|Home | About | Journals | Submit | Contact Us | Français|
Genetic screens for modifiers of activated Ras phenotypes have identified a novel protein, kinase suppressor of Ras (KSR), which shares significant sequence homology with Raf family protein kinases. Studies using Drosophila melanogaster and Caenorhabditis elegans predict that KSR positively regulates Ras signaling; however, the function of mammalian KSR is not well understood. We show here that two predicted kinase-dead mutants of KSR retain the ability to complement ksr-1 loss-of-function alleles in C. elegans, suggesting that KSR may have physiological, kinase-independent functions. Furthermore, we observe that murine KSR forms a multimolecular signaling complex in human embryonic kidney 293T cells composed of HSP90, HSP70, HSP68, p50CDC37, MEK1, MEK2, 14-3-3, and several other, unidentified proteins. Treatment of cells with geldanamycin, an inhibitor of HSP90, decreases the half-life of KSR, suggesting that HSPs may serve to stabilize KSR. Both nematode and mammalian KSRs are capable of binding to MEKs, and three-point mutants of KSR, corresponding to C. elegans loss-of-function alleles, are specifically compromised in MEK binding. KSR did not alter MEK activity or activation. However, KSR-MEK binding shifts the apparent molecular mass of MEK from 44 to >700 kDa, and this results in the appearance of MEK in membrane-associated fractions. Together, these results suggest that KSR may act as a scaffolding protein for the Ras-mitogen-activated protein kinase pathway.
Ras activation is an essential step in signaling in response to a variety of extracellular signals, including receptor tyrosine kinase ligands which bind and activate their corresponding tyrosine kinase receptors. Activation of receptor tyrosine kinases leads to activation of Ras via the action of specific guanine nucleotide exchange factors. Activated Ras can physically interact with numerous downstream targets and activate several different signaling pathways (15).
One of the best-characterized Ras signaling pathways is the Raf-MEK-ERK pathway, also known as the mitogen-activated protein (MAP) kinase cascade (20, 25). Ras directly binds Raf in a GTP-dependent manner, and this interaction appears to be critical for recruitment of Raf to the membrane, where it undergoes activation. Activated Raf directly phosphorylates and activates MEK, also known as MAP kinase kinase, which in turn directly phosphorylates and activates ERK (25). Activation of ERK is critical for numerous Ras-induced cellular responses, including transcriptional activation of a number of genes (11). Ternary complex factors (TCFs) are among the best-characterized physiological substrates of ERK. Activated ERK directly phosphorylates and thereby activates the transcription activation potential of TCFs (8, 12, 18). It appears that TCFs, in association with the serum response factor, play an essential role in the activation of many mitogen-inducible genes (11).
The Ras-MAP kinase pathway is highly conserved in eukaryotes. Genetic studies of Caenorhabditis elegans and Drosophila melanogaster indicate that this pathway is involved in many developmental programs (15). Genetic screens for mutations that suppress constitutively active Ras mutants have identified numerous components of the MAP kinase pathway. Such screens identified KSR (for kinase suppressor of Ras), a putative protein kinase with significant sequence identity to Raf (17, 30, 31). Genetic data indicated that KSR plays a positive role in Ras signaling and functions parallel to, or downstream of, Ras. Microinjection experiments using Xenopus oocytes showed that KSR is able to enhance Ras-induced germinal vesicle breakdown and MAP kinase activation, indicating that KSR has a positive role in Ras signaling in vertebrates (32). However, recent reports suggest that KSR may have a negative role in certain aspects of Ras signaling (6, 14, 29, 39). KSR was found to inhibit Ras-induced cellular transformation in NIH 3T3 cells and to inhibit MEK and/or ERK activation (6, 14, 39). However, the mechanism of this inhibition remains to be elucidated. We have previously reported that KSR selectively inhibits TCF phosphorylation and transcription activity, without significantly affecting ERK activation in COS1 cells (29). Therefore, the biochemical function of KSR in mammalian cells is rather perplexing, though KSR clearly plays a role in Ras-MAP kinase signal transduction.
The deduced amino acid sequence of KSR predicts it to be a protein kinase, and one report has suggested that KSR is a ceramide-activated protein kinase (40). However, whether KSR has intrinsic kinase activity remains to be confirmed. It has been proposed that KSR may function as a scaffold protein to assemble a signaling complex in mammalian cells (14, 39). Some limited data exist to support this notion. Previous studies have suggested that KSR can interact directly with 14-3-3 proteins, MEK, and possibly ERK (6, 14, 37, 39). In addition, KSR translocates to membrane fractions and associates with Raf (perhaps indirectly) in a Ras-dependent manner (22, 37).
We tested the hypothesis that KSR may function independently of its presumed protein kinase activity. Surprisingly, we observed that upon microinjection, kinase domain mutants of KSR complement KSR loss-of-function alleles in C. elegans vulval induction. With this in mind, we turned our attention to the mouse homolog of KSR (mKSR), testing whether kinase-independent functions of mKSR exist in mammalian cells and whether they are sensitive to mutations analogous to genetically isolated loss-of-function alleles. Consistent with this possibility, we observed that KSR is a component of a multimolecular complex in human embryonic kidney (HEK) 293T cells consisting of MEK1, MEK2, HSP90, HSP70, HSP68, p50CDC37, and 14-3-3 in addition to several other, unidentified proteins. ERK also associates with KSR, though this binding appears to be much weaker than KSR-MEK interactions. A loss-of-function mutant, KSR C809Y, specifically lacks the ability to interact with MEK yet retains the ability to bind other KSR-associated proteins. Interestingly, ectopic KSR expression alters the apparent molecular mass of MEK from 44 kDa to approximately 700 kDa and results in the translocation of MEK from a soluble to a membrane-associated fraction. Treatment of PC12 cells with nerve growth factor (NGF) resulted in induction of KSR protein levels and a concomitant increase in KSR-MEK association, suggesting that KSR may play a role in vertebrate differentiation. These data suggest that KSR may function as a scaffolding protein in vivo.
Standard methods were used for the handling and culture of animals (2). Mutations used were let-60(n1046) (G13E) (1, 9), ksr-1(ku68) (R531H) (29), and ksr-1(n2526) (W255STOP) (17). The ksr-1 transgenes pMS44 (wild type), pMS77 (K503M), pMS108 (D618A), and pMS153 (R531H) consist of genomic positions 1809 (BamHI) to 15401 (BamHI) from cosmid F13B9 (GenBank accession no. U39853), followed by cDNA positions 520 (BamHI) to 2394 (ClaI) from ksr-1 (GenBank accession no. U38820), followed by genomic positions 19965 (ClaI) to 21454 (SacII) from cosmid F13B9, in pBluescript SK(+). Point mutations were introduced into the cDNA fragment by PCR. For in vitro translation, C. elegans KSR-1 cDNAs were cloned into the BamHI site pcDNA3-HA (29) to create either pMS82 or pMS83. C. elegans KSR-1b (CeKSR-1b; pMS183) consists of the entire cDNA; CeKSR-1a (pMS182) begins from an alternate initiation methionine at position 189 and may be a naturally occurring variant (29a).
HEK 293T cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) containing 10% fetal bovine serum (FBS; Life Technologies). Transfections were performed by using Lipofectamine (Life Technologies) as directed by the manufacturer. Rat PC12 cells were cultured in DMEM containing 10% horse serum and 5% FBS and were induced to differentiate by incubating in DMEM containing 2% horse serum, 1% FBS, and 50 ng of NGF (Calbiochem) per ml for 5 days, with a change of medium every 48 h.
To construct a yeast expression vector for two-hybrid screening, the entire open reading frame of mKSR was subcloned into the bait vector plex-Ade (kindly provided by Anne Vojtek), creating an in-frame fusion with LexA. To construct Lex-Ade-KSR KD and Lex-Ade-KSR KD C809Y fusions, a 1,188-bp EcoRV fragment from pcDNA3-HA-KSR or pcDNA3-HA-KSR C809Y (see below) was ligated into plex-Ade that had been digested with BamHI and blunted with Klenow polymerase. MEK1-Gal4 activation domain fusions were constructed by ligating the entire open reading frame of human MEK1 or MEK1Δproline rich (kindly provided by Melanie Cobb) into pGAD10 (Clontech). Expression vectors encoding point mutants of murine KSR were constructed by site-directed mutagenesis of pcDNA3-HA-KSR (29), using a Quick Change kit (Stratagene). pcDNA3-HA-KSR 1-301 was constructed by amplifying amino acids 1 to 301 of mKSR via PCR and subcloning into the XbaI-EcoRI sites of pcDNA3-HA (29). HA-KSR KD was constructed by amplifying the region coding for amino acids 500 to the stop codon of mKSR and subcloning as an XbaI-EcoRI fragment into pcDNA3-HA. pcDNA3-HA-MEK2 was constructed by subcloning a BamHI fragment containing the entire open reading frame of human MEK2 (41) into the BamHI site of pcDNA3-HA. Expression vectors for Elk-1 have been described elsewhere (28).
Glutathione S-transferase (GST)–MEK1, GST-ERK1, and a kinase-dead mutant, GST-ERK1 KR, were expressed in Escherichia coli and purified as described previously (28, 41). GST-CeMEK2 was expressed and purified as described elsewhere (36b).
Subconfluent 293T cells in 10-cm-diameter dishes were transfected with 10 μg of the indicated expression vector; 48 h posttransfection, cells were labeled for 4 h in cysteine- and methionine-free DMEM–10% dialyzed FBS (Life Technologies) containing [35S]methionine/cysteine at 200 μCi/ml (Tran35S-label; ICN). Cells were lysed in buffer (10 mM Tris-Cl [pH 7.5], 150 mM NaCl, 50 mM NaF, 1 mM EDTA, 1% Nonidet P-40 [NP-40]) containing 1 mM dithiothreitol [DTT], 1 mM NaVO4, and a protease inhibitor cocktail (Complete; Boehringer Mannheim) for 20 min at 4°C. Precleared lysates were incubated for 3 h at 4°C with 10 μg of antibody precoupled to protein G-Sepharose (Pharmacia) or protein A-agarose (Pierce). For peptide competition, antibody was preincubated with hemagglutinin (HA) peptide (YPYDVPDYA) before addition to lysates. Immunocomplexes were collected by gentle centrifugation, washed five times in lysis buffer, boiled in sodium dodecyl sulfate (SDS) sample, and separated by SDS-polyacrylamide gel electrophoresis (PAGE). For KSR immunoprecipitation experiments using PC12 cells, 1 mg of extract from PC12 cells treated with or without NGF for 5 days was immunoprecipitated with 10 μg of KSR-specific antibody (α-KSR; Santa Cruz Biotechnology) in 1% NP-40–150 mM NaCl lysis buffer. Where indicated, antibody was incubated with a twofold excess of immunizing peptide prior to immunoprecipitation.
To determine the effects of geldanamycin on KSR protein levels, 293T cells transfected with HA-KSR were labeled for 30 min with 100 μCi of [35S]methionine/cysteine per ml with or without 10 μM geldanamycin (Sigma), followed by a chase with complete medium for the indicated times with methionine/cysteine-containing medium. Lysates were prepared and immunoprecipitated as described above with α-HA followed by SDS-PAGE and autoradiography.
For GST pull-down experiments, full-length CeKSR and mKSR cDNAs were translated in vitro (Promega) in the presence of [35S]methionine. A portion of the translation reaction mixture was incubated with 5 μg of either GST, GST-human MEK1, or GST-CeMEK2 bound to glutathione-Sepharose (Pharmacia) for 1 h at 4°C in 1% NP-40–150 mM NaCl buffer followed by four washes, SDS-PAGE, and autoradiography.
For gel filtration experiments, two 10-cm-diameter dishes of 293T cells, transfected as described above, were lysed by homogenization in phosphate-buffered saline (PBS) plus protease and phosphatase inhibitors. Cellular debris was pelleted by centrifugation in a microcentrifuge at 6,000 × g for 15 min at 4°C. The resulting supernatant (about 1 mg of protein) was fractionated on a Superose 6-HR column (Pharmacia) equilibrated in PBS at a flow rate of 0.5 ml/min. Fractions of 333 μl, 15 μl of which was subjected to SDS-PAGE and immunoblot analysis with the indicated antibodies, were collected.
For subcellular fractionation, transfected 293T cells were lysed by homogenization in PBS containing protease and phosphatase inhibitors. Cellular debris was removed by centrifugation at 6,000 × g for 15 min at 4°C. The resulting supernatant was centrifuged for 1 h at 100,000 × g at 4°C. The microsomal pellet (P100) was resuspended in buffer containing 0.5% NP-40 followed by centrifugation at 13,000 × g to fully remove remaining insoluble material. Equal volumes of each fraction were examined by SDS-PAGE and immunoblotting with the indicated antibodies.
α-HA immunocomplexes from approximately 5 × 108 293T cells transfected with HA-KSR were subjected to SDS-PAGE and Coomassie blue staining. Bands were excised, digested with lysyl endopeptidase, purified via reversed-phase high-performance liquid chromatography (HPLC), and sequenced as described previously (38).
KSR or MEK immunoprecipitates were equilibrated in buffer (25 mM HEPES [pH 8.0], 0.5 mM EDTA, 0.25% β-mercaptoethanol) before addition of 1 μg of GST-ERK1 or GST-ERK1 KR in 10 mM HEPES (pH 8.0)–10 mM MgCl2–1 mM DTT–50 μM ATP in a volume of 30 μl. GST-ERK1 or GST-ERK1 KR activation was allowed to proceed for 15 min at 30°C, after which 5-μl aliquots were removed and assayed for ERK activity for 15 min at 30°C in 10 mM HEPES (pH 8.0)–10 mM MgCl2–1 mM DTT–50 μM ATP–10 μCi of [γ-32PO4]ATP (ICN) and 10 μg of myelin basic protein (MBP; Sigma). Reactions were terminated by adding EDTA (pH 8.0) to 50 mM, and reaction mixtures were spotted on P81 phosphocellulose paper (Whatman), washed, and counted in a scintillation counter.
Two-hybrid screening using mKSR as a bait was carried out essentially as described elsewhere (35). α-14-3-3 and α-KSR were from Santa Cruz Biotechnology; α-HA was from Babco; α-phospho-MEK antibody was from New England Biolabs; α-MEK and α-ERK have been described elsewhere (42); and α-HSP90 was from Transduction Laboratories. Blots were developed by using enhanced chemiluminescence (Amersham).
Mutations in the C. elegans ksr-1 gene were isolated as suppressors of the Multivulva (Muv) phenotype caused by activated Ras (17, 30), suggesting that in C. elegans, the KSR-1 protein normally plays a positive role in Ras-mediated signaling. Of the 12 ksr-1 mutations originally described (17, 30), 8 are missense mutations affecting the putative kinase domain; this finding suggested that kinase activity could be important for KSR-1 function. To test directly the importance of kinase activity, we constructed ksr-1 transgenes bearing substitutions predicted to eliminate kinase activity and tested these transgenes for the ability to complement the suppressor phenotype caused by ksr-1 loss-of-function mutations. In these assays, complementation is observed as a restoration of the activated Ras Muv phenotype. We created a mutant in the Mg2+-ATP binding motif (K503M) as well as a mutant lacking the catalytic nucleophile aspartic acid (D618A). These residues are highly conserved in protein kinases, and mutations analogous to these in other protein kinases have been used to create kinase-inactive molecules. Surprisingly, both of the kinase-dead transgenes, KSR-1 K503M and KSR-1 D618A, retained complementing activity comparable to that of wild-type KSR-1 (Table (Table1).1). However, a KSR-1 R531H transgene (bearing the same substitution caused by the endogenous ksr-1 allele ku68) lacked complementing activity. These data argue that KSR-1 does not require kinase activity to promote Ras signaling and are inconsistent with models in which KSR acts as a protein kinase. They do, however, raise the possibility that KSR has other biochemical functions.
The above results, in addition to repeated failed attempts to detect intrinsic kinase activity in vitro, led us to test whether any other biochemical function could be attributed to KSR. We therefore performed immunoprecipitation experiments using transfected HEK 293T cells expressing HA-tagged mKSR to identify KSR-associated proteins. When 35S-labeled α-HA immunocomplexes were subjected to SDS-PAGE and autoradiography, we observed that several proteins associated specifically with HA-KSR (Fig. (Fig.1A,1A, lane 3). Inclusion of HA peptide prior to immunoprecipitation competed all coprecipitated proteins, demonstrating the specificity of the immunoprecipitation (Fig. (Fig.1A,1A, lane 2). These coimmunoprecipitated bands were not due to C-terminal proteolytic degradation of HA-KSR, as immunoblot analysis with α-HA detected only KSR. In fact, several of the KSR-associated bands have been identified as distinct cellular proteins, as opposed to fragments of KSR (see below). In addition, certain KSR-associated proteins can be coprecipitated by N- or C-terminal truncations of KSR, which themselves migrate faster than some of the KSR-associated proteins (Fig. (Fig.2A).2A). We obtained similar results using COS1 cells or with a Myc-tagged KSR expression vector. These results demonstrate that KSR specifically associates with numerous cellular proteins.
We reproducibly detect several major HA-KSR-associated proteins in immunoprecipitation experiments; a number of these proteins (p90, p70, p68, p60, p50, p46, p44, and p30) are present in near-stoichiometric amounts with respect to one another and can be detected by Coomassie blue staining of as few as 107 transfected cells (Fig. (Fig.1C).1C). Size exclusion chromatography demonstrated that transfected KSR has a molecular mass of about 106 Da (Fig. (Fig.3A),3A), suggesting the possibility that KSR is a constituent of a multiprotein complex in vivo. These observations are supported by the fact that a significant portion of KSR-associated proteins remains bound even under high-salt conditions (1 M NaCl) or in the presence of 0.1% SDS (data not shown).
Previous reports have indicated that KSR is capable of associating with MEK1 and -2 as well as 14-3-3 proteins (6, 37, 39). We have also isolated 14-3-3 β and MEK2 in yeast two-hybrid screens using as baits full-length mKSR and the kinase domain of KSR, respectively (data not shown). We therefore tested whether MEK1, MEK2, or 14-3-3 proteins were present in our KSR immunoprecipitates. Immunoblot analysis using α-MEK1/2 clearly detected the presence of a tightly spaced pair of immunoreactive bands corresponding to MEK1 and MEK2, respectively (Fig. (Fig.1A,1A, lane 4). The presence of MEK in the KSR complex was also demonstrated by direct peptide sequencing (Table (Table2).2). The specificity of the KSR-MEK interaction was confirmed by reciprocal coimmunoprecipitation. α-MEK specifically precipitated HA-KSR, which was not precipitated by a nonspecific antibody (Fig. (Fig.1D,1D, lanes 3 to 6).
Similarly, we also probed HA-KSR immunoprecipitates with α-14-3-3. We detected immunoreactive bands that correspond well with the predicted molecular mass of 14-3-3 proteins, approximately 30 kDa (Fig. (Fig.1A,1A, lane 4). Immunoprecipitation with α-14-3-3 also demonstrated the associated HA-KSR (Fig. (Fig.1D,1D, lane 2). By peptide sequence analysis, we demonstrated that the 30-kDa KSR-associated protein is a 14-3-3 protein (Table (Table22).
In marked contrast to HA-KSR-transfected cells, no specific proteins were associated with HA-MEK2 in immunocomplexes in the absence of KSR transfection under identical conditions (Fig. (Fig.1B,1B, lane 1). Likewise, Myc-14-3-3 β immunocomplexes consisted primarily of 14-3-3 (Fig. (Fig.1B,1B, lane 2). The fact that no endogenous KSR was visible in HA-MEK2 and Myc-14-3-3 β immunocomplexes is consistent with our observation that 293T cells express little or no endogenous KSR (Fig. (Fig.2F).2F). Thus, we conclude that the large number of proteins copurifying with HA-KSR are not MEK- and/or 14-3-3-associated proteins but rather appear to be highly specific and dependent on the presence of HA-KSR.
To determine the identity of KSR-associated proteins, we attempted direct protein sequencing from α-HA immunoprecipitates. We were successful in obtaining amino acid sequences from peptides corresponding to p90, p70, p68, p44/43, and p30. The sequences of two peptides from p90 were perfect matches to the human heat shock protein HSP90 (Table (Table2).2). Peptide sequences of p70 and p68 were identical to those of human HSP70 and HSP68, respectively (Table (Table2).2). We also tested whether p50CDC37 (4, 7, 16, 27), a protein kinase-targeting subunit of the HSP90 complex, was present in KSR immunocomplexes. Immunoblot analysis revealed the presence of a 50-kDa protein using α-p50CDC37, suggesting that p50CDC37 is capable of associating with KSR (Fig. (Fig.11A).
The functions of HSP90 and p50CDC37 have been implicated in signal transduction and cell cycle regulation through assembly of protein kinase complexes (4, 7, 16, 23, 24, 26, 27, 34). Therefore, we tested the effects of the HSP90 inhibitor geldanamycin on KSR protein levels since it has been observed that geldanamycin treatment reduces levels of Raf, which also associates with the HSP90 complex (26). Treatment of HA-KSR-transfected cells with geldanamycin significantly reduced steady-state levels of KSR yet had no effect on either MEK or ERK (Fig. (Fig.1E).1E). Consistent with this finding brief treatment of HA-KSR-transfected 293T cells with geldanamycin nearly abolished the association between KSR and HSP90 and p50CDC37, as determined by coimmunoprecipitation experiments (Fig. (Fig.1F).1F). To test whether geldanamycin destabilizes KSR, we performed pulse-chase labeling experiments in the presence or absence of geldanamycin. Treatment of HA-KSR-transfected cells with geldanamycin significantly reduced (>50%) the half-life of HA-KSR (Fig. (Fig.1G),1G), suggesting that the HSP90 complex may act to stabilize KSR in vivo.
Several missense mutations in CeKSR-1 were identified as loss-of-function alleles in genetic screens (17, 30). Some of these are mutations at positions absolutely conserved in all protein kinases and would therefore be likely to affect the overall kinase structure of KSR (10). Other mutations, however, are found in residues conserved only within the Raf and KSR subfamily of protein kinases. Since our data indicated that kinase activity was not required for KSR function in C. elegans, we tested whether mutants analogous to genetically derived loss-of-function alleles would alter formation of the multimolecular KSR complex. We constructed HA-KSR expression vectors bearing the substitution G580E, R615H, or C809Y in the kinase domain of mKSR corresponding to C. elegans mutant alleles ku83, ku68, and ku148, respectively (30). 35S metabolic labeling and immunoprecipitation experiments were performed as described above. These mutant proteins were efficiently expressed and generally associated with all KSR-associated proteins exception MEK1 and MEK2 (Fig. (Fig.2A).2A). No detectable MEK1 or MEK2 was observed in HA-KSR C809Y immunoprecipitates, while the amount of MEK1/2 associated with HA-KSR G580E and R615H mutants was significantly reduced (Fig. (Fig.2A2A and B). Thus, in these cases, loss of ksr-1 function in C. elegans correlates with reduced MEK binding of the corresponding murine KSR mutant protein. We also tested HA-KSR-L56G,R57S, containing a substitution in the CA1 domain corresponding to the Drosophila weak loss-of-function allele S-548 (31). HA-KSR L56G,R57S displayed wild-type binding to all KSR-associated proteins.
All protein kinases contain a conserved lysine residue in the ATP binding domain (10). Surprisingly, both mouse and human KSRs contain an arginine residue in place of the conserved lysine at this position, whereas C. elegans and Drosophila KSRs contain a lysine. HA-KSR R589M, which contains a methionine in place of this arginine and is presumably kinase dead, was also tested for its ability to associate with MEK proteins. HA-KSR R589M also displayed a decreased association with MEK1 and MEK2 (roughly 50% of wild-type level) but still interacted with other KSR-associated proteins (Fig. (Fig.2A2A and B, lanes 3). This appears to contradict our observation that a ksr-1 transgene encoding a K503M substitution can still function in vulva induction. However, it is not clear whether the two mutant proteins in question, mKSR R589M and CeKSR-1 K503M, are biochemically equivalent.
These observations were expanded upon by immunoblotting the same immunoprecipitates with α-MEK1/2 and α-14-3-3. Our results demonstrate that the KSR C809Y mutant is completely defective in MEK1/2 association yet retains wild-type affinity for 14-3-3 proteins (Fig. (Fig.2B).2B). Similarly, HA-KSR G580E, R615H, and R589M showed a significant decrease in MEK1/2 association, while the association with 14-3-3 proteins was not changed (Fig. (Fig.22B).
We tested which regions of KSR mediate binding and complex formation. To accomplish this, we expressed either the N-terminal amino acids 1 to 301 of KSR (HA-KSR 1-301), a region which contains the N-terminal CA1 and CA2 domains (31), or the kinase domain (HA-KSR KD; amino acid 500 to the stop codon) of KSR. In vivo labeling and immunoprecipitation experiments demonstrated that the isolated kinase domain of KSR was sufficient for binding to most of the proteins detected in full-length HA-KSR immunoprecipitates (Fig. (Fig.2A,2A, lane 9). Specifically, HSP70, HSP68, p60, and p50CDC37 were capable of associating with the isolated kinase domain of KSR. Immunoblot analysis with αHSP90 revealed that HSP90 is also easily detectable in HA-KSR KD immunocomplexes, though it does not appear particularly dramatic on this gel (data not shown). In contrast, the only proteins that were able to associate with the amino-terminal 301 amino acids were p36, p33, and in a reduced amount, p34 (Fig. (Fig.2A,2A, lane 8). These results suggest that different domains of KSR can bind different KSR-associated proteins in vivo.
We performed experiments in the yeast two-hybrid system to determine if KSR directly interacts with MEK. We observed that LexA-mKSR construct showed little interaction with MEK1 in the yeast two-hybrid assay (Table (Table3).3). In contrast, the isolated kinase domain of KSR fused to LexA interacts with MEK1 strongly in yeast. Mutation of cysteine 809 to tyrosine completely eliminated the interaction with MEK (Table (Table3).3). These results indicate that cysteine 809 of KSR may be directly involved in its interaction with MEK1 and MEK2. MEK1 and MEK2 contain a proline-rich region between the kinase subdomains IX and X (5, 10, 13). This proline-rich domain has been implicated in interaction between Raf and MEK and has a role in the biological function of MEK (5, 13). Our results indicate that the proline-rich region of MEK1 is not required, however, for interaction with KSR, as deletion of it had no effect on KSR KD-MEK1 interaction (Table (Table3).3).
The presence of ERK was reproducibly detected in the KSR complex (Fig. (Fig.2C),2C), consistent with previous observations (39). However, the amount of ERK associated with KSR appears to be far less than the amounts of MEK1 and MEK2, since we cannot detect ERK by Coomassie blue staining, although MEK1 and MEK2 are easily detectable (Fig. (Fig.1D).1D). The association between ERK and KSR was confirmed by reciprocal immunoprecipitation. α-ERK consistently precipitated far less KSR than α-MEK (Fig. (Fig.2D;2D; compare long and short α-HA exposures). These results indicate that KSR may weakly or indirectly interact with ERK. Interestingly, the KSR C809Y mutant failed to precipitate detectable amounts of ERK (Fig. (Fig.2C,2C, D), suggesting that MEK1/2 binding may affect this interaction.
The results shown in Fig. Fig.2A2A and B indicate that KSR interacts with MEK and 14-3-3 proteins via different domains. We tested whether KSR could mediate interactions between MEK and 14-3-3 proteins. α-14-3-3 precipitated MEK1 and MEK2 only upon transfection of HA-KSR (Fig. (Fig.2E).2E). Furthermore, MEK–14-3-3 interaction was not observed in the absence of KSR expression or when the MEK binding-defective KSR C809Y mutant was expressed (Fig. (Fig.2E,2E, lane 3). The above results support the hypothesis that the interaction between MEK and 14-3-3 is facilitated by KSR and suggest that KSR can serve as a link between MEK and 14-3-3 proteins.
We also tested whether endogenous KSR was associated with MEK, HSP90, p50CDC37, and 14-3-3. Immunoblot analysis with αKSR revealed that mouse brain and PC12 cells express detectable amounts of KSR (Fig. (Fig.2F,2F, left panel). α-KSR-reactive bands were competed by preincubation with peptide antigen, demonstrating the specificity of the antibody (Fig. (Fig.2F,2F, right panel). We also observed that the level of KSR was significantly increased during NGF-induced differentiation of PC12 cells. Immunoprecipitation experiments from differentiated PC12 cell extracts revealed that KSR antibody coprecipitated endogenous MEK1 and MEK2, in addition to HSP90, p50CDC37, and 14-3-3 protein (Fig. (Fig.2G).2G). KSR immunoprecipitates from differentiated PC12 cells contain considerably more MEK than those from undifferentiated cells, presumably due to increased KSR protein levels in differentiated cells (Fig. (Fig.2G).2G). We reasoned that C. elegans MEK and KSR homologs should be capable of direct interaction. Therefore, both C. elegans and murine KSRs were in vitro translated and tested for MEK binding in a GST pull-down assay using either GST, GST-CeMEK2 (36b), or GST-human MEK2. Our data clearly indicate that both nematode and mammalian KSRs can specifically associate with MEKs in vitro (Fig. (Fig.2H)2H) and appear to bind with comparable affinities. These results demonstrate that MEK-KSR binding is an evolutionarily conserved function of KSR.
If KSR and MEK form a large, stable complex, this may be reflected in their molecular weights and/or subcellular distribution. In 293T cells, MEK1 and MEK2 exhibit an apparent molecular mass of 44 kDa based on Superose 6 size-exclusion chromatography (Fig. (Fig.3A).3A). This is in good agreement with the predicted molecular weights of monomeric MEK1 and MEK2. However, in KSR-transfected cells, the apparent molecular mass of MEK shifts to approximately 700 kDa (Fig. (Fig.3A).3A). In fact, endogenous MEK quantitatively resides in a high-molecular-weight complex in KSR-transfected cells. We tested whether this alteration in the apparent molecular mass of MEK requires its interaction with KSR. Identical experiments performed with the KSR C809Y mutant did not alter the apparent molecular mass of MEK (Fig. (Fig.3A).3A). We also tested the possibility that the large KSR complex that we observed was due to KSR self-oligomerization. HA-KSR and a Myc-tagged variant of mKSR were coexpressed in 293T cells. Lysates were immunoprecipitated with either α-HA or α-Myc, and the extent of KSR self-oligomerization was examined by immunoblotting with α-Myc or α-HA. No detectable KSR oligomerization was observed with either HA- or Myc-tagged KSR (data not shown). These results support our hypothesis that the KSR complex contains KSR and other KSR-associated proteins, including MEK.
MEK normally exists as a soluble cytoplasmic protein (42). KSR has been reported to exist in both cytoplasmic and membrane-associated forms (22, 37). To test whether association with KSR altered the subcellular distribution of MEK, KSR-transfected cells were fractionated into soluble and particulate fractions by ultracentrifugation. Immunoblot analysis indicated that KSR exists in both soluble (S100) and particulate (P100) fractions (Fig. (Fig.3B).3B). Immunoblotting with α-MEK indicated that a significant portion of MEK existed in the P100 fraction in KSR-transfected cells (Fig. (Fig.3B).3B). Membrane association of MEK required KSR binding, as the KSR C809Y mutant did not noticeably alter the subcellular distribution of MEK. As with KSR C809Y, KSR KD had no detectable effect on MEK localization. The above observation suggests that a potential physiological role of KSR is to alter the subcellular distribution of MEK. It is interesting that the KSR C809Y mutant exists almost exclusively in the P100 fraction whereas a significant portion of wild-type KSR remains in the S100 fraction (Fig. (Fig.3B).3B). This observation suggests that MEK binding can, in turn, affect the subcellular distribution of KSR. Thus, it appears that the localization of MEK and KSR is a cooperative process. Immunoblotting with α-c-Raf revealed that c-Raf distribution between S100 and P100 fractions was not significantly altered by expression KSR (Fig. (Fig.3B),3B), which is consistent with our observation that little endogenous c-Raf is associated with KSR (data not shown).
We tested the effects of expressing wild-type KSR or KSR C809Y on the ability of MEK1 and MEK2 to undergo phosphorylation and activation in response to growth factors. Epidermal growth factor (EGF) stimulated a dramatic increase of active, phosphorylated MEK1 and MEK2, which can be detected by α-phospho-MEK. Cotransfection of either HA-KSR or KSR C809Y had no effect on the ability of EGF to stimulate MEK1 and MEK2 phosphorylation (Fig. (Fig.4A).4A).
To directly examine the phosphorylation state of MEK1 and MEK2 in the KSR complex, we transfected cells with either HA-KSR, HA-KSR C809Y, or HA-MEK2; then lysates from cells grown in the presence of serum were precipitated with α-HA and blotted with α-MEK and α-phospho-MEK (Fig. (Fig.4B).4B). Our results clearly indicate that the MEK1/2 in the KSR complex is phosphorylated. In fact, the relative phosphorylation of MEK1/2 in the KSR complex is no less than that in HA-MEK2 immunoprecipitates without KSR transfection (Fig. (Fig.4B).4B). We also directly examined the ability of growth factors to induce MEK activity in the KSR complex by a coupled kinase assay. Our results indicate that coprecipitated MEK1/2 in the KSR complex is fully capable of activating recombinant ERK1 (Fig. (Fig.4C4C and D). Furthermore, the specific activity of MEK in the KSR complex was no different than that of HA-MEK2 alone (Fig. (Fig.4C4C and D). HA-KSR C809Y immunocomplexes contained virtually no detectable MEK activity. This finding is consistent with our observation that the C809Y mutant is unable to associate with MEK and also demonstrates the high specificity of this kinase assay (Fig. (Fig.4D).4D). These results demonstrate that KSR-bound MEK1/2 can be phosphorylated and activated by Raf, and, in turn, can phosphorylate and activate GST-ERK1 in vitro. Our results support the hypothesis that KSR inhibits neither the activity nor the activation of MEK but rather modulates its localization in the cell. However, it remains possible that the effect of KSR on MEK activation is cell type dependent or dependent on relative expression levels.
We have observed that KSR can specifically block phosphorylation and activation of Elk-1, a physiological substrate of MAP kinases (29). To test the relationship between the ability of KSR to block Elk-1 phosphorylation and its ability to interact with MEK, KSR C809Y was used. Although C809Y may affect other functions of KSR, a plausible explanation is that KSR-MEK interaction is needed for its ability to block Elk-1 activation in these cells. We tested this hypothesis. KSR effectively blocked EGF-stimulated Elk-1 phosphorylation, while KSR C809Y had no effect on Elk-1 phosphorylation (Fig. (Fig.5A).5A). This result suggests that the ability of KSR to bind MEK1 and MEK2 may correlate with its ability to block Elk-1 activation by MAP kinases. We also tested truncated mutants of KSR in the same assay to determine which regions of KSR are necessary and sufficient to inhibit Elk-1. Our results show that the full-length KSR protein is required to inhibit Elk-1 activation (Fig. (Fig.5B).5B).
KSR was originally identified by genetic means to be a regulator of Ras-MAP kinase signaling pathways controlling Drosophila photoreceptor differentiation and C. elegans vulval induction (17, 30, 31). Subsequent studies confirmed that KSR indeed plays a role in Ras-MAP kinase signaling in Xenopus and mammalian cells as well (6, 14, 29, 32, 37, 39). However, the mechanism of KSR function is not well understood in any system. Although the sequence of KSR predicts it to be a protein kinase, to date we are unable to demonstrate any kinase activity intrinsic to KSR whether it is expressed in bacteria, insect cells, or mammalian cells as either a full-length protein or the isolated C-terminal kinase domain. Our C. elegans data argue that if KSR is a protein kinase, this activity is not required for its positive signaling function during vulval induction.
We observed that mutation of the Mg2+-ATP-coordinating lysine residue (KSR-1 K503M) or the catalytic nucleophile aspartic acid residue (KSR-1 D618A) did not compromise the function of ksr-1 in C. elegans vulval induction, although these mutations are frequently used to render a protein kinase dead. Since these complementation assays rely on transgenes, which are likely overexpressed, it remains possible that endogenous levels of a KSR-1 (kinase-dead) protein are signaling deficient. Nevertheless, our data clearly indicate that a kinase-independent function of KSR-1 can promote vulval induction. Our data are consistent with previous reports that a kinase-independent function of KSR can enhance Ras-induced germinal vesicle breakdown in Xenopus oocytes (22, 31) but are inconsistent with a report that KSR acts as a ceramide-activated protein kinase (40). One likely kinase-independent function of KSR is suggested by our finding that murine KSR forms a multimolecular complex in vivo.
The KSR complex displays an apparent molecular mass of roughly 106 Da, as determined by size-exclusion chromatography. Immuneprecipitation of HA-KSR revealed numerous cellular proteins that specifically coprecipitate with KSR. We consistently observed the following proteins in the KSR complex affinity purified from HEK 293T cells: HSP90, HSP70, HSP68, p50CDC37, p60, MEK1, MEK2, p36, p34, p33, and 14-3-3. Data derived from experiments using deletion and point mutants of KSR suggest that different regions of KSR may mediate interactions with distinct signaling proteins. For example, the N-terminal domain of KSR (amino acids 1 to 301) interacts with p36, p34, and p33. Similarly, the C-terminal kinase domain of KSR interacts with a unique and distinct set of proteins, notably, HSP90, HSP70, HSP68, and p50CDC37. It is not clear whether all KSR-associated proteins directly interact with KSR, nor do we know that each of the associated proteins exists in the same complex in vivo. However, our data strongly suggest that MEK directly binds KSR, as this interaction is observed in yeast and in vitro binding assays. Furthermore, KSR C809Y, which corresponds to a loss-of-function allele in C. elegans, cannot interact with MEK yet can still associate with other KSR-associated molecules. Similarly, KSR is likely to directly interact with 14-3-3. 14-3-3 proteins have been shown to interact with numerous cellular proteins, including Raf, KSR’s closest relative.
Previous reports have demonstrated interactions between KSR and other proteins, including Raf, MEK, ERK, and 14-3-3 (6, 14, 22, 37, 39). These studies were based on yeast two-hybrid and/or coimmunoprecipitation experiments and have revealed little information regarding the relative strength or functional significance of these interactions. Similarly, it has not been demonstrated whether KSR is capable of simultaneous interactions with different signaling molecules. In this report, we have identified several novel KSR-associated proteins (HSP90, HSP70, HSP68, p50CDC37, p36, p34, and p32) in addition to those previously reported. Furthermore, we demonstrated that KSR is capable of forming a multimolecular complex, although binding affinities among known KSR-associated proteins appear to vary considerably. Specifically, the three kinases of the MAP kinase cascade, Raf, MEK, and ERK, display markedly different abilities to associate with KSR in HEK 293T cells. We consistently observe that MEK stoichiometrically associates with KSR, where as a small fraction of ERK it is detected in KSR immunocomplexes. The fraction of Raf that associates with KSR appears to be smaller still. The data presented here significantly advance the concept of KSR as a scaffolding protein in the Ras-MAP kinase pathway.
It is noteworthy that HSP90, HSP70, HSP68, and p50CDC37 are components of the KSR complex. HSP90 has been observed to interact with numerous proteins, including steroid hormone receptors and protein kinases (24). Protein-protein interactions with HSP90 appear to be important for maintaining both protein stability and biological function. In addition, p50CDC37 has been proposed to be a protein kinase-targeting subunit of HSP90 (27). Consistent with this, inhibition of HSP90 with geldanamycin abolishes association between KSR and p50CDC37. In addition, geldanamycin destabilizes KSR in vivo, suggesting a role for these proteins in maintaining levels of KSR. HSPs are also known to facilitate protein folding, and it remains possible that KSR exists in both folded and unfolded forms in our experimental setting. Therefore, it may be difficult to determine whether KSR-HSP90 interactions are biologically significant. Our data do support the hypothesis that KSR-MEK interactions are physiological in nature, as KSR and MEK appear to associate at endogenous protein levels in PC12 cells (Fig. (Fig.2G2G and reference 39). It is noteworthy that KSR protein levels are induced during PC12 cell differentiation. The Ras-MAP kinase pathway plays a pivotal role in PC12 cell differentiation, and further experiments seem warranted to determine whether KSR is involved in this system. Based on several lines of evidence, our data support the model in which HSP90-KSR interactions are also physiologically relevant. First, almost equal amounts of HSPs, MEK, and 14-3-3 proteins are found associated with KSR. Second, distinct regions of KSR mediate protein-protein interactions with HSPs. Third, low-level expression of KSR yields a similar set of coprecipitated molecules. Furthermore, we show here that HSP90 plays a positive role in maintaining KSR protein levels. Finally, mutations in both HSP90 and p50CDC37 have been identified by using genetic screens very similar to those in which KSR was isolated, suggesting that these molecules may function in the same pathway (4, 34).
Scaffolding proteins of other MAP kinase pathways have been shown to maintain specificity and to enhance signaling efficiency (3, 19, 36a). The best example is the Ste5 protein, which tethers components of the Fus3 MAP kinase pathway in the mating pheromone response in budding yeast. Ste5p is essential for mating and maintains specificity of the Fus3p MAP kinase cascade (3, 19). Recently, a scaffolding protein of the mammalian stress-activated MAP kinase cascade has been reported (36a). Although the molecular structure of the KSR complex requires further biochemical characterization, we favor the model in which KSR can simultaneously and directly interact with multiple cellular proteins to form a large signaling complex. Some of the components, such as MEK, HSPs, and 14-3-3, associate with KSR with high stoichiometry. Other components, such as ERK, may transiently (or in a regulated manner) associate with KSR with low stoichiometry. KSR has been reported to weakly interact with Raf in a Ras-dependent manner (22, 32, 37), and we detect Raf-KSR association only upon overexpression of both proteins (data not shown).
We propose that KSR functions in vivo to recruit numerous molecules to form a large signaling complex. Consistent with this model, several loss-of-function mutants exhibited reduced MEK association in vivo. Furthermore, MEK binding is a conserved function of both murine and C. elegans KSRs, though it is not known whether the two are functionally interchangeable. A plausible explanation is that KSR’s interaction with MEK is important for its physiological functions. KSR appears to be a core component of this complex and may function as a scaffold protein to maintain specificity and to increase or restrict the signaling kinase cascade. Other components of the MAP kinase cascade may transiently interact with the complex, depending on intracellular signaling conditions. For example, Raf might transiently interact with the complex and activate the associated MEK. Similarly, ERK may temporally associate with the KSR complex, where it can undergo activation by the tightly bound MEK. Finally, our data suggest that the ability of KSR to form a signaling complex may be more important than its putative kinase activity for its physiological function. The precise role of the KSR signaling complex should be clarified by the identification of the remaining KSR-associated proteins and by the identification of loci that interact genetically with ksr-1 in C. elegans and Drosophila.
We thank Anne Vojtek for reagents and advice on two-hybrid screening; Gerald Rubin and Melanie Cobb for mKSR and MEK plasmids, respectively; Kim Orth and Haris W. Vikis for advice and discussion; and Tianquin Zhu for technical assistance.
This work was supported by the Cancer Biology Training Program, NIH (grant 5T32 CA09676 to S.S.), a Life Sciences Research Foundation/Boehringer Mannheim postdoctoral fellowship (M.S.), American Cancer Society (M.H.), and Public Health Service grant GM51586 and a MacArthur Foundation fellowship (K.-L.G.).