Apoptosis Stimulated by the 91-kDa Caspase Cleavage MEKK1 Fragment Requires Translocation to Soluble Cellular Compartments*

MEKK1, a 196-kDa mitogen-activated protein kinase (MAPK) kinase kinase, generates anti-apoptotic signaling as a full-length protein but induces apoptosis when cleaved by caspases. Here, we show that caspase-dependent cleavage of MEKK1 relocalizes the protease-generated 91-kDa kinase fragment from a particulate fraction to a soluble cytoplasmic fraction. Relocalization of MEKK1 catalytic activity is necessary for the pro-apoptotic function of MEKK1. The addition of a membrane-targeting signal to the 91-kDa fragment inhibits caspase activation and the induction of apoptosis but does not change the activation of JNK, ERK, NFκB, or p300. These results identify the caspase cleavage of MEKK1 as a dynamic regulatory mechanism that alters the subcellular distribution of MEKK1, changing its function to pro-apoptotic signaling, which does not depend on the currently described MEKK1 effectors.

Programmed cell death or apoptosis is a genetically coded (1) cellular mechanism by which cells activate pathways that promote suicide. It is defined by morphological characteristics, including cytoplasmic shrinkage, nuclear condensation, and DNA fragmentation (2). Apoptosis is vital at many stages of development in higher organisms and remains important for homeostasis throughout their lifetime. It has recently become clear that signal transduction pathways influence and control apoptosis. Signaling pathways controlling apoptosis may be implicated in the aging process (3)(4)(5) and several diseases, including cancer and neurodegenerative diseases (6,7).
A crucial biochemical event required for most apoptotic responses is the activation of proteases of the caspase subfamily (8 -11). Finding the activators and effectors of caspases is therefore essential to an understanding of the mechanisms by which caspases, and ultimately apoptosis in general, are controlled. Only a subset of signaling proteins is cleaved by caspases during apoptosis (12,13). One of these proteins is the MAPK 1 kinase kinase MEKK1, which regulates the ERK and the JNK MAPK pathways, as well as the transcription factor NFB and the p300 transcriptional co-activator (14,15).
Previous evidence indicates that MEKK1 positively regulates apoptosis. Expression of the kinase domain of MEKK1 into cells induces apoptosis in a manner that depends on a functional kinase activity (16,17). MEKK1 is necessary for apoptosis caused by detachment from the extracellular matrix (anoikis) in Madin-Darby canine kidney cells (18) or in response to UV-C irradiation and several chemotherapeutic drugs (19). In these situations, MEKK1 is cleaved by caspases into a 91-kDa kinase-containing fragment that further stimulates the activation of caspases and, consequently, apoptosis. The kinase domain of MEKK1 may also favor apoptosis by inducing an increased expression of Fas and Fas ligand (20,21).
On the other hand, genetic evidence indicates that fulllength MEKK1 induces anti-apoptotic signals. Targeted disruption of both MEKK1 alleles causes a selective loss of ERK and JNK activation in response to serum factors and specific stress insults (22). Functionally, full-length MEKK1 Ϫ/Ϫ embryonic stem cells are extremely fragile and have enhanced sensitivity to apoptotic agents. It is important to note that the putative protective function of full-length MEKK1 is not observed when the protein is overexpressed in cells, because this leads to its cleavage into the 91-kDa pro-apoptotic fragment (19,23).
Collectively these studies are consistent with the hypothesis that MEKK1 functions as a molecular switch to regulate apoptosis and that the conversion from anti-to pro-apoptotic signaling is mediated by a caspase-mediated cleavage of MEKK1, resulting in the release of the 91-kDa MEKK1 carboxyl-terminal domain from the regulatory amino-terminal domain (19,24). The nature of the pro-apoptotic signals generated by the carboxyl-terminal domain of MEKK1 is unknown.
Here, we determined that the switch that converts MEKK1 into a pro-apoptotic protein is controlled by the subcellular localization of its catalytic domain. The endogenous 196-kDa MEKK1 is associated with insoluble structures, but the caspase cleavage product behaves as a soluble cytoplasmic protein. A change in location of the active 91-kDa kinase fragment of MEKK1 is required for the activation of caspases and induction of apoptosis. We also show that this kinase fragment, whether membrane-tethered or not, similarly activates p300, NFB, or JNK, indicating that these pathways are not involved or not sufficient to mediate the apoptotic response induced by cytoplasmic MEKK1.

EXPERIMENTAL PROCEDURES
Cell Lines-Human embryonic kidney 293 cells (HEK293) were maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% bovine calf serum and 100 units/ml penicillin/streptomycin. Swiss 3T3 cells were maintained in DMEM supplemented with 5% bovine calf serum, 5% fetal calf serum, and 100 units/ml penicillin/streptomycin. T47D human breast carcinoma cells were maintained in DMEM supplemented with 10% fetal calf serum and 100 units/ml penicillin/streptomycin. HeLa cells were maintained in RPMI 1640 containing 10% newborn calf serum (Invitrogen) at 37°C and 5% CO 2 . Cells were transfected using LipofectAMINE (Invitrogen) as described (25). Briefly, 2 ϫ 10 6 cells, plated the previous day in 10-cm Petri dishes, were incubated for 5 h with a DNA (6 g)-LipofectAMINE (10 l) mixture in 5 ml of RPMI at 37°C in 7.5% CO 2 . The total amount of DNA was kept constant using empty vectors when required.
Plasmids-The mammalian expression vector pcDNA3 is from Invitrogen. The plasmid pEGFP-C1 encoding the red-shifted variant of wild-type GFP (excitation maximum at 488 nm; emission maximum at 507 nm) is from CLONTECH. Construction of amino-terminally hemagglutinin (HA) tagged full-length (196 kDa) and 91-kDa mouse MEKK1 cDNAs in pcDNA3 has been described previously (25). Construction of the kinase-inactive full-length mouse MEKK1 cDNA in pcDNA3 has been previously described (19). Construction of cDNAs of the kinase domain of mouse MEKK1 (⌬MEKK1) and its kinase-inactive counterpart in mammalian expression vectors has been described previously (16).
The carboxyl termini of full-length and 91-kDa mouse MEKK1 cDNAs in pcDNA3 were modified by the addition of the Ha-Ras prenylation and palmitoylation signal (the COOH-terminal 20 amino acids of Ras) by first using the PCR to modify this sequence of Ha-Ras by the addition of a COOH-terminal BstXI restriction site and an NH 2 -terminal sequence identical to that of the COOH-terminal sequence of the full-length and 91-kDa MEKK1 constructs in pcDNA3. The PCR product was then used in a second PCR reaction with a primer complementary to the sequence of MEKK1 from base 3638 to base 3657. The resulting PCR product was digested with NheI and BstXI then ligated into the MEKK1 cDNA that had been similarly digested. This resulted in in-frame addition of the COOH-terminal 60 bases of Ha-Ras to MEKK1 (full-length and 91 kDa), as confirmed by DNA sequencing.
Plasmid pEGFP-F encodes the 20-amino acid farnesylation signal from c-Ha-Ras fused to the COOH terminus of EGFP. This farnesylation signal directs EGFP-F to membranes (particularly the inner face of the plasma membrane) (26,27). Plasmid GFP-LAMP encodes a fusion protein between GFP and lysosome-associated membrane protein 1 (LAMP1, also called LGP120) that localizes to late endosomes and lysosomes (28). Plasmids GFP-C2 and GFP-C3 encode GFP fusion proteins that specifically label the ER/nuclear membranes and the mitochondria, respectively. 2 Plasmid Gal4-p300 is a pRc/Rous sarcoma virus-derived plasmid encoding a fusion protein between the DNA binding domain of Gal4 and human p300-(2-2414) with two FLAGencoding sequences at the carboxyl terminus. Plasmid Gal4-luc contains four Gal4 binding sequences adjacent to a minimal thymidine kinase promoter that control expression of the luciferase cDNA (16). prLUC contains two NFB binding sites upstream of the luciferase cDNA. Plasmid MEK1-DN.dn3 encodes a kinase-dead (Lys-97 3 Met) form of mouse MEK1 that functions as a dominant-negative mutant for ERK activation when overexpressed in cells (29).
Cell Fractionation-Cells were placed on ice, and the growth medium was aspirated and saved. Cells were washed 2ϫ with cold PBS, and the washes were added to the growth medium. This solution was centri-fuged at 150 ϫ g for 5 min to collect any detached cells. Cells were lysed in hypotonic buffer (10 mM Tris-HCl, pH 7.5, 10 mM NaCl, 1 mM EDTA, 0.5 mM EGTA, 1 g/ml aprotinin, 5 g/ml leupeptin, 1 mM Na 3 VO 4 ) for 15 min on ice, then scraped off the dishes and homogenized in a Dounce homogenizer until intact cells could not be detected by trypan blue exclusion. Lysates were then centrifuged at 1000 ϫ g for 5 min to separate nuclei. The supernatant was then subjected to centrifugation at 200,000 ϫ g for 2 h to pellet insoluble material. The pellet was resuspended in EB buffer (10 mM Tris, pH 7.4, 5 mM EDTA, 50 mM NaCl, 50 mM NaF, 1% Triton X-100, 1 mM phenylmethylsulfonyl fluoride, 1 mM Na 3 VO 4 , 20 g/ml aprotinin). The supernatant was concentrated by vacuum centrifugation.
Western Blot Analysis-Cells were fractionated as above or lysed in EB buffer. Lysate containing equivalent amounts of total protein was resolved by SDS-PAGE (8% acrylamide), transferred to nitrocellulose in 10 mM CAPS, pH 11, 20% methanol using a transphor apparatus (Hoeffer, San Diego, CA) for 1 h at 1 A. Blots were incubated in Tris-HCl, pH 7.5, buffered saline with 5% powdered milk then incubated with a primary antibody in the same 5% milk solution. The primary antibodies used in this study were: rabbit antisera (95012) raised against a peptide encoding the last 13 amino acids of MEKK1 (19), C-22 anti-MEKK1 antibody (Santa Cruz Biotechnology), anti-MEKK1 COOH-terminal (CT) antibody (Upstate Biotechnology, Inc.), anti-HA antibody (12CA5, Berkeley Antibody Co., anti-poly-(ADP-ribose) polymerase (PARP) antibody (Upstate Biotechnology Inc.), or anti-MKP1 antibody (V-15, Santa Cruz Biotechnology). Blots were then incubated with secondary antibody coupled with horseradish peroxidase (HRP) or with protein A-HRP (Zymed Laboratories Inc.). The blots were then developed with Western blot detection reagents (enhanced chemiluminescence, Amersham Biosciences, Inc., or Renaissance, PerkinElmer Life Sciences).
In Vitro Kinase Assays-ERK and JNK assays were performed as previously described (19). The in vitro ERK assay results represent the mean of triplicate determinations ϮS.E. The in vitro JNK assay represents one experiment. Phosphorimaging analysis of triplicate determinations varied by less than 5%.
Immunofluorescence and Apoptosis Assays-Cells were grown on glass coverslips and transfected with LipofectAMINE. One (for localization) or two (for apoptosis) days post-transfection cells were fixed in 3% paraformaldehyde and 3% sucrose in PBS for 10 min at room temperature. After three washes with PBS, cells were permeabilized with 0.2% Triton X-100 in PBS for 5 min at room temperature. After 3 PBS washes, cells were incubated with DMEM containing 10% serum and 0.02% sodium azide for 20 min at room temperature. For apoptosis assays, cells were incubated for 1 h in terminal deoxytransferase reaction mix (200 mM potassium cacodylate, 25 mM Tris-HCl, pH 6.6, 50 g/ml bovine serum albumin, 5 mM CoCl 2 , 0.25 units of terminal deoxytransferase (Roche Molecular Biochemicals)/l, 10 M Biotin-dUTP (Roche Molecular Biochemicals)) at 37°C in a humidified atmosphere. After three PBS washes or without the terminal deoxytransferase incubation for the localization studies, the coverslips were incubated for 1 h at room temperature with primary antibody diluted in DMEM, 10% newborn calf serum (C-22 antibody was used at a 1/100 dilution, MEKK1 COOH-terminal antibody was used at a 1/100 dilution, anti-HA antibody was used at 1/100 dilution). The coverslips were washed 6 times over 30 min in PBS then incubated with secondary antibody coupled to a fluorescent marker diluted in DMEM, 10% newborn calf serum (donkey anti-rabbit indocarbocyanine (Cy3)-conjugated antibody (Jackson Immunological) was diluted 1/1000, donkey anti-mouse fluorescein (fluorescein isothiocyanate)-conjugated antibody (Jackson Immunological) was diluted 1/100, 5 g/ml Cy3-conjugated streptavidin (for apoptosis assays)), and 40 g/ml DAPI. The coverslips were washed 3 times in PBS and soaked for 16 -20 h in PBS before being mounted in 20 mg/ml o-phenyldiamine dihydrochloride (Sigma) in 0.1 M Tris, pH 8.5, 90% glycerol. Images were captured with DMRXA microscope and analyzed with SlideBook version 2.0 software (Intelligent Imaging Innovations, Denver, CO). In Figs. 6C, 8, and 9, apoptosis was measured by scoring the number of transfected cells stained with Hoechst 33342 that displayed a pyknotic nucleus (30). At least 200 cells per condition were counted.
NFB and p300 Activity Assays-Cells were transfected with the different MEKK1 constructs in the presence of 0.5 g of the NFB reporter plasmid prLUC (for the measure of NFB activity) or in the presence of 1 g of Gal4-p300 and 1 g of Gal4-luc (for measurement of p300 activity). The amount of DNA per transfection was kept constant by the addition of an empty vector (pcDNA3) if required. Two days after transfection, the cells were lysed, and the luciferase levels were measured with a Promega kit (E1500) using 50 -100 g of lysates. The basal 2 C. Borner, unpublished results. A, T47D human breast carcinoma cells were serum-starved overnight and challenged for 10 min with 100 ng/ml EGF or left resting. Cells were fixed in 3% paraformaldehyde and stained with rabbit antisera recognizing the COOH terminus of MEKK1 (95-012 anti-serum). Fluorescently labeled secondary antibody was used for detection of MEKK1, and DNA was stained with DAPI. The fluorescent image was overlaid with a Nomarski image of the same cell. B, HeLa cells were transfected with a plasmid encoding the full-length MEKK1 protein together with various plasmids encoding fusion proteins between GFP and subcellular markers (GFP-F for the plasma membrane (first row), GFP-C2 for nuclear membranes and the endoplasmic reticulum (second row), GFP-C3 for mitochondria (third row), and GFP-LAMP for late endosomes and lysosomes (fourth row); see "Experimental Procedures" for details). The cells were then fixed and stained with a carboxylterminal MEKK1-specific rabbit antibody (C22 from Santa-Cruz) labeled with a Cy3-conjugated secondary anti-rabbit antibody. Alternatively, HeLa cells transfected with a plasmid encoding the fulllength MEKK1 protein were fixed, stained for MEKK1 using the C22 antibody, and incubated with either a mouse monoclonal antibody specific for tubulin (fifth row) or with phycoerythrin-labeled phalloidin that binds to actin fibers (sixth row). The appropriate combination of fluorescein isothiocyanate-or Cy3-conjugated secondary antibodies was used for the fluorescent detection of the indicated proteins. Long thin closed arrows indicate membrane ruffles, thick closed arrows depict actin fibers associated with focal adhesions, thick open arrows show actin stress fibers, and thin arrowheads indicate MEKK1 immunoreactivity on cortical actin fibers.
(except for the 196 kDa CAAX, which represents two experiments), analyzed for the S.E. (error bars).

RESULTS
As a first step to investigate how full-length MEKK1 is converted into a pro-apoptotic 91-kDa fragment, we determined whether these two MEKK1 forms could be differentially localized in cells. Digital confocal immunofluorescence using antibodies directed against MEKK1 showed that endogenous MEKK1 in T47D and the ectopically expressed protein in HeLa cells are associated with punctate cytoplasmic structures predominantly in perinuclear regions and more diffusely at the periphery of the cell (Fig. 1, A and B; see also Refs. 31 and 32). As a means to better visualize peripheral membrane structures, T47D cells were stimulated with EGF, which led to the formation of membrane ruffles (right panel of Fig. 1A). In this condition, a fraction of MEKK1 immunoreactivity was localized at the periphery of the cells in the membrane ruffles (Fig. 1A). Subcellular localization of MEKK1 was further analyzed in HeLa cells transfected with a MEKK1-encoding plasmid and with various plasmids encoding GFP-tagged subcellular markers. As was the case for T47D cells, when membrane ruffles could be detected, MEKK1 was found to be associated with these structures (long thin closed arrows; Fig. 1B, fifth and sixth rows). In the absence of ruffles, MEKK1 was not detected at the plasma membrane (Fig. 1B, first row). There was no co-localization of MEKK1 with late endosomes/early lysosomes, mitochondria, the endoplasmic reticulum, or nuclear membranes (Fig. 1B, second through fourth rows). There was no strict association of MEKK1 with microtubules ( Fig. 1B, fifth row). MEKK1 did not colocalize with actin stress fibers (thick open arrows; Fig. 1B, sixth row) or with the actin fibers associated with focal adhesion (thick closed arrows; Fig. 1B, sixth row). However, a few spots of MEKK1 immunoreactivity could be detected on cortical actin fibers (arrowheads; Fig. 1B, sixth row). In some cells (e.g. Madin-Darby canine kidney cells), MEKK1 has been shown to associate with actin fibers, but this only concerned a fraction of the total MEKK1 immunoreactivity (32). Therefore, it appears that the majority of MEKK1 proteins in cells is not associated with actin fibers. These results together with previously published studies (31,32) indicate that MEKK1 does not appear to be localized exclusively within a given subcellular compartment. Among the different intracellular membrane structures analyzed, only membrane ruffles seem to always colocalize with MEKK1. This colocalization, however, only concerns a fraction of the total MEKK1 cellular pool.
These observations indicate that at least a fraction of MEKK1 in cells can associate with membranes and cytoskeleton structures. To further analyze this possibility, lysates from HEK293 and Swiss 3T3 cells, which were either resting or stimulated for 10 min with EGF, were fractionated by differential centrifugation. The totality of the endogenous 196-kDa MEKK1 protein resided in the particulate fraction in both resting and growth factor-stimulated cells (Fig. 2). This sug-gests that the 196-kDa MEKK1 is constitutively associated with either membranes or the cytoskeleton or both. Detergent treatment using Triton X-100 or Nonidet P-40 relocalized MEKK1 to the soluble fraction (not shown).
The kinase domain of MEKK1 is located in the COOHterminal ϳ37 kDa of MEKK1 (33,34). The remainder of the protein encodes regulatory functions including 14-3-3 binding (35), a putative pleckstrin homology domain, and Ras (36) and Cdc42/Rac binding domains (31). Caspase-3 cleavage of MEKK1 during apoptotic signaling generates a 91-kDa COOHterminal kinase-active fragment that no longer includes the 14-3-3 binding and pleckstrin homology domains. The localization of the 91-kDa MEKK1 caspase-generated cleavage fragment therefore may be different from the 196-kDa full-length MEKK1 protein. To test whether the 91-kDa cleavage fragment generated after apoptotic stress has a different cellular distribution than the 196-kDa MEKK1, HEK293 cells were left untreated or exposed to 100 J/m 2 UV-C irradiation, then incubated for varying periods of time (Fig. 3). The cells were lysed, HEK293 and Swiss 3T3 cells were serum-starved overnight and then challenged with EGF as described in Fig. 1, panel A. Cells were lysed in hypotonic medium using a Dounce homogenizer and fractionated by centrifugation. Particulate and soluble fractions were immunoblotted with the MEKK1-specific antibody used in Fig. 1, panel A.   FIG. 3. The 91-kDa caspase cleavage product of MEKK1 behaves as a soluble cytoplasmic protein. Log phase cultures of HEK293 cells were exposed to 100 J/m 2 UV-C and then incubated at 37°C for the indicated times. Alternatively, cells were serum-starved overnight and challenged with 100 ng/ml EGF for 10 min. Cells were lysed using a Dounce homogenizer, and the lysates were fractionated, resolved by SDS-PAGE, and immunoblotted with the anti-MEKK1 CT antibody. Two different antisera to the MEKK1 COOH terminus (95-012 and C12) gave similar results in independent experiments.
FIG. 4. MEKK1-CAAX proteins are membrane-associated. The wild-type (WT) and CAAX-modified MEKK1 proteins were transiently expressed in T47D cells. Twenty-four h post-transfection the cells were fixed and stained with anti-HA-tag antibody followed by a secondary fluorescently labeled antibody and DAPI to label DNA. Cells were analyzed by immunofluorescence microscopy (A-D). Alternatively (E and F), transfected cells were lysed using a Dounce homogenizer, and lysates were fractionated by centrifugation. Soluble and particulate fractions were analyzed by immunoblotting using anti-HA tag antibody or anti-MEKK1 CT antibody. The results are representative of four independent experiments. and particulate and soluble fractions resolved by differential centrifugation and analyzed by SDS-PAGE. Immunoblots using antisera directed against the COOH terminus of MEKK1 indicated that the presence of the 91-kDa fragment increased with time after exposure of cells to UV-C. The 91-kDa MEKK1 fragment was in the soluble fraction, as determined by subcellular fractionation (Fig. 3). This contrasts with the localization of the full-length 196-kDa MEKK1 protein found in the particulate fraction (Fig. 2). These data suggest that caspase cleavage releases the 91-kDa MEKK1 fragment to a soluble cytoplasmic location in the cell.
To determine whether the cytoplasmic redistribution of MEKK1 is necessary for the induction of apoptosis, we fused the COOH terminus of the murine MEKK1 cDNA in both full-length and 91-kDa-fragment forms with the COOH-terminal 20 amino acids of Ha-Ras. The sequence from Ha-Ras encodes the prenylation and palmitoylation sites and therefore targets proteins to membranes (37)(38)(39). The membrane-targeted MEKK1 constructs are referred to as CAAX because this sequence of Ras contains a CAAX box. Our hypothesis predicted that the tethering of the MEKK1 91-kDa kinase domain via its COOH terminus to membranes would inhibit its ability to induce apoptotic signaling by keeping the active fragment membrane-bound, not allowing relocalization to a soluble cytoplasmic fraction that could result in a change in the substrates available to MEKK1 for phosphorylation. Two techniques were used to assess whether the CAAX constructs differed from the wild-type constructs in their subcellular localization when transiently expressed in cells. First, T47D breast carcinoma cells were transfected with 196-kDa MEKK1 (196 kDa), 91-kDa fragment of MEKK1 (91 kDa), 196 kDa MEKK1-CAAX (196-kDa CAAX), or 91-kDa MEKK1-CAAX (91-kDa CAAX) and stained with a monoclonal mouse antisera (12CA5, Berkeley Antibody Co.) directed against the NH 2 -terminal HA tag encoded in each construct. A clear difference between the location of the wild-type and CAAX proteins was observed. The 196-kDa and 91-kDa MEKK1 proteins had a diffuse cytoplasmic distribution (Fig. 4, A and C). In contrast, most of the 196-kDa and 91-kDa CAAX-MEKK1 proteins were detected in peripheral membrane structures (Fig. 4, B and C).
The localization of the different MEKK1 proteins was analyzed by fractionation of transfected cells by differential centrifugation. After fractionation, proteins were resolved by SDS-PAGE, transferred to nitrocellulose, and immunoblotted with anti-HA or anti-MEKK1 antibodies. The transiently expressed full-length MEKK1 proteins were distributed between the soluble and particulate fractions (Fig. 4E), unlike the endogenous 196-kDa MEKK1 protein that is only particulate (Fig. 2). The increase in 196-kDa MEKK1 in the soluble fraction is most likely due to its overexpression and a fraction of the protein being loosely associated with insoluble structures. Both 196-kDa MEKK1 constructs appear as doublets, which is characteristic of a gel shift that represents phosphorylation and activation of the slower migrating MEKK1 protein (19). A much more dramatic difference was seen between the 91-and the 91-kDa CAAX proteins. The 91-kDa MEKK1 protein was exclusively in the soluble fraction, whereas 90 -95% of the 91-kDa CAAX protein appeared in the particulate fraction (Fig. 4F). The data show that the addition of the membrane-targeting prenylation signal of Ras to full-length MEKK1 changes the localization of the transfection-expressed kinase to a predominantly peripheral membrane component, and the addition of this signal to the 91-kDa fragment completely alters its location in the cell, from a soluble fraction to a particulate fraction. The immunofluorescence results indicate that the 196-kDa CAAX and 91-kDa CAAX constructs are primarily localized at the cell periphery.
Having altered the subcellular distribution of transiently expressed 196-kDa and 91-kDa MEKK1 by membrane targeting, we tested whether the ability to induce apoptosis by expression of the different MEKK1 constructs was similar or altered when membrane association was stabilized. Cells were transfected with one of the MEKK1 constructs, fixed with paraformaldehyde, incubated with terminal deoxytransferase in the presence of biotin labeled dUTP, then incubated with streptavidin bound to a fluorescent marker and viewed by fluorescence microscopy. The cells were also stained with an antibody directed against the HA tag of the transfected protein and DNA stained using DAPI (Fig. 5). Apoptotic cells were then counted by the intensity of their terminal deoxytransferase signal and confirmed by nuclear morphological changes associated with apoptosis as seen by staining the nuclei with DAPI. Cells transfected with the 196-kDa MEKK1 underwent apoptosis ϳ3 times more often than cells expressing 196-kDa CAAX or a 196-kDa MEKK1 construct rendered kinase-inactive with a lysine to methionine mutation in the ATP binding site (34). The results of four independent experiments showed no difference between the apoptosis induced by the 196-kDa CAAX or the 196-kDa K-M proteins (Fig. 5B). Cells expressing the 91-kDa MEKK1 proteins were used in a similar experiment with the addition of the kinase domain of MEKK1 (⌬MEKK1) and its kinase-inactive counterpart (K-M) as positive and negative controls. The 91-kDa MEKK1 protein induced ϳ5- fold more apoptosis than the membrane-tethered 91-kDa CAAX MEKK1 protein, which was similar to ⌬MEKK1 K-M in inducing apoptosis (Fig. 5B).
The addition of a membrane-targeting signal to either the 196-or 91-kDa MEKK1 proteins eliminated induction of apoptosis after their overexpression. Therefore, we investigated whether different MEKK1 signal functions were altered among the various constructs. MEKK1 has been shown to activate caspases as part of its pro-apoptotic signaling, and inhibition of these caspases inhibits the apoptotic response mediated by MEKK1 (18,19). Therefore, the effect of overexpression of these various constructs on caspase activity was investigated. HEK293 cells were transfected with each of the MEKK1 constructs, and 24 h later, caspase activity was measured in cell lysates. Fig. 6A shows that the prenylated versions of the 196and 91-kDa MEKK1 proteins had a reduced ability to stimulate caspase activity compared with the non-prenylated constructs.
Nuclear PARP has been shown to be a specific target for caspase cleavage in apoptotic responses, and its disappearance has been shown to directly correlate with caspase activity (40 -42). Therefore, as a second measure of caspase activation after overexpression of the different MEKK1 constructs, cells were lysed 24 h after transfection, and the lysates were resolved by SDS-PAGE, transferred to nitrocellulose, and immunoblotted with antiserum directed against PARP. The results show that levels of the 126-kDa full-length PARP protein are drastically reduced after expression of the 91-kDa MEKK1 protein (Fig.  6B). Compared with control cells, ϳ40% of the PARP protein was lost after expression of 196-kDa MEKK1, whereas the levels of PARP did not diminish when cells were transfected with plasmids encoding the 91-kDa CAAX or the 196-kDa CAAX proteins (Fig. 6B). It should be noted that the 85-kDa fragment of PARP, which is often seen after activation of caspases, was difficult to detect after transfection and lysis, perhaps because it is quickly degraded. However, disappearance of the 126-kDa full-length PARP protein was seen consistently after transfection with either full-length MEKK1 or the 91-kDa fragment of MEKK1. Probing the same blot with antisera against the MAPK phosphatase, MKP-1, a protein not cleaved by caspases (12), indicated that similar amounts of protein were loaded on the gel. Therefore, the differences in apoptotic signaling between the wild-type and membrane-targeted MEKK1 proteins correlate with their ability to induce caspase activity.
To assess which of the two main upstream caspases (caspase 8 or 9) could be involved in the apoptotic response induced by FIG. 6. Caspase activation resulting from the expression of different MEKK1 constructs. HEK293 cells were transfected with the indicated MEKK1 constructs. Twenty four h after transfection cells were lysed and analyzed for caspase activity (A) or immunoblotted using antisera recognizing PARP or MKP-1 (B). DEVD-amidomethylcoumarin was used as a substrate for caspase activity. The fluorescence of cells transfected with an empty vector was subtracted in each case. Values are the mean Ϯ S.E. of triplicate determinations. The asterisks indicate statistically significant differences (p Ͻ 0.05; t test). In panel C, HEK293 cells transfected with a plasmid encoding the 91-kDa fragment of MEKK1 were incubated with the indicated concentrations of caspase 8-and 9-specific inhibitors (benzyloxycarbonyl-IETD-aminofluoromethylcoumarin and benzyloxycarbonyl-LEHD-aminofluoromethylcoumarin, respectively (Enzyme Systems Products, FK-012 and FK-022)). Apoptosis was scored 48 h later (the mean of duplicate determinations Ϯ S.E.; representative of three independent experiments). WT, wild type.

FIG. 7. Regulation of MAPK pathways in HEK293 cells expressing different MEKK1 constructs.
Cells were transfected with the indicated MEKK1 constructs and twenty-four h later assayed for JNK and ERK activity (panels A and B, respectively). Expression of each construct was shown to be similar by immunoblotting using the HA-tag antibody. Results are expressed as the fold increase over the basal response obtained in cells transfected with the control empty vector (pcDNA3) only. For the JNK assays, the phosphorimaging analysis of triplicate determinations varied by less than 5% (for clarity, the corresponding error bars were omitted). The ERK experiment shown corresponds to the mean of triplicate determinations Ϯ S.E. and is representative of three independent experiments. WT, wild type. the carboxyl-terminal MEKK1 caspase fragment, HEK293 cells, transfected with plasmids encoding 91-kDa MEKK1, were incubated with 2 doses of caspase 8-and 9-specific inhibitors. Fig. 6C shows that both inhibitors were equally potent in their ability to block 91-kDa MEKK1-induced cell death. It seems, therefore, that both caspases are required for the 91-kDa fragment of MEKK1 to induce apoptosis.
Some of the downstream effectors of MEKK1, including MAPKs (e.g. ERK and JNK) and transcriptional regulators (e.g. p300 and NFB), can affect apoptotic signaling in some systems (15,43). The role of the ERKs is generally protective against apoptosis, whereas the role of specific JNKs in apoptosis is not completely understood and appears to vary depending on the conditions and cell lines used (17, 19, 21, 44 -49). The activation of JNK after expression of the 196-kDa MEKK1, 91-kDa MEKK1, and their complementary membrane-targeted proteins appeared to be similar (Fig. 7A). This indicates that the regulatory NH 2 -terminal domain of MEKK1 is not necessary for activation of the JNK pathway and suggests that JNKs can be activated when MEKK1 kinase activity is either soluble or particulate. ERKs were activated strongly by both the 196-kDa wild-type and 196-kDa CAAX proteins. Neither the 91-kDa wild-type nor the 91-kDa CAAX MEKK1 proteins significantly activated ERKs (Fig. 7B). This confirms previous data showing that the NH 2 -terminal domain of MEKK1 is necessary for significant ERK activation (23).
Mouse genetics has revealed that the full-length MEKK1 protein generates survival signals in cells (22). However, in transfected cells, a fraction of the ectopically expressed MEKK1 is cleaved into a 91-kDa fragment that corresponds to the pro-apoptotic MEKK1 form generated during anoikis or when cells are treated with genotoxic agents (18,19). The generation of this 91-kDa fragment most likely accounts for the cell death response observed in cells transfected with MEKK1encoding plasmids. The full-length MEKK1 protein, in contrast to the 91-kDa fragment, efficiently stimulates the ERK MAPK pathway ( Fig. 7B and Ref. 23), which could explain why in some experiments ectopically expressed full-length MEKK1 generates a weaker apoptotic response than the 91-kDa fragment. To determine whether the ERK MAPK pathway could modulate the ability of MEKK1 to induce apoptosis, cells were transfected with plasmids encoding the full-length MEKK1 or the 91-kDa fragment in the presence or in the absence of a plasmid encoding MEK1 DN, a kinase-dead form of MEK1, that functions as a dominant-negative mutant and blocks the activation of ERK1/2 (29). Fig. 8A shows that MEK1 DN had no effect on the ability of the different MEKK1 constructs to induce apoptosis. Similarly, a constitutively active Raf protein, BxB Raf, known to strongly stimulate the ERK MAPK pathway (50), did not significantly reduce the ability of either MEKK1 constructs to induce apoptosis (Fig. 8B). Therefore, the ERK MAPK pathway does not seem to regulate the ability of MEKK1 to induce apoptosis and cannot account for the differences in the apoptotic responses observed in some experiments between cells expressing either the full-length MEKK1 or the 91-kDa fragment.
NFB can regulate apoptosis both in a positive or negative manner depending on the cellular system and the stimuli used (51,52). For example, NFB is required for p53-induced apoptosis (53), whereas it mediates the anti-apoptotic signals initiated by Akt (54). A recent study suggests that the transcriptional co-activator p300 mediates the ability of the kinase domain of MEKK1 to induce apoptosis (15). To determine whether NFB or p300 could mediate the 91-kDa MEKK1induced apoptosis, cells were transfected with the wild-type and the CAAX forms of 91-kDa MEKK1 together with reporter plasmids for NFB or p300 activity. Fig. 9 shows that both the 91-kDa wild-type and the 91-kDa CAAX MEKK1 proteins activated NFB and p300 to a similar extent, whereas only wildtype 91-kDa MEKK1 induced apoptosis. The expression levels of each construct were found to be similar as measured by Western blot analysis (data not shown). Taken together these results show that the differential ability of wild-type and prenylated MEKK1 constructs to initiate apoptosis is not due to a difference in activation of the JNKs, ERKs, NFB, or p300. DISCUSSION Our results demonstrate that caspase cleavage of MEKK1 releases the 91-kDa kinase-active fragment from a particulate fraction to a soluble subcellular localization. This change in localization allows MEKK1 to activate caspases and, conse- FIG. 9. Membrane localization does not alter the ability of 91-kDa MEKK1 to activate NFB and p300. HEK293 cells were transfected with the indicated amounts of wild-type (WT) or prenylated 91-kDa MEKK1-encoding plasmids together with 1 g of a GFP-encoding plasmid (to visualize the transfected cells) and either 1 g of an NFB reporter plasmid (prLUC) or p300 reporter plasmids (Gal4-luc and Gal4-p300, 1 g each). Forty-eight h later apoptosis and NFB or p300 activity were measured. The luciferase data are the mean of duplicate determinations Ϯ S.E. (when not visible, error bars are within the width of the symbols). These experiments have been repeated at least three times with similar findings. quently, apoptosis. Inhibitors specific either for caspase 8 or for caspase 9 were equally potent in blocking 91-kDa MEKK1induced cell death, indicating that both caspases are required for this apoptotic response. Because caspase 9 can be activated after cleavage of Bid by caspase 8 (55), it is possible that the MEKK1 91-kDa fragment induces the activation of the death receptors (e.g. Fas), which results in caspase 8 activation, which in turn activates caspase 9. There is in fact evidence in Jurkat cells that MEKK1 kinase activity induces cell death in a Fas-dependent manner (20,21). However, mouse embryonic fibroblasts derived from lpr mice that do not express functional Fas receptors, are killed by the 91-kDa fragment of MEKK1 as efficiently as wild-type cells. 3 Therefore, the requirement for Fas activation to mediate the 91-kDa MEKK1-induced apoptosis may be cell type-specific. Besides the possible involvement of caspase 8, it is clear that the mitochondrial pathway leading to caspase 9 activation plays a role in the apoptotic response mediated by the 91-kDa fragment of MEKK1. As mentioned above, caspase 9-specific inhibitors suppressed 91-kDa MEKK1induced cell death. Moreover, Bcl2 overexpression, which inhibits the release of cytochrome c from the mitochondria and the activation of caspase 9 (56), blocks the ability of 91-kDa MEKK1 to induce apoptosis (57). The observation that the kinase domain of the 91-kDa MEKK1 fragment (⌬MEKK1) induces the proapoptotic conformation of Bak, a member of the Bcl2 family of proteins (58), is also consistent with the notion that the mitochondrial death pathway mediates MEKK1-induced cell death.
Activation of the JNK MAPK pathway, NFB, or p300 by MEKK1 does not depend on membrane localization, as the soluble and membrane-localized 91-kDa MEKK1 forms activate JNK, NFB, and p300 to a similar degree. The activation of these signaling pathways is thus not required (or at least not sufficient) to mediate 91-kDa MEKK1-induced apoptosis. Additionally, the ERK MAPK does not seem to be involved because its inhibition does not affect the ability of MEKK1 to induce apoptosis and because neither the wild-type nor the membrane-targeted forms of 91-kDa MEKK1 significantly activate the ERK MAPK pathway. We have also found recently that there is no correlation between the ability of a given MEKK to induce apoptosis and its ability to stimulate NFB, the ERK MAPK pathway, or the JNK MAPK pathway (23). Therefore, the mechanism used by the released 91-kDa MEKK1 to induce apoptosis remains unknown.
It is likely that the NH 2 -terminal domain of MEKK1 is responsible for targeting MEKK1 to insoluble structures since its removal relocalizes MEKK1 to a soluble fraction. The nature of the proteins or structures (membranes or cytoskeleton) that tether full-length MEKK1 to the particulate fraction is, however, unclear. MEKK1 binding to Ras is apparently not sufficient to retain MEKK1 on membranes, because the Ras binding domain on MEKK1 is present within the sequence coding for the soluble 91-kDa MEKK1 carboxyl-terminal fragment (36). A protein that could sequester MEKK1 away from the soluble compartment of the cell is the actin-bundling protein ␣-actinin that binds MEKK1 in its amino-terminal domain (32), but further studies are required to determine whether ␣-actinin is involved in the regulation of the apoptotic activity of MEKK1.
The NH 2 terminus of MEKK1 binds to 14-3-3 proteins (35,59). Because 14-3-3 proteins bind phosphorylated Bad to protect cells from apoptosis, it is possible that 14-3-3 binding to MEKK1 also protects cells from the apoptotic potential of MEKK1. During apoptosis, the cleavage of MEKK1 would release the 91-kDa catalytic domain from sequestration by 14-3-3, unmasking its pro-apoptotic function. It remains to be determined whether 14-3-3 proteins can modulate MEKK1 subcellular location.
Caspases can modify the functions of their substrates in many different ways (13,60,61). The inactivation of proteins such as PARP and DNA protein kinase by caspase cleavage inhibit signals for DNA repair (41). In contrast, caspase cleavage of kinases such as MEKK1, PAK2, and protein kinase C␦ activate and/or alter their regulation including interaction with other proteins (18,62,63). The cleavage of these kinases functions to convert their behavior into pro-apoptotic stimuli. MEKK1 activation and cleavage occurs with the same time course as the activation of caspases (19). MEKK1 therefore is an immediate and early substrate for caspases, consistent with its importance in the cascade of events mediating apoptosis. The requirement for the translocation of the MEKK1 kinase domain away from insoluble structures for its pro-apoptotic function defines a new role for caspases during the apoptotic response.