Phosphorylation by p34cdc2 Protein Kinase Regulates Binding of the Kinesin-related Motor HsEg5 to the Dynactin Subunit p150Glued *

The kinesin-related motor HsEg5 is essential for centrosome separation, and its association with centrosomes appears to be regulated by phosphorylation of tail residue threonine 927 by the p34cdc2 protein kinase. To identify proteins able to interact with the tail of HsEg5, we performed a yeast two-hybrid screen with a HsEg5 stalk-tail construct as bait. We isolated a cDNA coding for the central, α-helical region of human p150Glued, a prominent component of the dynactin complex. The interaction between HsEg5 and p150Glued was enhanced upon activation of p34CDC28, the budding yeast homolog of p34cdc2, provided that HsEg5 had a phosphorylatable residue at position 927. Phosphorylation also enhanced the specific binding of p150Glued to the tail domain of HsEg5 in vitro, indicating that the two proteins are able to interact directly. Immunofluorescence microscopy revealed co-localization of HsEg5 and p150Glued during mitosis but not during interphase, consistent with a cell cycle-dependent association between the two proteins. Taken together, these results suggest that HsEg5 and p150Glued may interact in mammalian cells in vivo and that p34cdc2 may regulate this interaction. Furthermore, they imply that the dynactin complex may functionally interact not only with dynein but also with kinesin-related motors.

Microtubule-based motors of the kinesin and dynein families are implicated in many different cellular processes, including organelle transport, mitotic spindle assembly, and chromosome segregation (for recent reviews see Refs. [1][2][3][4][5][6]. Molecular studies have greatly improved our understanding of the structures of these motors, and mechanistic aspects of microtubule-motor interactions are beginning to emerge (for reviews see Refs. 7 and 8). In contrast, it remains largely unknown what mechanisms control the subcellular distribution and activity of individual motors in time and space (9).
Accessory proteins with a potential role in targeting have been identified for both dynein and kinesin family members. Specifically, cytoplasmic dynein has been shown to interact with a multimolecular complex termed dynactin (10 -12), and it has been proposed that dynactin might anchor cytoplasmic dynein to specific subcellular structures (9,13,14). Dynactin is composed of at least nine polypeptides, including an actinrelated protein termed centractin or Arp1 (15)(16)(17), a 50-kDa protein termed dynamitin (18), and a 150-kDa protein homologous to the Drosophila gene product Glued (10,19,20). Genetic data from Saccharomyces cerevisiae, Neurospora crassa, and Drosophila melanogaster support a functional association between dynein and dynactin complexes (21)(22)(23)(24)(25)(26)(27)(28). By immunocytochemistry, both cytoplasmic dynein and dynactin localize to membranous organelles, centrosomes, and kinetochores, indicating that the two complexes display at least partially overlapping subcellular distributions (10, 16, 18, 29 -31). Kinesin function also requires accessory proteins, both in vivo and in vitro. One prominent protein interacting with kinesin, termed kinectin (32), may promote the association between kinesin and cargo vesicles (33)(34)(35). Many KRPs 1 may require accessory proteins for a specific association with appropriate subcellular structures, but the identification of partners for KRPs has proven difficult, most likely because the interactions between these proteins are labile and highly regulated.
Prominent among the KRPs implicated in different aspects of mitotic spindle assembly and function are the BimC family members (reviewed in Refs. 3, 4, and 6). These motors are required for spindle pole separation and hence for bipolar spindle formation in organisms ranging from yeast to human. Named after the BimC gene product of Aspergillus nidulans (36), likely functional homologs of BimC have been described in S. cerevisiae (Cin8p and Kip1p) (37), Schizosaccharomyces pombe (cut7) (38,39), D. melanogaster (KLP61F) (40), Xenopus laevis (XlEg5) (41,42), and Homo sapiens (HsEg5) (43). In a previous study, we showed that HsEg5 associates with centrosomes in early prophase (43), and data obtained for both XlEg5 and HsEg5 indicate that this event depends on phosphorylation of tail residue threonine 927 (Thr-927) by p34 cdc2 /cyclin B (43,44). Prompted by these findings, we have now used a yeast two-hybrid screen to identify proteins that would interact with the HsEg5 tail domain. We report that the HsEg5 tail interacts specifically with the human homolog of the dynactin component p150 Glued . Moreover, we present evidence that this interaction is enhanced by phosphorylation of residue Thr-927 of HsEg5, both in yeast cells in vivo and in vitro. These provocative results have implications not only for the function of HsEg5, but they also provide a new perspective on the role of the dynactin complex.

EXPERIMENTAL PROCEDURES
Plasmid Constructions-A cDNA encoding the stalk and tail domains of HsEg5 (Eg5ST) was fused to the GAL4 DNA binding domain (DBD) by inserting a NdeI-SalI fragment of the original HsEg5 cDNA (43) into plasmid pAS2 digested with NdeI-SalI (45). Similarly, for fusion of Eg5ST to the GAL4 activation domain (AD), a (NdeI)-EcoRI fragment excised from the HsEg5 cDNA was inserted into plasmid pACT2 digested with SmaI and EcoRI (45). (Some restriction enzymes are marked in parentheses to indicate that, before ligation, Klenow polymerase was used to blunt end 5Ј-protruding DNA fragments). The HsEg5 tail domain (Eg5T) was fused to GAL4 AD by inserting a HincII-BamHI fragment of the cDNA into plasmid pACT2 digested with (NcoI) and BamHI. The resulting constructs, Eg5ST and Eg5T, thus comprise the C-terminal moieties of HsEg5 starting at residues 245 and 764, respectively. For control, the Nek2 protein kinase (46) was fused to GAL4 DBD by inserting a (NaeI)-(XbaI) fragment excised from the full-length cDNA into pAS2 digested with SmaI. Nek2 was also fused to GAL4 AD by inserting a NcoI-BamHI fragment into pACT2 digested with NcoI-BamHI. The HsGlu-1 cDNA isolated in the course of the two-hybrid screen was fused to an Myc epitope tag by inserting an EcoRI-BglII fragment into plasmid pBSmyc digested with EcoRI and BglII (47). The whole HsEg5 sequence was fused to the GST protein by inserting the EcoRI fragment of the cDNA into plasmid pGEX-3X (Pharmacia Biotech, Inc.).
Two-hybrid Techniques-A yeast two-hybrid screen was performed essentially as described (45,48). In brief, yeast strain Y190 expressing the fusion protein DBD-Eg5ST was transformed with a human cDNA library constructed in plasmid pACT (48) and plated on selective medium (lacking tryptophan, leucine, and histidine) supplemented with 25 mM 3-aminotriazole. From 4 ϫ 10 6 transformants, 141 His ϩ colonies were obtained and assayed for LacZ expression. After being cured from the bait plasmid, the resulting 50 positive colonies were mated with Y187 yeast strains expressing DBD-Eg5ST, -cdc2, -cdk2 or -lamin fusion proteins. Seventeen colonies showed a strong specific interaction with DBD-Eg5ST only. Of 11 colonies without rearrangements in the initial pACT1 vector, two contained the same insert, termed HsGlu-1. Specific interactions between HsGlu-1 plasmids and the DBD-Eg5ST bait were also observed when both plasmids were retransformed into Y190 yeast cells. For ␤-galactosidase activity assays on filters, yeast colonies were replica-plated onto nitrocellulose filters, and these were frozen for 20 s in liquid nitrogen. Then they were placed onto a Whatman No. 3MM paper soaked with Z buffer (60 mM Na 2 HPO 4 , 40 mM NaH 2 PO 4 , 10 mM KCl, 1 mM MgSO 4 , 50 mM ␤-mercaptoethanol, pH 7) containing 5-bromo-4-chloro-3-indolyl ␤-D-galactopyranoside (1 mg/ml) and incubated at 30°C. For ␤-galactosidase liquid assays, 10 ml of mid-log phase yeast cultures (A 600 ϳ 0.5) were centrifuged and resuspended in 1 ml of Z buffer. Cells were lysed by the addition of 100 l of CHCl 3 and 100 l of 0.1% SDS, followed by a 20-s vortex. After equilibration at 30°C in a water bath, 0.2 ml of a 4 mg/ml o-nitrophenyl ␤-D-galactopyranoside substrate solution was added. Reactions were stopped by the addition of 0.5 ml of 1 M Na 2 CO 3 . The ␤-galactosidase activity was then estimated by measuring the absorbance at 420 nm, and units of activity were calculated using the expression, 1000 ϫ A 420 )/(t ϫ v ϫ A 600 ), where t is the reaction time in minutes and v is the volume of the culture in ml. For nocodazole arrest, yeasts were grown to mid-log phase. Then cultures were supplemented with 1% Me 2 SO and 15 g/ml nocodazole, and incubation was extended for another 90 min. ␤-Galactosidase activity was then measured as described above. The cell cycle stages of the cultures were determined by microscopic observation of budding.
In Vitro Protein-Protein Binding Assays-Myc-tagged proteins were translated in rabbit reticulocyte lysates according to manufacturer instructions (Promega). For in vitro interaction assays, 100 ng of GST or MalE fusion proteins, purified according to manufacturer instructions (Pharmacia and New England Biolabs, Inc.), were mixed with 10 l of in vitro translation reactions, and volumes were adjusted to 100 l with immunoprecipitation buffer (0.25% Nonidet P-40, 150 mM NaCl, 50 mM Tris-HCl, pH 8). Samples were incubated for 60 min at 30°C before 5 l of monoclonal 9E10 anti-Myc antibody (49) adsorbed to protein G-Sepharose beads were added. After a 60-min incubation at 4°C on a rotating wheel, beads were rinsed three times in immunoprecipitation buffer, boiled in loading buffer, and run on an SDS-polyacrylamide gel. Immunoblotting was then performed as described, using anti-MalEg5T serum (43). For in vitro phosphorylation of MalEg5T, 10 g of purified protein (43) was incubated for 1 h at 37°C with 1 l of p34 cdc2 /cyclin B kinase (New England Biolabs) in a total volume of 100 l according to manufacturer instructions. For coimmunoprecipitation assays, onehundredth of the sample (i.e. 100 ng of phosphorylated MalEg5T protein) was incubated with in vitro translation products.

RESULTS
HsEg5 and p150 Glued Interact in Yeast-To identify cellular proteins that might interact with HsEg5, a yeast two-hybrid screen was performed (51,52). Since KRP function may often require oligomerization (53-56), we considered it important to use a bait protein that would be able to oligomerize in yeast. Overexpression of the full-length HsEg5 protein in S. cerevisiae was toxic; thus, the motor domain was deleted from HsEg5, and the remaining stalk-tail moiety (termed Eg5ST) was fused to either the DBD or the AD of GAL4. Coexpression of DBD-Eg5ST and AD-Eg5ST in yeast activated the lacZ reporter gene (48), indicating that these two fusion proteins were able to interact (Fig. 1A). Deletion of the stalk domain from one of the two proteins (shown for the pair DBD-Eg5ST and AD-Eg5T) or coexpression of Eg5ST constructs with control fusion proteins (AD-Nek2 and DBD-Nek2) did not result in any interaction (Fig. 1A). These data show that HsEg5 fusion proteins can oligomerize in yeast provided that they contain stalk domains.
Using the DBD-Eg5ST construct as bait, a human cDNA library was screened (48), and library clones able to activate both the selection marker His3 and the lacZ reporter gene specifically in the presence of DBD-Eg5ST were selected. Among the clones showing a specific interaction with Eg5ST, two were identical and contained a 1179-base pair cDNA insert. Data base searches revealed that the corresponding translation product was closely related to a fragment of the rat p150 Glued protein (10,19); hence, this cDNA was termed Hs-Glu-1. As shown in Fig. 1B, AD-Glu-1 activated the lacZ reporter gene specifically in the presence of DBD-Eg5ST but not in the presence of unrelated bait constructs (DBD-cdc2, DBD-cdk2, or DBD-lamin). An alignment between HsGlu-1 and the central ␣-helical domain (encompassing residues 420 -811) of rat p150 Glued is shown in Fig. 1C. Over this region, the two proteins display 97% identity, indicating that HsGlu-1 almost certainly represents part of the human homolog of rat p150 Glued . Fig. 1D shows a schematic view of p150 Glued and indicates the regions identified as potential interaction domains for HsEg5 (this study) and the intermediate chain of dynein (17,57). There is potentially a partial overlap between these two putative binding sites, but it is noteworthy that the HsEg5 interaction site falls outside of the predicted coiled coil domain of p150 Glued .
The interaction of HsEg5 and HsGlu-1 in Yeast Is Modulated by Phosphorylation-As shown previously, phosphorylation of tail residue Thr-927 by the p34 cdc2 protein kinase promotes the association of HsEg5 with early prophase centrosomes (43). To determine whether the interaction between Eg5ST and Hs-Glu-1 observed in yeast might be favored by phosphorylation of Thr-927, DBD-Eg5ST mutants containing an alanine, serine, or aspartic acid in place of Thr-927 were constructed. These mutants were then coexpressed in yeast together with AD-Glu-1, and protein extracts were assayed for ␤-galactosidase activities. As shown in Fig. 2A, the average ␤-galactosidase activity was higher in yeasts expressing either the wild-type (T) or the serine mutant (S) of Eg5ST than in yeasts expressing the corresponding alanine (A) or aspartic acid (D) mutants. These data indicate that the interaction between Eg5ST and HsGlu-1 is reduced by the mutation of Thr-927 to a nonphosphorylatable residue (alanine or aspartic acid) but not by a mutation to another phosphorylatable residue (serine). It is noteworthy that the replacement of threonine 927 by aspartic acid did not enhance binding of Eg5ST to HsGlu-1, suggesting that the additional negative charge introduced by this substitution cannot mimic phosphorylation of HsEg5. This result is entirely consistent with our previous finding that a T927D mutation did not confer constitutive centrosomal association to HsEg5 in vivo (43).
Exponentially growing yeast populations contain only a small proportion of cells in G 2 and M phase (approximately 20%). Therefore, the extent of phosphorylation of the Eg5ST bait proteins by p34 CDC28 protein kinase is expected to be low, and it is not surprising that analyses of exponentially growing cells revealed only a modest influence of the identity of residue 927 on the interaction with HsGlu-1 ( Fig. 2A). However, considering that p34 CDC28 protein kinase is maximally active during mitosis, we predicted that binding of wild-type Eg5ST to HsGlu-1 should be substantially increased in mitotically arrested cells, whereas binding of the alanine mutant should not be enhanced. Indeed, compared with exponentially growing cells (20% in G 2 or M), ␤-galactosidase activity was doubled when yeast cells expressing wild-type Eg5ST were blocked in mitosis with nocodazole (80% in G 2 or M) but was almost unchanged in the case of the alanine mutant (Fig. 2B). These results indicate that phosphorylation of Thr-927 by p34 CDC28 enhances the interaction between Eg5ST and HsGlu-1, and this in turn implies that HsGlu-1 most probably binds to the tail region of the motor.
Phosphorylation Enhances a Direct in Vitro Interaction be- Yeast two-hybrid screen identifies p150 Glued as interacting with HsEg5. A, HsEg5 stalk-tail protein (Eg5ST) is able to dimerize in yeast. Y190 yeast cells were transformed with two plasmids encoding various fusion proteins containing either the GAL4 DBD or the GAL4 AD. Fusion partners were derived from HsEg5 (Eg5ST, stalk and tail domains; Eg5T, tail domain) or the human Nek2 protein kinase (46) for control. Interactions between the different proteins were examined by monitoring transactivation of the lacZ reporter gene, using a filter assay for ␤-galactosidase activity. B, specificity of HsEg5 and HsGlu-1 interaction. Yeast Y190 cells expressing library clone HsGlu-1 as an AD fusion protein were mated with yeast Y187 cells expressing either Eg5ST, cdc2 kinase, cdk2 kinase, or lamin as DBD fusion proteins. Interactions between the different proteins were visualized by ␤-galactosidase assays on filters. C, alignment of the amino acid sequences of HsGlu-1 (top line) and rat p150 Glued (bottom line) (19). Stars mark difference in the amino acid sequences, whereas circles indicate conservative substitutions (R/K, D/E). D, comparison between the HsEg5 and dynein intermediate chain (IC) interaction domains on p150 Glued . The open box shows the HsEg5 interaction domain with p150 Glued , as defined by the yeast two-hybrid interaction data described here; the black box represents the interaction domain of dynein IC as reported earlier (12,57). Numbers indicate amino acid positions in the rat p150 Glued sequence (19). Several domains in p150 Glued are marked: MTB, microtubule binding site; C-Coil, potential coiled coil structure, defined by heptad repeats of hydrophobic residues; KKEK, putative centractin-Arp1 binding site (adapted from 57).
tween the HsEg5 Tail and HsGlu-1-To determine whether HsEg5 and p150 Glued interact directly, in vitro binding studies were performed. For this purpose, the entire HsEg5 protein was fused to GST (GST-Eg5), and the tail domain was fused to the MalE protein (MalEg5). Both fusion proteins were overexpressed in Escherichia coli and purified. The HsGlu-1 cDNA was Myc-tagged (myc-Gl1), and the corresponding RNA was translated in a rabbit reticulocyte lysate. For control, a second protein isolated during the two-hybrid screen, and thus potentially interacting with HsEg5, was also tagged (myc-Q2) and translated. In a first experiment (Fig. 3A), myc-Gl1 (lane 1) and myc-Q2 (lane 2) translation products, as well as control reticulocyte lysate (lane 3), were incubated with GST-Eg5 protein; then immunoprecipitations were performed with anti-Myc antibodies, and the isolated complexes were probed by immunoblotting with anti-Eg5 antibodies. Efficient immunoprecipitation of the GST-Eg5 fusion protein by anti-Myc antibodies was observed only in the sample in which myc-Gl1 protein was present (Fig. 3A, compare lane 1 with lanes 2 and 3). These results show that an interaction between HsGlu-1 and HsEg5 occurs not only in yeast cells in vivo, but also in vitro, implying that HsGlu-1 is able to bind directly to HsEg5. In contrast, when tested under identical conditions, the protein encoded by Q2 did not bind to HsEg5 in vitro, suggesting that its isolation in the yeast two-hybrid screen may have been the result of an indirect interaction.
To rule out a role of the GST moiety in binding to HsGlu-1 and to determine more precisely which part of HsEg5 interacted with HsGlu-1, a similar type of experiment was performed with MalEg5T using MalE protein for control. Moreover, prompted by the finding that phosphorylation of Thr-927 promotes the interaction of HsEg5 with the spindle apparatus in vivo (43) and the interaction with HsGlu-1 in yeast cells (Fig.   2), we examined the influence of prior phosphorylation of the MalEg5T fusion protein by recombinant p34 cdc2 /cyclin B on its ability to interact with myc-Gl1 in vitro. In the presence of the myc-Gl1 translation product, anti-Myc antibodies efficiently precipitated the MalEg5T fusion protein but not MalE alone (Fig. 3B, lanes 1 and 2), indicating that p150 Glued can interact directly and specifically with the tail domain of HsEg5. Prior phosphorylation by p34 cdc2 /cyclin B clearly enhanced the association of the MalEg5T fusion protein with myc-Gl1 (Fig. 3B,  lane 3). To prove that this enhanced binding involved phosphorylation of residue Thr-927, the same experiment was repeated, this time comparing the effect of phosphorylation on the binding ability of either the wild-type MalEg5T fusion protein or the corresponding T927A mutant construct (Fig. 3C). The wildtype Eg5T protein exposed to p34 cdc2 /cyclin B displayed strong binding to myc-Gl1 (lane 2), whereas both the mutant construct (lane 3) and the unphosphorylated wild-type tail protein (lane 1) showed markedly lower binding. All our attempts to coimmunoprecipitate HsEg5 and p150 Glued directly from mammalian cell extracts or to extract soluble HsEg5⅐p150 Glued complexes from cells have failed (data not shown, see "Discussion"). This is not entirely unexpected, however, considering that similar approaches have also failed to reveal the purported interaction between cytoplasmic dynein and p150 Glued (for discussion see Ref. 18). In fact, it may be extremely difficult (and perhaps impossible) to find conditions affording the solubilization of the mitotic spindle apparatus while at the same time preserving particular protein-protein interactions to a sufficient extent for visualization by coimmunoprecipitation. In support of this view, we found that efficient solubilization of p150 Glued required treatment of cells with 1% Nonidet P-40, whereas detergent concentrations Ͼ0.25% abolished the in vitro interaction between p150 Glued and HsEg5 (data not shown).
Co-localization of HsEg5 and p150 Glued during Mitosis-To further explore the possibility that HsEg5 and p150 Glued might interact in mammalian cells in vivo, we have performed indirect immunofluorescence microscopy on HeLa cells using antibodies specific for HsEg5 and p150 Glued . If, as suggested by the above results, HsEg5 and p150 Glued undergo a cell cycle-regulated interaction, one would in fact predict that the two proteins should co-localize during mitosis but not during interphase. Immunofluorescent staining of HeLa cells showed that, during interphase of the cell cycle, p150 Glued is associated with centrosomes (Fig. 4b), whereas HsEg5 is distributed throughout the cytoplasm (Fig. 4a), consistent with previous data (10,43,58). Beginning in early prophase, however, HsEg5 also associates with the centrosomal region, presumably as a consequence of phosphorylation of Thr-927 by p34 cdc2 /cyclin B (43,44). As a result, the distributions of HsEg5 and p150 Glued overlap at least partially as soon as mitosis begins (Fig. 4, c-e). Both proteins then remain associated with the mitotic spindle throughout mitosis (Fig. 4, f-k). Although of limited resolution, these immunocytochemical data support the idea that HsEg5 and p150 Glued may interact not only in yeast cells and in vitro but also in mammalian cells in vivo; furthermore, they are consistent with the hypothesis that the interaction between the two proteins is controlled by cell cycle-dependent phosphorylation of Thr-927 in the tail domain of HsEg5. DISCUSSION In human cells, the KRP HsEg5 is required for centrosome separation and formation of a bipolar spindle (43). At the onset of mitosis, the association of this mitotic motor with the spindle apparatus is promoted by phosphorylation of a critical residue (Thr-927) in the tail domain of HsEg5, and this phosphorylation is most likely brought about by p34 cdc2 (43,44). Thus, HsEg5 provides an interesting example for studying the regulated targeting of a microtubule-dependent motor to a particular subcellular structure. Here, we have used a yeast twohybrid screen to search for proteins interacting with HsEg5 and have isolated a cDNA encoding the central part of a human homolog of p150 Glued , a subunit of the dynactin complex. Moreover, a recombinant HsEg5 tail protein could also bind to p150 Glued in vitro, indicating that the two proteins are able to interact directly. We may conclude, therefore, that the central domain of p150 Glued (corresponding to residues 420 -811 in the rat sequence) contains a binding site for the tail domain of HsEg5 (residues 764 -1057). Most remarkably, the association between the two proteins was enhanced by phosphorylation of Thr-927, both in vitro and in yeast cells in vivo. Finally, immunofluorescence microscopy showed that HsEg5 and p150 Glued display a striking co-distribution during mitosis, whereas, as expected, they do not co-localize during interphase of the cell cycle.
Taken together, these observations raise the intriguing possibility that HsEg5 may undergo a cell cycle-regulated interaction with the dynactin complex in mammalian cells. However, we caution that the available data do not constitute definitive proof for a direct in vivo interaction between HsEg5 and p150 Glued . All our attempts at coimmunoprecipitating the two proteins directly from mammalian cell extracts have been unsuccessful. This negative result may be explained by our observation that the in vitro association between HsEg5 and p150 Glued is readily broken under the conditions that are required to efficiently extract these proteins from cells. Also, we note that similar difficulties have been encountered in the case of the purported association between dynein and dynactin (58). Support for a physiologically significant interaction between dynein and dynactin rests not only on biochemical data (12,57) but also on genetic studies showing that mutations in the two complexes produce similar phenotypes (21)(22)(23)(24)(25)(26)(27)(28). Both HsEg5 and cytoplasmic dynein are clearly required for bipolar spindle formation (43,59), but we are not aware of genetic evidence that suggests a direct interaction between BimC family members and dynactin. One possible explanation for the lack of such genetic data is that mitotic KRPs display a significant degree of functional redundancy (60). Thus, genetic interactions between KRPs and dynactin may be difficult to reveal without mutational inactivation of multiple genes. Our present findings suggest that a deliberate search for genetic interactions between BimC homologs, other KRPs, and dynactin components might be rewarding.
What could be the physiological significance of a cell cycleregulated interaction between HsEg5 and p150 Glued ? Previously, the function of p150 Glued has been considered primarily in the context of the purported interaction between dynactin and cytoplasmic dynein, a minus end-directed motor (13). Our identification of p150 Glued as a potential partner of a plus end-directed KRP therefore constitutes a surprise. Potentially, our findings have important implications not only for the functions of HsEg5 but also for those of the dynactin complex. The exact function of dynactin in relation to cytoplasmic dynein is not definitively established, and several possibilities have been considered (reviewed in Refs. 9, 13, and 14). One intriguing, albeit poorly explained observation is that both cytoplasmic dynein and the p150 Glued component of dynactin are able to bind to microtubules. Specifically, the N terminus of p150 Glued harbors a putative microtubule binding domain that is clearly able to mediate binding to microtubules in vitro (17). Interest- ingly, this domain resembles a motif identified in CLIP-170, a protein implicated in linking endosomes to microtubules (61). Thus, it has been proposed that dynactin might act as a tether among cytoplasmic dynein, microtubules, and cargoes (e.g. vesicles or kinetochores) during those stages of the dynein ATPase cycle when the motor domain is not in contact with microtubules (for discussion see Refs. 14, 17, and 18). Conceptually, it is not difficult to extend this model to mechanochemical motors other than cytoplasmic dynein (see below).
An essential role for HsEg5 and other BimC family members in centrosome and spindle pole body separation is well established, but the precise function of these motors remains unknown, and different scenarios have been considered (for discussion see Refs. 4 and 42). One specific model proposes that plus end-directed motors, while attached to some structure in the vicinity of one centrosome, might move toward the plus end of a microtubule emanating from the other centrosome, thereby promoting spindle pole separation (42). Thus, p150 Glued might serve as an anchor for HsEg5, mediating its association with near-centrosomal structures during prophase and perhaps with additional elements of the spindle during later stages of mitosis. Potentially relevant to this model, the dynein-dynactin complex has recently been reported to functionally interact with the protein NuMA, and this protein may also contribute to tether microtubules to the poles (62). An alternative model proposes that BimC family motors could act as tetramers and promote centrosome separation by moving toward the plus ends of microtubules while being bound to two microtubules emanating from distinct centrosomes (4). In this scenario, p150 Glued might mediate the association of HsEg5 oligomers with microtubules throughout their ATPase cycles, as proposed previously for the dynein-dynactin complex (see above).
The idea that accessory proteins might function in relation to both dynein-and kinesin-related motors is not without precedent. Specifically, we note that kinectin, an accessory protein implicated in targeting of KRPs to endomembranous vesicles, has been postulated to interact also with cytoplasmic dynein (32,33). If, as proposed here, dynactin complexes are able to interact with both minus and plus end-directed motors, this would imply that the role of dynactin is more general than hitherto surmised. One obvious possibility is that dynactin complexes might tether multiple classes of microtubule-dependent motors to microtubules, as discussed above. Another possibility would be that dynactin-motor interactions might be important for the recycling of motor complexes. In the context of spindle assembly, for example, the available evidence is consistent with the idea that dynein-dynactin complexes may function at kinetochores (reviewed in Refs. 6, 9, 13, and 14). However, migration toward the minus ends of microtubules will in time be expected to cause these complexes to accumulate at centrosomes, a scenario supported by immunocytochemical data (10,18,29,30,58). It is attractive to speculate, therefore, that one role of HsEg5 (and perhaps of other centrosomeassociated plus end-directed motors) could be to prevent the excessive accumulation of dynein-dynactin complexes at centrosomes by moving them back toward microtubule plus ends. Consistent with such a model, recent studies suggest that mitotic aster formation in vitro requires precisely balanced plus and minus end-directed motor activities (63).
In conclusion, the results described here raise the intriguing possibility that the multiprotein complex dynactin may interact not only with cytoplasmic dynein but also with at least one member of the KRP family, HsEg5. Thus, much like the integral membrane protein kinectin may play a dual role in mediating the association of both dynein and kinesin-related motors with membrane vesicles (33), we propose that the dynactin complex may also play a dual role in mediating the association of cytoplasmic dynein and HsEg5 (and perhaps other KRPs) with the mitotic spindle apparatus.