|Home | About | Journals | Submit | Contact Us | Français|
The mitogen-activated protein kinase (MAPK) module, composed of a MAPK, a MAPK kinase (MAPKK), and a MAPKK kinase (MAPKKK), is a cellular signaling device that is conserved throughout the eukaryotic world. In mammalian cells, various extracellular stresses activate two major subfamilies of MAPKs, namely, the Jun N-terminal kinases and the p38/stress-activated MAPK (SAPK). MTK1 (also called MEKK4) is a stress-responsive MAPKKK that is bound to and activated by the stress-inducible GADD45 family of proteins (GADD45α/β/γ). Here, we dissected the molecular mechanism of MTK1 activation by GADD45 proteins. The MTK1 N terminus bound to its C-terminal segment, thereby inhibiting the C-terminal kinase domain. This N-C interaction was disrupted by the binding of GADD45 to the MTK1 N-terminal GADD45-binding site. GADD45 binding also induced MTK1 dimerization via a dimerization domain containing a coiled-coil motif, which is essential for the trans autophosphorylation of MTK1 at Thr-1493 in the kinase activation loop. An MTK1 alanine substitution mutant at Thr-1493 has a severely reduced activity. Thus, we conclude that GADD45 binding induces MTK1 N-C dissociation, dimerization, and autophosphorylation at Thr-1493, leading to the activation of the kinase catalytic domain. Constitutively active MTK1 mutants induced the same events, but in the absence of GADD45.
Living organisms are frequently exposed to cellular stresses, which are defined as diverse environmental conditions that are detrimental to the normal growth and survival of the cells. Typical cellular stresses include UV, ionizing radiation, genotoxins, hyperosmolarity, oxidative stress, low-oxygen supply (hypoxia), and inhibition of protein synthesis by antibiotics and plant toxins. In coping with the barrage of these and other cellular stresses, multicellular eukaryotic organisms have developed strategies for how damaged cells will respond to stresses. In general, if the intensity of damage is moderate, the affected cell will seek to repair the damage. If, however, the damage to a cell is too severe for a complete repair, the affected cells are eliminated by apoptosis. This reduces the risk to the organism as a whole, such as the development of a cancer. Such crucial decision making between repair and death is, at least in part, mediated by the stress-activated mitogen-activated protein kinase (SAPK) pathways (for general reviews on mitogen-activated protein kinase [MAPK] and SAPK, see references 8, 20, 24, 29, and 30).
As the name implies, the SAPK pathways are homologous to and share many characteristics with the classic (extracellular signal-regulated kinase 1/2 [ERK1/2]) MAPK pathway. Eukaryotic MAPK pathways are conserved signaling modules that transmit signals from the cell surface to the nucleus. The core of any MAPK pathway is composed of three tiers of sequentially activating protein kinases, namely, MAPK kinase kinase (MAPKKK), MAPK kinase (MAPKK), and MAPK. The activation of MAPKs is achieved by the phosphorylation of threonine and tyrosine residues within a conserved Thr-Xaa-Tyr motif in the activation loop (also called the T-loop) catalyzed by MAPKKs. MAPKKs, in turn, are activated by any of several MAPKKKs via the phosphorylation of serine and/or threonine residues within their activation loops.
All eukaryotic cells possess multiple MAPK pathways, each of which is activated by a distinct set of stimuli. In the budding yeast Saccharomyces cerevisiae, for example, hyperosmotic stress activates the Hog1 MAPK pathway, whereas mating pheromones activate the Fus3/Kss1 MAPK pathway (13, 18). In mammalian cells, four different subfamilies of MAPKs are present, namely, ERK1/2, Jun N-terminal protein kinase 1/2/3 (JNK1/2/3), p38α/β/γ/δ, and ERK5. The ERK1/2 MAPK pathway is preferentially activated in response to mitogenic stimuli, such as growth factors and phorbol esters, and plays a role in cell growth and cell survival. The ERK1/2 pathway is regulated mainly by the monomeric GTPase Ras, which recruits MAPKKKs of the Raf family to activate the two downstream MAPKKs: MEK1 and -2. These MAPKKs, in turn, activate the ERK1/2 MAPKs. The JNK and p38 MAPKs (collectively called SAPKs), in contrast, preferentially respond to various cellular stresses and are thus called SAPK pathways. Besides cell stresses, the SAPK pathways are also activated by cytokines such as interleukin-1, tumor necrosis factor alpha, and transforming growth factor β (TGF-β). The JNK subfamily of MAPKs are activated mainly by the MKK4 and MKK7 MAPKKs, while the p38 subfamily MAPKs are activated by the MKK3 and MKK6 MAPKKs. In clear contrast to this limited number of MAPKKs in the SAPK pathways, there are numerous MAPKKKs that function upstream of the JNK and p38 MAPKs. These include MEKK1/2/3, MTK1 (also known as MEKK4), TAK1, ASK1/2, TAO1/2/3, MLKs, and perhaps others. This multiplicity at the level of MAPKKK surely reflects the vastly diverse stress stimuli that can recruit these SAPK pathways.
MTK1 is one of the human MAPKKKs belonging to the SAPK pathways, and the mouse ortholog is called MEKK4 (16, 33). The kinase domain of MTK1 (MEKK4) is homologous to other MAPKKKs and especially similar to mammalian MEKK1/2/3 and ASK1/2 and yeast SSK2/SSK22, but its N-terminal noncatalytic domain (regulatory domain) is unique (33). Analyses using MEKK4-deficient mice have shown that the MEKK4 signaling pathway integrates signals from both the T-cell antigen receptor and interleukin-12/STAT4 in developing Th1 cells and promotes STAT4-independent gamma interferon production (9), and MEKK4 is essential for normal neural and skeletal development (2, 10).
In a yeast two-hybrid screening aimed at identifying MTK1 activators, we found three growth arrest and DNA damage-inducible 45 (GADD45) family proteins to be strong binding partners of MTK1 (34). The GADD45 gene was originally identified as a UV-inducible gene in Chinese hamster cells (14). The human genome encodes three GADD45-like proteins, GADD45α (the original GADD45), GADD45β (MyD118), and GADD45γ (CR6 or OIG37) (25, 44). These will be referred to collectively as the GADD45 proteins. The GADD45 proteins share 55 to 58% sequence identity. The three GADD45 genes are all inducible by cellular stresses, although optimal stimuli for each gene appear to be different (25). The expression profiles of the three GADD45 genes are also distinct in various tissues (34). The GADD45 proteins interact with various intracellular molecules, such as proliferating cell nuclear antigen, Cdc2-cyclinB1 complex, p21Waf1/Cip1, and core histones, and play important roles in stress-adaptive processes, including growth control, maintenance of genomic stability, DNA repair, and apoptosis (3, 17, 25, 38). In other words, the GADD45 proteins are emergency calls in damaged cells.
Expression of transfected GADD45 genes strongly activates coexpressed MTK1 kinase and induces apoptosis in mammalian cells (34). TGF-β-induced GADD45β expression also activates p38 MAPK through MTK1 activation (35). MEKK4-deficient mice have lost GADD45-induced gamma interferon production (9). The activation of the SAPK pathway by MTK1 and GADD45 is temporally a slow process because it requires the induction of GADD45 gene expression prior to activation of MTK1. Thus, the activation of MTK1 by GADD45 differs from other modes of SAPK activation that occur within minutes. GADD45/MTK1-mediated SAPK activation may therefore serve as a more long-term adaptive mechanism for stressed cells.
We previously proposed that the binding of GADD45 to the N-terminal region of MTK1 counters the autoinhibitory effect of the MTK1 N-terminal segment on the kinase domain and, at the same time, enables the MTK1 kinase domain to bind its cognate MAPKKs (MKK3 and MKK6) via the latter's DVD docking sites (28, 36). The details of MTK1 activation by GADD45, however, remained obscure. In this report, we investigated the molecular mechanism by which GADD45 regulates MTK1 kinase activity.
Lysis buffer contains 20 mM Tris-HCl (pH 7.5), 1% Triton X-100, 10% glycerol, 137 mM NaCl, 2 mM EDTA, 50 mM β-glycerophosphate, 10 mM NaF, 1 mM sodium vanadate, 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, 10 μg/ml leupeptin, and 10 μg/ml aprotinin. Sodium dodecyl sulfate (SDS) loading buffer is 65 mM Tris-HCl (pH 6.8), 5% (vol/vol) 2-mercaptoethanol, 2% SDS, 0.1% bromophenol blue, and 10% glycerol. Kinase buffer contains 25 mM Tris-HCl (pH 7.5), 25 mM MgCl2, 0.5 mM sodium vanadate, 25 mM β-glycerophosphate, 2 mM dithiothreitol, and 2 mM EGTA.
The mammalian expression plasmids pFlag-MTK1, pFlag-MTK1-K/R, pcDNA3-GADD45β, pFlag-MTK1(L534Q), pFlag-MTK1(Q637P), pFlag-MTK1(V1300F), pFlag-MTK1(I1360M), and pGST-MKK6(K/A) were described previously (28, 32, 34). GADD45β deletion mutants and MTK1 mutants were generated by PCR mutagenesis. pEGFP-C1 (Clontech), pcDNA4Myc, and pcDNA3 vectors were used to generate green fluorescent protein (GFP)-tagged GADD45β, Myc-tagged MTK1, and GADD45α/GADD45γ expression plasmids, respectively. pcDNA3FKBP vector was used to generate FKBP-fused, hemagglutinin (HA)-tagged MTK1 expression plasmids. MTK1-N contains Met-22 through Ala-994, MTK1-C contains Ser-1247 through Glu-1607 (the C-terminal amino acid [aa]), and MTK1-NΔBD contains Ser-253 through Ala-994.
Glutathione S-transferase (GST)-MKK6(K/A) was expressed in Escherichia coli DH5 and purified using glutathione-Sepharose beads as previously described (33).
COS-7 cells were maintained in Dulbecco's modified Eagle medium supplemented with 10% fetal bovine serum, l-glutamate, penicillin, and streptomycin. For transient transfection assays, cells grown in 60-mm dishes were transfected with the appropriate expression plasmids using the Effectene transfection reagent (QIAGEN). The total amount of plasmid DNA was adjusted to 1 μg per plate with vector DNA (pcDNA3).
Immunoblotting analyses were carried out as described previously (36). The following antibodies were used: anti-Flag monoclonal antibody (MAb) M2 (Sigma), anti-HA MAb 3F10 (Roche), anti-GFP MAb B-2 (Santa Cruz), polyclonal anti-Myc (Santa Cruz), anti-GADD45α MAb 4T-27 (Santa Cruz), and goat polyclonal antisera to GADD45β and GADD45γ (Santa Cruz). An anti-MTK1 antiserum has been described previously (9). An anti-phospho-T1493 rabbit antiserum (αP-T1493; lot no. A2340PE) was made in house.
Two plasmids that encode a Flag-tagged MTK1 N-terminal segment (residues 22 to 994) and a Myc-tagged MTK1 C-terminal segment (residues 1247 to 1607) were constructed (Fig. (Fig.1A).1A). Flag-MTK1-N and Myc-MTK1-C were individually expressed in COS-7 cells, and cell lysates were prepared 24 h after transfection. The lysates were mixed in vitro and incubated for 4 h at 4°C. Flag-MTK1-N was precipitated from the mixture by an anti-Flag antibody conjugated to protein G-Sepharose, and coprecipitated Myc-MTK1-C was detected by immunoblotting using an anti-Myc antibody.
Cell lysates, prepared in lysis buffer with 0.5% deoxycholate, were incubated with an appropriate antibody for 2 h at 4°C to precipitate epitope-tagged MTK1. Endogenous MTK1 was precipitated using anti-MTK1 antibody. Immune complexes were recovered with the aid of protein G-Sepharose beads, washed three times with lysis buffer containing 500 mM NaCl and 0.5% deoxycholate and twice with lysis buffer only, resuspended in SDS loading buffer, and separated by SDS-polyacrylamide gel electrophoresis (PAGE) for immunoblotting analyses using αP-T1493.
Cell lysates were incubated with protein G-Sepharose beads at 4°C for about 20 h. Then, precleared lysates were incubated with anti-Flag MAb M2 bound to protein G-Sepharose beads at 4°C for 4 h with gentle rotation. Immunoprecipitates were collected by centrifugation, washed six times with lysis buffer containing 500 mM NaCl and 0.5% deoxycholate, and subjected to SDS-PAGE.
Transfected cells were lysed in lysis buffer. Cell lysates were incubated with the appropriate antibody for 2 h at 4°C. Immune complexes were recovered with the aid of protein G-Sepharose beads, and washed twice with lysis buffer containing 500 mM NaCl and 0.5% deoxycholate, twice with lysis buffer, and twice again with kinase buffer. Immunoprecipitates were resuspended in 30 μl of kinase buffer containing 4 μg of GST-MKK6(K/A). The kinase reaction was initiated by the addition of 6 μl of [γ-32P]ATP (3.3 μCi/μl, 20 μM ATP). Following a 30-min incubation at 30°C, reactions were terminated by the addition of SDS loading buffer. Samples were boiled, separated by SDS-PAGE, dried, and visualized by autoradiography. The amounts of phospho-GST-MKK6(K/A) were quantified using the imaging analyzer FLA-3000.
The ARGENT regulated homodimerization kit (version 2.0) and the ARGENT regulated transcription retrovirus kit (version 2.0) were obtained from ARIAD Pharmaceuticals, Inc. COS-7 cells were transfected with pFKBP-HA-MTK1, and 36 h later, the medium was replaced with fresh medium containing the appropriate amount of AP20187 and the cells were incubated for 2 h at 37°C prior to lysis and assay.
Previously, we have shown that the GADD45 family proteins bind to a GADD45-binding domain (BD) (Fig. (Fig.1A)1A) in the N-terminal regulatory region of MTK1 (34). An MTK1 mutant protein that lacks the GADD45 BD cannot be activated by GADD45 proteins. In the following work, we used GADD45β as a representative member of the GADD45 family because GADD45β is most strongly induced by methyl methanesulfonate (MMS) stress and by TGF-β. Whenever we compared the three GADD45 proteins, however, there were no qualitative differences among them.
GADD45β is a small protein of 160 amino acid residues (Fig. (Fig.1B).1B). To determine the region of GADD45β proteins that is directly involved in MTK1 activation, we constructed a series of deletion mutants, each of which has a deletion of ~10 amino acids. COS-7 cells were cotransfected with a plasmid encoding GFP-tagged GADD45β (GFP-GADD45β) or one of its deletion derivatives and another plasmid encoding full-length Flag-tagged MTK1 (Flag-MTK1). After transfection for 36 h, kinase activity of Flag-MTK1 was assayed in vitro. Coexpression of the full-length GADD45β increased Flag-MTK1 kinase activity more than 10-fold (Fig. (Fig.1C).1C). Deletions at the N terminus (residues 1 to 12) or C terminus (residues 133 to 160) had relatively little effect on MTK1 activation. In contrast, the series of 10-aa deletion mutants between residue 13 and residue 132 was nonactivating, indicating that this region of GADD45β is necessary for MTK1 activation.
To determine whether the GADD45 mutants' failure to activate MTK1 reflected an inability to bind MTK1, binding was assessed by two-hybrid analysis (see Fig. S1 in the supplemental material) and coprecipitation assays (Fig. (Fig.1D1D and data not shown). Whereas the full-length GADD45β and the activating GADD45β mutants could bind to MTK1, the nonactivating GADD45β mutants failed to do so. Neither the expression levels of the GADD45β deletion mutants (Fig. (Fig.1C)1C) nor their subcellular localizations (data not shown) differed significantly from those of wild-type (WT) GADD45β. These results thus support our hypothesis that a direct binding of GADD45 to MTK1 is necessary to activate the MTK1 kinase domain.
The deletion of the entire N-terminal regulatory segment of MTK1 constitutively activates its C-terminal kinase domain (33), implying that the N-terminal segment contains an autoinhibitory domain (AID) that binds to and inhibits the kinase domain. If so, a possible mechanism by which GADD45 might activate MTK1 is by disruption of the N-C interaction in an MTK1 molecule. As a first step in testing this model, we examined whether the N-terminal and C-terminal segments of MTK1 can associate stably with each other in an in vitro binding assay (see Materials and Methods for details). The results in Fig. Fig.1E1E (lane 2) clearly demonstrate that Flag-MTK1-N (residues 22 to 994) bound to Myc-MTK1-C (residues 1247 to 1607). The N-C interaction, however, was completely disrupted when another cell lysate containing GADD45β was included in the incubation mixture (Fig. (Fig.1E,1E, lane 3). The addition of a lysate containing GADD45βΔ(53-62) (lacking residues 53 through 62), which cannot bind to MTK1, did not inhibit the interaction (Fig. (Fig.1E,1E, lane 4). Flag-MTK1-NΔBD, which lacks the GADD45 binding domain (residues 147 to 250), could bind Myc-MTK1-C (Fig. (Fig.1E,1E, lane 5), but the interaction was no longer susceptible to interference by GADD45β (lane 6). In summary, it is concluded that the binding of GADD45β to the N-terminal segment of MTK1 disrupts MTK1 N-C association.
The activation of many protein kinases entails phosphorylation at a specific residue(s) in the activation loop (T-loop) between subdomains VII and VIII (21). There are five potential phosphorylation sites in the MTK1 activation loop region (Ser-1484, Thr-1493, Ser-1500, Thr-1501, and Thr-1504) (Fig. (Fig.2A).2A). To examine whether phosphorylation at any of these residues is required for MTK1 activation, we first individually mutated each of these residues by substitution with Ala. These Flag-MTK1 Ala substitution mutants were coexpressed with GADD45β, and their kinase activities were measured in vitro (Fig. (Fig.2B).2B). Kinase activities of MTK1-S1484A, S1500A, and T1501A were only moderately lower than that of wild-type MTK1. In contrast, MTK1-T1504A was completely inactive, and MTK1-T1493A had only very weak activity.
To examine whether Thr-1493 and/or Thr-1504 is phosphorylated in vivo, we developed specific antisera to the MTK1 T-loop peptides phosphorylated at Thr-1493 or at Thr-1504. To date, however, by using two different antisera, we have obtained no evidence that Thr-1504 is phosphorylated. It is possible that the mutation of Thr-1504 disturbs an essential secondary structure element unrelated to phosphorylation (45). In contrast, we could detect stimulation-dependent phosphorylation at Thr-1493 using an anti-phospho-T1493 (αP-T1493) rabbit antibody. The αP-T1493 antibody reacted with MTK1 only when the latter was activated by GADD45β coexpression; unstimulated MTK1 was unreactive (Fig. (Fig.2C,2C, lanes 1 and 2). Furthermore, the antibody did not react with the MTK1-T1493A mutant protein whether it was stimulated by GADD45β coexpression or not (Fig. (Fig.2C,2C, lanes 4 and 3, respectively). These results indicate, on the one hand, that the αP-T1493 antibody is specific to phospho-T1493 and, on the other hand, that the activation of MTK1 by GADD45β induces Thr-1493 phosphorylation. Occasionally, but not always, a substitution of Thr or Ser by an acidic amino acid (Asp or Glu) mimics the effect of phosphorylation. Thus, we tested the substitution of Thr-1493 by Asp or Glu, but both T1493D and T1493E were inactive.
Next, we asked whether the phosphorylation of MTK1 at Thr-1493 is affected by the autophosphorylation of MTK1 or by another kinase using the catalytically inactive pFlag-MTK1-K/R construct. As shown in Fig. Fig.2D,2D, no phosphorylation at Thr-1493 was observed for MTK1-K/R under the same conditions in which wild-type MTK1 was strongly phosphorylated. This result indicates that Thr-1493 phosphorylation is dependent on MTK1 kinase activity. Because Thr-1493 phosphorylation is also dependent on MTK1 homodimerization (see below), we conclude that the phosphorylation at Thr-1493 is mediated by the MTK1 kinase itself.
To test whether Thr-1493 phosphorylation plays a role in the activation of MTK1 in response to stress, we analyzed Thr-1493 phosphorylation following the addition of MMS to the human embryonic kidney HEK293 cells. MMS is a well-known cell stressor and a potent inducer of the GADD45 genes (34). Strong Thr-1493 phosphorylation was observed after 3 h of exposure of the cells to MMS (Fig. (Fig.3A3A and see Fig. Fig.8).8). Furthermore, this phosphorylation was dependent on MTK1 kinase activity because no Thr-1493 phosphorylation occurred when the kinase-defective Myc-MTK1-K/R mutant was used instead of wild-type Myc-MTK1. We also tested whether the endogenous MTK1 molecule is phosphorylated at Thr-1493 when cells are stimulated by stress. This is a difficult experiment because the amount of the endogenous MTK1 molecules is small, and the affinity of the available anti-MTK1 antibody is weak. Nonetheless, we could detect the MMS-induced phosphorylation of Thr-1493 (Fig. (Fig.3B).3B). Thus, the phosphorylation at Thr-1493 is a physiologically relevant reaction to external stress stimuli.
We next determined whether MTK1 phosphorylates Thr-1493 by an intramolecular (cis) reaction or by an intermolecular (trans) reaction. We transfected COS-7 cells simultaneously with two different MTK1 constructs, namely, Flag-tagged, wild-type MTK1 and Myc-tagged, kinase-dead MTK1 (Myc-MTK1-K/R), together with pGADD45β or the empty vector (Fig. (Fig.4A,4A, lanes 1 and 2). Myc-MTK1-K/R contains five repeats of the Myc tag and is easily distinguishable from Flag-MTK1 by SDS gel electrophoresis. If the Thr-1493 phosphorylation occurs only in cis, Myc-MTK1-K/R will not be phosphorylated, even in the presence of the activated Flag-MTK1 protein. Thr-1493 phosphorylation was detected, however, both in Myc-MTK1-K/R (Fig. (Fig.4A,4A, lane 2) and in Flag-MTK1. When the tags were reversed and Myc-MTK1 and Flag-MTK1-K/R were used (Fig. (Fig.4A,4A, lanes 3 and 4), an essentially identical result was obtained. These data indicate that Thr-1493 phosphorylation is likely to be an intermolecular (trans) reaction.
We tested whether MTK1 can form a stable homo-oligomer by coexpression of Myc-MTK1 and Flag-MTK1. Flag-MTK1 was immunoprecipitated from cell lysates, and the presence of coprecipitated Myc-MTK1 was probed by immunoblotting. As shown in Fig. Fig.4B,4B, lane 2, Flag-MTK1 and Myc-MTK1 coprecipitated only very weakly from unstimulated cells. However, when GADD45α/β/γ was also expressed in the same cells, interaction between Myc-MTK1 and Flag-MTK1 was much enhanced (Fig. (Fig.4B,4B, lanes 3 to 5). GADD45βΔ53-62, which cannot bind to MTK1 (Fig. (Fig.1B),1B), could not enhance the Myc-MTK1/Flag-MTK1 interaction (Fig. (Fig.4C,4C, lane 4). These results prove that MTK1 oligomerizes when stimulated by GADD45. Because full-length MTK1 is a very large molecule, it is difficult to measure the extent of oligomerization with any accuracy by gel filtration or other methods. Gel filtration analyses of a shorter fragment that contains the oligomerization domain (see below) were consistent with a dimer formation, but a trimer formation could not be excluded (data not shown). With this caveat, we conclude that MTK1 dimerizes when stimulated by GADD45.
GADD45 proteins are known to form homo- and hetero-oligomeric complexes (23). In principle, therefore, MTK1 dimerization could be a consequence of dimerization of the associated GADD45. Alternatively, a GADD45-induced conformational change might unmask a latent dimerization domain in MTK1. To distinguish these possibilities, we undertook the experiments described below.
Initially, using a full-length Flag-MTK1 and a series of Myc-MTK1 truncation constructs, we mapped the region in MTK1 that is responsible for dimerization. Full-length Myc-MTK1 coprecipitated with full-length Flag-MTK1 in the presence of GADD45β (see Fig. S2 [lane 2] in the supplemental material). Most notably, Myc-MTK1(853-1341), in which both the C-terminal kinase catalytic domain (amino acids 1342 to 1607) and the N-terminal 852 amino acids are absent, could also bind to full-length Flag-MTK1 in the presence of GADD45 (see Fig. S2 [lane 10] in the supplemental material). Because Myc-MTK1(853-1341) lacks the entire GADD45-binding domain (amino acids 147 to 250), this result disproves the hypothesis that MTK1 dimerization is an indirect consequence of the GADD45 dimerization.
The MTK1 dimerization domain was further mapped using additional deletion constructs. The progressive deletion of MTK1(853-1341) from its N-terminal side revealed that MTK1(941-1341) can bind full-length MTK1, but MTK1(953-1341) cannot, indicating that the sequence between amino acid residues 941 and 953 is essential for dimerization (Fig. (Fig.5A,5A, left panel). The deletion of MTK1(941-1341) from its C-terminal side showed that the shortest segment that can bind full-length MTK1 is MTK1(941-1272) (Fig. (Fig.5A,5A, right panel). Thus, we conclude that the minimal region required for MTK1 dimerization is the region between amino acid residues 941 and 1272, although a slightly longer segment (residues 941 to 1321) binds full-length MTK1 more efficiently.
Because the dimerization assay in the above mapping was performed entirely in vitro, we verified that MTK1(941-1321) and full-length MTK1 interact in vivo, by coexpressing Myc-MTK1(941-1321) and full-length Flag-MTK1 in COS-7 cells in either the presence or the absence of GADD45β (Fig. (Fig.5B).5B). In this in vivo experiment, as in the in vitro experiments, Myc-MTK1(941-1321) and full-length Flag-MTK1 dimerized, but only in the presence of GADD45β.
A database search revealed no significant candidates that are structurally similar to the MTK1 dimerization domain (residues 941 to 1272), except the corresponding regions in MTK1 orthologs of vertebrates, insects, and nematodes. The COILS program (26), however, predicted that the short region of residues 982 to 1012 may form a coiled-coil structure (Fig. (Fig.6A).6A). A coiled-coil is made of two α-helices that wrap around each other to form a twisted supercoil structure and is often involved in protein-protein assemblage. The MTK1 coiled-coil motif is conserved in the MTK1 orthologs of evolutionarily diverse organisms (see Fig. S3 in the supplemental material).
To examine whether the coiled-coil motif plays a role in MTK1 dimerization and/or activation, we constructed in vivo coimmunoprecipitation assays using four MTK1 mutants in the putative coiled-coil motif. The first mutant, MTK1-ΔCC, lacks the entire coiled-coil motif (amino acids 982 to 1012). In the other three mutants, one or more positions (Leu-997, Ile-1001, and Val-1008) are substituted by a helix-disrupting proline; the L997P/I1001P double mutant is abbreviated as PP, and the L997P/I1001P/V1008P triple mutant is abbreviated as PPP. When the entire coiled-coil region was deleted (MTK1-ΔCC) or two critical amino acid residues were replaced by proline (MTK1-PP), very little interaction occurred between the Flag- and Myc-tagged constructs, even in the presence of GADD45β (Fig. (Fig.6B6B).
From these results, we conclude that the coiled-coil formation is critical for GADD45-induced MTK1 dimerization. We also conclude that the dimerization domain in full-length MTK1 is usually masked and is uncovered only in the presence of the GADD45 proteins.
Both dimerization and Thr-1493 autophosphorylation of MTK1 are induced by the coexpression of GADD45, suggesting that these events might be causally related. More specifically, we hypothesized that MTK1 dimerization is a prerequisite for autophosphorylation at Thr-1493. To test this hypothesis, we examined whether dimerization-defective MTK1 mutants can autophosphorylate at Thr-1493. GADD45-stimulated phosphorylation of Thr-1493 was undetectable in MTK1-ΔCC, compared to the robust Thr-1493 phosphorylation of the wild-type MTK1 molecule (Fig. (Fig.7A).7A). PP and PPP substitution mutants are also completely defective in autophosphorylation, although the L997P single mutant could weakly autophosphorylate at Thr-1493 (Fig. (Fig.7B).7B). Perhaps consistent with these findings, the L997P single mutant did not inhibit MTK1 dimerization to an appreciable degree (data not shown). Thus, Thr-1493 autophosphorylation requires MTK1 dimerization.
Earlier, we suggested that Thr-1493 autophosphorylation is an intermolecular, or trans, reaction between a pair of MTK1 molecules in a dimer. To further examine whether Thr-1493 phosphorylation is an intermolecular reaction, we transfected COS-7 cells with two MTK1 constructs: a Myc-tagged, catalytically active, full-length molecule (Myc-MTK1) and a Flag-tagged, catalytically defective molecule, without or with a mutation in the coiled-coil motif (Flag-MTK1-K/R or Flag-MTK1-PP-K/R, respectively). Because Flag-MTK1-K/R has no catalytic activity, the only way its Thr-1493 can be phosphorylated by the coexpressed Myc-MTK1 is in trans. When Myc-MTK1 and Flag-MTK1-K/R were coexpressed, immunoprecipitated Flag-MTK1-K/R was phosphorylated at Thr-1493 when GADD45β was also expressed (Fig. (Fig.7C,7C, lanes 1 and 2). However, when Myc-MTK1 and the dimerization-defective Flag-MTK1-PP-K/R construct were coexpressed, Flag-MTK1-PP-K/R was not phosphorylated at all by the coexpressed Myc-MTK1 kinase, even in the presence of GADD45β (Fig. (Fig.7C,7C, lane 4). From these results, we conclude that the MTK1 dimerization is required for Thr-1493 phosphorylation in trans between two MTK1 molecules.
So far, we have shown separately that MTK1 dimerization is required for Thr-1493 phosphorylation and that Thr-1493 phosphorylation precedes the full activation of the MTK1 kinase. Thus, we can predict that dimerization is required for the full activation of the MTK1 kinase catalytic domain. This prediction was tested using an in vitro kinase assay (Fig. (Fig.7D).7D). MTK1-L997P, which weakly autophosphorylates in the presence of GADD45β (Fig. (Fig.7B),7B), also had a weaker kinase activity than wild-type MTK1 did (Fig. (Fig.7D,7D, lanes 4 and 2, respectively). MTK1-PP and MTK1-PPP, which are both completely defective in dimerization (Fig. (Fig.6B6B and data not shown) and in autophosphorylation at Thr-1493 (Fig. (Fig.7B),7B), are completely defective in kinase activity (Fig. (Fig.7D,7D, lanes 6 and 8, respectively). Thus, there is a very strong correlation between dimerization, autophosphorylation, and catalytic activation of MTK1.
Next, we tested whether the dimerization of MTK1 by itself is sufficient to achieve kinase domain activation by using FK506-binding protein 12 (FKBP12)-mediated dimerization of MTK1. An FKBP12 domain can be induced to dimerize by adding the dimeric FK506 reagent AP20187 to the medium (31, 40). MTK1 phosphorylation at Thr-1493 could not be induced by AP20187 at any concentration between 0 and 100 nM (see Fig. S4 in the supplemental material) The coexpression of GADD45β induced Thr-1493 phosphorylation of FKBP-MTK1, showing that the fusion protein is capable of autophosphorylation, but only after binding of GADD45. Thus, dimerization is necessary but not sufficient for MTK1 activation. Perhaps the AP20187-induced dimer has a less optimal orientation of the coiled-coil segments for activation.
To test whether GADD45 binding and MTK1 dimerization play roles in the phosphorylation of MTK1 Thr-1493 in response to stress, we analyzed Thr-1493 phosphorylation following the addition of MMS to the cells (Fig. (Fig.3A).3A). HEK293 cells were stably transfected with a wild type, GADD45-binding site deletion (ΔBD) mutant, or coiled-coil defective (PP) mutant version of a Myc-MTK1 construct. MTK1 phosphorylation at Thr-1493 was monitored before and after the exposure of these cells to MMS (Fig. (Fig.8).8). Strong Thr-1493 phosphorylation of the wild-type MTK1 was observed after 3 h of exposure to MMS. In contrast, Thr-1493 phosphorylation was completely absent in ΔBD and PP mutants, indicating that both GADD45 binding and MTK1 dimerization are important for phosphorylation at Thr-1493 by extracellular stress stimuli.
Previously, we described constitutively active MTK1 mutants that can bind to and phosphorylate MKK6 in the absence of GADD45 (28). The mechanism of activation of these MTK1 mutants in the absence of GADD45 binding should shed further light on how GADD45 activates the MTK1 kinase. Thus, we investigated whether these constitutively active mutations can mimic MTK1 activation by GADD45 binding. Specifically, we tested three constitutively active mutants (L534Q, Q637P, and I1360M) that can phosphorylate the MKK6 substrate, both in vivo and in vitro, in the absence of GADD45 expression. The fourth mutant, V1300F, can bind but does not phosphorylate MKK6 in the absence of GADD45.
First, we tested MTK1 Thr-1493 phosphorylation in these mutants. As shown in Fig. Fig.9A,9A, MTK1-L534Q, MTK1-Q637P, and MTK1-I1360M were constitutively phosphorylated at Thr-1493 in the absence of any GADD45 protein (lanes 3, 4, and 6, respectively), whereas the wild-type MTK1 or MTK1-V1300F was not (lanes 1 and 5, respectively). Next, we tested whether these mutants were dimerized by coexpressing Myc-tagged and Flag-tagged versions of each mutant. The three constitutively active mutants (L534Q, Q637P, and I1360M) were dimerized in the absence of GADD45 (Fig. (Fig.9B,9B, lanes 3, 4, and 6), whereas MTK1-V1300F was dimerized only at the level of the wild-type MTK1 protein without GADD45 (lane 5). Thr-1493 phosphorylation of the constitutively active mutants was dependent on the MTK1 dimerization domain because the dimerization-defective PP mutation completely abolished Thr-1493 phosphorylation of the L534Q mutant (Fig. (Fig.9C).9C). Lastly, we tested the effects of constitutively active mutations on MTK1 N-C interaction. As we showed earlier in this paper, N- and C-terminal segments of MTK1 bind to each other and this association is disrupted by GADD45 binding to the N-terminal fragment (Fig. (Fig.9D,9D, lanes 2 and 3). The constitutively active mutation L534Q in the MTK1 N-terminal fragment inhibited the N-C interaction, and this inhibition does not require GADD45 (Fig. (Fig.9D,9D, lane 4). Thus, the constitutively active MTK1 mutations mimic the effects of GADD45 binding, including the Thr-1493 autophosphorylation, MTK1 dimerization, and disruption of the N-C interaction, suggesting that they are key residues in GADD45-mediated MTK1 activation.
The data from this study lead to the following model of GADD45-mediated activation of the MTK1 MAPKKK. The activation of MTK1 by GADD45 occurs through a series of molecular steps I through VI, as summarized in Fig. Fig.10.10. In brief, each step is as follows. For step I, in unstimulated cells, MTK1 is in a closed (inhibited) conformation in which the N-terminal AID blocks the C-terminal kinase catalytic domain. In step II, extracellular stimuli, such as MMS exposure, induce the expression of stress-inducible GADD45 proteins, which bind to the MTK1 N-terminal GADD45-BD. In step III, GADD45 binding to MTK1 dissociates the latter's AID from the C-terminal kinase catalytic domain. In step IV, at the same time, GADD45 binding unmasks the MTK1 dimerization domain, inducing homodimer formation. At this stage, the MTK1 kinase domain is in open conformation (i.e., not actively inhibited) but not yet fully active as a kinase. In step V, dimerized MTK1 becomes fully activated when Thr-1493 is trans autophosphorylated. In step VI, GADD45 binding also unmasks a site in the MTK1 kinase domain that interacts with the MAPKK DVD docking sites, allowing MTK1 to interact with and phosphorylate the cognate MAPKKs, namely, MKK3 and MKK6 (36).
Because the GADD45 proteins are highly conserved (34), it is likely that all three members of the GADD45 family activate MTK1 by the same mechanism. Indeed, in our investigation of GADD45-mediated MTK1 activation, we did not find any qualitative differences among the three GADD45 proteins. We did find, however, that there are quantitative differences among them; e.g., GADD45α does not stimulate MTK1 as efficiently as GADD45β or GADD45γ does (34). It is possible that they have slightly different affinities for MTK1. In terms of function, however, their varied expression patterns are perhaps more important. The expression of each member of the GADD45 gene family is induced by a distinct set of environmental stresses and cytokines in different tissues. For example, the expression of GADD45β, but not GADD45α or -γ, is induced by TGF-β (35) and the expression of GADD45α, but not GADD45β or -γ, is modulated by p53 (25). Consistently, the GADD45 proteins (GADD45α/β/γ) serve overlapping but nonidentical functions in different apoptotic and growth inhibitory pathways (25).
Full activation of MTK1 by GADD45 entails four different molecular mechanisms: removal of the autoinhibitory domain, dimerization, phosphorylation of the activation loop, and unmasking of the docking site for MAPKKs. Individually, these mechanisms are used by other MAPKKKs. However, the details are different for each MAPKKK, reflecting different physiological roles. Therefore, in order to appreciate its physiological function, it is important to analyze how the binding of one protein (GADD45) orchestrates these mechanisms, thereby converting an inert enzyme (MTK1) to a fully active one.
The autoinhibition of the kinase catalytic domain by N-terminal (or sometimes C-terminal) regulatory sequences is a mechanism employed by many MAPKKKs. Artificial deletion of such autoinhibitory domains converts the kinase into a constitutively active one (5, 6, 11, 43, 46). In physiologically relevant activation processes, however, release from autoinhibition is effected by a different mechanism for each kinase. For example, MEKK1 is activated by caspase-3-mediated shedding of its N-terminal inhibitory domain (41) and Ste11 is activated by Ste20-mediated phosphorylation of the autoinhibitory domain (39). In the case of MTK1, the binding of GADD45 to MTK1 seems to compete with the adjacent autoinhibitory domain for interaction with the MTK1 catalytic domain.
Dimerization is also frequently utilized to activate protein kinases, including MAPKKKs. For example, it has been reported that forced dimerization of full-length MEKK1 induces its autophosphorylation and the subsequent phosphorylation of the MKK4 (SEK1) MAPKK (7). It was also reported recently that the dimerization of the MEKK4 (MTK1) kinase domain (without the N-terminal autoinhibitory domain) by only AP20187 increases MEKK4 kinase activity (1). Unlike these cases, however, artificial dimerization of full-length MTK1 using FKBP and AP20187 did not enhance its in vitro catalytic activity and did not lead to Thr-1493 phosphorylation in vivo. Therefore, it is likely that dimerization is necessary but not sufficient for the activation of full-length MTK1. Perhaps it is important for MTK1 activation that GADD45 dissociates the N-terminal autoinhibitory domain from the C-terminal catalytic domain. The specific configuration of the coiled-coil-mediated dimerization might also be important.
Since GADD45 proteins can themselves form homo- and heterodimers (or oligomers) (23), one possibility was that MTK1 indirectly dimerized through the dimerization of bound GADD45 molecules. This is unlikely, however, for two reasons. First, MTK1 fragments that do not contain the GADD45-BD can bind MTK1. Second, constitutively active MTK1 mutants dimerize in the absence of GADD45 proteins (Fig. (Fig.9B).9B). Nonetheless, it is possible that GADD45 dimerization might serve to concentrate MTK1 molecules, hence enhancing the latter's dimerization. Finally, it should be noted that MTK1 is dimerized by other stimuli, such as TRAF4, presumably via a GADD45-independent mechanism (1).
The phosphorylation of the activation loop is perhaps the most common activation mechanism for diverse families of protein kinases (4, 15, 19, 42). However, different kinases might use different mechanisms to achieve phosphorylation of the activation loop. Many kinases are phosphorylated by their upstream activating kinase in a manner similar to that for the phosphorylation of MAPKKs by MAPKKKs or the phosphorylation of MAPKs by MAPKKs. Other kinases autophosphorylate the activation loop (12, 22). Thr-1493 phosphorylation of MTK1 is induced by the expression of the GADD45 proteins or by extracellular stimuli, such as MMS, that are known to induce GADD45 expression. Dimerization-defective MTK1 mutants, however, cannot autophosphorylate Thr-1493. Thus, stable dimerization induced by GADD45 binding is responsible for efficient intermolecular Thr-1493 phosphorylation of MTK1.
Finally, the GADD45 proteins indirectly control the interaction between activated MTK1 and its substrates, i.e., MKK3 and MKK6. Both MKK3 and MKK6 contain an ~20-amino-acid peptide, termed the DVD domain, at their C termini (36). The deletion and point mutations of the MKK3/MKK6 DVD sequence inhibit stable interaction between MTK1 and MKK3/MKK6 and, consequently, inhibit the activation of MKK3/MKK6 by MTK1. The DVD-mediated interaction between MKK6 and full-length MTK1 does not take place unless GADD45 is also present (28). It is likely that the MTK1 N-terminal autoinhibitory domain prevents the MTK1 catalytic domain from binding the MKK3/MKK6 DVD docking site.
With the detailed activation mechanism of MTK1 now available, it will be possible to study this important but underexplored area of cellular signal transduction.
We thank P. O'Grady for critical reading of the manuscript, M. Kitamura and N. Yoshida for excellent technical assistance, and ARIAD Pharmaceuticals, Inc., for the gift of the ARGENT regulated homodimerization kit and the ARGENT transcription retrovirus kit.
This work was supported in part by several Grants-in-Aid from the Ministry of Education, Culture, Sports, Science and Technology of Japan (to H.S. and M.T.) and a PRESTO program from the Japan Science and Technology Agency (to M.T.).
Published ahead of print on 22 January 2007.
†Supplemental material for this article may be found at http://mcb.asm.org/.