Axin Utilizes Distinct Regions for Competitive MEKK1 and MEKK4 Binding and JNK Activation*

Axin is a multidomain protein that plays a critical role in Wnt signaling, serving as a scaffold for down-regulation of β-catenin. It also activates the JNK mitogen-activated protein kinase by binding to MEKK1. However, it is intriguing that Axin requires several additional elements for JNK activation, including a requirement for homodimerization, sumoylation at the extreme C-terminal sites, and a region in the protein phosphatase 2A-binding domain. In our present study, we have shown that another MEKK family member, MEKK4, also binds to Axin in vivo and mediates Axin-induced JNK activation. Surprisingly MEKK4 binds to a region distinct from the MEKK1-binding site. Dominant negative mutant of MEKK4 attenuates the JNK activation by Axin. Activation of JNK by Axin in MEKK1–/– mouse embryonic fibroblast cells supports the idea that another MEKK can mediate Axin-induced JNK activation. Expression of specific small interfering RNA against MEKK4 effectively attenuates JNK activation by the MEKK1 binding-defective Axin mutant in 293T cells and inhibits JNK activation by wild-type Axin in MEKK1–/– cells, confirming that MEKK4 is indeed another mitogen-activated protein kinase kinase kinase that is specifically involved in Axin-mediated JNK activation independently of MEKK1. We have also identified an additional domain between MEKK1- and MEKK4-binding sites as being required for JNK activation by Axin. MEKK1 and MEKK4 compete for Axin binding even though they bind to sites far apart, suggesting that Axin may selectively bind to MEKK1 or MEKK4 depending on distinct signals or cellular context. Our findings will provide new insights into how scaffold proteins mediate ultimate activation of different mitogen-activated protein kinase kinase kinases.

Axin is a highly modular protein, possessing numerous protein-binding domains that include adenomatous polyposis coli (APC), 1 GSK-3␤, ␤-catenin, casein kinases, Axam, PP2A, and Dishevelled to coordinate Wnt signaling (1)(2)(3)(4)(5)(6)(7)(8)(9)(10). More recently, Axin has also been shown to interact with low density lipoprotein receptor-related protein 5 (LRP5) that acts as a co-receptor for Wnt (11), linking the upstream signaling components to downstream signaling transducers. The interaction of LRP5 with Axin in fact may provide a new mechanism for the Axinbased regulation of the intracellular levels of ␤-catenin, which is independent of the action of GSK-3␤ (12). In this pathway, LRP recruits Axin to the membrane, leading to the degradation of Axin. As a consequence, ␤-catenin is no longer bound by Axin and is stabilized, resulting in elevated nuclear signaling by ␤-catenin. Clearly Axin is a multifunctional protein, and its functions are highly regulated. In addition to the Wnt pathway, Axin has also been shown to activate the JNK mitogen-activated protein kinase pathway (13). Other components in the Wnt pathway, Dishevelled and Diversin, also participate in the regulation of JNK activity (14,15). It was convincingly shown that a defect in the activation of JNK by Dishevelled, resulting from a point mutation in the DEP domain, leads to abnormality of planar polarity in Drosophila. Rescue analysis reveals different protein domain requirements in Dishevelled for the Wnt and JNK pathways; the C-terminal DEP domain is essential to rescue planar polarity defects and induce JNK signaling (14). Interestingly, as with Axin, the Dishevelled/JNK pathway is discriminatory from the canonical Frz/Dvl Wnt pathway in that distinct domains are utilized in the two pathways (14,16,17). In addition, the slipper gene that encodes a mixed lineage kinase (MLK), a JNK kinase kinase or MAP3K (18,19), is critical for dorsal closure in Drosophila (20). These findings all point to the significance of the JNK pathway in development in that it controls cell fate and cell sheet morphogenesis (21,22).
In the JNK pathway, we previously showed that Axin interacts with MEKK1 and identified the MEKK1-binding domain as the MID domain flanked by the binding sites for APC on the N-terminal side and GSK-3␤ on the C-terminal side (13). Several intriguing phenomena on the Axin-induced JNK activation prompted us to further define the exact domains required for Axin to activate JNK. First of all, we observed that, in addition to the MEKK1-binding domain, Axin required the entire region C-terminal to the ␤-catenin-binding site. Although we now know that the DIX domain is required for homodimerization of Axin (23) and that the extreme C-terminal six amino acids for sumoylation (24), it remains unclear why the region in the PP2A-binding region is also required. Furthermore MEKK1 binding is highly regulated by conformational changes elicited by either GSK-3␤ binding or casein kinase I binding to Axin (5,16). We reasoned that further deciphering the molecular mechanism whereby Axin activates the JNK pathway would yield invaluable insights into how the scaffold protein Axin mediates MAP3K and subsequently JNK activation. Among them, we shall be able to understand the structural and biochemical basis for the differential complex formations based on Axin.
In the course of further delineating the binding region for MEKK1, we found that newly created deletion mutants of Axin, which could no longer bind to MEKK1, still fully effectively activated JNK. We wondered whether the previously created JNK activation-defective Axin⌬MID happened to adopt a conformation that disallows Axin to activate JNK. To clarify this, we thoroughly examined all the critical Axin regions that could contribute to JNK activation. Through detailed analysis, we found that Axin also interacts with MEKK4 in addition to MEKK1. Surprisingly MEKK4 binds to a separate region inside the PP2A binding area far apart from the MEKK1-binding site. Dominant negative MEKK4-K/M mutant effectively attenuates JNK activation by wild-type Axin. Axin can activate JNK in mouse embryonic cells null for MEKK1 in accordance with the presence of MEKK4 in these cells. Furthermore small interfering RNA specific for MEKK4 inhibits JNK activation by Axin mutants that are defective in MEKK1 binding in 293T cells and by wild-type Axin in MEKK1 Ϫ/Ϫ cells. The finding that two MAP3Ks can participate in the Axin/JNK pathway via competitive binding to two different sites adds a new level of complexity of the Axin-based differential complex formations. Our current data also suggest that Axin may receive distinct signals via MEKK1 or MEKK4 to activate JNK.

EXPERIMENTAL PROCEDURES
Cell Culture-Human embryonic kidney 293T cells were maintained in RPMI 1640 medium as described previously (13). Mouse embryonic MEKK1 Ϫ/Ϫ fibroblast cells, a gift from M. Karin (University of California, San Diego), were generated as described previously (25) and were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, 100 IU of penicillin, 100 g/ml streptomycin, and 2 mM glutamine.
Construction of Expression Plasmids-Expression vectors for wildtype mouse HA-Axin, Myc-Axin, Axin M5, C2, HA-MEKK1, and FLAGtagged JNK1 were generated as described previously (23). Wild-type full-length human MEKK2 and mouse MEKK3 were generated from fragments by polymerase chain reaction using primers 5Ј-ccatggatgatcagcaagctttgaac-3Ј and 5Ј-tctagactagtgataatgcacaaacatgtgc-3Ј for MEKK2 and 5Ј-ccatggatgaacaagaggcattagac-3Ј and 5Ј-tctagatcagtacactagctgtgcaaagtg-3Ј for MEKK3. The cDNA sequence for MEKK4-C starts from the BamHI site to the 3Ј-end derived from Clone KIAA0213 (courtesy of Kazusa DNA Research Institute) encoding aa 755-1605. All PCR products were verified by sequencing. MEKK2, MEKK3, and MEKK4 were fused in-frame with HA and Myc tag and cloned into the ClaI and XbaI sites of the mammalian expression vector pCMV5. Axin deletion mutants D4, D19, DR2, D20, M4, M6, M8, M9, M10, and M11 were created by subcloning PCR-generated fragments that contained convenient restriction sites followed by fusing the respective deletion mutant fragments with their appropriate flanking wild-type cDNA. Axin deletion mutant D19/M8 was created by fusing the ClaI-XbaI fragment of Axin D19 with the XbaI-BamHI fragment of Axin M8 followed by cloning into the ClaI and BamHI sites of pCMV5. Detailed information is available upon request.
Transient Transfection and Immunokinase Assays-Transfections were performed in 60-mm dishes using Dosper Liposomal Transfection Reagent (Roche Applied Science) according to the manufacturer's instructions. The total amount of transfected DNA of each plate was adjusted with the empty vector pCMV5 where necessary. Cells were harvested at 40 h posttransfection and lysed in a lysis buffer (20 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1% Triton, 2.5 mM sodium pyrophosphate, 1 mM ␤-glycerol phosphate, 1 mM sodium orthovanadate, 1 g/ml leupeptin, 1 mM phenylmethylsulfonyl fluoride). FLAG-tagged JNK1 was immunoprecipitated using mouse monoclonal anti-FLAG M2 beads (Sigma); the JNK activity was determined as described previously using 1 g of GST-c-Jun-(1-79) (Stratagene) as substrate (13)   were generated based on our previous finding that the region of aa 210 -352 in mutant M2 encompasses an MEKK1-interacting domain. These mutants were assayed for their ability to bind MEKK1 by co-immunoprecipitation and their JNK activation by immunokinase assay. Cells were transfected with 2 g each of HA-tagged Axin constructs together with 2 g of HA-MEKK1-C or 1 g of FLAG-JNK1. Cells were harvested at 36 h posttransfection; cell lysates were subjected to immunoprecipitation (IP) with anti-MEKK1 for MEKK1 and anti-FLAG for JNK1 followed by immunoblotting (IB) with anti-HA for Axin proteins. Following immunoprecipitation of FLAG-JNK1, kinase activities were assayed using GST-c-Jun as a substrate. The amount of the kinase in each immunoprecipitate was quantified by immunoblotting. Data are expressed as -fold kinase activation compared with basal kinase activity in vector-transfected cells. The values represent the means Ϯ S.E. from three separate experiments.

Co-immunoprecipitation and Western
Blotting-Transiently transfected 293T or MEKK1 Ϫ/Ϫ embryonic fibroblast cells in 60-mm dishes were lysed in the same lysis buffer as described above, sonicated 10 times for 1 s each, and centrifuged at 13,000 rpm for 30 min at 4°C. HAor Myc-tagged Axin proteins or MEKK proteins were immunoprecipitated from the cell lysate with anti-HA (Roche Applied Science), anti-Myc (9E10), or anti-MEKK1 (C-22) (Santa Cruz Biotechnology, Inc.) antibodies and Protein A/G Plus agarose beads (Santa Cruz Biotechnology, Inc.). The rabbit polyclonal antibody against endogenous Axin was a generous gift of Dr. David M. Virshup (University of Utah); anti-MEKK4 was purchased from Abgent (San Diego, CA). Immunoprecipitates or total cell lysates were analyzed by Western blotting as described previously (13). The boiled samples were separated on 10% SDS-polyacrylamide gels and transferred to polyvinylidene difluoride membrane (Roche Diagnostics). After blocking with 5% skim milk in phosphate-buffered saline with 0.1% Tween 20 for 1 h, the membranes were probed with either anti-Myc (9E10), anti-MEKK1 (C-22), anti-HA, or anti-FLAG antibodies. Bound antibodies were visualized by enhanced chemiluminescence (Amersham Biosciences) using horseradish peroxidase-conjugated antibodies.
RNA Interference-The pSUPER vector was used to knock down the expression of MEKK4 in 293T cells and in MEKK1 Ϫ/Ϫ cells. A human MEKK4-specific 19-mer nucleotide, 5Ј-tgttatgcacgttggcttg-3Ј, was inserted into pSUPER vector according to the instructions (26). To test for the efficiency of pSUPER-MEKK4, cells were co-transfected with 1 or 5 g of pSUPER-MEKK4 and 1 g of Myc-MEKK4-C plus 1 g of FLAG-JNK1 in each transfection as a control for transfection efficiency. After 40 h of additional culture, cells were lysed, and the expression of MEKK4-C was analyzed by immunoblotting using anti-Myc. To examine JNK activation by Axin deletion mutant D19 in pSUPER-MEKK4treated 293T cells and that by wild-type Axin in MEKK1 Ϫ/Ϫ cells, cells were transfected with 5 g of pSUPER-MEKK4 and 1 g of HA-tagged Axin D19 or wild-type Axin together with 1 g of FLAG-JNK1. Immunokinase assays were performed as described above.

Axin Mutants Lacking the MEKK1-binding Site Retain the
Ability to Activate JNK-It was previously shown that MEKK1, but not TAK or ASK, functionally interacted with Axin on a domain termed MID (aa 210 -352) (13). We have also shown that GSK-3␤ or casein kinase I␣ and I⑀ binding to Axin excludes MEKK1 binding (16), suggesting that JNK activation by Axin is at least in part regulated by conformational changes of the Axin protein. As part of efforts to address the structural basis underlying JNK activation by Axin, we fine mapped the region required for MEKK1 binding (Fig. 1, lower left panel). Deletion mutant D4 had aa 210 -243 removed and was intact in MEKK1 binding. DR2 mutant lacked aa 330 -352 and was also capable of binding to MEKK1, indicating that the region between aa 244 and 329 is responsible for MEKK1 binding. Surprisingly, when the newly created MID domain mutants were assayed for their ability to activate JNK, we found that these MEKK1 binding-defective mutants (i.e. D19 and D20) activated Endogenous Axin, MEKK1, and MEKK4 were immunoprecipitated from the total cell lysate using anti-Axin, anti-MEKK1, and anti-MEKK4, respectively. Protein A/G Plus agarose beads with no antibody linked were used in immunoprecipitation (IP) as a negative control. Immunoprecipitates were analyzed by immunoblotting (IB) using anti-Axin, anti-MEKK1, and anti-MEKK4 antibodies for Axin, MEKK1, and MEKK4, respectively. Middle panel, 293T cells were transfected with 1 g of HA-Axin D19 that lacks the MEKK1-binding site. Cell lysates were subjected to immunoprecipitation using anti-Axin, anti-MEKK1, or anti-MEKK4. Immunoprecipitates were then analyzed by immunoblotting as described above. Right panel, 293T cells were transfected with 1 g of HA-Axin M8 that lacks the MEKK4-binding site. Immunoprecipitation and immunoblotting were carried out as above.
MEKK4 Interacts with Axin via a Novel Domain Distinct from the MEKK1-binding Site-As MEKK2, -3, and -4 are structurally related to MEKK1, we first tested whether they also interacted with Axin, which might account for the JNKactivating function of the Axin mutants defective in MEKK1 binding. Myc-tagged Axin was co-transfected with HA-tagged MEKK1-C, -2, -3, or -4-C. Immunoprecipitation was carried out reciprocally with anti-Myc and anti-HA; the immunoprecipitates were analyzed by Western blotting with anti-HA for MEKKs and anti-Myc for Axin. As shown previously, the Cterminal fragment of MEKK1 containing the kinase domain bound to wild-type Axin ( Fig. 2A, Ref. 13). Interestingly MEKK4-C also strongly interacted with Axin, whereas MEKK2 or MEKK3 was not co-immunoprecipitated with Axin ( Fig. 2A).
To map the region of Axin for MEKK4 binding, a series of Axin deletion mutants were created, including the ones indi-cated on Fig. 2B, top. We separately co-transfected different HA-tagged Axin mutants with Myc-MEKK4-C and performed co-immunoprecipitation. Results revealed that a domain, distinct from the MEKK1-binding region, binds to MEKK4, which is located around aa 679 -746 as deduced from MEKK4 binding assays with constructs M6 and C2 (Fig. 2B). Further deletion of aa 507-672 (M6) or removal of the C-terminal aa 747-832 (C2) did not affect the ability of Axin to bind MEKK4, whereas further removal of aa 673-705 (M5) rendered Axin incapable to bind MEKK4. Consistently deletion of the 33 aa from 679 to 711 (M8) abolished Axin binding to MEKK4. This MEKK4binding site is inside the region previously known also to interact with PP2A (10). It is interesting to note that the two regions for MEKK1 and MEKK4 binding are far apart.
To verify the Axin-MEKK4 interaction, we tested whether the interaction occurred in their physiologic concentrations. We lysed untransfected 293T cells and pulled down endogenous Axin, MEKK1, or MEKK4 by using anti-Axin, anti-MEKK1, or anti-MEKK4 polyclonal antibodies, respectively. Axin was detected in the anti-MEKK1 and anti-MEKK4 immunoprecipitates; MEKK1 and MEKK4 were both detected in the anti-Axin immunoprecipitated complexes. Interestingly MEKK1 was not detected in the immunocomplexes pulled down by the anti-MEKK4 antibody. Similarly MEKK4 was not present in the anti-MEKK1 immunocomplexes (Fig. 3, left panel). The mutually exclusive presence of MEKK1 and MEKK4 in the Axin complexes is consistent with our finding that they compete against each other in binding to Axin (see below). In parallel, we also performed single transfection experiments to test Axin site dependence for MEKK binding. Axin deletion mutants that lack the binding site of either MEKK1 (D19) or MEKK4 (M8) were transfected into 293T cells followed by immunoprecipitation using antibodies against Axin, MEKK1, or MEKK4. Immunoblotting analysis of the immunoprecipitates showed that Axin D19 could no longer interact with the endogenous MEKK1 and that Axin M8 could not interact with the endogenous MEKK4.
Dominant Negative MEKK4 Diminished Axin-mediated JNK Activation-To test whether MEKK4 could indeed mediate the Axin-induced JNK activation, we substituted the conserved lysine (K) residue in MEKKs with methionine (M) to create their respective dominant negative forms, MEKK-K/Ms. As expected, Axin alone activated JNK to greater than 10-fold. In contrast, when Axin was co-transfected with MEKK4C-K/M, the JNK activation was drastically diminished as in the cells co-transfected with MEKK1C-K/M. In contrast, MEKK2-K/M or MEKK3-K/M failed to attenuate the Axin-induced JNK activation (Fig. 4). These observations are in accordance with their binding activity toward Axin.

MEKK1 and MEKK4 Require a Common Region of Axin in Addition to Their Respective Binding Sites for JNK Activa-
tion-During the course of fine deletion mapping for both MEKK1 and MEKK4 interaction regions in Axin, we found another region of Axin that is required by both MEKK1 and MEKK4 for JNK activation (Fig. 5). M6 (⌬aa 507-672) could bind to both MEKK1 and MEKK4 but failed to activate JNK (Fig. 5), indicating that the region of aa 507-672 is required by both MEKK1 and MEKK4 for JNK activation. The requirement for this novel domain is further supported by observations that other mutants such as M4 (⌬aa 507-730), in which part of the region of aa 507-672 is deleted, could bind to MEKK1 as strongly as did the wild-type Axin but failed to activate JNK (Fig. 2 and data not shown). In contrast, M8 (⌬aa 679 -711) that retains the region of aa 507-672 activated JNK well. To further define this important region of Axin for JNK activation, we generated another three deletions, M9, M10, and M11, in the region of aa 610 -720 as schematically diagrammed in Fig. 4. JNK assay and MEKK4 binding experiment were carried out with these mutants. The results showed that the region encompassing aa 642-673 N-terminal to the MEKK4binding site is indispensable for Axin to activate JNK (Fig. 5), suggesting that MEKK1 and MEKK4 independently require this novel domain for JNK activation. As the Axin regions of aa 210 -338 and aa 679 -711 are required for MEKK1 and MEKK4 binding respectively, we constructed an Axin deletion mutant, D19/M8, that removes both of the two MEKK-binding sites and tested whether it could activate JNK. The result in Fig. 5 (5,16). We therefore asked whether MEKK1 and MEKK4 could mutually affect their binding to Axin. As dimerization-defective Axin is fully capable of binding to both MEKK1 and MEKK4 (Ref. 13 and Fig. 2B), we used HA-Axin C2 that lacks the DIX domain to test whether MEKK1 and -4 compete for binding to the same monomeric Axin molecules. The 293T cells were co-transfected with 1 g each of Axin C2 and MEKK1 together with increasing amounts of MEKK4-C (0, 0.5, 1.0, and 2.0 g). Results showed that in the presence of increasing amounts of MEKK4, MEKK1 binding to Axin was diminished and was completely abolished when co- Immunoprecipitation and immunoblotting were carried out as described above. C, kinase activity of MEKK4 is not required for its competition against MEKK1 for Axin binding. Cells were co-transfected with increasing amounts of Myc-MEKK4-C-K/M, 1 g of HA-Axin C2, and 1 g of MEKK1-C. Immunoprecipitation and Western blotting were carried out as described above. D, MEKK4 binding to Axin is indispensable for the competition of MEKK4 and MEKK1 in Axin binding. Cells were transfected with 1 g of HA-tagged Axin M8 that lacks the MEKK4-binding site and 1 g of MEKK1 together with increasing amounts of Myc-MEKK4-C. Immunoprecipitation and Western blotting were carried out as described above. transfected with 2.0 g of MEKK4-C (Fig. 6A). Similarly MEKK1 competed against MEKK4 binding to Axin, although MEKK4 seemed to have higher affinity than MEKK1 as cotransfection of 2.0 g of MEKK1 could not as effectively diminish MEKK4 binding as did MEKK4 to MEKK1 (Fig. 6B). These results indicate that MEKK1 and MEKK4 binding may change the conformation of Axin so that MEKK1 and MEKK4 cannot co-exist in the complex. We then tested whether the kinase activity of MEKK4 is required for its competition against MEKK1 in Axin binding. We co-transfected increasing amounts of the kinase-inactive form of MEKK4, MEKK4-K/M, with 1 g each of Axin C2 and MEKK1. Results showed that MEKK1 binding to Axin was gradually decreased in the presence of increasing amounts of MEKK4-K/M (Fig. 6C). However, MEKK4 requires its binding domain in Axin to compete against MEKK1 binding as it could not attenuate MEKK1 binding to Axin M8 that lacks the MEKK4-binding site (Fig. 6D).
Axin Activates JNK in the MEKK1 Ϫ/Ϫ Embryonic Fibroblast Cells-With the advent of MEKK4 as a new Axin-binding protein, we wished to test whether Axin could activate JNK in the embryonic fibroblast cells null for MEKK1, which were generated by gene targeting (25). MEKK4 was detected in the MEKK1 Ϫ/Ϫ cells, and as expected, MEKK1 was not expressed in the cells (Fig. 7B). Axin and FLAG-JNK1 were transfected into the MEKK1 Ϫ/Ϫ cells and into 293T cells as a comparison. As shown in Fig. 7A, Axin activated JNK in the MEKK1 Ϫ/Ϫ cells to a level comparable to that in 293T cells. This finding is particularly significant in that MEKK1 and MEKK4 in fact compete for binding to Axin.
Small Interfering RNA of MEKK4 Attenuates JNK Activation by MEKK1 Binding-defective Axin Mutant-As shown above, when the MEKK1-binding domain of Axin is removed (D19 and D20, Fig. 1), the Axin mutant could utilize MEKK4 to activate JNK. We reasoned that, in the absence of MEKK1, a decrease of the endogenous MEKK4 should result in reduced JNK activation by Axin. In particular, it would indicate whether MAP3Ks other than MEKK1 and MEKK4 could also play a role in the Axin-mediated JNK activation. The small interfering RNA technique has been utilized to successfully knock down a variety of genes in mammalian cells (26 -32). We therefore constructed a pSUPER vector that contains a human MEKK4specific 19-mer oligonucleotide to knock down the expression of MEKK4 in 293T cell (Fig. 8A) and tested whether it could diminish JNK activation. Axin D19 co-transfected with the blank pSUPER vector activated JNK as usual; however, when co-transfected with pSUPER-MEKK4, Axin D19 virtually lost its ability to activate JNK (Fig. 8B). As expected, pSUPER-MEKK4 only slightly reduced JNK activation by wild-type Axin (data not shown). We also co-transfected wild-type Axin and pSUPER-MEKK4 into MEKK1 Ϫ/Ϫ cells. The result showed that JNK activation of Axin was drastically reduced by specific knock-down of MEKK4 in MEKK1 Ϫ/Ϫ cells (Fig. 8C), confirming that MEKK4 can mediate Axin activation of JNK independently of MEKK1. DISCUSSION MAP kinases are instrumental in integrating numerous signals to biological processes including embryogenesis, cell differentiation, cell proliferation, and cell death (33)(34)(35)(36)(37). It is generally known that a MAP kinase cascade contains at least three components, namely MAP3K, MAPKK, and MAP kinase. However, relatively little is known about how MAP3Ks are activated by upstream signals. In the present study, through detailed mapping by creating a series of deletion mutants of Axin and utilizing MEKK1 Ϫ/Ϫ cells and RNA interference, we have demonstrated that MEKK4 also plays a role in Axininduced JNK activation. We have found several peculiar aspects in the Axin-mediated JNK activation. First of all, it was found that MEKK4 interacts with Axin on a distinct domain that is far apart from the site for MEKK1 binding despite the fact that MEKK1 and MEKK4 share significant similarity with each other, ϳ55% in the catalytic domain (38). Second, like MEKK1, the C-terminal region of MEKK4 containing its kinase domain is involved in Axin binding instead of its long N-terminal regulatory regions that have been known to interact with multiple proteins (38,39). Third, as previously shown, Axin-mediated JNK activation requires homodimerization via the DIX domain and sumoylation at the C terminus (23,24). Fourth, in addition to their respective binding domains, both MEKK1 and MEKK4 require an additional common domain, operationally termed domain X (Fig. 9) outside of their binding sites to trigger JNK activation.
It is unclear why the originally generated MEKK1 bindingdefective deletion mutant Axin⌬MID (M2 removing aa 210 -352) is inactive in JNK activation as tested on numerous occasions. Axin⌬MID is intact in binding to MEKK4, homodimerization, and sumoylation. It should be pointed out that it is unlikely that any other factor may bind to the regions flanking the MEKK1-binding site as other deletion mutants that combine to remove the region of aa 210 -352 are all capable of activating JNK (data not shown). In fact, larger deletions to both sides that flank the originally defined MID domain did not lose their ability to activate JNK. It is therefore likely that the original deletion Axin⌬MID M2 based on two convenient restriction sites happened to create a conformation that disallows Axin to activate JNK. This is especially true given that different Axin complex formations seem to be controlled by its conformational changes (16).
In addition to identification of MEKK4 as a new MAP3K mediating Axin activation of JNK, we found that MEKK1 and -4 are mutually exclusive in binding to Axin based on the following observations. First of all, using Axin ⌬C-250 that is not able to form dimmers with increasing concentrations of MEKK4, MEKK1 binding is gradually abolished from binding to the Axin protein. Conversely increasing amounts of MEKK1 abolished MEKK4 binding to Axin. These results indicate that MEKK1 and -4 could not bind to the same molecule of Axin. Furthermore, from co-immunoprecipitation assays for endogenous MEKK-Axin interactions, we found that anti-MEKK1 could not co-immunoprecipitate MEKK4 and vice versa. That anti-Axin could pull out both MEKK1 and MEKK4 is because the anti-Axin antibody could precipitate the whole pool of Axin-MEKK1/4 complexes. Moreover, although MEKK1 or MEKK4 alone seems to be sufficient for Axin-mediated JNK activation, introduction of either MEKK1-K/M or MEKK4-K/M alone blocked virtually the entire JNK activation by Axin in 293T cells. The data are also consistent with the notion that preoccupation of Axin by either MEKK prevents the other MEKK from binding to Axin. These observations that MEKK1 and MEKK4 bind to two distinct regions of Axin and that they are competitive in binding to Axin raise an interesting possibility that they may participate in different signaling pathways. Through targeted gene disruption experiments, MEKK1 is known to be required for JNK activation in response to microtubule disruption, viral infection, and double-stranded RNA (25,40,41). MEKK1 is also essential for induction of embryonic cell migration by serum factors (25,41). However MEKK1 is not required for tumor necrosis factor-␣ or interleukin-1 regulation of JNK or NF-B activation in macrophages or fibroblasts. Relatively little is known about MEKK4 except that it can also activate JNK (42). Nevertheless it is evident that MEKK1 and MEKK4 may have functional specificity as well as functional redundancy depending on cell types. MEKK1 is associated with the plasma membrane, while MEKK4 is localized to the perinuclear membrane and vesicular compartments (38,43). In addition, they contain diverse N-terminal regions and functional domains (38,44,45). The possibility that MEKK1 and -4 may have distinct functions is also supported by our observations that they compete for Axin binding and that in MEKK1-null cells Axin activated JNK equally well as in 293T cells. Furthermore 14-3-3 and 14-3-3⑀ were found to interact with MEKK1, -2, and -3 but not MEKK4 (46).
Another interesting feature of Axin in JNK activation is that there seems to be another domain critical for Axin-mediated FIG. 8. MEKK4 small interfering RNA attenuates JNK activation by MEKK1 binding-defective Axin mutant D19 and by wild-type Axin in MEKK1 ؊/؊ cells. A, pSUPER-MEKK4 effectively knocks down MEKK4 expression. The expression plasmid for MEKK4 small interfering RNA, pSUPER-MEKK4, was co-transfected with pCMV5-MEKK4-C into 293T cells to test for its efficiency to knock down MEKK4 expression. The total amount of DNA per 60-mm plate was 7 g using empty pSUPER vector where necessary. FLAG-JNK was included in each transfection as an internal control for similar transfection efficiency. B and C, pSU-PER-MEKK4 blocks JNK activation by mutant D19 (B) and by wild-type Axin in MEKK1-null cells (C). Cells were transfected with 2 g each of Axin constructs together with 5 g of pSUPER-MEKK4 followed by immunokinase assays as described above. IB, immunoblotting. JNK activation, which is located between the MEKK1-and MEKK4-binding sites. Removal of either the MEKK1-binding site or MEKK4-binding site did not affect JNK activation by Axin. However, removal of a small region encompassing aa 642-673, N-terminal to the MEKK4-binding site, completely abolished Axin activation of JNK, although these mutants are intact in MEKK1 or MEKK4 binding. As it was formally possible that removal of this novel domain could cause a conformational change that somehow renders Axin incapable of activating JNK as seen with Axin⌬MID, we created multiple deletion mutants of this region, including M10 and M11. M10, M11, and several other mutants not shown all failed to activate JNK even though they all bound to MEKK1 well. These data indicate that MEKK binding alone does not suffice to activate JNK. This is reminiscent of a requirement of the extreme C-terminal six amino acid residues that comprise two sumoylation sites as well as homodimerization via its DIX domain in Axin-mediated JNK activation. This observation adds a new complexity to the mechanism by which Axin activates JNK. It remains an interesting issue as to how and why Axin-mediated JNK activation requires this region. Based on our current and previous findings, we summarize domain requirements for the Axin/JNK pathway in Fig. 9. Specifically either MEKK1-or MEKK4-binding domain combines with the newly defined domain X (aa 642-673), DIX homodimerization domain, and the C-terminal sumoylation motifs to activate JNK. Implied in the model is also that MEKK1-and MEKK4-binding sites each independently combine with the common domains for JNK activation. In support of this notion are the observations that, when either the MEKK1-or MEKK4-binding site is removed, Axin is fully capable of activating JNK, whereas removal of any of the other domains abolishes JNK activation. How these domains coordinate with one another in the Axin/JNK pathway requires further investigation.
A number of scaffold proteins for different MAP3Ks have been identified, including Ste5p, JIPs, and MIP-1 (47)(48)(49). They seem to associate with all three different components in the trikinase cascades. Ste5p simultaneously interacts with the MAP3K Ste11p, MAPKK Ste7p, and MAPK Fus3p, whereas JIP assembles a MAPK module formed by the MAP3K Mlk3, MKK7, and JNK (44). Axin seems to differ from these scaffold proteins in that it does not interact with either MKK4/7 or JNK directly (13). Moreover, from the new interacting factors we have pulled out from yeast two-hybrid screens, Axin is found to interact with membrane-associated proteins that may in fact be linked to receptors. 2 It is exciting to note that Axin may directly receive signals from membrane-localized receptors and in turn transmit signals to MEKK1 or MEKK4 for subsequent JNK activation.