INTRODUCTION

To achieve faithful transmission of the genetic material fromone cell to its daughters, cells assemble special devices calledmitotic spindles during cell division. The highly coordinatedmitotic events occur concomitantly with continuous changes inspindle structure. The major framework of the spindle iscomposed of microtubules that have been demonstrated to bein a highly dynamic state. Since the dynamic nature of micro-tubules is a key to the regulation of spindle morphogenesis andmitotic function (Kirschner and Mitchison, 1986), it isimportant to identify and characterize molecules controllingthe dynamics of mitotic microtubules to understand themechanism of mitosis and its regulation.

A growing body of evidence suggests that microtubule-

based motor proteins are included in the spindle structure andplay an essential role during mitosis (McIntosh and Pfarr,1991; Gelfand and Scholey, 1992; Sawin and Endow, 1993).In vivo (Rieder et al., 1990; Gorbsky et al., 1987; Nicklas,1989) and in vitro (Mitchison and Kirschner, 1985; Hyman andMitchison, 1991; Hyman et al., 1992) observations of mitoticcells strongly suggested the presence of ATPase activity withinthe kinetochore. Localization of cytoplasmic dynein (Pfarr etal., 1990; Steuer et al., 1990), kinesin (Neighbors et al., 1988;Leslie et al., 1987) and kinesin-like proteins (Sawin et al.,1992a; Yen et al., 1992) was also shown by immunofluores-cence microscopy at the kinetochore, pole and spindle fibers.Genetic analysis of yeast and

Drosophila led to identificationof mitosis-specific kinesin-like motor molecules, includingcut7 (Hagan and Yanagida, 1990), BimC (Enos and Morris,

3485

The CHO1 antigen is a mitosis-specific kinesin-like motorlocated at the interzonal region of the spindle. The humancDNA coding for the antigen contains a domain withsequence similarity to the motor domain of kinesin-likeprotein (Nislow et al.,

Nature 359, 543, 1992). Here wecloned cDNAs encoding the CHO1 antigen by immuno-screening of a CHO Uni-Zap expression library, the samespecies in which the original monoclonal antibody wasraised. cDNAs of CHO cells encode a 953 amino acidpolypeptide with a calculated molecular mass of 109 kDa.The N-terminal 73% of the antigen was 87% identical tothe human clone, whereas the remaining 27% of the codingregion showed only 48% homology. Insect Sf9 cells infectedwith baculovirus containing the full-length insert produced105 and 95 kDa polypeptides, the same doublet identifiedas the original antigen in CHO cells. Truncated polypep-tides corresponding to the N-terminal motor and C-terminal tail produced a 56 and 54 kDa polypeptide in Sf9cells, respectively. Full and N-terminal proteins co-sedi-mented with, and caused bundling of, brain microtubules

in vitro, whereas the C-terminal polypeptide did not. Cellsexpressing the N terminus formed one or more cytoplasmicprocesses. Immunofluorescence as well as electron micro-scopic observations revealed the presence of thick bundlesof microtubules, which were closely packed, forming amarginal ring just beneath the cell membrane and a corein the processes. The diffusion coefficient and sedimenta-tion coefficient were determined for the native CHO1antigen by gel filtration and sucrose density gradient cen-trifugation, respectively. The native molecular mass ofoverinduced protein in Sf9 cells was calculated as 219 kDa,suggesting that the antigen exists as a dimer. IntrinsicCHO1 antigen in cultured mammalian cells forms a largernative complex (native molecular mass, 362 kDa), whichmay suggest the presence of additional molecule(s) associ-ating with the CHO1 motor molecule.

Key words: mitosis, spindle, baculovirus, kinesin-like protein,mitotic motor

SUMMARY

Heterogeneity and microtubule interaction of the CHO1 antigen, a mitosis-

specific kinesin-like protein

Analysis of subdomains expressed in insect sf9 cells

Ryoko Kuriyama, Sasa Dragas-Granoic, Takami Maekawa*, Alexei Vassilev, Alexey Khodjakov†

and Hitoshi Kobayashi‡

Department of Cell Biology and Neuroanatomy, University of Minnesota, 4-135 Jackson Hall, 321 Church St SE, Minneapolis,Minnesota 55455, USA

*Present address: Hokkaido Red Cross Blood Research Center, Sapporo, Hokkaido, Japan†Present address: Department of Histology and Cytology, Moscow State University, Moscow, Russia‡Present address: Iwate Biotechnology Research Center, Kitakami, Japan

Journal of Cell Science 107, 3485-3499 (1994)Printed in Great Britain © The Company of Biologists Limited 1994

3486

1990), KAR3 (Meluh and Rose, 1990), CIN8 (Hoyt et al.,1992), KIP1 (Roof et al., 1992; Hoyt et al., 1992), ncd (Endowet al., 1990; McDonald and Goldstein, 1990) and nod (Zhanget al., 1990). Recent progress in PCR (polymerase chainreaction) technologies has allowed the identification of morekinesin-like proteins, including KLPA (O’Connell et al.,1993), KLP61F (Heck et al., 1993) and KIP2 (Roof et al.,1992). Mutations of some of these kinesin-like proteins led todeficiency in mitotic processes, and the importance of thekinesin-like motors for organization of functional spindles hasbeen further implicated by in vivo analysis of antibodymicroinjection into living cells (Yen et al., 1991) and in vitroassembly of spindle structures in Xenopus egg extracts (Sawinet al., 1992b).

Using the CHO1 monoclonal antibody raised against mitoticspindles isolated from CHO cells, we have identified a novel95/105 kDa spindle constituent (Sellitto and Kuriyama, 1988).The antibody was originally screened as a specific probe forcentrosomes in interphase CHO cells. During mitosis,however, the antigen localizes throughout the spindle andrelocates to the interzonal region at anaphase. It is thendiscarded as a midbody component at the end of cell division(Sellitto and Kuriyama, 1988; Kuriyama and Nislow, 1992).Microinjection of the purified CHO1 antibody into mammalianmitotic cells resulted in mitotic inhibition in a stage-specificand dose-dependent manner, indicating that the CHO1 antigenis essential for mitotic progression (Nislow et al., 1990). Tofurther extend the study of this unique spindle component, wehave cloned the human gene by screening a HeLa expressionlibrary using CHO1 as a probe (Nislow et al., 1992). Analysisof the nucleotide and deduced amino acid sequence showedthat the N-terminal half of the CHO1 antigen contains a regionof 350 amino acids that has significant identity with the motordomain shared among kinesin-like proteins. The C-terminalhalf of the antigen, however, shows little homology to otherkinesin-like proteins, suggesting that the CHO1 antigen is anovel member of the kinesin superfamily (Nislow et al., 1992).Since the CHO1 antigen is a kinesin-like motor moleculespecific to mitotic cells, it is also called mitotic kinesin-likeprotein-1, MKLP-1 (Nislow et al., 1992).

cDNA encoding the CHO1 antigen was also cloned from aCHO expression library, the same species in which the originalantibody was raised. Here we report sequence comparison ofthe protein between the two species. Full coding as well astruncated polypeptides were expressed in insect Sf9 cells usingthe baculovirus expression system. Overexpression of themotor domain resulted in characteristic reorganization ofmicrotubule arrays in Sf9 cells. Molecular properties ofinduced CHO1 polypeptides were also compared with those ofthe intrinsic CHO1 antigen in cultured mammalian cells to gainmore information regarding structure and function of thisunique mitotic motor molecule.

MATERIALS AND METHODS

Isolation and sequence analysis of cDNA clones encodingthe CHO1 antigen in CHO cellsA commercially available Chinese hamster ovary (CHO) cell cDNAexpression library cloned in

λUni-Zap (Stratagene, La Jolla, CA) wasimmunoscreened with the monoclonal CHO1 antibody (Sellitto and

Kuriyama, 1988), according to the procedure described previously(Kuriyama et al., 1990; Maekawa and Kuriyama, 1993). Three out offour positive clones obtained cross-hybridized and showed identicaloverlapping regions (see Fig. 1A). The largest of these clones (CHO1-8a clone), which spans 3.2 kb, was chosen for further nucleotidesequence analysis. A set of nested deletions of both strands of the 8aclone in pBluescript KS+ (Stratagene, La Jolla, CA) was generatedusing a ExoIII/Mung bean deletion kit (Stratagene). Double-strandedDNA was then sequenced by the dideoxy chain termination method(Sanger et al., 1977), using the Sequenase kit (US Biochemical Corp.,Cleveland, OH). The DNA sequence was assembled and analyzedwith the IntelliGenetics Software Package (Maekawa and Kuriyama,1993).

Preparation of CHO1 antigen expressed in Sf9 cells Expression of full/truncated CHO1 antigens in insect Sf9 cellsThe CHO1-8a clone was used to express the entire coding sequenceof the CHO1 antigen, and the N- and C-terminal halves of the proteinswere made by digesting the 8a clone at EcoRV site (nucleotideposition 1,465; arrow in Fig. 1A). Purified cDNAs were subclonedinto multi-cloning sites of pVL1392 or pVL1393 (PharMingen, SanDiego, CA), and introduced into moth ovarian Sf9 cells by co-trans-fection with partially deleted linearized baculovirus DNA as describedpreviously (Maekawa and Kuriyama, 1993). Viruses released by thetransfected cells were collected and used for further amplification.

Protein preparationSf9 cells expressing the different domains of CHO1 antigen werewashed twice with phosphate buffered saline (PBS) and resuspendedin 5-10 vols of 100PEM buffer (100 mM PIPES at pH 6.8, 1 mMEGTA, 1 mM MgCl2) containing a mixture of protease inhibitors (20µM leupeptin, 1% aprotinin and 0.1 mg/ml PMSF) at 0°C. After son-ication, disrupted cells were fractionated into supernatants and pelletsby centrifugation at 13,000 g for 15 minutes. As shown in Results, amajor part of the full-length is recovered in the pellet fraction. To sol-ubilize the protein, CHO1 antigen-containing pellets were resus-pended in 100PEM supplemented with 0.6 M NaCl and 5 mMMgATP. After 10 minutes of incubation on ice, the sample was cen-trifuged at 100,000 g for 60 minutes in a Beckman TL-100 table-topcentrifuge to recover supernatants, which were further dialyzedagainst an excess volume of 100PEM. Insoluble materials wereseparated by sedimentation, and the supernatant was used as thepartially purified full-length CHO1 antigen.

Preparation of CHO1 antigen from HeLa cellsHeLa cell lysates containing the intrinsic CHO1 antigen wereprepared as described previously (Maekawa et al., 1991). Briefly,monolayer cultures of cells were treated with 0.1 µg/ml of nocoda-zole for 24 hours to prepare synchronous cells arrested at mitosis.After washing twice in cold PBS, cells were resuspended in 5-10 volsof 100PEM plus protease inhibitors. Cells were disrupted by sonica-tion for about 1 minute, supernatants were recovered after sedimen-tation at 15,000 g for 15 minutes and stored at −80°C until use.

Analysis of physical parameters of the native CHO1protein complexTo determine the native size of CHO1 antigen, the intrinsic HeLaCHO1 antigen as well as recombinant proteins of full and truncatedN- and C-terminal CHO1 antigen in the cell supernatant weresubjected to FPLC on a Superdex 200 HR 10/30 column (PharmaciaLKB Biotechnology, Piscataway, NJ). The column was run accordingto the manufacture’s protocol with 100PEM as the elution buffer. Fordetermination of the sedimentation coefficient (s20,w ×10−13 seconds),5% to 15 or 5% to 20% sucrose density gradients in 100PEM weremade in Beckman 14 mm × 89 mm centrifuge tubes. Samples of 0.5ml of cell supernatants were loaded, and the gradients were cen-

R. Kuriyama and others

3487Baculovirus expression of a kinesin-like protein

trifuged at 200,000 g for 12 (for 5% to 20% gradient) or 18 hours (5%to 15% gradient) at 4°C in a Beckman SW41Ti rotor. Fractions of 0.5ml were collected and analyzed on Coomassie-stained SDS-PAGEand immunoblots. The markers used for calculation of the apparentmolecular mass, diffusion coefficient (D20,w ×10−7 cm2 s−1) and sed-imentation coefficient are thyroglobulin (669 kDa, D20,w = 2.6, s20,w= 19.4), ferritin (440 kDa, D20,w = 3.6, s20,w = 17.6), catalase (232kDa, D20,w = 4.1, s20,w = 11.3), aldolase (158 kDa, D20,w = 4.6, s20,w= 7.4), brain α/β heterodimer (100 kDa), BSA (66 kDa, D20,w = 6.1,s20,w = 4.4) and ovalbumin (43 kDa, s20,w = 3.6). The native molecularmass, Stokes’ radius and axial ratio were calculated as described byBloom et al. (1988).

Assays of microtubule interaction with the CHO1 antigenSupernatants of Sf9 cells expressing either the full-length, or the N-or C-terminal half of the protein were first dialyzed against excessvolume of 100PEM. After clarifying by centrifugation at 13,000 g for15 minutes at 4°C, the samples were mixed with taxol-stabilizedMAP-free brain microtubules for 10 minutes at 0°C in a buffer con-taining 100PEM plus 0.5 mM GTP and 10 µg/ml taxol (Maekawa etal., 1991; Maekawa and Kuriyama, 1993). Supernatants as well aspellets of microtubules plus associated proteins were separated bycentrifugation, mixed with SDS sample buffer, and run on 7.5% SDS-polyacrylamide gels. For assay of the microtubule bundling activity,a sample of the CHO1 antigen-containing fraction and stabilized brainmicrotubule mixtures was mounted on a slide glass for observationby phase-contrast microscopy or Formvar-coated 200-mesh grids forwhole-mount electron microscopy as described previously (Kuriyamaand Borisy, 1981).

Immunological techniquesAntibody preparation and purificationThe recombinant CHO1 antigen with an apparent molecular mass of95 kDa was purified by excising from gels. The gel bands were mixedwith Freund’s complete adjuvant, and used to immunize rabbits bysubcutaneous injections. Sera collected from the animals were affinitypurified by binding to the protein immobilized on nitrocellulose blots(Maekawa and Kuriyama, 1993).

Immunofluorescence stainingImmunofluorescence staining of CHO and Sf9 cells, and isolatedCHO mitotic spindles was performed as before (Sellitto andKuriyama, 1988; Maekawa and Kuriyama, 1993). Briefly, CHO cellsgrown on coverslips in F-10 medium plus 7.5% fetal calf serum (FCS)were fixed in methanol for 5 minutes at −20°C. Mitotic spindlesisolated from synchronized CHO cells (Kuriyama et al., 1984) weremounted on polylysine-coated coverslips, then fixed with absolutemethanol as above. Sf9 cells were cultured in a TNM-FH mediumsupplemented with 10% FCS (Summers and Smith, 1987), and fixedwith methanol first, then transferred into absolute acetone. After incu-bation for 5 minutes at −20°C, coverslips were air-dried and rehy-drated with phosphate buffered saline containing 0.05% Tween-20(PBS-Tw20). The polyclonal antibody raised against 95 kDa CHO1antigen was used as a primary antibody. Double immunofluorescencestaining was performed by incubation of coverslips in a mixture ofpolyclonal 95 kDa antibody and either a mouse monoclonal anti-chicken β-tubulin antibody (Amersham, Arlington Height, IL) forCHO cells, or a mouse monoclonal CHO1 antibody for isolated CHOspindles. Microscopic observation was made on an Olympus BH2microscope with epifluorescence optics, and photographs were takenon Tri-X film.

Immunoblot analysisProteins in whole cell lysate, supernatant and pellet fractions as wellas isolated CHO spindles were separated in 7.5% SDS-polyacryl-amide gels, transferred to nitrocellulose membranes as before(Maekawa et al., 1991), then probed with monoclonal CHO1, poly-

clonal 95 kDa CHO1 antibody, and affinity-purified polyclonal anti-bodies raised against 10 amino acid residues (LNLVDLAGSE) thatare conserved among kinesin-like molecules (generous gift from DrLinda Wordeman) (Sawin et al., 1992a). Immunoreactive bands werevisualized by alkaline phosphatase-conjugated secondary antibodies(HyClone, Logan, UT) with NBT/BCIP as the chromogen (Maekawaet al., 1991).

Thin-section electron microscopySf9 cells cultured on coverslips coated with either polylysine orVectaBond (Vector Laboratory, Burlingame, CA) were infected withrecombinant baculovirus. After 2-3 days, cells were fixed with 2%glutaraldehyde in 100PEM for 30 minutes at room temperature, rinsedin PBS, then postfixed in 1% OsO4. After washing with distilledwater, the samples were next dehydrated through an ethanol series,infiltrated and embedded in Epon-Araldite plastic. Cells with cyto-plasmic processes were marked with a diamond scribe. Ultrathinsections were cut with LKB Nova Ultramicrotome and stained withuranyl acetate and lead citrate. The specimen were examined in aJEOL 100CX electron microscope.

RESULTS

Sequence analysis of the CHO1 antigen in CHOcells; comparison with the human geneBy screening a λUni-Zap expression library using CHO1 as aprobe, we isolated three immunopositive clones out of 106

plaques screened. They shared common restriction maps, andanalysis of their partial sequence showed them to be identicalin the overlapping regions (Fig. 1A). The longest clone (8a),which contains 3,248 nucleotides, overlapped with the othertwo DNA fragments (12a and 14a clones) at the C terminus,implying that the epitope to the CHO1 antibody resides on theC-terminal half of the protein. The nucleotide sequence of the8a clone is shown in Fig. 1B. The clone contains the entirecoding sequence (2,859 base pairs) of the CHO1 antigen alongwith 379 nucleotides of 3′-untranslated region in which apotential polyadenylation signal and a long poly(A)+ tract areincluded. This open reading frame predicts a protein of 953amino acids in length with a calculated molecular mass of 109Da, which is in good agreement with 105/95 kDa of the CHO1antigen determined from SDS-PAGE (Sellitto and Kuriyama,1988). The isoelectric point of the CHO1 antigen was calcu-lated as 8.4; the C-terminal half of the protein is much morebasic (pI = 9.8) than the N terminus (pI = 6.9). Particularlybasic regions are included in the N terminus at amino acidpositions 600-625 (pI = 11.6), 650-675 (pI = 10.3), 800-825(pI = 12.3), 900-910 (pI = 12.8) and 940-953 (pI = 10.2).

Fig. 1B also compares the amino acid sequence of the CHOclone with that of HeLa cells (Nislow et al., 1992). As reportedbefore, the CHO1 antigen possesses a conserved region of 350amino acid residues corresponding to the putative motordomain at the N terminus (Vale and Goldstein, 1990;Goldstein, 1993) (Fig. 1C). It contains sequences similar to theconsensus sequence proposed for ATP binding proteins atamino acid positions 114 to 118, as well as several oligopep-tide sequences, such as SSRSH and DLAGSE at amino acidpositions 298 and 337 (underlined), respectively, which areconserved among different classes of kinesin-like molecules.In contrast, the C-terminal two-thirds of the protein shows littlehomology to any kinesin-related molecules, suggesting that the

3488 R. Kuriyama and others

A

B

3489Baculovirus expression of a kinesin-like protein

Fig. 1. Schematic diagram of cDNA clones and theprimary sequence of the CHO1 antigen in CHOcells. (A) Map of the cDNA clones coding for theCHO1 antigen. The open and filled trianglesindicate the position of the start and stop codons,respectively. The numbers underneath at the end ofeach clone represent nucleotide numbers encoded

by the clones. An arrow shows the position (nucleotide number: 1465) where the CHO1-8a clone was cut with EcoRV to prepare N-terminalmotor and C-terminal tail domains for expression in Sf9 cells. (B) Complete DNA sequence of the CHO1 antigen in CHO cells. The deducedamino acid sequence indicated by the single letter code below the DNA sequence was also compared with that of HeLa clone (Nislow et al.,1992). Consensus nucleotide binding domain and the amino acid sequences conserved among the members of kinesin superfamily areunderlined. Double-underlining shows the position of a potential polyadenylation signal. (C) Sequence comparison between CHO and HeLaclones. Stippled areas indicate the regions with high amino acid seqeuence similarity. An arrowhead shows the position where the CHO1 clonewas cut to prepare N- and C-terminal domains.

B (cont)

C

3490

CHO1 antigen is a novel member of the kinesin superfamily.The N-terminal 73% of the CHO1 antigen is 87% identical tothe human clone (Fig. 1C). In addition, a region at amino acidposition 797 to 938 in the CHO clone corresponded well to theHeLa sequence between 698 and 839 (87% homology); that is,there is a shift of nearly 300 nucleotides in the alignedsequence. In contrast, the region from amino acid 692 to 796,showed no similarity to the HeLa sequence. Thus theremaining 27% of the coding sequence in the C-terminaldomain of the CHO clone showed only 48% homology to thehuman clone. Since the sequence of two other clones (CHO1-12a and CHO1-14a) are identical to the CHO1-8a clone in theC-terminal region (Fig. 1A), it suggests that either the CHO1antigens are heterogeneous among different species, ordifferent forms of CHO1 antigen molecules may exist in asingle species, or both.

In vivo expression of full and truncatedpolypeptides of the CHO1 antigen in insect Sf9 cells The CHO1 antigen was originally identified as a spindlecomponent localized at the equatorial region of the anaphasemitotic spindle. It possesses a plus-end-directed motor activityin vitro (Nislow et al., 1992) and is essential for mitotic pro-gression (Nislow et al., 1990). Towards the goal of analyzingthe regulatory mechanism of the CHO1 function duringmitosis, we have expressed subregions of the antigen in insectovarian Sf9 cells using the baculovirus expression system. TheCHO1 antigen was split in the middle of the protein betweennucleotide positions 1465 and 1466, indicated by an arrow inFig. 1A. The cDNA fragments coding for the full-length aswell as N- and C-terminal halves of the protein were insertedinto the genome of the Autographa californica nuclear poly-hedrosis virus at the polyhedrin gene locus by homologousrecombination. SDS-PAGE analyses revealed the presence of

major 105 and 95 kDa polypeptides in Sf9 cells expressing thefull-length CHO1 antigen (Fig. 2, lanes 3 and 3′ in CBBcolumn), the same as the doublet identified as the CHO1antigen in CHO cells (Sellitto and Kuriyama, 1988; lane m inSpindle column). Amounts of the upper 105 kDa polypeptidevary among different preparations as is shown in lanes 3 and3′. It was also noted that, as the antigen was further purified,the upper band of the expressed full-length CHO1 antigeneventually disappeared (data not shown). This indicates thatthe lower 95 kDa band could be a degraded product of the 105kDa molecule. Expression of clones corresponding to N- andC-terminal halves of the protein resulted in production of 56and 54 kDa polypeptides, respectively (lanes 2 and 4 in CBBcolumn). The epitope for the CHO1 monoclonal antibodyresides at the C terminus (Fig. 1A); thus the original mono-clonal antibody recognized 95/105 kDa full-length and 54 kDaC-terminal fragment (Fig. 2 lanes 3m and 4m in IB column),but not the 56 kDa N-terminal half of the protein (data notshown). In contrast, affinity-purified antibodies raised againsta decapeptide corresponding to a conserved region from thekinesin motor domain (LDVLAGSE: amino acid positions 335to 342) (Sawin et al., 1992) reacted with the amino-terminal(lane 2L in IB column) and full-length CHO1 antigen, but notthe C-terminal fragment (data not shown).

The lower molecular mass 95 kDa of the full-length proteinwas excised from gels and injected into rabbits as animmunogen (Fig. 2, lane 3p in IB column). Polyclonal anti-bodies affinity-purified from such immune sera reacted withthe same antigen in isolated CHO mitotic spindles (Fig. 2, lanep in Spindle column, arrow) as the original monoclonalantibody. They also recognized truncated N- as well as C-terminal proteins (Fig. 2, lanes 3p and 4p in 1B column). Fig.3 shows whole CHO cells (A) and isolated spindles andmidbodies (B) at different mitotic stages double-stained with

R. Kuriyama and others

Fig. 2. Overexpressed full-length, N-terminalmotor and C-terminal tail domains of theCHO1 antigen in Sf9 cells identified byprotein gel and immunoblot analyses. Lane M,molecular mass markers containing brainMAP1 and MAP2, myosin, β-galactosidase,phosphorylase B, BSA, brain α- and β-tubulins, and ovalbumin; lane 1, whole celllysates of uninfected Sf9 cells; lane 2, Sf9 celllysates expressing N-terminal motor domainof the CHO1 antigen; lane 3 and 3′, Sf9 cellsexpressing full-length CHO1 antigensequence; lane 4, Sf9 cell lysates expressingC-terminal tail domain of the CHO1 antigen.CBB, 7.5% polyacrylamide gel stained withCoomassie Blue. Dots indicate the position ofinduced protein bands and arrows show theposition of α- and β-tubulin; IB, immunoblotanalysis of nitrocellulose strips incubated withcrude (lane 2p) or affinity-purified (lanes 3pand 4p) polyclonal anti-95 kDa CHO1 antigenprepared from infected Sf9 cells, monoclonalCHO1 antibody (lanes 3m and 4m), and

affinity-purified polyclonal anti-kinesin oligopeptide antibody (LAGSE) (lane 2L); Spindle, mitotic spindles isolated from CHO cells were runon 7.5% gel and stained with Coomassie Blue (lane CBB). Blotted nitrocellulose filters were stained with monoclonal (lane m) and affinity-purified polyclonal (lane p) CHO1 antibodies. Bars indicate the position of molecular mass markers; myosin (200 kDa), β-galactosidase (116kDa), phosphorylase B (93 kDa), BSA (66 kDa) and ovalbumin (45 kDa).

3491Baculovirus expression of a kinesin-like protein

polyclonal CHO1 antibodies (A,a′ to d′ and B,b) and eithertubulin (A,a″ to d″) or the original monoclonal CHO1 antibody(B,c). The polyclonal antibody labeled interphase centrosomes(arrowheads in A,a′) and interzonal region of mitotic spindle

structures (A,c′ and B,b); those are the structures typicallyrevealed by the original monoclonal antibody (Sellitto andKuriyama, 1988; Fig. 3B,c). Although the spindle pole was notstained by the monoclonal antibody at all, faint staining of

Fig. 3. Localization ofthe CHO1 antigen inwhole CHO cells (A)and isolated mitoticspindles (B). The samecells/spindles are seenby phase-contrast (A,ato d and B,a) andfluorescence microscopyafter double-stainingwith affinity-purifiedpolyclonal CHO1antibody (A,a′ to d′ andB,b) with either tubulin(A,a′′ to d′′) ormonoclonal CHO1 (B,c)antibody. Thepolyclonal and originalmonoclonal antibodiesstained the samesubcellular structures.Arrowheads in a′ andarrows in b′ representinterphase centrosomesand spindle poles,respectively. Bars, 10µm.

A

B

3492

spindle poles by the polyclonal antibody was sometimes noted(indicated by arrows in A,b′). Antibodies raised against HeLacell fusion protein were also able to label spindle poles (Nislowet al., 1992). These immunofluorescence staining and westernblotting results indicate that the isolated clones indeed encodethe authentic CHO1 antigen molecule.

Molecular properties of the CHO1 antigenIn order to examine the CHO1 antigen in more detail, weanalyzed the molecular properties of the polypeptides corre-sponding to different regions of the CHO1 antigen andcompared them with those of intrinsic CHO1 antigen preparedfrom cultured mammalian cells.

Full-length and truncated polypeptides expressed in Sf9cellsOverexpressed N-terminal fragment was recovered in thesupernatant fraction, whereas the major part of the full-lengthas well as C-terminal half of the protein came to the pellet aftercentrifugation at 13,000 g for 15 minutes (Table 1). Todetermine the physical properties of the CHO1 antigen,proteins in the supernatant were analyzed by gel filtration andsucrose density gradient centrifugation. As summarized inTable 1, FPLC analysis on a Superdex 200 column showed thatthe full-length, N- and C-terminal CHO1 antigen in super-natants eluted in a peak corresponding to a molecular mass of440, 100 and 200 kDa, respectively. Sucrose density gradientcentrifugation allowed us to calculate the sedimentation coef-ficient (s20,w) as 8.0, 4.6 and 5.5 S, respectively. Based on thevalue of s20,w and the diffusion coefficient (D20,w×10−7 cm2

s−1) estimated from elution profiles of Superdex 200 columns,we calculated the native molecular mass of full-length, N- andC-terminal CHO1 antigens as 219, 74 and 114 kDa, respec-tively (Table 1). These results indicate that the overinducedCHO1 antigen in Sf9 cells forms a dimer complex. The proteinis expected to be asymmetric with an axial ratio of approxi-mately 17.

Comparison of the CHO1 antigen between Sf9 andHeLa cellsTo compare the size of native protein in Sf9 cells with that ofintrinsic CHO1 antigen, we prepared cytosolic fractions fromcultured mammalian cells and analyzed by both FPLC andsucrose density gradient centrifugation. HeLa cells wereemployed for these analyses because: (1) large quantities ofthese cells were available; (2) immunofluorescence/immunoblotting pattern of the antigen is identical to that inCHO cells (Nislow et al., 1990); (3) despite the fact that there

is ~300 nucleotide shift in the C terminus, the HeLa CHO1antigen shows high homology to the protein in CHO cells.

The CHO1 antigen is localized at both the centrosome andnucleus in interphase cells, and the antigen is recovered innuclear as well as cytosolic fractions (Nislow et al., 1990)(Table 1). When the soluble fraction was applied to the FPLCSuperdex 200 sizing column, the protein fractionated into asingle peak with an apparent molecular mass of ~800 kDa.Sucrose density gradient centrifugation provided evidence thatthe intrinsic CHO1 antigen possesses a sedimentation value of10.3, which is in good agreement with the value of 11 S pre-viously reported by Nislow et al. (1990) in HeLa cells. Theintrinsic CHO1 antigen has the native molecular mass of 362kDa with approximate radial axis of 18.3. The protein in HeLacells is apparently larger than that in Sf9 cells, indicating thepossibility that additional molecule(s), such as the light chainsassociated with the conventional kinesin motor (Bloom et al.,1988; Kuznetsov et al., 1988), is (are) included in the CHO1antigen fraction.

In vitro interaction of the expressed CHO1 antigen withbrain microtubulesThe CHO1 antigen in HeLa cells co-sedimented with brainmicrotubules in vitro, and the corresponding fusion proteinexpressed in bacteria has also been shown to cross-bridgeantiparallel microtubules (Nislow et al., 1990, 1992). In orderto identify the domains responsible for interaction with micro-tubules, we have characterized overinduced CHO1 antigenmolecules. Full-length CHO1 antigen in either crude high-speed supernatants or a partially purified fraction was mixedwith taxol-stabilized microtubules polymerized from purifiedMAP-free brain tubulin. During the 15 minute incubation onice, microtubules became cross-linked to form large bundlesthat were easily detected by phase-contrast microscopy (datanot shown). Fig. 4 illustrates such bundles of microtubules asseen by whole-mount electron microscopy. The bundles aretypically composed of 10 to 20 microtubules, and the individ-ual microtubules are closely associated side-by-side along theirentire length. The N-terminal half of the CHO1 antigen wasalso able to induce microtubule bundles (data not shown).

To further demonstrate protein interaction with micro-tubules, we recovered microtubules plus associated proteins bycentrifugation, and determined the molecular species presentin supernatants and pellets by immunoblot analysis. Fig. 5shows results of microtubule association with full-length(column F), truncated N- (column N) and C-terminal (columnC) fragments of the CHO1 antigen, respectively. Althoughsmall portions of all three polypeptides were recovered in the

R. Kuriyama and others

Table 1. Physical properties of the CHO1 antigen in Sf9 and HeLa cells

Cell App. mol. mass (kDa) D20,w s20,w Stokes’ Axial Native mol. Oligomer- fractionation SDS-PAGE FPLC (×10−7 cm2 s−1) (×10−13 s) radius ratio mass (kDa) ization

Sf9 cellsFull-length Insoluble 95/105 440 3.2 8.0 6.9 16.6 219 Dimer N-terminal Soluble 56 100 5.5 4.6 3.9 7.1 74 DimerC-terminal Insoluble 54 200 4.2 5.5 4.9 6.3 114 Dimer

HeLa cellsCHO1 antigen Nuclei/cytosol* 100/115* 800 2.5 10.3 8.7 18.3 362 Dimer

*Nislow et al. (1990).

3493Baculovirus expression of a kinesin-like protein

pellet fraction, even in the absence of microtubules (lanes 6and 8), the majority of full as well as N-terminal polypeptidesspecifically co-sedimented with the taxol-stabilized micro-tubules (lanes 4 in column F and N). In contrast, the polypep-tide corresponding to the C-terminal half of the CHO1 antigenstayed in the supernatant, as in the control without micro-tubules (lanes 4 and 8 in column C), indicating that the antigenassociates with microtubules through the globular motordomain at the N terminus. When the full-length and/or the N-terminal half of the protein was incubated with brain micro-tubules in the presence of 5 mM MgATP, the amount ofimmunoreactive polypeptides that co-sedimented with micro-tubules decreased to some extent (lanes 2 in columns F and N).These results suggest that the N-terminal motor domaincontains both nucleotide-dependent and -independent micro-tubule binding sites. Interaction of the full-length, C- and N-terminal fragments with microtubules is summarized in Table2.

Reorganization of microtubule arrays in Sf9 cells byexpression of N-terminal motor domain of the CHO1antigenDuring the course of immunofluorescence screening of Sf9cells, we noticed reorganization of cytoplasmic microtubules incells overexpressing the motor domain of the CHO1 antigen(Fig. 6). While only diffuse tubulin staining was seen in non-

infected control cells (data not shown), Sf9 cells overexpress-ing the N-terminal fragment contain bundled microtubulesrunning in different directions (a′). The majority of the cells,however, included only a very thick microtubule ring justbeneath the cell membrane (b). Since these kinds of microtubulearrays were never detected in control cells, it is concluded thatthe motor domain of CHO1 antigen modified microtubuleorganization inside the Sf9 cells. Fig. 7 shows electron micro-graphs of cells expressing the N-terminal motor domain, con-firming the presence of thick bundles of microtubules, runningin different directions (a) or forming a ring just beneath the cellmembrane (b). A higher magnification of a microtubule ring isseen in d; although the bundle is packed too tightly to distin-guish individual tubular structures, there is an area wheremicrotubules are seen splaying off the bundle (d). In cross-

Fig. 4. Whole-mount electron micrographs of microtubules bundledin vitro by the presence of full-length CHO1 antigen produced in Sf9cells. Partially purified full-length CHO1 antigen was incubated withtaxol-stabilized brain microtubules for 10 minutes on ice; 10 to 20microtubules are cross-linked together and associate along theirlength. Bars, 1 µm (a) and 0.25 µm (b).

Fig. 5. Co-sedimentation of taxol-stabilized brain microtubules withfull-length (column F), N-terminal motor domain (column N), and C-terminal tail domain (column C) of the CHO1 antigen assayed byimmunoblot analysis. Lanes 1,3,5,7, supernatant after centrifugationat 13,000 g for 15 minutes; lanes 2,4,6,8, pellets after centrifugation.Each full-length, N-terminal and C-terminal half of the proteinfraction was incubated for 10-15 minutes at 0°C in the presence(lanes 1 to 4) or absence (lanes 5 to 8) of brain microtubules, and inthe presence (lanes 1,2,5,6) or absence (lanes 3,4,7,8) of 5 mMMgATP. Arrow in C indicates the position of an intact 54 kDa C-terminal tail fragment. Full-length and N-terminal motor domain, butnot C-terminal tail fragment, bind to microtubules in both ATP-dependent and -independent manners.

Table 2. Interaction of the CHO1 antigen withmicrotubules

In vitro In vitro Process MT-binding MT-bundling in Sf9 cells

Sf9 cellsFull-length + + −N-terminal + + +C-terminal − ND −

HeLa cellsCHO1 antigen +* ND NA

ND, not determined; NA, not applicable.*Nislow et al. (1990).

3494

sectional views, it is seen that the bundles consist of 10-20microtubules embedded in a kind of electron-dense material (c).

Although the microtubule ring generally had a nice roundedshape during the early stage of virus infection (Fig. 6b), at laterstages it started to bend or to have an uneven surface. Subse-quently changes in cell shape were observed (Fig. 6c-c′). Oncemicrotubule bundles find a site where they can extend, the ringsstart unwinding to elongate (Fig. 6d), ultimately other bundlesof microtubules appeared to join this outlet (Fig. 6e), resultingin the formation of long cytoplasmic processes (Fig. 6f-f′). Thenumber of cytoplasmic extensions per cell ranged from one tomany (Fig. 6g-i). No tubulin/microtubules appear to remain inthe body of cells with long processes; as a result we detectedvery low background of tubulin fluorescence in the cell body.Electron microscopic analyses of thin-sectioned processes showbundles of microtubules present in the center, which appears toprovide a structural core for the cytoplasmic extensions (Fig.

8a). In cross-sectional view, a group of microtubules arebundled together at the center of the process (Fig. 8b). Individ-ual microtubules are packed tightly and linked to adjacentmicrotubules by bridge-like structures (arrows in Fig. 8c), whichmight represent the N-terminal motor domain of the antigen.

In contrast to the cells overproducing the N-terminal motordomain, the Sf9 cells that expressed the full or C-terminal taildid not show any prominent morphological alterations (Table2). It is, however, sometimes possible to detect microtubulebundles in cells expressing the full-length molecule, byimmunofluorescence microscopy (data not shown). Suchbundles of microtubules run in different directions inside thecells, and their fluorescence was far less intense than those inmotor domain-expressing cells. Fig. 8d is a thin-section electronmicrograph of a cell expressing the full-length CHO1 antigen.Although there are parallel arrays of microtubules (arrows inFig. 8d), individual microtubules are not packed as tightly as in

R. Kuriyama and others

Fig. 6.Immunofluorescencestaining of Sf9 cellswith anti-tubulinantibody. The samecells are seen byphase-contrast (a,c,f)andimmunofluorescence(a′,c′,f′) microscopy.Cells expressing theN-terminal motordomain of theantigen form eithermicrotubule bundlesrunning in differentdirections (a′), orthick microtubulerings just beneaththe cell membrane(b,c′), andeventually produce amicrotubuleextension andcytoplasmic process(f,f′). Multipleextensions are alsoformed (g,h,i). Bars,10 µm.

3495Baculovirus expression of a kinesin-like protein

N-terminal motor-expressing cells. On very rare occasions, wefound examples in which Sf9 cells expressing the full-lengthCHO1 antigen formed cytoplasmic extensions. As in thecellular processes induced in motor domain-expressing cells,these extensions appeared to contain a core of microtubules;however, the spacing between individual microtubules is muchwider than that found in N-fragment expressing cells (Fig. 8e).

DISCUSSION

The gene encoding a kinesin-like motor protein, the CHO1

antigen, which is also called mitosis-specific kinesin-likeprotein-1 (MKLP-1) in HeLa cells (Nislow et al., 1992), wascloned and sequenced in CHO cells. Comparison of thededuced amino acid sequence of the protein between Chinesehamster and human cells showed over 87% homology in theN-terminal half of the protein (amino acid position: 1 to 691in total 953 amino acid residues). Although over half of theremaining C-terminal sequence in CHO cells (amino acidposition: 797 to 938) is well-aligned with the human sequencebetween 698 and 839 with 87% homology, the remainingsequence between 692 and 796 does not match any regions ofthe HeLa clone (Fig. 1C). It is not clear why the homologous

Fig. 7. Thick bundles of microtubules formed in N-terminal fragment-expressing Sf9 cells seen by thin-section electron microscopy.Microtubule bundles running in different directions (a), or a microtubule ring beneath the cell membrane (arrows in b). (c and d) Cross- andlongitudinal sections of the microtubule rings seen at higher magnification. Bars, 1 µm (a,d), 5 µm (b), and 0.1 µm (c).

3496

region in the C-terminal domain is shifted between the twospecies; this may, however, represent heterogeneity of theCHO1 antigen.

There are good reasons to believe that the CHO1 antigenplays an essential role during mitosis. Microinjection of theCHO1 antibody into living cells results in mitotic inhibition ina stage-specific and dose-dependent manner (Nislow et al.,1990). The protein displays a unique subcellular localizationin dividing cells, shifting from diffuse spindle staining at

metaphase to an increasingly narrow band of microtubuleswithin the interzone during anaphase (Sellitto and Kuriyama,1988; Kuriyama and Nislow, 1992). Immunofluorescenceanalysis of injected cells revealed that those that completemitosis display normal localization of the CHO1 antigen,whereas arrested cells show no specific localization within thespindle (Nislow et al., 1990). These results suggest the impor-tance of translocation of the protein from throughout thespindle to the midplane for its function in mitotic cells. The

R. Kuriyama and others

Fig. 8. Thin-section electron micrographs of Sf9 cells expressing full-length (d,e) and N-terminal motor domain (a to c) of the CHO1 antigen.Longitudinal (a) as well as cross-sectional (b) view of the cytoplasmic process formed in N-terminal fragment-expressing cells show thepresence of highly packed microtubule bundles. Area outlined in b is shown in c at higher magnification. Small arrows in c indicate the cross-bridge structure formed between neighboring microtubules. Full sequence-expressing Sf9 cells contain microtubule bundles in both the cellbody (arrows in d) and process (e), but spacing between neighboring microtubules was wider than in N-terminal fragment-expressing cells.Bars, 0.1 µm (a-c,e), and 5 µm (d).

3497Baculovirus expression of a kinesin-like protein

CHO1 antigen is a plus-end-directed motor molecule (Nislowet al., 1992), thus it may be possible for it to move along thelength of spindle microtubules toward the plus end at thespindle midzone. It is suggested that the normal size of thebipolar spindle is maintained by the balance of forces actingon the spindle structure. Motors located in the metaphase kine-tochore exerting force toward the spindle poles pull the spindlepoles toward each other. In order to prevent the spindle fromshrinking, therefore, this force must be opposed by one or moreforces that keep the spindle poles apart. The CHO1 antigencould be a candidate for producing such force by interactionbetween microtubules of opposite polarities in the center of thespindle. In support of this hypothesis, we have alreadyobserved that short and stubby spindle structures are formed individing cells microinjected with the CHO1 antibody (Nislowet al., 1990). Alternatively, since the CHO1 antigen is presentin the interzonal region of the spindle where antiparallel micro-tubules overlap, the motor may be involved in the generationof force necessary for sliding of the pole-to-pole microtubulesduring anaphase B.

On the basis of comparison of amino acid sequences, anumber of kinesin-like proteins are now classified into severalsubclasses, including the bimC family and the KAR3 family(Goldstein, 1993). The human gene corresponding to theCHO1 antigen is an apparent ‘orphan’; that is, it cannot beplaced into any of the previously identified families. Althoughfunctional specificity of the kinesin-like motors appears to beregulated by the tail domain of the molecule (Vale andGoldstein, 1990), there is a certain degree of functional redun-dancy among different types of kinesin-like proteins despitethe fact that the primary amino acid sequences of the tailregions are apparently not well conserved (Roof et al., 1992;Hoyt et al., 1992; Saunders and Hoyt, 1992; O’Connell et al.,1993). Localization of the CHO1 antigen at the midzone andpole in the spindle structure resembles that of cut7, a kinesin-like protein in Schizosaccharomyces pombe (Hagan andYanagida, 1992). It is thus of interest to determine whether theCHO1 antigen is the homolog to cut7 and whether they areable to complement each other functionally during the processof spindle formation and maintenance.

Like conventional kinesin as well as many other membersof the kinesin superfamily, the motor domain of the antigen islocated in 350 amino acid residues at the N terminus.Expression of different domains in Sf9 cells has revealedseveral characteristic features. When Sf9 cells overproducingthe CHO1 antigen were extracted in a detergent-containingbuffer, it was found that the N-terminal half of the protein wasrecovered in the supernatant. However, major amounts of full-length as well as C-terminal polypeptides sedimented in thepellets. These results may suggest that the C-terminal domaincould be sticky and has a tendency to form aggregates, whichis also suggested by the fact that some full-length as well asC-terminal proteins expressed in Sf9 cells eluted in the voidvolume on a sizing column (data not shown). Alternatively,sedimentation into the cell pellet fraction may be due to thepresence of a nuclear localization signal (NLS) in the Cterminus, since the CHO1 antigen is known to be localizedinside the nucleus in interphase cells (Sellitto and Kuriyama,1988); thus part of the molecule is recovered in the nuclearfraction (Nislow et al., 1990).

The predicted secondary structure of the CHO1 antigen is

composed of two globular heads at N and C termini connectedby a central stalk portion. The expressed full-length as well asN- and C-terminal polypeptides appear to exist as a dimer,which could be due to the presence of an α-helical coiled-coilcentral stalk. While the full-length as well as N-terminal halfof the protein can bind to and bundle microtubules, the Cterminus failed to interact with microtubules. The very basiccarboxy domains of CHO1 apparently are not involved informing electrostatic interactions with the acidic regions oftubulin. Dimer formation of the full-length and N-terminalpolypeptides allows the globular heads with microtubule-binding capacity to extend outward, which might allow theprotein to cross-link microtubules. The CHO1 antigen islocated in the interzonal region of the spindle where antiparal-lel microtubules originating from opposite poles overlap. Sincemicrotubules in the midzone region/midbody are packedclosely together, it would be of interest to compare the lengthof the molecular complex of CHO1 antigen and the spacingbetween neighboring microtubules in the midbody.

Intrinsic CHO1 antigen in HeLa cells always forms a largermolecular complex (native molecular mass = 362 kDa) thanthe protein expressed in Sf9 cells (native molecular mass = 219kDa). One possible explanation is that, like the conventionalkinesin protein, the CHO1 motor could be associated withadditional molecule(s), such as kinesin light chains (Bloom etal., 1988; Kuznetsov et al., 1988). Suppose each 110 kDaCHO1 motor in HeLa cells binds to a second molecule in anequimolar ratio to form a 362 kDa dimer complex, themolecular mass of the associated protein of the CHO1 antigenwould be calculated as 72 kDa, which is quite close to themolecular mass of kinesin light chain (62-84 kDa) identifiedin several different species (Bloom et al., 1988; Kuznetsov etal., 1988; Johnson et al., 1990).

As they overproduce the motor domain of the kinesin-likeprotein, rounded Sf9 cells alter their shape to extend long cyto-plasmic processes. Nerve-specific MAPs, such as tau (Knopset al., 1991; Chen et al., 1992) and MAP2 (Chen et al., 1992;LeClerc et al., 1993), have also been shown to produce longcytoplasmic processes in infected Sf9 cells. The appearance ofsuch extensions is similar to the neuronal outgrowth.Moreover, both the processes are supported by a core of micro-tubule bundles, and microtubules formed in Sf9 cells exten-sions are organized in a pattern similar to that in neuronalprocesses, from the point of polarity (Baas et al., 1991; Chenet al., 1992) and spacing between microtubules (Chen et al.,1992). Cytoplasmic extensions in Sf9 cells might therefore beconsidered as a model system for the study of neuronal processformation. However, processes in Sf9 cells have organellar dis-tribution patterns similar to those in the cell body (Knops etal., 1991; Baas et al., 1991), which is in sharp contrast to theaxonal process. Extensions in Sf9 cells could also be inducedby expression of the catalytic subunit of cAMP-dependentprotein kinase (Cheley et al., 1992). Nuclear polyhedrosis viralpolypeptide p10 (Kuzio et al., 1984) becomes phosphorylatedby the action of overinduced kinase enzyme, which in turncauses p10 to associate with microtubules and cells to produceprocesses. These results clearly indicate that formation ofcellular outgrowth results from modification of microtubulearchitecture, thus it is not restricted to cells infected withrecombinant baculovirus expressing neuron-specific MAPs.

We are puzzled by the fact that cells expressing the full-

3498

length protein form neither cytoplasmic processes norprominent microtubule bundles. The major part of the full-length protein expressed in Sf9 cells was found in the pelletafter centrifugation at 13,000 g for 15 minutes. Therefore theamount of protein remaining in the cytosol may be insufficientto induce microtubule reorganization. Both the full-length andN-terminal domain contain ATP-dependent and -independentmicrotubule binding sites, and can cross-link microtubules invitro. Since a major difference between cells overexpressingthe full-length and N-terminal CHO1 antigen is the presenceor absence of highly packed bundles of microtubule, it is rea-sonable to speculate that Sf9 cells must induce densely packedmicrotubule bundles, to a certain extent, in order to extendcellular processes. In fact spacing between microtubules in abundle present inside the cell body was much wider than thatin the process (Fig. 8d and e). It is likely that microtubule inter-actions with the intact CHO1 antigen may be blocked by somefactor(s) through association with the C-terminal tail domainof the antigen. In natural conditions, such inhibition could beprevented by the presence of additional specific molecules,such as light chain component(s), to form a stoichiometriccomplex with the heavy chain of CHO1 kinesin-like molecule.

The authors express their thanks to Dr Linda Wordeman and MrRod Kuehn for kindly providing the anti-kinesin oligopeptideantibody and for electron microscopic assistance, respectively. We arealso grateful to Dr Mary Kimble for critical reading of the manuscript.This work was supported by grants from the National Institutes ofHealth (GM41350) and the Council for Tobacco Research Inc. (no.3157) to R.K., and the Japan Society for the Promotion of Science(no. 2154) to T.M.

REFERENCES

Baas, P. W., Pienkowski, T. P. and Kosik, K. S. (1991). Processes induced bytau expression in Sf9 cells have an axon-like microtubule organization. J.Cell Biol. 115, 1333-1344.

Bloom, G. S., Wagner, M. C., Pfister, K. K. and Brady, S. T. (1988). Nativestructure and physical properties of bovine brain kinesin and identification ofthe ATP-binding subunit polypeptide. Biochemistry 27, 3409-3416.

Cheley, S., Kosik, K. S., Paskevich, P., Bakalis, S. and Bayley, H. (1992).Phosphorylated baculovirus p10 is a heat-stable microtubule-associatedprotein associated with process formation in Sf9 cells. J. Cell Sci. 102, 739-752.

Chen, J., Kanai, Y., Cowan, N. J. and Hirokawa, N. (1992). Projectiondomains of MAP2 and tau determine spacings between microtubules indendrites and axons. Nature 360, 674-677.

Endow, S. A., Henikoff, S. and Soler-Niedziela, L. (1990). Mediation ofmeiotic and early mitotic chromosome segregation in Drosophila by aprotein related to kinesin. Nature 345, 81-83.

Enos, A. P. and Morris, N. R. (1990). Mutation of a gene that encodes akinesin-like protein blocks nuclear division in A. nidulans. Cell 60, 1019-1027.

Gelfand, V. I. and Scholey, J. M. (1992). Every motion has its motor. Nature359, 480-482.

Goldstein, L. S. B. (1993). With apologies to Scheherazale: tails of 1001kinesin motors. Annu. Rev. Genet. 27, 319-351.

Gorbsky, G. J., Sammak, P. J. and Borisy, G. G. (1987). Chromosomesmove poleward in anaphase along stationary microtubules that coordinatelydisassemble from their kinetochore ends. J. Cell Biol. 104, 9-18.

Hagan, I. and Yanagida, M. (1990). Novel potential mitotic motor proteinencoded by the fission yeast cut7+ gene. Nature 347, 563-566.

Hagan, I. and Yanagida, M. (1992). Kinesin-related cut7 protein associateswith mitotic and meiotic spindles in fission yeast. Nature 356, 74-76.

Heck, M. M. S., Pereira, A., Pesavento, P., Yannoni, Y., Spradling, A. C.and Goldstein, L. S. B. (1993). The kinesin-like protein KLP61F is essentialfor mitosis in Drosophila. J. Cell Biol. 123, 665-679.

Hoyt, M. A., He, L., Loo, K. K. and Saunders, W. S. (1992). TwoSaccharomyces cerevisiae kinesin-related gene products required for mitoticspindle assembly. J. Cell Biol. 118, 109-120.

Hyman, A. A. and Mitchison, T. J. (1991). Two different microtubule-basedmotor activities with opposite polarities in kinetochores. Nature 351, 206-211.

Hyman, A. A., Middleton, K., Centola, M., Mitchison, T. J. and Carbon, J.(1992). Microtubule-motor activity of a yeast centromere-binding proteincomplex. Nature 359, 533-536.

Johnson, C. S., Buster, D. and Scholey, J. M. (1990). Light chains of seaurchin kinesin identified by immunoadsorption. Cell Motil. Cytoskel. 16,204-213.

Kirschner, M. and Mitchison, T. (1986). Beyond self-assembly: frommicrotubules to morphogenesis. Cell 45, 329-342.

Knops, J., Kosik, K. S., Lee, G., Pardee, J. D., Cohen-Gould, L. andMcConlogue, L. (1991). Overexpression of tau in a nonneuronal cellinduces long cellular processes. J. Cell Biol. 114, 725-733.

Kuriyama, R. and Borisy, G. G. (1981). Centriole cycle in Chinese hamsterovary cells as determined by whole mount electron microscopy. J. Cell Biol.91, 814-821.

Kuriyama, R., Keryer, G. and Borisy, G. G. (1984). The mitotic spindle ofChinese hamster ovary cells isolated in taxol-containing medium. J. Cell Sci.66, 265-275.

Kuriyama, R., Savereide, P., Lefebvre, P. and Dasgupta, S. (1990). Thepredicted amino acid sequence of a centrosphere protein in dividing seaurchin eggs is similar to elongation factor (EF-1a). J. Cell Sci. 95, 231-236.

Kuriyama, R. and Nislow, C. (1992). Molecular components of the mitoticspindle. BioEssays 14, 81-88.

Kuzio, J., Rohel, D. Z., Curry, C. J., Krebs, A., Carstens, E. B. andFaulkner, P. (1984). Nucleotide sequence of the p10 polypeptide gene ofAutographa californica nuclear polyhedrosis virus. Virology 139, 414-418.

Kuznetsov, S. A., Vaisberg, E. A., Shanina, N. A., Magretova, N. N.,Chernyak, V. Y. and Gelfand, V. I. (1988). The quaternary structure ofbovine brain kinesin. EMBO J. 7, 353-356.

LeClerc, N., Kosik, K. S., Cowan, N., Pienkowski, T. P. and Baas, P. W.(1993). Process formation in Sf9 cells induced by the expression of amicrotubule-associated protein 2C-like construct. Proc. Nat. Acad. Sci. USA90, 6223-6227.

Leslie, R. J., Hird, R. B., Wilson, L., McIntosh, J. R. and Scholey, J. M.(1987). Kinesin is associated with a nonmicrotubule component of sea urchinmitotic spindles. Proc. Nat. Acad. Sci. USA 84, 2771-2775.

Maekawa, T., Leslie, R. and Kuriyama, R. (1991). Identification of a minusend-specific microtubule-associated protein located at the mitotic poles incultured mammalian cells. Eur. J. Cell Biol. 54, 255-267.

Maekawa, T. and R. Kuriyama. (1993). Primary structure and microtubule-interacting domain of the SP-H antigen: a mitotic MAP located at the spindlepole and characterized as a homologous protein to NuMA. J. Cell Sci. 105,589-600.

McDonald, H. B. and Goldstein, L. S. B. (1990). Identification andcharacterization of a gene encoding a kinesin-like protein in Drosophila. Cell61, 991-1000.

McIntosh, J. R. and Pfarr, C. M. (1991). Mitotic motors. J. Cell Biol. 115,577-585.

Meluh, P. B. and Rose, M. D. (1990). KAR3, a kinesin-related gene requiredfor yeast nuclear fusion. Cell 60, 1029-1041.

Mitchison, T. J. and Kirschner, M. W. (1985). Properties of the kinetochorein vitro. II. Microtubule capture and ATP-dependent translocation. J. CellBiol. 101, 766-777.

Neighbors, B. W., Williams, R. C. Jr and McIntosh, J. R. (1988).Localization of kinesin in cultured cells. J. Cell Biol. 106, 1193-1204.

Nicklas, R. B. (1989). The motor for poleward chromosome movement inanaphase is in or near the kinetochore. J. Cell Biol. 109, 2245-2255.

Nislow, C., Sellitto, C., Kuriyama, R. and McIntosh, J. R. (1990). Amonoclonal antibody to a mitotic microtubule-associated protein blocksmitotic progression. J. Cell Biol. 111, 511-522.

Nislow, C., Lombillo, V. A., Kuriyama, R. and McIntosh, J. R. (1992). Aplus-end-directed motor enzyme that moves antiparallel microtubules invitro localizes to the interzone of mitotic spindles. Nature 359, 543-547.

O’Connell, M. J., Meluh, P. B., Rose, M. D. and Morris, N. R. (1993).Suppression of the bimC4 mitotic spindle defect by deletion of klpA, a geneencoding a KAR3-related kinesin-like protein in Aspergillus nidulans. J. CellBiol. 120, 153-162.

Pfarr, C. M., Coue, M., Grissom., P. M., Hays, T. S., Porter, M. E. and

R. Kuriyama and others

3499Baculovirus expression of a kinesin-like protein

McIntosh, J. R. (1990). Cytoplasmic dynein is localized to kinetochoresduring mitosis. Nature 345, 263-265.

Rieder, C. L., Alexander, S. P. and Rupp, G. (1990). Kinetochores aretransported poleward along a single astral microtubule during chromosomeattachment to the spindle in newt lung. J. Cell Biol. 110, 81-95.

Roof, D. M., Meluh, P. B. and Rose, M. D. (1992). Kinesin-related proteinsrequired for assembly of the mitotic spindle. J. Cell Biol. 118, 95-108.

Sanger, F., Nicklen, S. and Coulson, A. R. (1977). DNA sequencing withchain-terminating inhibitors. Proc. Nat. Acad. Sci. USA 74, 5463-5467.

Saunders, W. S. and Hoyt, M. A. (1992). Kinesin-related proteins required forstructural integrity of the mitotic spindle. Cell 70, 451-458.

Sawin, K. E., Mitchison, T. J. and Wordeman, L. G. (1992a). Evidence forkinesin-related proteins in the mitotic apparatus using peptide antibodies. J.Cell Sci. 101, 303-313.

Sawin, K. E., LeGuellec, K., Philippe, M. and Mitchison, T. J. (1992b).Mitotic spindle organization by a plus-end-directed microtubule motor.Nature 359, 540-543.

Sawin, K. E. and Endow, S. A. (1993). Meiosis, mitosis and microtubulemotors. BioEssays 15, 399-407.

Sellitto, C. and Kuriyama, R. (1988). Distribution of a matrix component ofthe midbody during the cell cycle in Chinese hamster ovary cells. J. Cell Biol.106, 431-439.

Steuer, E. R., Wordeman, L., Schroer, T. A. and Sheetz, M. P. (1990).Localization of cytoplasmic dynein to mitotic spindles and kinetochores.Nature 345, 266-268.

Summers, M. D. and Smith, G. E. (1987). A Manual of Methods forBaculovirus Vector and Insect Cell Culture Procedures. San Diego:Pharmingen.

Vale, R. D. and Goldstein, L. S. B. (1990). One motor, many tails: anexpanding repertoire of force-generating enzymes. Cell 60, 883-885.

Yen, T. J., Compton, D. A., Wise, D., Zinkowski, R. P., Brinkley, B. R.,Earnshaw, W. C. and Cleveland, D. W. (1991). CENP-E, a novel humancentromere-associated protein required for progression from metaphase toanaphase. EMBO J. 10, 1245-1254.

Yen, T. J., Li, G., Schaar, B. T., Szilak, I. and Cleveland, D. W. (1992).CENP-E is a putative kinetochore motor that accumulates just before mitosis.Nature 359, 536-539.

Zhang, P., Knowles, B. A., Goldstein, L. S. B. and Hawley, R. S. (1990). Akinesin-like protein required for distributive chromosome segregation inDrosophila. Cell 62, 1053-1062.

(Received 6 July 1994 - Accepted 16 August 1994)