-Demethoxy geldanamycin was purchased from InvivoGen. 293 cells were transiently transfected by calcium phosphate precipitation using standard protocols. HeLa cells were transfected with DNA constructs using Lipofectamine 2000 (Invitrogen). Synthetic siRNAs were transfected in final concentrations between 0.5 and 10 nM. Lipid-mediated transfection was carried out with atufect01 (36). For sequence information, see the supplemental data. Where indicated, cells were additionally transfected with DNA constructs 48 h after siRNA transfection using Lipofectamine 2000. Cells were lysed 72 h after siRNA transfection. Fulllength human IKK␣, IKK␤, and IKK␥ cDNAs were cloned with N-terminal FLAG epitope into pcDNA3 or with N-terminal Myc epitope into pRK5. Kinase-dead mutants of IKK␣ and IKK␤ carry a K44A mutation (37). N-terminally FLAG-tagged full-length, truncated, or point-mutated human Cdc37 was cloned into pcDNA3.
Immunoprecipitation, Western Blotting, Kinase Assay, and Electrophoretic Mobility Shift Assay (EMSA)-Immunoprecipitations and Western blots were performed in accordance with standard procedures. Cells were lysed in 100 mM NaCl, 25 mM Tris, pH 8, 0.2% Nonidet P-40, 10% glycerol, supplemented with complete protease inhibitor mixture (Roche Applied Science), 10 mM NaF, 8 mM ␤-glycerophosphate, 0,2 mM Na 3 VO 4 , 1 mM dithiothreitol. For antibodies, see the supplemental data. Kinase assays and electrophoretic mobility shift assays were performed as described previously (38). Intensities of signals for FIGURE 1. Cdc37, Hsp90, FKBP51, and ELKS are not constitutively associated components of the IKK complex. A, HeLa cells were pulse-labeled with [ 35 S]methionine, and cell extracts were immunoprecipitated using IKK␣, IKK␥, or isotype control antibodies. B, HeLa cells were transfected with different siRNAs as indicated and pulse-labeled with [ 35 S]methionine. Cell extracts were immunoprecipitated using IKK␥ or isotype control antibodies. C, HeLa cell extracts were supplied to a Superose 6 gel filtration column, and the fractions were analyzed by Western blotting. Migration of marker proteins is indicated. D, HeLa cells were transfected with siRNAs as indicated, and extracts were analyzed by gel filtration and subsequent Western blotting (WB) for IKK␥. E, down-modulation of the indicated proteins was analyzed by Western blotting. The specific bands for ELKS and FKBP51 in C and E are marked by an asterisk.
Reporter Gene Assay-293 cells were transfected in a 6-well plate with 0.25 g of pRL-TK (Promega), 3 ng of 6ϫ NF-B-luc (39), and 2.5 g of pCDNA3. Where indicated, 0.8 g of Cdc37 expression construct and 0.7 g of FLAG-IKK␤ was transfected together with pcDNA3 vector up to 2.5 g. 24 h post-transfection, the cells were stimulated with 40 ng/ml TNF␣ for 4 h, as indicated, and luciferase activity was determined using a dual luciferase assay kit (Promega).
Metabolic Labeling-HeLa cells were labeled for 1 h with 70 Ci/ml [ 35 S]methionine (38) and lysed in 50 mM HEPES, pH 7.5, 150 mM NaCl, 1% Triton X-100, 1 mM EDTA, 1 mM dithiothreitol, 10 mM NaF, 8 mM ␤-glycerophosphate, 0.1 mM orthovanadate, 10% glycerol plus complete protease inhibitor mixture (Roche Applied Science). Extracts were precleared with protein A-Sepharose for 1 h and incubated with anti-IKK␥ antibodies overnight at 4°C. Finally, protein A-Sepharose was added for 3 h. Precipitates were washed five times with lysis buffer and boiled SDS loading buffer. The supernatant was applied on a SDS-PAGE and analyzed by autoradiography.

Hsp90-Cdc37 Interacts with the Core IKK Complex in a Transitory
Manner-To determine the role of Hsp90 and Cdc37 in IKK signaling, we analyzed the mode of their association with the IKK holocomplex. IKK␥ and IKK␣ precipitation from lysates of [ 35 S]methionine-labeled HeLa cells revealed only three prominent interacting species (Fig. 1A). Silencing of single IKK components with specific siRNAs verified the observed species as IKK␣, IKK␤, and IKK␥ (Fig. 1B). In contrast, silencing of Cdc37 or FKBP51, suggested as stoichiometric components (Hsp90-Cdc37) or regulatory co-factor (FKBP51) of the IKK complex (14, 24) did not inter-

FIGURE 2. Cdc37 preferentially binds to IKK␣ and connects Hsp90 and FKBP51 to the IKK complex.
A, HeLa cells were transfected with siRNA against the indicated proteins. The IKK complex was precipitated with an IKK␥ antibody, and co-immunoprecipitation of Hsp90 and Cdc37 was analyzed by Western blotting. Equal amounts of cell extracts (Input) were immunoblotted to determine protein expression levels. B, 293 cells were transfected with FLAG-Cdc37 and different Myc-IKK constructs as indicated. K/A, kinase-inactive K44A mutant. Lysates were precipitated with FLAG antibody. Precipitates were analyzed for Myc-IKK, Hsp90, and FLAG-Cdc37 by Western blotting. Expression levels in the lysates were also determined (Input). C, HeLa cells were transfected with the indicated siRNAs, and lysates were precipitated with an IKK␥ antibody. Co-immunoprecipitation of FKBP51 was analyzed by Western blotting. Cell extracts (Input) were immunoblotted to determine expression levels of IKK␣, IKK␤, and Cdc37. The specific band for FKBP51 is marked by an asterisk. D, 293 cells were transfected with FLAG-Cdc37 wild type (WT) or deletion constructs along with Myc-IKK␣. Lysates were precipitated with FLAG antibody. Precipitates were analyzed for Myc-IKK, Hsp90, and FLAG-Cdc37 by Western blotting.
fere with the pattern of IKK␥ co-precipitated proteins from metabolically labeled cells (Fig. 1B).
As a second approach, cell lysates were analyzed by size exclusion chromatography. IKK␣, IKK␤, and IKK␥ were identically eluted in fractions 12-16 with a peak in fraction 14 ( Fig.  1C) with an apparent molecular mass between 700 and 900 kDa, as described before (8,40,41). Upon gel filtration, Hsp90 (fractions 18 -22), Cdc37 (fractions 24 -28), and FKBP51 (fractions 22-26) eluted with distinct smaller sizes compared with the IKK complex (Fig. 1C). Moreover, TNF␣ stimulation did not change this gel filtration profile of Cdc37 and Hsp90 (Fig. S1). Because small amounts of Cdc37 and Hsp90 trailed into the IKK fractions, a stoichiometric contribution of Cdc37 and Hsp90 to the IKK complex could still be possible if Cdc37 and Hsp90 were present in high molar excess over the IKK components. However, quantification of the molar ratio of the endogenous proteins indicated only a roughly 2-3-fold molar excess of Cdc37 over IKK␤ (Fig. S2). These data reveal that only a small fraction of cellular IKK complexes interacts with Cdc37 and Hsp90 at a given time, considering that Hsp90 is largely recruited to IKKs via Cdc37 (see below). Surprisingly, our data gave no indication that the previously proposed essential subunit ELKS (105 kDa) (11) is part of the IKK complex ( Fig. 1, A-C). In fact, further investigations, addressing IKK binding and NF-B activation (data not shown), could not support ELKS as an essential, generally associated regulatory subunit of the IKK complex.
As expected for a stoichiometric component, silencing of IKK␣ (in contrast to ELKS, Cdc37, or FKBP51) caused a clearly discernible decrease of the apparent size of the residual IKK␥containing complex (Fig. 1D), indicating a different structure or oligomeric composition. The siRNA-mediated knockdown of indicated proteins (Fig. 1, B and D) was controlled by Western blotting (Fig. 1E). Taken together, we conclude that the chaperones are not stable stoichiometric components of the IKK complex but rather associate transiently.
Cdc37 Preferentially Interacts with IKK␣ to Recruit Hsp90 and FKBP51 to the IKK Complex-Although the importance of Hsp90 for IKK activation pathways seems unambiguous, it is not understood how the chaperone complex interacts with the IKK complex. Previous studies, using ectopically expressed IKK deletion mutants and GST pull-down assays, indicated that Cdc37-Hsp90 complexes bind to the kinase domains of IKK␣ and -␤ (14). Indeed, both Cdc37 and Hsp90 were co-precipitated with anti-IKK␥ antibodies, indicating association with the endogenous IKK complex ( Fig. 2A). Silencing of Cdc37 expression disrupted binding of Hsp90 to IKKs. Residual Hsp90 may reflect incomplete Cdc37 depletion. Notably, loss of IKK␣ but not IKK␤ interfered with Hsp90-Cdc37 recruitment, suggesting that preferentially IKK␣ bridges Cdc37 and Hsp90 to the IKK complex. Protein interaction studies in cells transiently expressing FLAG-tagged Cdc37 together with Myc-tagged IKK␣ and Myc-IKK␤ confirmed the highly preferential binding of Cdc37 to IKK␣ (Fig. 2B). Remarkably, a kinase domain mutant (K44A) of both IKKs bound with much higher affinity to Cdc37 compared with wild type (Fig. 2B), suggesting that structural alterations of the kinase domain influence Cdc37 interaction. A double band of Cdc37 was detected when cotransfected with wild type IKKs, suggesting IKK-dependent phosphorylation. Further examination revealed that FLAG-Cdc37 was phosphorylated at serine 300 by ectopically expressed IKKs (Fig. S3, A-C), in turn confirming that Cdc37 interacts with both IKKs. Moreover, TNF␣-stimulated endogenous IKK complexes phosphorylated Cdc37 in an in vitro kinase assay (Fig. S3D). Future analyses will be important to clarify the physiological function of IKK-dependent phosphorylation of Cdc37.
In addition to Cdc37 and Hsp90, we could show that the co-chaperone FKBP51, which has been suggested as kinase cofactor (24), also co-precipitates with endogenous IKK complexes (Fig. 2C). As observed for Hsp90, down-modulation of either Cdc37 or IKK␣ disrupted its binding. Thus, given that FKBP51 is a known Hsp90 binding partner (18,25), its recruitment to the IKK complex is indirect through Hsp90, Cdc37, and IKK␣. At present, the function of FKBP51 for IKK signaling is unclear, since downmodulation of FKBP51 interfered with neither TNF␣-induced IKK activity nor IB␣ degradation and p65 translocation (data not shown).
To investigate the mode of Hsp90-Cdc37-IKK␣ interaction in more detail, we transiently expressed FLAG-tagged Cdc37 mutant constructs together with Myc-tagged IKK␣. In co-immunoprecipitation studies, all deletion mutants were defective in Hsp90 association (Fig. 2D). Surprisingly, N-terminal (⌬N) as well as C-terminal deletion mutants (⌬C1, ⌬C2) could efficiently bind to IKK␣, indicating two independent IKK␣ binding sites on Cdc37 (Fig.  2D). One of these seems to be located in a central region of Cdc37, since binding of ⌬C1 to IKK␣ was diminished. In sum, these data confirm Cdc37 as an adaptor between Hsp90 and IKK␣.
Strict Requirement of Hsp90-Cdc37 for Activation of Ectopically Expressed IB Kinases-To define the functional role of Hsp90 and Cdc37 for the IKK activation process, we directly manipulated Hsp90 and Cdc37 levels in cells. Co-transfection of IKK␤ together with Cdc37 robustly increased NF-Bdependent transcription (Fig. 3A). Moreover, Cdc37 amplified activation loop phosphorylation of Myc-IKK␣ or IKK␤ (Fig. 3B). In turn, silencing of endogenous Cdc37 abolished T-loop phosphorylation and kinase activity of transfected IKK␤ without affecting IKK expression levels (Fig. 3C). Down-modulation of Hsp90␣ and Hsp90␤ likewise resulted in reduced kinase activity, but to a minor extent, due to incomplete knockdown (data not shown). . Cdc37 is required for the activity of de novo synthesized IB kinases. A, HeLa cells were transfected with siRNAs as indicated and treated with TNF␣ for 10 min or left untreated. Left, cell extracts were used for EMSA and Western blotting. Right, IB␣ protein expression levels before and after TNF stimulation were determined by Western blotting and quantified. Expression relative to unstimulated controls (percentage) is indicated. B, two-step transfection of HeLa cells with different siRNAs was performed as indicated. 4 -6 days after transfection, cells were treated with TNF␣ for 10 min. Cell extracts were immunoprecipitated with an IKK␥ antibody and analyzed for kinase activity. The -fold TNF induction of phosphorylated GST-I〉␣ compared with nonstimulated cells is indicated. Protein expression levels (Input) were determined by Western blotting. C, the indicated cell lysate was applied to a Superose 6 gel filtration column, and the fractions were analyzed for IKK␤ and -␥ expression.

JOURNAL OF BIOLOGICAL CHEMISTRY 32315
Ectopically expressed IKKs respond in a much more sensitive manner to modulation of Cdc37 expression than preformed cellular IKK complexes in TNF␣-induced cells, as judged by luciferase reporter (Fig. 3A) and kinase assays (Fig. S4). Taken together, these observations strongly suggest that newly synthesized IKKs require Hsp90-Cdc37 to achieve an inducible conformation.

Cdc37 Is Required for the Formation of Functionally Competent Endogenous de Novo Synthesized IKKs-The fact that
Hsp90-Cdc37 mediates maturation of ectopically expressed IKKs into a catalytically competent conformation alludes to similar functions for de novo synthesized endogenous IKKs. Thus, we analyzed Hsp90-Cdc37-dependent IKK activity in a situation where the role of de novo synthesis of endogenous IKK␣ and IKK␤ can be addressed.
In agreement with prior findings (Figs. 3A and S4), downmodulation of Cdc37 alone did not affect TNF␣-induced NF-B activation, as judged by NF-B DNA binding activity and IB␣ degradation (Fig. 4A). Although silencing of IKK␤ caused a robust inhibition of in vitro kinase activity (Fig. S4), only a moderate inhibition of TNF␣ induced IB␣ degradation, and concurrently a medium decrease in NF-B DNA binding could be observed (Fig. 4A). This discrepancy might be explained by different sensitivities of these assays (e.g. different contribution of residual IKK␤ toward endogenous substrate compared with excess GST-IB␣ used in the in vitro kinase assay). Intriguingly, combined siRNA-mediated depletion of Cdc37 and IKK␤ significantly inhibited IB␣ degradation and NF-B DNA-binding activity in a synergistic manner. We assume that under these conditions, a major portion of residual IKKs results from de novo synthesis, which in turn requires the contribution of Hsp90-Cdc37 as shown for ectopically expressed IKKs (Fig. 3).
For further verification, a two-step transfection strategy (at day 0 and day 3) was used to achieve long term knockdown of Cdc37 (Fig. 4B, lanes 4, 5, 8, 9, 12, and 13). IKK␣ and IKK␤ were silenced in parallel in a single step (day 0). In general, protein expression is recovered 5 or 6 days after siRNA transfection, thereby allowing the observation of IKK de novo synthesis and complex assembly. Hence, cells were stimulated with TNF␣ at days 4 -6, and lysates were examined for protein expression or kinase activity (Fig.  4B). At day 4, IKK␣ and IKK␤ protein expression was strongly reduced, thereby preventing kinase activation (lanes 3-6). At days 5 and 6 IKK␣ and IKK␤ expression levels increased (lanes 7-14), which in fact led to the production of new, TNF␣inducible IKK complexes in control cells (lanes 7, 10, 11, and 14). In contrast, loss of Cdc37 strongly diminished kinase activity of newly formed IKK complexes (lanes 8, 9, 12, and 13).
To test if chaperone activity is required for IKK complex assembly, we analyzed lysates of day 6 (lane 13) by gel filtration. Indeed, silencing of Cdc37 expression did not affect IKK complex assembly (Fig. 4C). In addition, we silenced Cdc37 expression and co-transfected FLAG-IKK␣, -␤, and -␥ in parallel. Neither IKK protein expression (Fig. S5A) nor complex assembly (Fig. S5B) was affected. Nevertheless, loss of Cdc37 again resulted in a significant reduction of IKK autophosphorylation and kinase activity (Fig. S5A). Taken together, these results establish an important function for Hsp90-Cdc37 in IKK maturation. Hsp90-Cdc37 interacts with de novo synthesized IB kinases to generate functionally competent IKK complexes.

Inhibition of Signal-induced Activation of Steady State IKKs Requires a Robust Block of Hsp90-Cdc37 Activity-Ectopic
Cdc37 expression only marginally interfered with TNF␣-induced NF-B activation (Fig. 3A). In line with this, neither silencing of Cdc37 and Hsp90 nor of IKK␣, which simultaneously abrogates binding of Cdc37 and Hsp90 ( Fig. 2A), strongly affected IKK activation (Figs. 4A, 5A, and S4). At first sight, these observations are in conflict with studies suggesting that GA inhibits IKK activation (14,29). However, significantly lower concentrations of the geldanamycin derivative 17-DMAG are required to inhibit NF-B DNA binding activity and IB␣ degradation (Fig. 5A) upon silencing of either Hsp90␣/␤ or Cdc37 (Fig. 5B). Apart from confirming the target specificity of 17-DMAG, these data also demonstrate a cooperative function of Hsp90 and Cdc37 in TNF␣-induced activation of mature IKKs. Due to the abundant expression level of Hsp90, a more robust inhibition of Hsp90-Cdc37 activity is needed to down-modulate signal-induced IKK and NF-B activation.
Hsp90-Cdc37 Acts Downstream of T-loop Phosphorylation in TNF␣ or DNA Damage-induced IKK Activation-To investigate whether Hsp90-Cdc37 acts preferentially on receptor-induced IKK cascades, we analyzed a mechanistically different signaling pathway, which promotes IKK and NF-B activation in response to DNA damage (6). Remarkably, pharmacological inhibition of Hsp90 nearly abolished DNA damage-induced NF-B DNA binding activity under conditions where protein expression of clients such as IKKs or RIP1 was unaffected (Fig.  6A). As shown for receptor-mediated signaling, singular knockdown of Cdc37 or Hsp90 was inefficient to disrupt DNA damage-induced NF-B activation (data not shown).
The fact that inhibition of Hsp90 interferes with both receptor-mediated and DNA damage-induced NF-B activation suggests that the chaperone directly acts on the IKK complex. Consequently, we assayed for IKK T-loop phosphorylation, which is the prerequisite for IKK activation in both experimental settings (Fig. 6B). Short pretreatment (4 or 2 h) with 17-DMAG significantly impaired TNF␣or DNA damage-induced IKK activity. In contrast, stimulus-dependent IKK T-loop phosphorylation was unaffected. These data reveal that kinase activity is not spontaneously augmented upon T-loop phosphorylation. Rather, a subsequent Hsp90-Cdc37-dependent process is required to generate a catalytically activated IKK complex, probably by mediating conformational alterations.

DISCUSSION
The chaperone complex Hsp90-Cdc37 has been suggested to stably associate with the IKK signalosome and to confer important regulatory functions in NF-B signaling, such as IKK recruitment to TNF receptors (14). Here we present evidence for a more general requirement of Hsp90-Cdc37 in IKK signaling (Fig. 7). Hsp90-Cdc37 interacts with the IKK complex in a transitory manner to promote two distinct functions. De novo synthesized IKKs depend on Hsp90-Cdc37 to achieve an enzymatically competent state. Upon stimulation either by receptor-mediated signaling or in response to DNA damage, mature IKKs again require Hsp90-Cdc37 following their T-loop phosphorylation to attain full catalytic activation. RNA interference of Cdc37 and Hsp90 interfered with IKK maturation, whereas inhibition of the final IKK activation step was only achieved by a more robust pharmacological block, indicating different threshold levels for the requirement of chaperone activities.
As a starting point, we investigated the composition of the proposed IKK holocomplex with approaches that allow stoichiometric predictions, such as gel filtration chromatography and IKK co-precipitation from metabolically labeled cells. Thus far, only IKK␣, -␤, and -␥ have been conclusively demonstrated as IKK subunits (8,9,41). In fact, our studies gave no evidence for Hsp90, Cdc37, or ELKS as bona fide components of the IKK complex (Fig. 1). These findings are contradictory to prior gel filtration experiments, showing co-elution of Hsp90-Cdc37 and ELKS with the IKK complex (11,14). The latter studies analyzed purified IKK complexes and did not address the stoichiometric composition of endogenous complexes. The assumption that IKK and Hsp90-Cdc37 form stoichiometric, stable complexes was supported by gel filtration experiments, where GA treatment resulted in a decrease of the apparent size of the IKK complex (14). However, extended exposure to GA causes ubiquitin-mediated proteasomal degradation of IKK␣ and -␤ (29), which results in altered migration of the IKK complex, as observed after down-modulation of IKK␣ (Fig. 1D). Therefore, we conclude that the suggested IKK holocomplex consists exclusively of the established IKK subunits, which transiently interact with a broad range of regulatory proteins, such as Hsp90-Cdc37.
The present study indicates a preferential interaction between Hsp90-Cdc37 and IKK␣ (Fig. 2). Structural domains, defining the assembly of Hsp90-Cdc37-client complexes have been identified recently. The central region of Cdc37 (amino acids 128 -282) contains the Hsp90 binding domain (reviewed in Refs. 23 and 25). In fact, partial deletion of this region disrupts Hsp90 recruitment to IKKs (Fig. 2D). The binding of Cdc37 to client proteins is more complex, since either the N-terminal portion (amino acids 1-126) or a central region (amino acids 191-195) has been proposed as crucial domains (42)(43)(44). Binding of IKK␣ is conferred by N-terminal as well as C-terminal regions of Cdc37, suggesting two independent binding sites (Fig. 2D). For the recognition of client proteins by Hsp90-Cdc37, two mechanisms have been suggested. Based on phage display analysis, a GXFG motif has been identified that is part of the canonical glycine-rich loop (GXGXXG) of protein kinases (45). Hsp90-Cdc37 was also shown to bind preferentially to positively or neutrally charged ␣C-␤4 loop regions of client kinases (32). Both regions are in proximity within the kinase domains of IKKs, which contain the binding site for the Hsp90-Cdc37 complex (14). Interestingly, kinase-dead mutants of both IKKs bind with strongly increased affinity to Cdc37 compared with the wild type proteins (Fig. 2B). In both kinases, the mutated residue (Lys-44) might alter the local structure and affect recognition of the flanked GXFG and ␣C-␤4 motifs at position 22 and 61, respectively. Although the co-immunoprecipitation data suggest highly preferential association with IKK␣, it is obvious that Cdc37 is capable of interacting functionally with IKK␤ as well, because FLAG-Cdc37 is equivalently phosphorylated at serine 300 upon co-expression with either kinase (Figs. 2 and S3). Phosphorylation of Cdc37 might function as part of a feedback mechanism, which alters the mode of interaction of Cdc37 with Hsp90 or with functional, correctly folded IKKs.
Hsp90 chaperone complexes primarily trigger functional maturation and activation of client proteins, although the molecular mechanisms are rather enigmatic (18,23,25). The present study reveals two modes of action for how Hsp90-Cdc37 acts on the IKK signalosome, namely maturation of de novo synthesized IB kinases and assistance during the catalytic cycle (Fig. 7). It has to be noted that modulation of Cdc37 expression affects T-loop autophosphorylation of ectopically expressed IKKs, which mechanistically differs from the signal-induced T-loop phosphorylation of mature endogenous IKKs by IKK kinases (Fig. 3). Ectopically expressed IKKs require Hsp90-Cdc37 for proper maturation, which allows T-loop autophosphorylation and subsequent full catalytic activation. Likewise, de novo synthesized endogenous IKKs require Cdc37 to gain a functionally active conformation, which then allows activation by upstream kinases (Fig. 4).
Previous studies, utilizing the Hsp90 inhibitor geldanamycin, have suggested that Hsp90-Cdc37 is required for TNF␣-dependent recruitment of the IKK complexes to TNF receptor 1 (14). However, extended Hsp90 inhibition strongly decreases expression of RIP1 (14,31,32), an essential component in this process (12). Notably, short term inhibition of Hsp90 impaired IKK kinase activation by various receptor-mediated signaling cascades without affecting expression levels of IKK subunits or signaling proteins (Fig. 7) (29). Chaperone activity is needed not only for receptor-mediated but also for DNA damage-induced IKK activation, which is a mechanistically distinct pathway (Fig. 6) (6,  7). Thus, Hsp90-Cdc37 acts as a generally required component in the activation process of the IKK complex. Signal-induced IKK activation depends on T-loop phosphorylation as an initial essential event. We could demonstrate that Hsp90 inhibition interferes with the activation of IKK kinase activity at a level downstream of T-loop phosphorylation (Fig. 6). In fact, the activation loop phosphorylation may not be the immediate activation step but rather the entry into a catalytic cycle, which primes IKK for activation. Induction of Ser/Thr kinases in general may involve conformational alterations (46,47), and a contribution of Hsp90-Cdc37 to these processes has been discussed (48). Accordingly, we propose that Hsp90-Cdc37 assists the IKK signalosome to undergo conformational changes that take place during the kinase activity cycle, after the T loops have been phosphorylated. Future structural analyses of the Hsp90-Cdc37-IKK interaction will be important to reveal further mechanistic details.