14-3-3 proteins interact with specific MEK kinases.

MEK (mitogen-activated protein kinase/extracellular signal-regulated kinase kinase) kinases (MEKKs) regulate c-Jun N-terminal kinase and extracellular response kinase pathways. The 14-3-3zeta and 14-3-3epsilon isoforms were isolated in a two-hybrid screen for proteins interacting with the N-terminal regulatory domain of MEKK3. 14-3-3 proteins bound both the N-terminal regulatory and C-terminal kinase domains of MEKK3. The binding affinity of 14-3-3 for the MEKK3 N terminus was 90 nM, demonstrating a high affinity interaction. 14-3-3 proteins also interacted with MEKK1 and MEKK2, but not MEKK4. Endogenous 14-3-3 protein and MEKK1 and MEKK2 were similarly distributed in the cell, consistent with their in vitro interactions. MEKK1 and 14-3-3 proteins colocalized using two-color digital confocal immunofluorescence. Binding of 14-3-3 proteins mapped to the N-terminal 393 residues of 196-kDa MEKK1. Unlike MEKK2 and MEKK3, the C-terminal kinase domain of MEKK1 demonstrated little or no ability to interact with 14-3-3 proteins. MEKK1, but not MEKK2, -3 or -4, is a caspase-3 substrate that when cleaved releases the kinase domain from the N-terminal regulatory domain. Functionally, caspase-3 cleavage of MEKK1 releases the kinase domain from the N-terminal 14-3-3-binding region, demonstrating that caspases can selectively alter protein kinase interactions with regulatory proteins. With regard to MEKK1, -2 and -3, 14-3-3 proteins do not appear to directly influence activity, but rather function as "scaffolds" for protein-protein interactions.

We have cloned four MEK (mitogen-activated protein kinase/ extracellular signal-regulated kinase kinase) kinases (MEKKs), 1 referred to as MEKK1, -2, -3, and -4 (1)(2)(3). Each is capable of activating the JNK pathway, and MEKK1, -2, and -3 are also capable of regulating the ERK pathway (1)(2)(3)(4)(5). MEKK1 and MEKK4 bind Cdc42 and Rac, whereas MEKK2 and MEKK3 do not; the subcellular localization of the MEKKs is different (6). MEKK1 is capable of inducing apoptosis (7,8) and is a substrate for proteolytic cleavage by caspase-3 (9). 2 Caspase-3 cleavage of MEKK1 releases the 91-kDa kinase domain from the N-terminal region, which encodes a proline-rich sequence and a predicted pleckstrin homology domain (for review, see Ref. 11). The cleavage of MEKK1 enhances its ability to induce apoptosis, and mutation of the MEKK1 cleavage site suppresses loss of adherence-induced apoptosis (9). MEKK2, -3, and -4 are not caspase substrates, and they are not very effective at inducing apoptosis. 3 14-3-3 proteins were first isolated as highly abundant acidic proteins in brain extracts, and at least seven highly conserved 14-3-3 isoforms have been identified (for review, see Ref. 13). 14-3-3 proteins associate with a number of different signaling proteins and have been proposed to be important in controlling mitogenic signaling pathways (Refs. 14 -19; for review, see Ref. 20). Relevant to the regulation of signal transduction pathways, 14-3-3 proteins have been shown to interact with Raf-1, Bcr-Abl, polyoma middle tumor antigen, KSR (kinase suppressor of Ras), the Bcl family member BAD, the platelet adhesion receptor, glycoprotein Ib-IX, insulin receptor substrate-1, and protein-tyrosine phosphatase H1 (14 -17, 19, 21-24). The biological consequence of binding of 14-3-3 proteins is controversial, but the importance of 14-3-3 proteins in controlling signal transduction pathways is beginning to emerge. For Raf-1, 14-3-3 binding has been reported to enhance, suppress, or play no role in regulating kinase activity (Refs. 14, 17-19, and 25; for review, see Ref. 26). Furthermore, 14-3-3 binding has been shown to protect phosphoserine residues from phosphatases and function as a scaffold to promote association with other proteins (27)(28)(29). Expression of 14-3-3 proteins in Xenopus oocytes activates the ERK pathway, and 14-3-3 proteins can potentiate ERK activation by KSR (22). In Drosophila, D14-3-3 is necessary for photoreceptor development and genetically maps upstream of Raf and downstream of Ras (30,31). In the fission yeast Schizosaccharomyces pombe, 14-3-3 proteins regulate a DNA-damage checkpoint (32), and in the budding yeast Saccharomyces cerevisiae, 14-3-3 proteins are essential for regulation of the Ras/mitogen-activated protein kinase pathway during pseudohyphal development (33). Structurally, 14-3-3 proteins dimerize with each subunit having a cleft that could function as a binding site for proteins and could induce the close proximity of different proteins to alter their interaction * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  and regulation (34). Thus, 14-3-3 proteins are essential for regulated signal transduction and may not directly activate or inhibit kinases, but rather behave as a "scaffold" or "anchor" to localize protein kinase activity.
In this report, we demonstrate that 14-3-3 proteins interact with MEKK1, -2, and -3, but not MEKK4. These findings implicate 14-3-3 proteins in the control of MEKK proteins as "scaffold-like" proteins since association does not directly influence kinase activity. In addition, we demonstrate that caspase-3 cleavage of MEKK1 releases the active kinase domain from the N-terminal 14-3-3-binding region, defining a new function for caspases, namely the release of active kinase from its cellular scaffold.

EXPERIMENTAL PROCEDURES
Yeast Two-hybrid Analysis-Yeast two-hybrid interaction analysis was performed as essentially described by Vojtek and Hollenberg (35). Plasmids pVP16, pBTM116, and pBTM116-lamin and the L40 yeast strain were kindly provided by Drs. Anne B. Vojtek and Jonathan A. Cooper. The N-terminal coding region of MEKK3 (amino acids 4 -361) was subcloned into pBTM116 (pBTM116-MEKK3(NH 2 )), and sequence was confirmed using an Applied Biosystems Model 377 automated DNA sequencer. pBTM116-MEKK3(NH 2 ) was transformed into the L40 yeast strain, propagated under the appropriate selection, and expression of the fusion protein (LexA-MEKK3) was determined in protein extract preparations (36) by immunoblotting with a LexA monoclonal antibody (kindly provided by Drs. John A. Printen and George F. Sprague, Jr.). The L40 yeast strain containing pBTM116-MEKK3(NH 2 ) was transformed with a murine day 9.5 and 10.5 embryonic library in pVP16 (kindly provided by Dr. Stanley M. Hollenberg). Histidine prototrophy was determined on plates containing either 1 or 5 mM 3-aminotriazole to screen for proteins with lower and higher affinities for MEKK3(NH 2 ). ␤-Galactosidase activity was utilized as a secondary screen. Clones that tested positive for both interaction screens were sequenced and identified using the BLAST algorithm to search the nucleotide data base at the National Library of Medicine (37).
Cell Culture and Transfection-Cells were maintained in a humidified CO 2 environment in Dulbecco's modified Eagle's medium supplemented with 100 units/ml penicillin and 100 g/ml streptomycin (Life Technologies, Inc.). PC12 cells were cultured in medium containing 10% fetal bovine serum and 5% horse serum. COS cell growth medium contained 5% newborn calf serum and 5% calf serum. HEK293 cells were cultured in medium containing 10% fetal bovine serum. Where indicated, cells were transfected with LipofectAMINE (Life Technologies, Inc.).
Affinity Precipitation-Full-length 14-3-3⑀ cDNA was obtained by polymerase chain reaction from a mouse brain cDNA library (CLON-TECH, Palo Alto, CA) using primers encompassing the 5Ј-start site and the 3Ј-stop site. Nucleotide sequence was confirmed by comparison with the reported sequence for 14-3-3⑀ (GenBank™ accession number Z19599). 14-3-3⑀ was subcloned into pGEX-5X-1 (Pharmacia, Uppsala, Sweden), expressed in Escherichia coli strain JM109, and purified with glutathione-Sepharose beads according to the manufacturer's instructions. Either untransfected PC12 cells or HEK293 cells transfected with plasmids encoding the indicated MEKK family member were lysed in buffer containing 0.5% Triton X-100, 10 mM Tris, 5 mM EDTA, 50 mM sodium fluoride, 50 mM sodium chloride, and 20 g/ml aprotinin and incubated with 10 g of immobilized GST or GST-14-3-3⑀ at 4°C for 3 h. Following washing of each reaction, the samples were run on an SDSpolyacrylamide gel, transferred to nitrocellulose, and Western-blotted with the indicated antibody.
Caspase Cleavage Assay- 35 S-Labeled 196-kDa MEKK1 was generated using a coupled reticulocyte lysate expression system (Promega, Madison, WI). Equal amounts of 196-kDa MEKK1 were incubated for 1 h in cleavage buffer (50 mM Tris (pH 7.4), 1 mM EDTA, and 10 mM EGTA) with 1 g of bacterially expressed and purified GST or GST-14-3-3 eluted from the glutathione-Sepharose beads. Jurkat cells were used as a source of caspase activity when 6 g of lysate derived from cells that were either unstimulated or stimulated with Fas was added, and the reaction was incubated for 2 h at 37°C. Samples were separated by SDS-PAGE, and bands were visualized by autoradiography.
Kinase Assays-For analysis of JNK activity, cells were transfected with the indicated plasmids along with a hemagglutinin (HA)-tagged form of JNK. After 48 h, the cells were lysed, and JNK activity was assayed following immunoprecipitation of HA-JNK with an anti-HA antibody (Berkeley Antibody Co., Richmond, CA) using c-Jun as a substrate as described by Fanger et al. (6). To assay MEKK activity, cells transfected with either empty plasmid or HA-tagged forms of 196-kDa MEKK1 or MEKK2 were lysed, and MEKK1 or MEKK2 was immunoprecipitated with an anti-HA antibody. Immunoprecipitates were incubated with 1 g of either GST or GST-14-3-3 for 1 h at 4°C before [ 32 P]ATP and a catalytically inactive form of JNK kinase (K 3 R) were added and incubated at 30°C for 30 min. Samples were sizefractionated by SDS-PAGE, and JNK kinase phosphorylation was detected by autoradiography.
Surface Plasmon Resonance Analysis-The Biacore 2000 was utilized to analyze the association of recombinant MEKK3 and 14-3-3 protein expressed in and purified from bacteria (for review, see Ref. 38). GST or GST-14-3-3⑀ was immobilized on a CM5 sensor chip via an anti-GST monoclonal antibody (Biacore, Inc., Piscataway, NJ). For the immobilized protein, concentration and flow rate were assayed to yield consistent 500 -1000 response units, which were utilized for each experiment. Recombinant MBP-MEKK3(NH 2 ) binding was assayed at a flow rate of 20 l/min at concentrations ranging from 10 to 1000 nM. Data were collected on a NEC Powermate V100 using the Biacore control software program. The relative affinity of MEKK3(NH 2 ) for 14-3-3 was determined using the BIAevaluation 2.0 software program by determining the K on and K off rates for this interaction by nonlinear curve-fitting methods.
Confocal Immunofluorescence-Cells were fixed with phosphatebuffered saline (pH 7.4) containing 3% paraformaldehyde and 3% sucrose and then permeabilized with 0.2% Triton X-100. The cells were incubated with rabbit polyclonal antibodies raised against peptides corresponding to MEKK1, MEKK2, 14-3-3␤ (Santa Cruz Biotechnology, Santa Cruz, CA), and 14-3-3⑀ (Santa Cruz Biotechnology) for 1 h, washed, incubated with 1.5 mg/ml Cy 3 -conjugated affinity-purified donkey anti-rabbit IgG (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA), and mounted onto slides with 20 mg/ml o-phenylenediamine in 1 M Tris (pH 8.5). To determine colocalization with 14-3-3 proteins, endogenous MEKK1 was localized with a monoclonal antibody raised against the C terminus (Santa Cruz Biotechnology). The colocalization index was calculated using deconvolution and segmentation analysis with slidebook imaging software (Intelligent Imaging Innovations Inc., Denver, CO) by determining the number of components that stained positive for either MEKK1 or 14-3-3 and calculating the percentage that stained for both molecules. Cells were visualized by digital confocal immunofluorescence, and images were captured with a cooled CCD camera mounted on a Leica DMR/XA microscope using a 63ϫ Plan Neo objective.

RESULTS
Yeast two-hybrid interaction analysis (39) was used to identify proteins that interact with MEKKs. In one set of screens, the N-terminal regulatory domain of MEKK3 (amino acids 4 -361) was fused in frame to the DNA-binding domain of LexA (pBTM116) and used to screen a mouse embryonic library subcloned into pVP16. Histidine prototrophy in the presence of 1 or 5 mM 3-aminotriazole and ␤-galactosidase activity were used for selection of proteins interacting with the N terminus of MEKK3. 1.7 ϫ 10 7 transformants were screened. Of the 100 clones selected for sequence analysis, one 14-3-3 clone and six 14-3-3⑀ clones were identified by comparison with the nucleotide data base at the National Library of Medicine using the BLAST algorithm. Four of the 14-3-3⑀ clones and the 14-3-3 clone were obtained from plates containing 5 mM 3-aminotriazole, suggesting a high affinity interaction between these proteins and the N terminus of MEKK3. Fig. 1 (A and B) shows that 14-3-3⑀ interacted with both the N-and C-terminal moieties as well as with full-length MEKK3. Identification of binding sites in both the MEKK3 N-terminal regulatory and C-terminal kinase domains was similar to that for Raf-1, which has 14-3-3 interaction sequences in the regulatory and catalytic domains (14,19).
The putative N-terminal regulatory domain of MEKK3 can be readily expressed in bacteria and purified as a fusion protein with MBP to produce MBP-MEKK3(NH 2 ). Surface plasmon resonance analysis demonstrated a K D of 90 nM for the interaction of MBP-MEKK3(NH 2 ) with immobilized GST-14-3-3⑀ (Fig. 2). Unfortunately, we have not succeeded in preparing functional full-length or C-terminal kinase domain of MEKK3 for similar analysis, but the high affinity interaction of 14-3-3 protein with the N terminus of MEKK3 suggests that this is an interaction relevant to the regulation of MEKK3 by 14-3-3 proteins in cells.
To determine if 14-3-3 proteins associated with other MEKK proteins, full-length MEKK1, -2, -3, and -4 with the HA tag at their N termini were expressed in HEK293 cells. Immunoblotting with the HA tag antibody (12CA5) was used to normalize for expression of each MEKK (Fig. 3). MEKK2 and MEKK3 migrate as ϳ80-kDa proteins, and MEKK4 is ϳ180 kDa. Fulllength MEKK1 is 196 kDa, but, when expressed in HEK293 cells, is partially proteolyzed to generate two large N-terminal fragments of ϳ134 kDa (fragment A) and 113 kDa (fragment B). We have demonstrated that the 113-kDa fragment B is the result of cleavage of MEKK1 by caspase-3 at Asp-874, which also generates a 91-kDa C-terminal activated kinase domain. 2 MEKK2, -3, and -4 are not caspase substrates. 3 Incubation of the HEK293 lysates expressing MEKK1, -2, -3, or -4 with GST-14-3-3⑀-Sepharose beads demonstrated that MEKK1, -2, and -3, but not MEKK4, associated with 14-3-3⑀ (Fig. 3). Fulllength MEKK2 and MEKK3 similarly bound to 14-3-3⑀. Both full-length MEKK1 and the N-terminal cleavage fragment of MEKK1 bound to 14-3-3⑀. In several experiments, no binding of MEKK4 to 14-3-3⑀ protein could be detected, although functional MEKK4 protein was expressed as determined by its stimulation of JNK activity (data not shown). The failure of MEKK4 to bind 14-3-3 proteins demonstrates a difference in the regulation of MEKK1, -2, and -3 relative to MEKK4. In this regard, transfection-mediated expression of MEKK1, -2, and -3 results in constitutively activated MEKK proteins that can strongly activate JNK. In contrast, overexpression of fulllength MEKK4 has a modest ability to activate the JNK pathway, whereas the MEKK4 catalytic domain strongly activates JNK (3). Cumulatively, these results indicate that the regulatory properties of MEKK4 are different from those of the other MEKKs.
One role of 14-3-3 proteins is to interact with phosphorylated amino acids, resulting in protection from dephosphorylation by phosphatases and potential alterations in overall protein activity (27,28,40). Although the phosphorylation sites of fulllength MEKK1, -2, and -3 have not been identified, phosphorylation sites have been mapped to the C-terminal catalytic domain of MEKK1 (41). These observations prompted us to focus on the kinase domains of MEKK1, -2, and -3 as potential sites for interaction with 14-3-3 proteins. The kinase domains of MEKK1, -2, and -3 show dramatically different abilities to bind 14-3-3⑀ (Fig. 4). Although there were equivalent amounts of protein expression in cell lysates, the kinase domain of MEKK2 consistently displayed a decreased ability to associate with 14-3-3⑀ relative to MEKK3. The kinase domain of MEKK1, which is serine/threonine-phosphorylated at multiple sites, was unable to bind to 14-3-3⑀. With respect to 14-3-3 interactions, these results indicate that the kinase domains of MEKK1, -2, and -3 are inherently different, even though their amino acid sequence is conserved. Thus, interaction of 14-3-3 proteins with the kinase domains of MEKK2 and MEKK3, and not the kinase domain of MEKK1, provides a mechanism to differentially regulate MEKK activity within the cell. , or lamin as a negative control in combination with either an empty plasmid (pVP16) expressing the activation domain alone or pVP16 expressing full-length 14-3-3⑀. Following selection on leucine/tryptophan-negative plates, five separate colonies were isolated and assayed for ␤-galactosidase activity. Error bars represent S.D. B, HEK293 cell lysates recombinantly expressing the kinase domain of MEKK3 (MEKK3c) were incubated with Sepharose beads conjugated to either GST alone as a negative control or GST-14-3-3⑀. Precipitates were analyzed by Western blot analysis with an anti-hemagglutinin antibody. C, wild-type PC12 cells were stimulated with epidermal growth factor (EGF) for 10 min prior to being lysed. Lysates were incubated with either GST-14-3-3⑀ or GST alone immobilized on Sepharose beads. Western blot analysis was performed with affinity-purified rabbit polyclonal antibodies raised against MEKK3(NH 2 ) as described under "Experimental Procedures." In B and C, representative examples of three separate experiments are shown for each panel. Immunoreactive proteins were visualized by enhanced chemiluminescence and autoradiography.
The inability of the kinase domain of MEKK1 to bind 14-3-3 proteins prompted us to determine if the 91-kDa C-terminal kinase domain (Fig. 5A), created by caspase-3 cleavage, was able to bind 14-3-3. Fig. 5B shows that, even though they were expressed at similar levels, the 196-kDa full-length MEKK1 protein associated with 14-3-3⑀ unlike the 91-kDa C-terminal MEKK1 protein, which displayed little associative properties for 14-3-3⑀. This result is consistent with the inability of the kinase domain (see Fig. 4) and the first 300 amino acids (data not shown) of the 91-kDa kinase domain of MEKK1 to bind 14-3-3⑀. Fig. 5C shows that multiple deletion products encoding different regions of the N terminus of 196-kDa MEKK1 bind to 14-3-3⑀. Minimally, the N-terminal 393 amino acids of MEKK1 are sufficient to bind 14-3-3⑀. This region is upstream of the predicted pleckstrin homology domain encoded in the MEKK1 sequence. Therefore, although 196-kDa MEKK1 binds to 14-3-3 proteins via a domain that is located in the first 393 amino acids, the 91-kDa C-terminal kinase domain does not associate with 14-3-3 proteins. Since 14-3-3 proteins associate with MEKK1 and caspase cleavage of MEKK1 is an important component of the apoptotic

FIG. 2. Analysis of 14-3-3 protein binding to MEKK3 by surface plasmon resonance.
The Biacore 2000 was used to analyze the interaction of 14-3-3 with MEKK3. Either GST or GST fused to 14-3-3⑀ was immobilized on a CM5 sensor chip, and surface plasmon resonance was utilized to measure the interaction with the N-terminal regulatory domain of MEKK3 (amino acids 4 -361). The response units, which are a relative indication of protein-protein interaction, were plotted as a function of time in seconds.

FIG. 3. Selective interaction of MEKK family members with 14-3-3 protein.
Lysates from HEK293 cells that were transiently transfected with HA-tagged full-length forms of MEKK1, -2, -3, and -4 were incubated with Sepharose-immobilized GST or GST fused to 14-3-3⑀. Lysate from cells transfected with empty vector (pCMV5) was included as a negative control. A nonspecific protein expressed in HEK293 cells, migrating at ϳ80 kDa (first through fifth lanes), was detected in cell lysates by the 12CA5 antibody. This nonspecific protein partially obscured the detection of MEKK2 and MEKK3 in cell lysates (third and fourth lanes). However, this protein was not precipitated by GST-14-3-3⑀ (seventh, ninth, eleventh, thirteenth, and fifteenth lanes). Western blot analysis was performed with an anti-HA antibody to determine relative binding of 14-3-3⑀ to each MEKK family member. Molecular mass markers (in kilodaltons) are indicated on the left. response (9), we next determined whether association of 14-3-3 protein would influence caspase cleavage in vivo and in vitro.
[ 35 S]Methionine-labeled MEKK1 was recombinantly expressed in rabbit reticulocyte lysate and combined with either GST as a negative control or GST-14-3-3. Following binding of 14-3-3 protein to MEKK1, caspase-mediated cleavage was induced by adding Jurkat lysates that were either unstimulated or stimulated with Fas. Although addition of the Fas-stimulated lysate, but not the unstimulated lysate, proteolytically cleaved the 196-kDa form of MEKK1 to a 113-kDa fragment (fragment B) and a 91-kDa fragment, no difference in cleavage was observed when 14-3-3 protein was bound to MEKK1 (Fig. 6A). Association [ 35 S]methionine-labeled MEKK1 and GST-14-3-3 was confirmed by precipitating a portion of the reaction with glutathione-Sepharose beads and analyzing radiolabeled MEKK1 by SDS-PAGE (data not shown). As another approach to determine if 14-3-3 association influenced caspase cleavage of MEKK1, HEK293 cells were cotransfected with plasmids expressing MEKK1 and 14-3-3 protein. Since overexpression of MEKK1 induces caspase activation, MEKK1 is proteolyzed in the absence of any other apoptosis-inducing stimuli (9). Following cell lysis, Western blot analysis with an anti-C-terminal MEKK1 antibody was used to determine the level of caspasemediated cleavage of MEKK1 in either the presence or absence of recombinantly expressed 14-3-3 protein. Although significant caspase-mediated cleavage of 196-kDa MEKK could be detected, there was no discernible difference in production of the 91-kDa MEKK1 fragment (Fig. 6B) when 14-3-3 was associated. Coimmunoprecipitation confirmed the association of recombinantly expressed MEKK1 and 14-3-3 protein (data not shown). Thus, as demonstrated by two different assays to measure proteolytic cleavage, 14-3-3 association did not influence caspase-mediated cleavage of MEKK1.
Previous reports indicate that in addition to 14-3-3 binding, which can alter enzymatic activity in vitro, 14-3-3 protein overexpression can activate Raf and ERK kinase activities (17,22,24,29,42). Therefore, we assessed whether 14-3-3 overexpression would influence MEKK activity in vivo by assaying JNK and ERK activation and determined whether in vitro association of 14-3-3 protein with MEKK1, -2, and -3 influenced kinase activity. As illustrated in Fig. 7A, overexpression of 14-3-3 alone was neither sufficient to stimulate JNK activation nor capable of influencing JNK activation that was induced by overexpression of MEKK1, -2, or -3. Furthermore, 14-3-3 protein overexpression did not alter ERK activation by MEKK1, -2, or -3, and it did not elicit ERK activation (data not shown), as has been reported in Xenopus oocytes (22). As determined by in vitro kinase assays, association of 14-3-3 protein with MEKK1 or MEKK2 did not alter their ability to phosphorylate a kinase-inactive form of JNK kinase (Fig. 7C), even though association was confirmed by Western blot analysis (data not shown). Similarly, when the activity of MEKK1 purified from rabbit reticulocyte lysate preparations was assayed in either the presence or absence of bound 14-3-3 protein, there was no difference in the ability of MEKK1 to phosphorylate JNK kinase (data not shown). These results provide strong evidence that 14-3-3 association does not stimulate or inhibit MEKK activity and are consistent with other reports indicating that 14-3-3 protein association with Raf-1 does not influence kinase activity (14,43).
Digital confocal immunofluorescence was used to localize 14-3-3 proteins in cells relative to the distribution of MEKK1 and MEKK2 (Fig. 8). MEKK1 and MEKK2 have been previously characterized for subcellular distribution (6). MEKK1 has a nuclear and punctate cytoplasmic distribution, whereas MEKK2 has a punctate cytoplasmic and Golgi localization. Indirect immunofluorescence with both antibodies for 14-3-3␤ (which recognizes 14-3-3␤ as well as other isoforms) and 14-3-3⑀ (specific for this isoform) indicated that 14-3-3 proteins have a Golgi and punctate cytoplasmic distribution and that at least 14-3-3⑀ was, in addition, localized in the nucleus. Thus, staining of 14-3-3 proteins shows a similar subcellular distribution of MEKK1 and MEKK2, consistent with an ability of 14-3-3 proteins to interact with MEKK proteins in cells. We do not have antibodies that selectively stain MEKK3 using indi- FIG. 6. In vivo and in vitro analyses of the effects of 14-3-3 association on caspase-mediated cleavage of 196-kDa MEKK1. A, 35 S-labeled 196-kDa MEKK1 was prepared using rabbit reticulocyte lysate and incubated with either GST or GST-14-3-3⑀ or without any recombinant protein (NONE). Lysates isolated from Jurkat cells either unstimulated (ϪFas) or stimulated with Fas (ϩFas) were incubated with the binding reactions prior to separation by SDS-PAGE analysis. Fragments were visualized by autoradiography. B, HEK293 cells were transfected with plasmids expressing the indicated genes. Following incubation for 48 h, cells were lysed, and to visualize proteolytic generation of the 91-kDa form of MEKK1, samples were separated by SDS-PAGE and subjected to Western blot analysis.

FIG. 7. Association of 14-3-3 does not influence MEKK activity.
A, HEK293 cells were transfected with either empty vector or vector expressing 14-3-3⑀ protein. Where indicated, plasmid expressing either MEKK1, -2, or -3 was included in the transfection, and MEKK activity was evaluated in intact cells by immunoprecipitating an HA-tagged form of JNK and measuring its ability to phosphorylate c-Jun in an in vitro kinase assay. B, recombinantly expressed 14-3-3⑀ was analyzed by Western blotting. C, cells were transfected with either empty plasmid or plasmid expressing either MEKK1 or MEKK2. Following immunoprecipitation with an anti-HA antibody, either GST alone or GST-14-3-3 was allowed to associate with the precipitation reaction, and MEKK activity was measured in an in vitro kinase assay in which a kinaseinactive form of JNK kinase (JNKK) was utilized as a substrate. rect immunofluorescence, but a functional green fluorescent protein-MEKK3 fusion protein selectively stains the Golgi region of the cell similar to 14-3-3 (data not shown). To conclusively demonstrate that the similar distributions of 14-3-3 proteins and MEKK were indeed due to colocalization of the two molecules, we utilized two-color immunofluorescence and simultaneously stained for both 14-3-3 and MEKK1. As illustrated in Fig. 9, a significant proportion of endogenous MEKK1 (red) and 14-3-3 protein (green) distinctly colocalized (yellow), particularly in the punctate structures located in the cytoplasm. Quantitative analysis indicated that the overlap index of the punctate structures located in the cytoplasm was 29.1 Ϯ 4.3%, a value determined by scoring the number of components labeled as green (14-3-3) and as red (MEKK1) compared with those components that were colocalized (yellow). In comparison, quantitative analysis of staining for the p85 subunit of phosphatidylinositol 3-kinase, a protein that does not biochemically interact with MEKK1, resulted in an average colocalization index of 3.8 Ϯ 0.8%. Thus, endogenous MEKK and 14-3-3 proteins interact in vivo. The combined two-hybrid interaction and biochemical and subcellular colocalization analysis indicates that 14-3-3 binding to MEKK1, -2, and -3 is a biologically relevant high affinity interaction. DISCUSSION In this report, we characterize the discovery that 14-3-3 proteins associate with specific members of the MEKK family of kinases. Our results indicate that MEKK1, -2, and -3 must be added to the growing list of signal transduction proteins that bind 14-3-3 proteins. Unlike MEKK1, -2, and -3, MEKK4 was unable to associate with 14-3-3 proteins and thus highlights the specificity inherent to 14-3-3 interactions. Furthermore, these results provide additional evidence that MEKK4 is regulated by mechanisms distinct from those of the other MEKK family members, a discovery that was only assumed by the dramatically different sequence of the N-terminal regulatory domains (for review, see Ref. 11). Additional binding specificity is apparent, as the kinase domain of MEKK1 was unable to bind to 14-3-3 proteins, unlike MEKK2 and MEKK3. Thus, although a large number of proteins have been described to interact with 14-3-3 proteins, results presented herein confirm that 14-3-3 proteins display a significant level of binding specificity.
Yeast two-hybrid analysis was used to initially discover the interaction of 14-3-3 with the N terminus of MEKK3. Approximately 7% of the clones that were capable of stimulating ␤-galactosidase activity and mediating histidine prototrophy represented 14-3-3 genes and were the largest population of interacting clones. By comparison, no 14-3-3 isoforms were isolated by two-hybrid analysis when the same library was screened with the C-terminal kinase domain of MEKK1, even though a similar number of positively interacting clones were sequenced. Thus, by two-hybrid analysis, a distinct level of binding specificity was apparent between these two different MEKK regions, highlighting the unique ability of the twohybrid system to discern between binding partners.
It has been difficult to define the biological role of 14-3-3 protein association (for review, see Ref. 20). One of the most extensively studied kinases with regard to 14-3-3 protein bind- FIG. 8. MEKK family members and 14-3-3 proteins display similar intracellular localization. Digital confocal immunofluorescence was utilized to determine the subcellular localization of 14-3-3 and MEKK proteins in COS-7 cells. The 14-3-3␤ antibody (A) recognizes multiple 14-3-3 isoforms, whereas the 14-3-3⑀ antibody (C) is specific for this isoform. As previously published (6), the anti-MEKK2 (B) and anti-MEKK1 (D) antibodies, both raised using peptides corresponding to divergent C-terminal regions of each protein, are specific for these isoforms and do not recognize other MEKK family members. Images were deconvolved to remove out-of-focus immunofluorescence.
FIG. 9. Subcellular colocalization of endogenously expressed 14-3-3 and MEKK. Shown are the results of digital confocal image analysis of MEKK1 and 14-3-3 protein in COS cells costained with antibodies directed against MEKK1 (red) and 14-3-3⑀ protein (green). Overlap of both MEKK1 and 14-3-3⑀ is indicated by yellow. A, the image was obtained at a magnification of ϫ40 and depicts a single cell. Bar ϭ 5 M. B, the image is of a portion of the cell and was obtained at a magnification of ϫ100. Bar ϭ 2 M. Images were deconvolved to remove out-of-focus immunofluorescence.
ing is Raf. 14-3-3 association has been shown to activate, reduce the activity of, or play no role in regulating Raf activity (Refs. 14, 17-19, and 25; for review, see Ref. 26). For other proteins such as protein-tyrosine phosphatase H1 and nitrate reductase, 14-3-3 binding decreases enzymatic activity, whereas for tryptophan hydroxylase, association increases activity (24,29,42). There is strong evidence to support the role of 14-3-3 binding to prevent dephosphorylation by phosphatases, as phosphorylated serine is in many instances important for 14-3-3 protein association (27)(28)(29). Since phosphorylation often can control the activation/inactivation state of an enzyme or kinase, the role of 14-3-3 in preventing dephosphorylation likely explains these results.
In addition to preventing dephosphorylation, structural and functional analysis demonstrates that 14-3-3 proteins dimerize, indicating that they also play a structural/organizational role in signal transduction pathways (34). All of the 14-3-3 clones that were isolated from our screen consisted of partial 14-3-3 amino acid sequences lacking the N-terminal dimerization domain, demonstrating the MEKKs bind to the C terminus of 14-3-3. This finding is consistent with the notion that proteins bind to the C terminus of 14-3-3 proteins and that dimerization of 14-3-3 proteins occurs via their N-terminal portion (34). This is highly analogous to the role of STE5, which also requires an intact dimerization domain for pheromone response pathway signaling (44). We also demonstrate that 14-3-3 binding does not directly influence MEKK activity, and thus, we conclude that rather than directly controlling the activation state of the MEKKs, 14-3-3 proteins likely play an organizational role that is important for assembly of macromolecular signaling complexes. Utilization of 14-3-3 proteins as tethering molecules via serine phosphorylation is a mechanism similar to that of the SH2 domain-containing proteins, which utilize phosphorylated tyrosines to bring proteins into proximity of each other (for review, see Ref. 45).
MEKK proteins function as mitogen-activated protein kinase kinase kinases (MKKKs) at a similar position in sequential protein kinase pathways as Raf. Thus, 14-3-3 proteins bind multiple but not all MKKKs, as MEKK4 displays no binding capacity. In S. cerevisiae, 14-3-3 homologs (BMH1 and BMH2) associate with STE20, which is genetically upstream of STE11 (33), the cognate MKKK in the mitogen-activated protein kinase pathway for mating and pseudohyphal development. STE20 can be considered as a potential MKKK kinase. There is no reported evidence for STE11 binding to 14-3-3 proteins, yet in mammalian cells, 14-3-3 proteins can bind other kinases that may function similar to STE20. For example, KSR positively mediates Ras signaling and also binds 14-3-3 proteins; KSR also binds Raf, but not in a 14-3-3-dependent manner (22). Thus, a possible structural theme begins to emerge where 14-3-3 proteins would bind an MKKK kinase (STE20, KSR) or an MKKK (Raf, MEKK). The dimeric structure of 14-3-3 proteins would then allow them to bind another protein that could be a scaffold-like protein or upstream activator. This is consistent with the hypothesis that 14-3-3 proteins could allow selective organization and subcellular localization of signal complexes (for review, see Ref. 10).
Based on these predictions, the findings in this report with MEKK1 are most intriguing. Unlike MEKK2 and MEKK3, the kinase domain of MEKK1 shows little ability to associate with 14-3-3 proteins. Full-length MEKK1 binds to 14-3-3 proteins via sequences in the N-terminal regulatory domain. MEKK1 is also a defined caspase-3 substrate; the consequence of caspase cleavage of MEKK1 is that the 91-kDa kinase domain is released from the N-terminal regulatory domain (9). 2 The 91-kDa MEKK1 kinase domain strongly induces apoptosis, whereas a caspase cleavage-resistant MEKK1 mutant expressed in cells suppresses apoptotic responses to external stimuli (9). Our findings indicate that caspase-3 cleavage of MEKK1 results in the kinase domain being released and no longer tethered to 14-3-3 proteins, potentially allowing for it to interact with and regulate proteins. Interestingly, MEKK2 and MEKK3 induce little or no apoptosis, 3 and unlike MEKK1, their catalytic domains strongly associate with 14-3-3 proteins. Thus, we propose that one function for caspase-3 cleavage of MEKK1 is the release of the MEKK1 kinase domain from 14-3-3 association so that it may interact with new substrates and participate in the apoptotic process. In contrast, when full-length MEKK1 is associated with 14-3-3 proteins, it functions positively to regulate cell processes in response to growth factors and cytokines (6,12). The association with 14-3-3 proteins, the activity of caspase-3, and cleavage of MEKK1 will therefore regulate the function of MEKK1 in cells.