p150Glued, dynein and microtubules are specifically required for activation of MKK3/6 and p38 MAPKs

were from Cell Signaling Technology; anti-p150Glued from BD Biosciences; anti- β -tublin, anti-DIC, and anti-Flag from Sigma; anti-MKK3/6 from Upstate; anti-HA, anti-Xpress, anti-GFP, anti-His-tag, anti-Akt, and anti-ERK1 from Biotechnology; anti-GST from Amersham Biosciences; and anti-giantin from Dr. HP. Hauri.


Introduction
MAPK-mediated intracellular signaling pathways play key roles in diverse cellular processes in eukaryotic organisms (1)(2)(3). The MAPK cascades transmit extracellular activating signals through sequential phosphorylation and activation of MKK kinase (MAP3K), MKK, and MAPK. At non-physiological concentrations in vitro, a purified MAP3K or MKK can directly bind and phosphorylate a particular MKK or MAPK respectively without a need for an additional regulatory molecule. However, it remains less clear in vivo how an activated upstream kinase finds and interacts with a specific inactive downstream kinase (among thousands of other cellular proteins). One can envision the involvement of either scaffold molecules or molecular motors that may bring MAP3K, MKK and MAPK to close physical proximity in cells. Indeed, several scaffolding molecules have been found which specifically group some or all components in a particular MAPK cascade to ensure signaling specificity and efficiency (4). In addition, certain MAPK components have also been found to bind cellular motors (5,6).
To look for potential regulators of MKK6, an MKK acting specifically upstream of the p38 MAPKs (9), we carried out a yeast two-hybrid screen. Unexpectedly, p150 Glued , a key component of the dynein/dynactin complex (7,8), was found to specifically interact with both MKK6 and MKK3 (a close homolog of MKK6 which also acts specifically upstream of the p38 MAPKs). Furthermore, we show that p150 Glued , dynein, and functional microtubules are indispensable for activation of MKK3/6 and p38 MAPKs in vivo.
Yeast two-hybrid screening. The yeast strain AH109 (Mat a) expressing the bait protein GAL4-DBD-human MKK6(KM) was mated with Y187 (Mat α) containing the MATCHMAKER human skeletal muscle cDNA library following manufacturers instruction (BD Biosciences).

Cell lysis, kinase assays and Western blot
These were carried out as described in (10).

Results p150 Glued specifically interacts with MKK3/6
Using MKK6 as bait to screen a human skeletal muscle cDNA library, we found several clones displaying specific interaction with MKK6(KM), a kinase-dead form of

MKK6 directly associates with tubulin
As p150 Glued is a microtubule-binding protein (8), we then tested whether MKK3/6 and p38 MAPK also associate with microtubules. Native microtubules were first polymerized and purified from rat brain ( Fig.2A) and the associated proteins were identified by Western blot. Like p150 Glued and dynein intermediate chain (DIC), a fraction of the endogenous MKK3/6 and p38 MAPK were also found to associate with microtubules ( Fig.2B). As a control, Akt was not present in our microtubule preparation (Fig.2B). Using microtubules polymerized in vitro with the purified tubulins, we found that only the recombinant MKK6, but not the p38 MAPK or ATF2, co-sedimented with microtubules (Fig.2C). This suggested that MKK6 but not the p38 MAPK directly binds to microtubules. To find out whether MKK6 also interacts with soluble tubulins, we carried out both in vivo and in vitro binding assays. In vivo, under conditions that tubulins did not polymerize, the endogenous MKK3/6 were specifically coimmunoprecipitated with the soluble tubulins as was p150 Glued dynactin (Fig.2D). In vitro, when mixed with the purified tubulins, only the recombinant MKK6, but not the p38 MAPK or ATF2, interacted with tubulins in the GST-pulldown assays (Fig.2E).

MKK3/6 and p38 MAPKs
To understand the cellular function of the interaction between MKK6 and p150 Glued , we first knocked down the expression of the endogenous p150 Glued with siRNA before cells were stimulated by Sorbitol. We found that the Sorbitol-induced phosphorylation of both MKK3/6 and p38 MAPK were specifically reduced by the p150 Glued -specific siRNA but not by a control siRNA (Fig.3A, lanes 5 and 6). Similarly, the p150 Glued siRNA could also specifically reduce the tumor necrosis factor α (TNFα)induced phosphorylation of MKK3/6 and p38 MAPK (our unpublished data). The decrease in the level of both phospho-MKK3/6 and phospho-p38 MAPK correlated with that of p150 G lued expression (Fig.3A, lanes 3 and 6). Interestingly, the p150 Glued -siRNA had no obvious effect on the epidermal growth factor (EGF)-induced phosphorylation of ERKs, the prototypic members of the MAPK family (our unpublished data). As p150 Glued is a key component of the dynein/dynactin complex, the reduction in p150 Glued level would be expected to adversely affect the dynein motor function. To find out whether the dynein motor was involved, we first resorted to erythro-9-[3-(2hydroxynonyl)]adenine (EHNA), a known inhibitor of the dynein ATPase activity (11,12). As a control, we first confirmed that EHNA caused dispersal of giantin in the Golgi apparatus (Our unpublished data), the aggregation of which being a known dynein-dependent process (13,14). Importantly, in cells pretreated with EHNA, the Sorbitol-induced phosphorylation of both MKK3/6 and p38 MAPKs was significantly reduced in a dose-dependent manner (Fig.3B). In contrast, the EGF-induced phosphorylation of ERKs was not affected by similar EHNA treatment (Fig.3C). To further implicate a role for dynein in MKK3/6-p38 MAPK activation, we overexpressed p50 dynamitin, which is known to disrupt the dynein/dynactin complex and interferes with the dynein function (7,8). Indeed, the Sorbitol-induced phosphorylation of the endogenous p38 MAPK was reduced in cells over-expressing p50 dynamitin but not the empty vector (Fig.3D).

Functional microtubules are required for signal-induced phosphorylation of MKK3/6 and p38 MAPKs
As dynein is a microtubule-dependent motor and a fraction of MKK6, the p38 MAPK, and 150 Glued were all found to associate with microtubules (Fig.2), we next examined the involvement of microtubules. Before stimulating cells with Sorbitol, we first pretreated cells separately with Nocodazole, Colchicine, and Taxol, all being microtubule-interfering drugs which specifically disrupt the normal function of microtubules via different mechanisms (15). Indeed, the Sorbitol-induced phosphorylation of both MKK3/6 and the p38 MAPK were significantly reduced in cells pretreated with the drugs (Fig.4A). Similarly, the TNFα-induced phosphorylation of p38 MAPK could also be inhibited by these microtubule-interfering drugs (our unpublished data). Consistent with our previous data (Fig.3C), the EGF-induced ERK phosphorylation and kinase activity were not affected by any of the three microtubuleinterfering drugs (Fig.4B).

Discussion
Associations between MAPK components and microtubules have long been observed. Both ERKs and JNKs were found to associate with microtubules in different cell types (16). However, the functional significance of such interactions remains poorly understood. Interestingly, links between MAPK components and microtubule-associated motors started to emerge in recent years. Mixed-lineage kinase 2 (MLK2), a member of the MAP3K superfamily, was found to interact with several members of the KIF3 kinesin motor complex (5). In addition, JNK-interacting proteins (JIPs), the scaffold proteins involved in organizing the JNK signaling complex (4), were found to be cargos for kinesin by interacting with the kinesin light chain (KLC) (6). In neuroblastoma cells, JIPs, together with other components of the JNK pathway, were transported to the neurite tips in a kinesin-dependent manner. In both cases above, however, it has not been addressed whether disruption of these interactions affects subsequent MAPK activation.
Similar to the role of KLC for kinesin motors, several components of the dynein/dynactin complex have been implicated in cargo-binding. For example, Arp-1 of the dynactin complex interacts with the Golgi-associated βIII spectrin (17). Dynein light chain Tctex-1 interacts with the cytoplasmic portion of rhodopsin and the rhodopsincontaining vesicles can be moved in vitro on microtubules in a dynein-dependent manner (18). As to p150 Gl ued, while its amino terminus binds microtubules, its carboxyl terminus is thought to participate in cargo-binding (7). Huntingtin-associated protein 1 (HAP1) was the first protein known to directly interact with the cargo-binding end of p150 Glued and was recently confirmed to be a bona fide cargo of dynein (19)(20)(21).
Interestingly, the binding site for HAP1 on p150 Glued overlaps with that for MKK3/6.
As p150 Glued , dynein, and functional microtubules are all required for activation of MKK3/6 and the p38 MAPKs, we suggest that MKK3/6 may also serve as cargos for the dynein motor. In contrast to the p38 MAPKs, we find that the activation of ERKs by EGF is not dependent on p150 Glued , dynein or functional microtubules. As ERKs can be activated by diverse stimuli in addition to EGF, it remains to be determined whether activation of ERKs under those conditions requires molecular motors and microtubules.
To our knowledge, the current study represents the first report of a functional link between MAPK components and dynein motors. Our data also suggest that the cellular motors can not only carry signaling complexes to different subcellular compartments but also play a more direct role in the process of signal transduction.