The MAP kinase kinase kinase MLK2 co‐localizes with activated JNK along microtubules and associates with kinesin superfamily motor KIF3

Koh‐ichi Nagata, Axel Puls, Clare Futter, Pontus Aspenstrom, Erik Schaefer, Takao Nakata, Nobutaka Hirokawa, Alan Hall

Author Affiliations

  1. Koh‐ichi Nagata2,
  2. Axel Puls2,
  3. Clare Futter1,
  4. Pontus Aspenstrom3,
  5. Erik Schaefer4,
  6. Takao Nakata5,
  7. Nobutaka Hirokawa5 and
  8. Alan Hall*,2,6
  1. 1 MRC Laboratory for Molecular Cell Biology, University College London, Gower Street, London, WC1E 6BT, GB
  2. 2 CRC Oncogene and Signal Transduction Group, University College London, Gower Street, London, WC1E 6BT, GB
  3. 3 Ludwig Institute of Cancer Research, Uppsala University, S‐751 24, Uppsala, Sweden
  4. 4 Signal Transduction Group, Promega Corporation, 2800 Woods Hollow Road, Madison, WI, 53711, USA
  5. 5 Department of Cell Biology and Anatomy, Graduate School of Medicine, University of Tokyo, Hongo, Tokyo, 113, Japan
  6. 6 Department of Biochemistry, University College London, Gower Street, London, WC1E 6BT, GB
  1. *Corresponding author. E-mail: Alan.Hall{at}


The MLK (mixed lineage) ser/thr kinases are most closely related to the MAP kinase kinase kinase family. In addition to a kinase domain, MLK1, MLK2 and MLK3 each contain an SH3 domain, a leucine zipper domain and a potential Rac/Cdc42 GTPase‐binding (CRIB) motif. The C‐terminal regions of the proteins are essentially unrelated. Using yeast two‐hybrid analysis and in vitro dot‐blots, we show that MLK2 and MLK3 interact with the activated (GTP‐bound) forms of Rac and Cdc42, with a slight preference for Rac. Transfection of MLK2 into COS cells leads to strong and constitutive activation of the JNK (c‐Jun N–terminal kinase) MAP kinase cascade, but also to activation of ERK (extracellular signal‐regulated kinase) and p38. When expressed in fibroblasts, MLK2 co‐localizes with active, dually phosphorylated JNK1/2 to punctate structures along microtubules. In an attempt to identify proteins that affect the activity and localization of MLK2, we have screened a yeast two‐hybrid cDNA library. MLK2 and MLK3 interact with members of the KIF3 family of kinesin superfamily motor proteins and with KAP3A, the putative targeting component of KIF3 motor complexes, suggesting a potential link between stress activation and motor protein function.


Activation of MAP kinases following ligand binding to cell surface receptors has been correlated with numerous cellular responses such as cell growth, differentiation and the regulation of metabolic pathways (reviewed in Marshall, 1995; Waskiewicz and Cooper, 1995; Kyriakis and Avruch, 1996; Robinson and Cobb, 1997). To date, three MAP kinase cascades have been well characterized in mammalian cells; ERK (extracellular signal‐regulated kinase), JNK (c‐Jun N‐terminal kinase or stress‐activated protein kinase, SAPK) and p38 (also known as cytokine‐suppressive anti‐inflammatory drug‐binding protein, CSBP, or reactivating kinase, RK). Activation of these MAP kinases (MAPKs) occurs through a cascade of upstream kinases; a MAP kinase kinase kinase (MAPKKK) first phosphorylates a dual‐specificity protein kinase (MAPKK), which in turn phoshorylates the MAPK. In the ERK cascade, c‐Raf, MEK and ERK correspond to MAPKKK, MAPKK and MAP kinase, respectively. c‐Raf activation, in response to growth factor stimulation of cells, is regulated by the Ras GTPase which induces translocation of c‐Raf to the plasma membrane, where it is thought to be activated (Leevers et al., 1994; Stokoe et al., 1994; Marais et al., 1995).

JNK and p38 are strongly activated in cells exposed to extracellular stress such as UV irradiation or osmotic shock (Galcheva‐Gargova et al., 1994) and by proinflammatory cytokines such as tumour necrosis factor α (TNFα) and interleukin‐1 (IL‐1) (Kyriakis and Avruch, 1996). Accumulating evidence suggests that these MAP kinase cascades are regulated by small GTPases of the Rho family (Coso et al., 1995; Minden et al., 1995; Olson et al., 1995). In contrast to Ras in the ERK cascade, however, the downstream targets of these GTPases and the nature of the intervening kinase cascades are far from clear. Numerous MAPKKK‐like proteins have been identified that can activate the JNK and p38 pathways, including ASK1 (Ichijo et al., 1997), Tpl‐2 (Salmeron et al., 1996), DLK (also known as MUK, Hirai et al., 1996), TAK1 (Moriguchi et al., 1996) and MEKK1, 2, 3 and 4 (Fanger et al., 1997; Gerwins et al., 1997). MEKK1 and MEKK4 can both interact directly with Cdc42 and Rac, while another Cdc42/Rac target, p65PAK, belongs to the family of MAP kinase kinase kinase kinases (MAPKKKK) and has also been reported to participate in JNK pathways (Manser et al., 1994; Bagrodia et al., 1995; Zhang et al., 1995; Brown et al., 1996; Joneson et al., 1996; Lamarche et al., 1996). Other possible mediators of JNK activation by Rac and Cdc42 are members of the mixed lineage kinase (MLK) family (Burbelo et al., 1995). The three known MLKs (MLK1, MLK2 and MLK3) each contain a potential binding site (CRIB motif) for Cdc42 and Rac, and their kinase domains most closely resemble MAPKKKs. It was reported recently that MLK3 activates both the JNK and p38 MAP kinase pathways when transfected into cells (Rana et al., 1996; Teramoto et al., 1996; Tibbles et al., 1996).

The three MLK kinases show ∼70–75% identity in their kinase domain and in their SH3 and leucine zipper domains (Katoh et al., 1995), suggesting strong functional conservation. Despite this, the C‐terminal regions of the proteins are highly divergent. We show here that MLK2 and MLK3 interact strongly with Cdc42 and Rac in a GTP‐dependent manner. Transfection of MLK2 into COS cells leads to strong and constitutive activation of the JNK pathway, as well as activation of ERK and p38 MAP kinases, but this is unaffected by co‐transfection with either constitutively active or dominant‐negative Rac or Cdc42. MLK2 co‐localizes with dually phosphorylated JNK to punctate structures associated with microtubules. We have identified several cellular proteins that interact with MLK2, including two members of the KIF3 family of microtubule motor proteins (Hirokawa, 1996) and KAP3, the putative targeting component of the KIF3 motor complex (Yamazaki et al., 1996).


MLKs are targets for Cdc42 and Rac

It has been reported previously that MLK3 contains a potential site (CRIB motif) of interaction for Cdc42 and Rac (Burbelo et al., 1995). To determine whether MLKs interact with Cdc42 or Rac and to compare their interaction with that of other known GTPase targets, we first made use of a yeast two‐hybrid analysis. The C‐terminal region of MLK2 (amino acids 338–953) or MLK3 (amino acids 348–847) (see Figure 1) fused to a GAL4 activation domain was introduced into yeast strains containing various GTPases fused to the GAL DNA‐binding domain. It can be seen in Figure 2A that MLK2 and MLK3 interact strongly with the GTP forms of both Cdc42 and Rac (L61 versions), but not with the GDP forms (N17 versions). Neither MLK interacts with Rho.

Figure 1.

Structure of MLKs. (A) Structure of MLK1, 2 and 3. The amino acid sequence identity between MLK1, MLK2 and MLK3 is shown. Structural domains of the proteins are abbreviated as follows: SH3, Src homology 3; Leu Z, leucine zipper; CRIB, Cdc42/Rac interactive binding motif. (B) Description of the MLK2 cDNA fragments constructed into expression vectors.

Figure 2.

Interaction of Rac and Cdc42 with MLK2 and MLK3. (A) The interaction of GAL4AD–MLK2 (amino acids 338–953) or GAL4AD–MLK3 (amino acids 348–847) fusion proteins with GAL4DB‐GTPase constructs in yeast. Four independent colonies from co‐transformants were patched on medium lacking histidine, tryptophan and leucine, but with 3‐aminotriazole, and analysed for growth. (B) Interaction of MLK2 with Rac or Cdc42 in dot‐blot analysis. GST protein (20 μg), GST–p65PAK (10 μg) and GST–MLK2 (Leu/CRIB/Div) (10 μg) were spotted onto a nitrocellulose filter in a volume of 10 μl and incubated with 0.5 μg of [γ‐32P]GTP‐loaded L61Rac or L61Cdc42. GTPase interactions were visualized by autoradiography.

To confirm this interaction in vitro, we have used a dot‐blot assay. Figure 2B shows that the MLK2‐Leu/CRIB/Div (amino acids 338–953, lacking the kinase domain) interacts strongly with the GTP‐bound form of Rac and Cdc42, with a slight preference for Rac. p65PAK, on the other hand, shows a slight preference for Cdc42.

Activation of MAP kinase pathways by MLK2

Rac and Cdc42 activate both the JNK and the p38 MAP kinase pathways in transfected cells (Bagrodia et al., 1995; Coso et al., 1995; Minden et al., 1995; Olson et al., 1995; Zhang et al., 1995). To determine whether MLK2 can activate MAP kinase cascades, MLK2 in the pRK5 vector was co‐transfected into COS cells with either ERK, JNK or p38 constructs. The three MAP kinases were immunoprecipitated and their kinase activities determined in an in vitro assay. As shown in Figure 3A, MLK2 strongly activates JNK and the activation is comparable with that induced by MEKK. As with MEKK, activation of JNK by MLK2 is constitutive in transfected cells. Using various amounts of MLK2 DNA, no significant inhibitory or stimulatory effect of co‐transfecting with either constitutively active or dominant‐negative versions of Cdc42 or Rac can be seen (data not shown). MLK2 causes moderate (11‐fold; Figure 3B) activation of co‐transfected ERK2 and p38 (8‐fold; data not shown).

Figure 3.

Activation of JNK, ERK and p38 MAP kinase cascades by MLK2. (A) MLK2 activates the JNK1 pathway. FLAG‐tagged JNK was expressed in COS‐1 cells either alone or with Myc‐tagged constructs for V12Rac, V12Cdc42, MLK2 (full‐length) or MEKK (C‐terminal domain) (Olson et al., 1995). JNK activity was assessed by immunocomplex kinase assays using c‐Jun as a substrate and revealed by autoradiography. The amount of JNK and of co‐transfected proteins, as determined by Western blotting using a mixture of M2 and 9E10 antibodies, is shown in the lower panel. (B) Activation of ERK2 kinase by MLK2. COS‐1 cells were transfected with vector control or pRK5‐Myc‐MLK2 together with pEXV‐Myc‐ERK2. pEXV‐Myc‐D12Ras was used as a positive control. MLK2 did not phosphorylate MBP directly in the conditions used. The levels of MLK2, ERK2 and Ras, visualized on Western blots using 9E10 antibody, are shown in the lower two panels. (C) Quiescent, serum‐starved Swiss 3T3 cells were fixed 2–3 h after injection of Myc‐tagged L61H‐Ras (a and d), MLK2 (b and e) and L61Rac (c and f). Cells expressing Myc‐tagged proteins were visualized by indirect immunofluorescence with 9E10 antibody (a–c). Activated dually phosphorylated ERK was visualized with anti‐active ERK antibody (Promega) (d–f). Three non‐injected, neighbouring cells can be seen in the bottom right corner of (b) and (e). Scale bar represents 30 μm.

To exclude the possibility that MLK2‐induced ERK activation might be indirect and caused by autocrine effects, cDNA was injected into serum‐starved quiescent Swiss 3T3 cells and activation of ERK visualized using an antibody specific for dually phosphorylated ERK. As shown in Figure 3C, MLK2 activates ERK only in injected cells and not in neighbouring cells, confirming that the effects are cell autonomous. We conclude that MLK2, expressed in COS‐1 cells, is a potent activator of JNK, but also activates the p38 and ERK pathways.

Microinjection of MLK2 into Swiss 3T3 cells

In addition to their effects on MAP kinase pathways, Rac and Cdc42 are important regulators of actin polymerization and the assembly of associated integrin complexes (Ridley et al., 1992; Kozma et al., 1995; Nobes and Hall, 1995). Microinjection of quiescent Swiss 3T3 cells with pRK5‐Myc‐MLK2, however, induced no detectable changes in the actin cytoskeleton (data not shown). Rac and Cdc42 have also been shown to stimulate G1 progression and entry into DNA synthesis when injected into quiescent Swiss cells. (Olson et al., 1995). To examine whether MLK2 can stimulate G1 progression, quiescent Swiss 3T3 cells were microinjected with pRK5‐Myc‐MLK2, and 6 h later ∼98% of injected cells strongly expressed Myc‐MLK2 (data not shown). However 20 h after injection, we were unable to assess the effects of MLK2 on DNA synthesis since most injected cells were dead (data not shown). Further analysis revealed that even when the experiment was repeated in the presence of 10% fetal calf serum (FCS), MLK2 induced cell death in >90% of injected cells by 20 h (data not shown). Time lapse video analysis revealed that cell death occurred with cytoplasmic shrinkage and nuclear condensation, characteristics of cells undergoing apoptosis. This was confirmed by incubating microinjected cells in the presence of the protease inhibitor Z‐Val‐Ala‐Asp(O‐Me)‐fluoromethylketone (100 μM), which prevented apoptosis when added in the medium just after microinjection (data not shown).

Immunostaining of MLK2 expressed in Swiss 3T3 cells

To determine the intracellular localization of MLK2, a Myc‐tagged expression construct was microinjected into the nuclei of Swiss 3T3 cells. As seen in Figure 4A, 2–3 h later MLK2 can be seen in punctate structures that appear to be organized radially within the cells. Co‐staining with an antibody to tubulin revealed that these structures were located along microtubules (Figure 4B). Further analysis using confocal laser microscopy (Figure 4C) confirmed that under these conditions the majority of MLK2 is found in a punctate distribution intimately associated with microtubules.

Figure 4.

Localization of MLK2 in Swiss 3T3 cells. Swiss 3T3 cells were fixed 2–3 h after injection of nuclei with pRK5‐Myc MLK2 (0.1 mg/ml). Myc‐MLK2 (A) and microtubules (B) were visualized by indirect immunofluoresence. (C) Confocal micrograph of a cell double‐labelled for microtubules (red) and Myc‐MLK2 (green). Scale bar represents 30 μm.

To determine whether these structures might have functional significance, the cellular distribution of endogenous, activated JNK in the MLK2‐injected cells was localized using an antibody that recognizes the dually phosphorylated active form of the JNK1 and 2 enzymes. As shown in Figure 5B, activated endogenous JNK partially co‐localizes with MLK2 (Figure 5A) to punctate structures along microtubules. In addition, activated JNK can also be found in the nucleus. MLK2‐Leu/CRIB/Div (lacking the kinase domain) also localized to these punctate structures (Figure 5C), but did not stimulate JNK activity (Figure 5D).

Figure 5.

Co‐localization of activated JNK with MLK2. Quiescent, serum‐starved Swiss 3T3 cells were fixed 2–3 h after injection of Myc‐tagged MLK2 (A and B) and MLK2‐Leu/CRIB/Div (C and D). The subcellular localization of expressed Myc‐tagged proteins was visualized by indirect immunofluorescence with 9E10 antibody (A and C). Endogenous dually phosphorylated JNK was visualized using anti–active JNK antibody (B and D). (E) FLAG‐tagged JNK was co–transfected into COS cells with pRK5 or pRK5‐Myc‐MLK2. Western blotting of the cell lysate was conducted with anti‐active JNK antibody (top panel) or M2 antibody (bottom). Scale bar represents 30 μm.

Two‐hybrid screen for MLK2‐interacting proteins

Since transfected MLK2 is constitutively active and is localized along microtubules, we have attempted to identify cellular proteins that might regulate this behaviour. A yeast two‐hybrid, human brain, cDNA library was screened with MLK2‐Leu/CRIB/Div (lacking the kinase domain, see Figure 1B). Thirteen positive clones, which showed a strong positive interaction with MLK2, were identified. Partial cDNA sequence analysis revealed that six clones corresponded to hippocalcin, five clones encoded 14‐3‐3ϵ and one clone encoded the C‐terminal region (starting at amino acid 551) of KAP3, a putative targeting molecule in the KIF3 microtubule motor complex (Yamazaki et al., 1996). The final clone encoded a novel cDNA which showed significant sequence similarity to the carboxy‐terminus of the Caenorhabditis elegans kinesin superfamily member, OSM‐3 (Tabish et al., 1995), and mammalian kinesin superfamily proteins, KIF3A and KIF3B (Aizawa et al., 1992; Yamazaki et al., 1995; Hirokawa, 1996). The DNA sequence of the yeast two‐hybrid clone (accession No. AF009624) corresponds to the carboxy‐terminal 233 amino acids of a novel member of the KIF3 family, which we tentatively name KIF3X. The deduced partial amino acid sequence of KIF3X shows ∼26% identity and 35% similarity to C.elegans OSM‐3 and 17% identity and 23% similarity to mammalian KIF3A (Figure 6A).

Figure 6.

MLK2 interacts with the KIF3 family of kinesin superfamily proteins. (A) Comparison of the KIF3X partial amino acid sequence with those of KIF3A and OSM‐3. The predicted rod α‐helix domain is underlined. Identical and conserved amino acids are denoted by boxes and shaded boxes, respectively. (B) The interaction of GAL4DB–MLK2 (Div) and GAL4DB–MLK3 (amino acids 348–847) with GAL4AD–KIF3X, KIF3A (amino acids 531–701), KIF3B (amino acids 524–747) and KAP3 (C‐terminal region from amino acid 511 obtained by yeast two‐hybrid screening).

To identify the binding site of hippocalcin, 14‐3‐3ϵ, KIF3X and KAP3 on MLK2, fragments of MLK2 were used in the two‐hybrid assay. KIF3X, KAP3 (Figure 6B), 14‐3‐3ϵ and hippocalcin (data not shown) interact only with the C‐terminal divergent region of MLK2. MLK3 also interacts with these four proteins (Figure 6B and data not shown). Since KIF3X appears to belong to the KIF3 subfamily of the kinesin superfamily of motor proteins, we wished to determine whether other members of the family could also interact with MLKs. The C‐terminal fragments of KIF3A (amino acids 531–701) and KIF3B (amino acids 524–747) were obtained by PCR from a mouse brain cDNA library. Figure 6B shows that MLK2 and MLK3 interact with KIF3A (more strongly in fact than with KIF3X), whereas KIF3B interacts weakly with MLK3 and not with MLK2.

Complex formation between MLK2 and hippocalcin, 14‐3‐3ϵ, KIF3A and KAP3A

To identify whether MLK2 can form complexes in mammalian cells with the proteins identified in the two‐hybrid screen, expression vectors containing FLAG‐hippocalcin, FLAG‐14‐3‐3ϵ, FLAG‐KAP3A and FLAG‐KIF3A were co‐transfected with Myc‐tagged MLK2 into COS cells. For these experiments, we have used cDNAs encoding full‐length versions of KIF3A and KAP3A (Aizawa et al., 1992; Yamazaki et al., 1995, 1996). Figure 7 shows that full‐length KAP3A, KIF3A, 14‐3‐3ϵ and hippocalcin can be co‐precipitated with MLK2 under these conditions. To determine whether any of the interacting proteins can affect MLK2 kinase activity, they were each co‐transfected into COS cells with MLK2 and JNK. In each case, there was no detectable inhibiton or enhancement of MLK2‐induced activation of JNK (data not shown).

Figure 7.

Interaction of MLK2 with KAP3A, KIF3A, 14‐3‐3ϵ and hippocalcin in COS cells. Co‐immunoprecipitation of Myc‐MLK2 with: (A) FLAG‐tagged, full‐length KAP3A; (B) FLAG‐tagged full–length KIF3A; (C) FLAG‐tagged 14‐3‐3ϵ; and (D) FLAG‐tagged hippocalcin. Lysates (∼90 μg of protein) from transfected COS‐1 cells were immunoprecipitated (IP) with M2 (FLAG) or 9E10 (MYC) antibody as indicated. Western blotting (WB) was carried out with a mixture of 9E10 and M2 antibodies. In (A) and (B), the first panel is a Western blot showing total protein levels in transfected cells and the second panel is a Western blot of immunoprecipitated material.

To determine the intracellular location of MLK2‐interacting proteins, the corresponding expression vectors were injected into the nuclei of Swiss 3T3 cells. 14‐3‐3ϵ and hippocalcin were found diffusely distributed throughout the cytoplasm (data not shown). KAP3A also appeared diffusely distributed (Figure 8A) as previously reported (Yamakazi et al., 1996), but KIF3A localized to punctate structures along microtubules (Figure 8B) indistinguishably from MLK2 (Figure 5A) and activated JNK (Figure 5B).

Figure 8.

Localization of KAP3A and KIF3A in Swiss 3T3 cells. Swiss 3T3 cells were fixed 2–3 h after injection of nuclei with 0.1 mg/ml each of (A) pRK5‐FLAG‐KAP3A (full‐length) and (B) pRK5‐FLAG‐KIF3A (full‐length). Expressed proteins were visualized by indirect immunofluoresence. Scale bar represents 30 μm.


In addition to their effects on the actin cytoskeleton, Rac and Cdc42 activate the JNK and p38 MAP kinase cascades when transfected into cells (Coso et al., 1995; Minden et al., 1995; Olson et al., 1995). Several groups have reported that p65PAK, a ser/thr kinase which is a target for both Rac and Cdc42, might mediate JNK activation, while others have reported that it does not (Bagrodia et al., 1995; Zhang et al., 1995; Brown et al., 1996; Lamarche et al., 1996; Joneson et al., 1996; Westwick et al., 1997). In a database search for additional Cdc42/Rac target proteins, the MLK family of ser/thr kinases was identified as having a potential GTPase‐interacting (CRIB) motif (Burbelo et al., 1995). The kinase domain of this family is most closely related to the MAP kinase kinase kinase family, and it has been reported recently that MLK3 can activate the JNK MAP kinase pathway, raising the possibilty that MLKs mediate Cdc42/Rac activation of MAP kinases (Rana et al., 1996; Teramoto et al., 1996; Tibbles et al., 1996).

We show here that MLK2 and MLK3 (through their CRIB motifs) interact strongly with both Rac and Cdc42 in a GTP‐dependent manner, suggesting that they are cellular targets of these GTPases. As previously reported for MLK3 (Tibbles et al., 1996), we could not observe complex formation of MLK2 with active Cdc42 or Rac, perhaps due to the transient nature of the interaction between these molecules. When transfected into COS cells with reporter plasmids, MLK2 is a potent activator of the JNK kinase cascade, though it also activates the ERK and p38 pathways. It appears, therefore, that MLK2 belongs to a growing number of MAPKKKs that can activate the JNK pathway, including MLK3, ASK1, Tpl‐2, DLK, TAK1 and MEKK1–4 (Hirai et al., 1996; Moriguchi et al., 1996; Salmeron et al., 1996; Fanger et al., 1997; Gerwins et al., 1997; Ichijo et al., 1997).

The mechanism by which MAPKKKs are activated is unclear. In addition to MLKs, MEKKs 1 and 4 (Fanger et al., 1997; Gerwins et al., 1997) have also been shown to contain a Cdc42/Rac interaction site. It has been reported by one group that MLK3 activity can be modified by co‐expression of Cdc42 (Teramoto et al., 1996), but in our hands MLK2 transfected into COS‐1 (this study), NIH‐3T3 or PC12 cells (data not shown) is constitutively active, and co‐transfection with activated or dominant‐negative versions of Cdc42 or Rac has no effect on its activity. The significance of the GTPase interaction site on MLKs is, therefore, unclear. Cdc42 has been reported to play a role in the regulation of MAP kinase pathways in yeast, though recent work has suggested that its role may not be in activation per se, but in localizing target kinases to distinct intracellular locations (Peter et al., 1996). Perhaps the role of mammalian Cdc42 and Rac is to localize MLKs to discrete cellular compartments or molecular complexes.

Since MLK2 and MLK3 are constitutively active after transfection, it seems likely that the endogenous MLKs are sequestered by cellular proteins. It was interesting to find, therefore, that MLK2 localizes to punctate structures along microtubules. The observation that dually phosphorylated JNK co‐localizes to the same structures, as well as in the nucleus, suggests that they are functionally significant. Interestingly, it has been shown recently that three components of the hedgehog signalling pathway, fused (a ser/thr kinase), cubitus interruptus (a transcription factor) and costal2 are sequestered in a complex along microtubules and released upon activation of the pathway (Robbins et al., 1997). The mechanism of microtubule attachment is through costal2, which is a kinesin‐related protein, though there is no evidence that any motor‐like activity is required.

To find proteins that might regulate the activity or localization of MLK2, we performed a yeast two‐hybrid screen and identified four proteins that interact with the divergent, C‐terminal region of the molecule. Two of these are 14‐3‐3ϵ and hippocalcin. The 14‐3‐3 family of proteins have received much attention since they associate with numerous signalling proteins including c‐Raf‐1 and the phosphatase cdc25 (reviewed in Aitken, 1996). The functional significance of these interactions is still unclear, though recent genetic analysis of the Ras/Raf pathway in Drosophila suggests that 14‐3‐3 is an essential component (Chang and Rubin, 1997; Kockel et al., 1997). It has been shown that 14‐3‐3 proteins interact with a phosphorylated consensus sequence, RGLpSPP; this motif is found in MLK2 (which interacts with 14‐3‐3), but not in MLK3 (which we find does not interact with 14‐3‐3, data not shown). However, we were unable to find any effect of 14‐3‐3 expression on MLK2 activity and, as expected since it interacts with a wide range of proteins, 14‐3‐3 showed a diffuse cytoplasmic localization pattern.

Hippocalcin is a 23 kDa Ca2+‐binding protein with three EF‐hand structures and is found predominantly in hippocampus (Kobayashi et al., 1992). It has a striking sequence homology to the recoverin family, members of which serve to regulate photo‐signal transduction systems (Dizhoor et al., 1991). It is possible that hippocalcin serves as an anchor for the specific intracellular localization of MLK2, since hippocalcin translocates to membranes in Ca2+‐dependent manner (Kobayashi et al., 1993). However, as with 14‐3‐3, hippocalcin showed a diffuse cytoplasmic location and co‐expression had no effect on MLK2 activity.

Two additional MLK2‐interacting clones were isolated in the two‐hybrid screen. The first, KAP3, associates with the non‐motor, tail domain of the kinesin superfamily microtubule motor protein, KIF3 (Yamazaki et al., 1996). The other interacting clone encodes a novel member of the KIF3 family of kinesin superfamily of motor proteins, tentatively referred to as KIF3X. Further analysis revealed that KIF3 itself also interacts with MLK2 and MLK3. Kinesin is a motor protein that transports vesicles along microtubules towards their plus end (Brady, 1995). It is composed of two heavy chains and two light chains that form a globular (ATP/microtubule‐binding) head domain, a stalk domain and a tail globular domain, which is thought to associate with membranous organelles via an associated protein, kinectin (Burkhardt, 1996). Interestingly, Rho, Rac and Cdc42 have each been found to associate with kinectin in a yeast two‐hybrid analysis, suggesting that these GTPases might influence the kinesin motor (Hotta et al., 1996). It is now clear that kinesin is one member of a large superfamily of microtubule motor proteins, referred to as KIFs (Hirokawa, 1996). KIF3A and KIF3B are two members of the kinesin superfamily that form a heterodimeric plus end‐directed motor. They are highly expressed in neurons and testis, but are expressed ubiquitously at lower levels (Aizawa et al., 1992; Yamazaki et al., 1995). Although the proteins contained in vesicles transported by KIF3A/KIF3B motors have not been identified, vesicular recognition is thought to be modulated by a KIF3 tail‐interacting protein, KAP3 (Yamazaki et al., 1996).

It appears, therefore, that MLK2 potentially can associate with at least two components of KIF motor complexes; the motor molecule itself (KIF3A and KIF3X) and the putative cargo recognition molecule (KAP3). When KAP3 was expressed in Swiss cells it localized in a diffuse cytoplasmic pattern, as previously reported (Yamakazi et al., 1996). KIF3A, however, localized predominantly in a similar punctate pattern to MLK2 and activated JNK, along microtubules. The nature of these structures is unclear at the moment; we have no evidence that they are active motor complexes.

The data reported here indicate that MLKs can associate with components of KIF3 motors and that they can activate JNK and to a lesser extent ERK/p38 cascades. This raises the intriguing possibility that microtubule motors and MAP kinase cascades are co‐ordinately controlled. One possibility is that motor components can be phosphorylated by active MLKs or JNKs in response to stress, and it is interesting that the closest known homologue to KIF3X (the new kinesin family member identified in our two‐hybrid screen) is the C.elegans protein OSM‐3. Mutants in this locus fail to avoid high osmotic concentrations of salts and sugar, conditions that are a classic stimulus for the JNK/p38 MAP kinase pathways in mammalian cells as well as in yeast (Kyriakis and Avruch, 1996). We would like to suggest the possibilty that MAP kinase pathways and the activity of microtubule motor complexes can be co‐ordinately regulated; whether MLKs are delivered by the motor complex to a site of JNK MAP kinase activation, or whether the activity of the motor complex is regulated by a MLK/Cdc42/Rac interaction is now open to investigation.

Materials and methods


The following constructs were kind gifts from colleagues; Dr M.Terada (National Cancer Center Research Institute, Tokyo), human MLK2; Dr C.J.Marshall (Chester Beatty Laboratories, London), pEXV‐Myc‐ERK2 and pEXV‐Myc‐D12Ras; Drs M.Karin (University of California) and J.Ham (Eisai London Research Laboratories), pCMV‐FLAG‐JNK1 and pGEX‐c‐Jun; Dr L.Zon (Dana Farber Cancer Institute, Boston), dominant–negative pMT2‐HA‐SEK‐1AL(S220A, T224L); Dr M.H.Cobb (University of Texas), pCEF‐HA‐p38; Dr B.Shamoon (Chiron Corp., CA), human brain cDNA library fused to the GAL4 activation domain (GAL4AD) in pACTII; Dr S.Moss (University College London), anti‐Myc antibody (9E10). Anti‐FLAG‐tag antibody (M2) and anti‐HA‐tag antibody (12CA5) were purchased from Kodak Inc. and Boehringer Inc., respectively. Anti‐tubulin antibody (rat) was from Dr J.Kilmartin (MRC Laboratory of Molecular Biology, Cambridge). Anti‐active ERK antibody, recognizing dual‐phosphorylated ERK1 and ERK2 (corresponding to T183 and Y185 of ERK2), and anti‐active JNK antibody, recognizing dual‐phosphorylated JNK1 and JNK2 (corresponding to T183 and Y185), are multistep affinity‐purified reagents from Promega Corp. (Madison, WI) (Khokhlatchev et al., 1997). 14‐3‐3ϵ, hippocalcin, mouse KAP3A, mouse KIF3A and KIF3X (C‐terminal region obtained in two‐hybrid screen) were subcloned into the pRK5 vector containing the FLAG‐tag. All constructs were verified by DNA sequencing. Human MLK3 (amino acids 348–847) was obtained by yeast two‐hybrid screening with an Epstein–Barr virus‐transformed B‐cell library and L61Cdc42 as bait.

Yeast two‐hybrid analysis

cDNAs encoding various GTPases were fused to the GAL4 DNA‐binding domain (GAL4DB) in the pYTH6 vector, and constructs were integrated into the genome of Saccharomyces cerevisiae, strain Y190. MLK2 (amino acids 338–953), MLK3 (amino acids 348–847) and RhoGAP were fused to the GAL4AD in the pACTII vector (Aspenstrom and Olson, 1995). After transformation of pACTII constructs into yeast strains containing integrated pYTH6‐GTPase constucts, cells were spread on medium lacking tryptophan, leucine and histidine but supplemented with 25 mM 3‐aminotriazole. Cells containing interacting fusion proteins appeared on this plate after 3 days. LacZ expression was assayed by the membrane transfer method (Aspenstrom and Olson, 1995).

To screen for MLK2‐interacting proteins, Y190:pYTH6‐MLK2 (amino acids 338–953) was transformed with a human brain cDNA library in pACTII and 1.2×106 transformants were screened essentially as described before (Aspenstrom and Olson, 1995). Plasmid DNA recovered from 13 positive clones retained the ability to induce the reporter gene when transformed into Y190:pYTH6‐MLK2. The obtained partial cDNA of KIF3X was sequenced by dideoxynucleotide chain termination methods.

Expression and purification of recombinant proteins

GTPases were expressed in Escherichia coli as glutathione S‐transferase (GST) fusion proteins and purified on glutathione–Sepharose beads. The recombinant proteins were released from the beads by cleavage with human thrombin as described (Lamarche et al., 1996). MLK2 [amino acids 338–953 for Leu‐zipper, CRIB motif and divergent region (MLK2‐Leu/CRIB/Div); amino acids 341–451 for Leu‐zipper (MLK2‐Leu); amino acids 338–539 for Leu‐zipper and CRIB motif (MLK2‐Leu/CRIB) and amino acids 497–953 for divergent region (MLK2‐Div)], p50rhoGAP (amino acids 198–439), full‐length p65PAK, c‐Jun and ATF–2 were also expressed as GST fusion proteins in E.coli. These fusion proteins were eluted from the beads with 5 mM reduced glutathione (Sigma Inc.). Purified proteins were dialysed against 15 mM Tris–HCl pH 7.5, 150 mM NaCl, 5 mM MgCl2, 0.1 mM dithiothreitol (DTT) and concentrated by ultrafiltration using Centricon‐10 (Amicon Inc.). Protein concentration was determined by the method of Bradford, and the purity of protein preparations was visualized on Coomassie blue‐stained SDS–polyacrylamide gels.

Dot‐blot assay

The interaction of target proteins with Rac and Cdc42 was assessed using a dot‐blot assay as previously described (Diekmann et al., 1995). Briefly, the indicated amounts of various GST–MLK2 fusion protein fragments (MLK2‐Leu/CRIB/Div, MLK2‐Leu/CRIB, MLK2‐Leu and MLK2‐Div) or p65PAK were spotted onto nitrocellulose membranes and the filter air dried and incubated with blocking buffer (1 M glycine, 5% milk powder, 1% ovalbumin and 5% FCS) for 2 h at room temperature. The membrane was washed in buffer A (50 mM Tris–HCl pH 7.5, 100 mM NaCl, 5 mM MgCl2 and 1 mM DTT) and incubated for 5 min at 4°C with [γ‐32P]GTP‐bound L61Rac or L61Cdc42 in buffer A. The filters were washed quickly three times with cold buffer A containing 0.1% Tween‐20, and interacting GTPases were visualized by autoradiography.

COS‐1 cell transfection

COS‐1 cells were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% FCS and antibiotics and were transfected by the DEAE–dextran method (Olson et al., 1995). Plasmid amounts per 10 cm Petri dish were as follows: 5 μg of pCMV‐FLAG‐JNK1 with or without 3 μg each of pRK5‐Myc, pRK5‐Myc‐L61Rac, pRK5‐Myc‐L61Cdc42, pMT‐Myc‐MEKK‐C terminal domain or pRK5‐Myc‐MLK2; 5 μg of pEXV‐Myc‐ERK2 with or without 3 μg each of pEXV‐Myc, pEXV‐Myc‐D12Ras or pRK5‐Myc‐MLK2; and 5 μg of pCEF‐HA‐p38 with or without 5 μg each of pRK5‐Myc, pRK5‐Myc‐L61Cdc42 or pRK5‐Myc‐MLK2. After 24 h, transfected cells were serum starved for 16 h and then harvested. For assessing complex formation in COS cells, 5 μg of pRK5‐Myc‐MLK2 were transfected with or without 2.5 μg each of pRK5‐FLAG‐hippocalcin, pRK5‐FLAG‐14‐3‐3ϵ, pRK5‐FLAG‐KAP3A or pRK5‐FLAG‐KIF3A.


The plasmids pRK5‐Myc‐MLK2, pRK5‐FLAG‐14‐3‐3ϵ, pRK5‐FLAG‐hippocalcin, pRK5‐FLAG‐KIF3A and pRK5‐FLAG‐KAP3A were transfected into COS‐1 cells in various combinations. The cells were harvested with 400 μl of lysis buffer C containing 40 mM Tris–HCl pH 7.5, 50 mM NaCl, 50 mM NaF, 100 μM Na3VO4 0.2% NP‐40, 10 μg/ml aprotinin and 10 μg/ml leupeptin. For experiments involving KAP3A and KIF3A, cells were lysed in buffer D containing 50 mM HEPES pH 6.9, 1 mM MgCl2, 5 mM EGTA, 10% glycerol, 0.2% NP‐40, 10 μg/ml aprotinin and 10 μg/ml leupeptin. Insoluble material was removed by centrifugation at 4°C for 10 min at 10 000 g, and 50 μl of lysate (50–90 μg of protein) were used for each assay. Myc‐MLK2, FLAG‐14‐3‐3ϵ, FLAG‐hippocalcin, FLAG‐KIF3A and FLAG‐KAP3A were immunoprecipitated using 9E10 or M2 antibodies. After washing the precipitates three times with the lysis buffer, the precipitates were subjected to SDS–PAGE (12 or 7.5% gel) and proteins were transferred to nitrocellulose filters. Western blotting was carried out with 9E10 and M2 antibodies for detecting Myc‐ and FLAG‐tagged proteins, respectively.

JNK, ERK and p38 MAP kinase assay

For JNK assays, transfected COS‐1 cells were lysed and harvested in buffer B (25 mM HEPES pH 7.6, 0.3 M NaCl, 50 mM NaF, 0.1 mM vanadate, 5 mM EGTA, 40 mM sodium pyrophosphate, 10 μg/ml aprotinin, 10 μg/ml leupeptin and 3 mM phenylmethylsulfonyl fluoride) containing 1% Triton X‐100. JNK1 kinase activity in cell extracts was measured after immunoprecipitation with anti‐FLAG M2 antibody, using 2 μg of GST–c‐Jun as a substrate (Olson et al., 1995). For ERK2 activity, the cells were harvested with buffer B containing 1% Triton X–100 and 0.5% deoxycholic acid, and kinase activity in cell extracts was measured after immunoprecipitation with anti‐Myc 9E10 antibody, using 5 μg of myelin basic protein (MBP) as a substrate (Olson et al., 1995). p38 kinase activity was assayed following a protocol almost identical to that of JNK assay. Briefly, the cells were harvested with buffer B containing 1% Triton X‐100 and 2 μM okadaic acid. p38 kinase activity in cell extracts was measured after immunoprecipitation with anti‐HA 12CA5 antibody, using 1 μg of bacterially expressed ATF‐2 as a substrate as described (Teramoto et al., 1996).

Immunocomplex kinase reactions were fractionated by SDS–PAGE (12 or 13.5% gel) followed by blotting onto nitrocellulose membrane and autoradiography. The relative phosphorylation levels of c‐Jun, MBP or ATF‐2 were determined by Phospho‐Imager (BioRad Inc.) analysis. The amount of immunoprecipitated JNK1, ERK2 and p38 was evaluated on Western blots using M2, 9E10 and 12CA5 antibodies, respectively as revealed by chemiluminescence detection using an ECL kit (Amersham Inc.).

Swiss 3T3 cell culture, microinjection and immunofluorescence

Swiss 3T3 cells were maintained in DMEM containing 10% FCS. Quiescent, serum‐starved Swiss 3T3 cells were prepared for microinjection as follows: cells were seeded at a density of 5×104 cells into 15 mm diameter wells containing 13 mm glass coverslips. Six to ten days after seeding, cells were serum starved overnight (16 h) in DMEM containing 2 g/l NaHCO3. For detection of Myc‐tagged‐MLK2 and microtubules, cells were fixed in 4% paraformaldehyde for 15 min, permeabilized with 0.2% Triton‐X 100 for 5 min and incubated with mouse monoclonal 9E10 and rat anti‐tubulin antibodies for 1 h followed by incubation with fluorescein isothiocyanate‐labelled anti‐mouse IgG and rhodamine‐labelled anti‐rat IgG. For the detection of activated ERK or activated JNK, rabbit anti‐active ERK or JNK antibody was used as the primary antibody, and rhodamine‐labelled anti‐rabbit IgG as the secondary antibody. Cells were viewed on a Zeiss Axioplan fluorescence microscope or a laser scanning confocal microscope (MRC‐1024, BioRad).


We thank Dr C.R.Hopkins (MRC‐LMCB) for valuable discussion and Dr M.Jacobson (MRC‐LMCB) for Z‐Val‐Ala‐Asp(O‐Me)‐fluoromethylketone. This work was generously supported by a Cancer Research Campaign (UK) programme grant (SP2249) to A.H. and by a Center of Excellence Grant from the Japan Ministry of Education, Science and Culture to N.H. K.N. was a recipient of fellowships from the Japan Society for Promotion of Science and the Uehara Memorial Foundation.