The Bipolar Kinesin, KLP61F, Cross-links Microtubules within Interpolar Microtubule Bundles of Drosophila Embryonic Mitotic Spindles

Oregon red strains of Drosophila melanogaster were maintained in large farms as previously described (Ashburner, 1989; Meyer et al., 1998). In some of the immunofluorescence and electron microscopic analyses described below, flies were collected from the University of California San Francisco fly facility.

High speed supernatant (HSS) from extracts of 0–2 h Drosophila embryos was generated as previously described (Meyer et al., 1998). Actin and myosin were depleted from HSS by incubation with hexokinase (10 U/ml) and 0.9% glucose for 30 min at room temperature and spinning at 35,000 g for 30 min. The supernatant was then collected and incubated with 1 mM GTP and 20 μM taxol for 45 min, then spun through a 15% sucrose cushion (containing 100 μM GTP, 100 μM ATP, and 5 μM taxol) and spun again at 20,000 g for 60 min at 10°C. The pellet was resuspended in PMEG, and immediately spun again for 20 min at 60,000 g at 10 C. The pellets were then resuspended in PMEG containing 10 μM ATP, 10 μM MgSO₄, 10 μM taxol, 100 μM GTP, and 0.5 M KCl for 4 h at 4°C and spun at 130,000 g for 45 min at 4°C. The pellets were resuspended in PMEG containing 100 μM GTP and 10 μM taxol aliquoted and stored at −80°C. Tubulin prepared in this manner was used for antibody generation and affinity purification, described below.

The taxol stabilized MTs were injected into three rabbits along with Freund's complete adjuvant and a week later 10 ml of blood was taken. This occurred once a month for 5 mo. Samples from each bleed were taken and tested for reactivity with tubulin by Western blotting Drosophila HSS and preparations of MTs and microtubule-associated proteins (MAPs). One of the three rabbits showed a strong response to the tubulin injections.

Anti-tubulin antibodies were affinity purified on columns of affigel 15 (Biogel) coupled to Drosophila embryonic tubulin. After binding of the tubulin to the columns, the columns were washed 10 vol of PBS-0.5% Tween-20 (PBS-Tween) and then blocked in PBS-Tween containing 5% BSA (Sigma) for 1 h at room temperature. Next, the columns were washed with 10 vol of PBS-Tween and incubated with the antiserum diluted 1:5 in blocking solution overnight at 4°C. The columns were then washed in 20 vol of PBS-Tween. Antibody was eluted from the column with 0.1 M glycine, pH 2.7, and collected in 0.8-ml fractions into clean tubes containing 0.2 ml of 1 M Tris, pH 8.0. The OD 280 of all elution fractions was determined against a standard of the tris/glycine mixture. All fractions with an OD 280 of >0.05 were kept and dialyzed into PBS overnight at 4°C.

Polyclonal anti-BCB peptide antibodies were raised using synthetic peptides of sequence CPTGTTPQRRDYA, corresponding to the highly conserved BCB of frog Eg5. The peptide was synthesized with and without a phospho-threonine residue at residue position 6 corresponding to the consensus cdk1 phosphorylation site (Heck et al., 1993; Sawin and Mitchison, 1995), and NH₂-terminal acetyl and COOH-terminal amide groups were included. Approximately 40 mg both peptides at >92% purity were obtained by HPLC at the UCSF biomolecular resource center where the masses of the p-BCB and BCB peptides were confirmed by mass spectroscopy. Both peptides were found to be highly water-soluble and were conjugated to KLH carrier protein (5 mg peptide/50 mg KLH). The anti-p-BCB and anti-BCB antibodies were produced by BabCo (Richmond) and specific antibodies were purified from immune sera on columns of BCB or p-BCB peptides coupled to affigel 10. Bound antibodies were eluted using 0.1 M glycine, pH 2.5, and rapidly neutralized using 1 M Tris buffer, pH 10. Antibodies were then incubated with a 100-fold excess of blocking peptide (p-BCB peptide in the case of the anti-BCB antibody and BCB peptide in the case of the anti-p-BCB antibody).

To test the reactivity of these antibodies against phospho- and unphospho-KLP61F, recombinant KLP61F (or its stalk-tail fragment) prepared as described by Barton et al. (1995) was bound batchwise to glutathione-agarose beads, the beads were washed in CSF extract buffer then incubated with M-phase or I-phase Xenopus egg extracts or corresponding buffer plus [³²P]ATP for 10 min (see below), then the beads were washed in cytostatic factor (CSF) extraction buffer, boiled in sample buffer, and analyzed by SDS-PAGE, immunoblotting with the BCB and p-BimC box antibodies, and autoradiography.

To analyze the hydrodynamic properties of phosphorylated and unphosphorylated native KLP61f from fly embryos, fractions prepared as described previously (Cole et al., 1994) were analyzed by immunoblotting with the anti-BCB and anti-p-BCB antibodies.

Bacterial strain BL21(DE3) was transformed with the pRSET vector containing a full length KLP61F insert. To isolate KLP61F from the transformant, a colony from a freshly streaked plate was inoculated into Luria broth medium (LB) containing 200 μg/ml ampicillin and incubated overnight at 37°C. The culture was then diluted 1:100 with fresh LB+ containing 200 μg/ml ampicillin and grown at 37°C to an optical density of A600 = 0.80. The culture was then cooled to 25°C for 30 min and then induced for 4 h at 25°C with 0.2 mM IPTG. After washing, the cells were resuspended in lysis buffer (50 mM NaH₂PO₄, 300 mM NaCl, 10 mM imidazole, pH 8.0, 100 μM MgATP) and protease inhibitors (PIs; PMSF, leupeptin, pepstatin A, benzamidine) and then lysed. The lysate was spun at 39,000 g for 45 min at 4°C. The supernatant was loaded onto a column containing Ni NTA superflow resin (Qiagen Inc.) then washed with lysis buffer and wash buffer (50 mM NaH₂PO₄, 300 mM NaCl, 20 mM imidazole, pH 6.0, 100 μM MgATP and PIs). rKLP61F was eluted with 250 mM imidazole, 5 mM magnesium acetate, 2 mM EGTA, 15 mM potassium acetate, pH 7.25, 100 μM MgATP and PIs. rKLP61F containing fractions were pooled, filtered through a 0.2–2 μm Millex-GV syringe filter, and loaded onto a Biogel A-1.5 m column preequilibrated with IMEK (100 mM imidazole, 5 mM magnesium acetate, 2 mM EGTA, 15 mM potassium acetate, pH 7.25) + 300 mM NaCl, 100 μM MgATP and PIs. rKLP61F was then concentrated with high trap Q column (Pharmacia) and eluted using a step gradient from 0.1–1.0 M NaCl in IMEK. The peak fraction from this concentration step was then used for the hydrodynamic studies.

While on the Ni-NTA column (see above), rKLP61F was washed with CSF-XB (10 mM Hepes, pH 7.7, 2 mM MgCl₂, 0.1 mM CaCl₂, 100 mM KCl, 5 mM EGTA, and 50 mM sucrose). 200 μl of Xenopus mitotic extract and 1.8 ml of CSF-XB + 2 mM ATP and 10 mM MgSO₄ were added followed by gentle rocking at room temperature for 20 min. The beads were then returned to 4°C and washed with wash buffer. The phosphorylated rKLP61F was then eluted and treated as stated above for further purification and concentration.

The Stokes radius, Rs, was determined for both phosphorylated and unphosphorylated rKLP61F by loading 500 μl of the peak high trap Q fraction onto a pharmacia superose-6 column preequilibrated with IMEK + 300 mM NaCl. Graphs of Rs vs. −0.5logKav were plotted and the Rs determined from these plots. The standard proteins and their stokes radii included cytochrome C (1.7 nm), carbonic anhydrase (2.0 nm), BSA (3.5 nm), beta amylase (5.4 nm), and thyroglobulin (8.5 nm). The sedimentation coefficient, S value, was determined for rKLP61F on 5-ml 5–20% linear sucrose gradients in IMEK + 300 mM NaCl, 100 μM ATP, and PIs. The gradients were overlaid with a solution containing the rKLP61F, 30 μg BSA (4.4 s), 30 μg aldolase (7.3 s), and 30 μg catalase (11.3 s). The gradients were centrifuged at 218,000 g for 12 h at 4°C. The gradients were determined to be linear by using a refractometer. A linear plot of the S value vs. the percent sucrose was used to calculate the S value for rKLP61F. The molecular mass of rKLP61F was calculated from the measured sedimentation values and Rs.

Drosophila embryos were prepared for immunofluorescence with a modification of the protocol described by Theurkauf (1994). 0–2 h Drosophila embryos were collected on food trays and rinsed from the tray with Triton-NaCl and collected onto a 200-μm nylon mesh rinsed with water. The chorions were removed by immersion in 50% bleach in Triton-NaCl for 2 min. The embryos were then extensively rinsed in double distilled water and Triton-NaCl. The vitelline was removed by immersion in 100% heptane at room temperature for 2 min and fixed by the addition of an equal volume of fresh 37% formaldehyde for 5 min. The formaldehyde was then removed and replaced with 100% methanol. The heptane and most of the methanol were removed, the embryos were incubated in fresh 100% methanol for 2–3 h, and then rehydrated by washes in a graded methanol series and two washes in 100% PBS. The embryos were then washed in PBS-0.01% Triton X-100 (wash solution) and blocked for 30 min to 1 h in PBS-0.01% Triton X-100 containing 5% BSA (blocking solution). Primary antibodies, either rabbit anti–Drosophila tubulin, rabbit anti-p-BCB, rabbit anti-BCB, or mouse anti-β-tubulin (N357; Amersham Life Sciences), were diluted to a concentration of 1 μg/ml in blocking solution and incubated with the embryos at 4°C overnight. In double label experiments, embryos were incubated with the anti-phospho-KLP61F antibody for 4 h at room temperature, then the mouse anti-β-tubulin was added and the incubation proceeded overnight at 4°C. After incubation with primary antibodies, the embryos were washed, blocked as above, and incubated with secondary antibodies, either goat anti–mouse conjugated to lissamine rhodamine, goat anti–mouse conjugated to cy5 or goat anti–rabbit cy3 (Jackson ImmunoResearch) diluted in blocking solution (1:100 for lissamine rhodamine and cy 5 and 1:500 for cy 3) and incubated with embryos for 2–3 h at room temperature. The embryos were washed then incubated in 4′,6′-diamidino-2-phenylindole (DAPI; Sigma) diluted to 0.1 μg/ml in PBS-Triton, for 4–10 min, rinsed washed again, rinsed with 100% PBS, and mounted in p-phenylenediamine (Sigma) 10 mg/ml in 90% glycerol/ PBS. Secondary antibody alone controls were used for all staining conditions. Embryos were visualized on a Leica TCS NT confocal microscope. Images were acquired by averaging 16–32 scans of a single optical section with the pinhole opened for optimal resolution.

Samples were prepared for EM by high pressure freezing and freeze substitution as described previously (McDonald, 1994). Fixed embryos were embedded in Eponate 12/Araldite (Ted Pella) for standard Transmission EM (TEM) or immunoEM or in LR-White (Ted Pella) for immunoEM. 35–50-nm sections were collected on an Ultracut T microtome (Leica) and picked up on copper grids coated with 0.3–0.5% formvar (Electron Microscopy Sciences) for standard TEM or on nickel grids (Ted Pella) coated with carbon and 0.5% formvar. For standard TEM, sections were post stained for 5 minutes in 2% uranyl acetate/70% methanol and 4 min in 0.5% lead citrate.

For immunoEM, sections were blocked in 5% BSA, 0.1% fish gelatin (Sigma) in PBS-0.05% Tween-20 for 30 min (blocking buffer), incubated with primary antibody (either rabbit anti-p-BCB or BCB antibodies diluted to 50 μg/ml or rabbit anti-tubulin diluted to 10 μg/ml in blocking buffer) for 1.5–2 h. Sections were rinsed in PBS-Tween then PBS and incubated in secondary antibodies (goat anti–rabbit IgG F(ab′)₂ [H+L]) conjugated to either 5 or 10 nm gold particle; Ted Pella) diluted 1:20 in blocking buffer for 1 h. Sections were washed as before, incubated in 0.5% glutaraldehyde in PBS for 5 min then rinsed in PBS and water. Sections were post stained in 2% uranyl acetate in 70% methanol for 5 min and in lead citrate for 4 min. Samples were visualized on a Philips 410 LS transmission electron microscope. Identically treated samples were stained with secondary antibody only as controls that in all cases revealed no labeling pattern. Sections labeled with the BCB antibodies revealed only background labeling as well. Digital image enhancement of gold-labeled spindles was done with Adobe Photoshop 5.0 on scanned photographs.

MTs were tracked using markers and plastic overlays as described in McDonald et al. (1977).

To understand the mechanism of action of the bipolar kinesin, KLP61F, in Drosophila early embryos it was important to determine the organization of the mitotic spindle components with which the motor might associate. To accomplish this, we used immunofluorescence and EM to analyze the organization of spindle MTs and the changes in overall spindle morphology that occur throughout mitosis (Figs. 1 and 2).

Fig. 1 shows confocal immunofluorescence of embryos immunostained with a rabbit polyclonal antibody raised against embryonic Drosophila MTs (see Materials and Methods) and the DNA stain DAPI. These images suggest that as prophase begins, duplicated centrosomes split and migrate away from the embryo surface, traveling down opposite sides of the nucleus (Fig. 1 A). The nuclear envelope remains intact throughout prophase (Oegema et al., 1995; Fig. 1 A, inset) and very little or no tubulin staining is observed in the nuclear region. By metaphase (Fig. 1 B), the nuclear envelope has fenestrated at the poles (Stahfstrom and Staehelin, 1984), microtubules have entered the nuclear region and the chromosomes have aligned at the metaphase plate. Dense bundles of MTs are observed in the half spindles while very few astral MTs are visible. The overall morphology of the metaphase spindle is fusiform and the average spindle length is ∼12 μm. Fig. 1, C and D show the tubulin staining pattern from an embryo in which anaphase A (C) has been completed in all nuclei, but only a subset have undergone anaphase B (D). In panel C, sister chromatids have moved completely to opposite poles, yet the length of the spindles, pole-pole, range from 12 to 14 μm, similar to metaphase. The spindles shown in Fig. 1 D, on the other hand, have clearly elongated, reaching lengths of ∼17–18 μm. Note that the decondensation of DNA begins during anaphase B. Also note the distinct change in the morphology of the interzone (region of the spindle lying between the separated chromosomes). Before spindle elongation, the region corresponding to the interzone is fusiform as in metaphase, however, as the spindle elongates the interzone becomes straight edged. We use this change in spindle geometry to distinguish anaphase A from B in the subsequent fluorescence analyses of KLP61F localization (below). One feature of the anti-tubulin immunofluorescence shown here and in other studies is that during both metaphase and anaphase, very few MTs cross the midzone and bridge the half spindles. However, the EM analyses shown below show abundant MTs crossing the midzone suggesting that antigenic sites are masked in immunofluorescence images.

An ultrastructural description of mitotic spindles from several stages in Drosophila syncytial blastoderms has been reported previously (Stahfstrom and Staehelin, 1984). However, the size of these embryos and rapid rate of mitosis during this stage of development results in a great deal of ultrastructural damage during fixation. Therefore, the use of EM to localize low abundance and easily extractable proteins requires preservation methods of a quality exceeding standard chemical fixation (McDonald, 1994). To minimize the potential extraction or mislocalization of KLP61F motors during specimen preservation, we have prepared our samples for EM by high-pressure freezing/ freeze substitution (HPF/FS; for a review of this technique see Steinbrecht and Mueller, 1987; McDonald, 1994; Kiss and Staehelin, 1995). In this method, whole embryos are frozen under pressure then dehydrated and fixed simultaneously at low temperatures to preserve cellular structures rapidly and evenly with minimal extraction. (In contrast to cryofixation, in which unfixed frozen embryos are embedded or viewed on cold stage, the preparation of embryos by HPF/FS allows structures such as MTs or the nuclear envelope to be visualized easily.)

Fig. 2 is shown to illustrate, in fine detail, the organization of spindle microtubule bundles and poles after preservation by HPF/FS. Fig. 2 A shows a bundle of MTs running from the pole to the chromosome at the metaphase plate. These MTs end on an electron-dense halo, presumably the kinetochore and its corona, extending as a bolus from the chromosomes. Interestingly, no kinetochores observed were overtly trilaminar as has been reported in other systems (for review see Brinkley, 1990). Fig. 2, B and C show bundles of MTs that bridge the half spindles during metaphase and anaphase, respectively. Note that in contrast to immunofluorescence, EM clearly reveals abundant midzonal microtubules suggesting that the lack of anti-tubulin fluorescence in the midzone, shown in Fig. 1, results from the masking of antigenic sites. Similar bundles have been identified in vertebrate spindles and have been termed interpolar MT bundles because the entire MT bundle runs from pole to pole (although individual MTs in the bundles do not; Mastronarde et al., 1993). Here, the term interpolar MT bundles will be used to describe bundles of MTs that traverse the spindle midzone and extend into each half spindle.

Fig. 2, D–G show the organization and overall preservation of spindle poles during different stages of mitosis. Throughout the cell cycle, the most prominent structural feature of spindle poles is two orthogonally positioned centrioles (Fig. 2 D). In prophase, centrioles consist of nine outer singlet MTs plus one central singlet (Fig. 2 E). By metaphase, they become increasingly electron dense and doublets can be seen forming on some of the outer MTs. In anaphase, more doublets can be seen on the outer nine MTs and the electron density remains similar to that observed in metaphase. The organization and preservation of MT bundles and spindle poles shown in Fig. 2 is characteristic of all the samples used in this study and reinforces our confidence in the use of HPF/FS for our immunoEM analyses.

Because of their position in relation to the spindle poles, the interpolar MT bundles in Drosophila early embryonic spindles (described above) may be sites for MT-MT interactions that play a role in centrosome positioning. To gain a sense of the organization of the MTs and extent of half spindle overlap within these bundles, a representative interpolar bundle was reconstructed from serial cross-sections through a mitotic spindle just beginning anaphase B, as determined by morphology and pole-pole length (Fig. 3). The results of this reconstruction are shown schematically in Fig. 3 A. Similar to other systems (McDonald et al., 1977; Euteneuer et al., 1982; Mastronarde et al., 1993; Winey et al., 1995), parallel MTs (MTs from the same pole) predominate near the poles, while antiparallel MTs are present in the midzone and extend ∼1–1.5 μm into each half spindle (Fig. 3 A). Fig. 3, B and C show cross-sectional profiles of the MTs in this bundle near the poles (B) and at the midzone (C), respectively. Surprisingly, MTs are more closely associated and appear more ordered near the poles while, in the midzone, MTs are organized into numerous small sets of 2–4 MTs surrounded by an electron-dense material. Antiparallel MTs spaced 50–100 nm apart often run adjacent to one another and electron-dense crossbridges between both parallel and antiparallel MTs are often visible.

To detect KLP61F motors in the mitotic spindle, we raised anti-peptide antibodies that react with a 12-residue segment of the Bim C box located within the tail of KLP61F (Heck et al., 1993). It has been proposed that a threonine residue within the Bim C box is phosphorylated by the mitosis-promoting kinase, cdk1/cyclin B that targets bipolar kinesins to spindles (Sawin and Mitchison, 1995; Blangy et al., 1995). We prepared antibodies using both BCB and p-BCB peptides and determined that these antibodies discriminate between the phosphorylated and unphosphorylated forms of KLP61F.

Both antibodies react specifically and with high affinity with a single band in Drosophila syncytial blastoderm cytosol at the predicted molecular mass for KLP61F (Fig. 4 A; antibodies raised against recombinant KLP61F [Barton et al., 1995] react with the same band on immunoblots [data not shown]). However, Fig. 4 B shows that only the BCB antibody reacts with untreated bacterially expressed KLP61F (rKLP61F), while the p-BCB antibody reacts with rKLP61F only after its phosphorylation by incubation with M-phase Xenopus extracts (lane M) and does not react with untreated rKLP61F (lane U) or with rKLP61F treated with matched interphase extracts (lane I). Parallel experiments using radiolabeled ATP revealed that the rKLP61F in lane M was phosphorylated ∼4.4-fold relative to the rKLP61F in Fig. 4, lane I. Based on these results, we will subsequently refer to the BCB antibody as anti-unphospho-KLP61F and the p-BCB antibody as anti-phospho-KLP61F.

Using these antibodies to probe the phosphorylation state of native and recombinant KLP61F, we determined that the phosphorylation of KLP61F does not appear to affect the homotetrameric state of KLP61F motors in vitro, as both phospho- and unphospho-KLP61F appear to be homotetrameric based on hydrodynamic analyses (Table I). This suggests that the oligomeric state of the motor that we immunolocalized using these antibodies (below) is likely to be homotetrameric.

To determine the general localization of both phospho- and unphospho-isoforms of KLP61F within Drosophila syncytial blastoderms, we used the anti-KLP61F antibodies described above for immunofluorescence analyses. Fig. 5 shows confocal immunofluorescence images of embryos stained with the anti-unphospho-KLP61F overlaid with DAPI staining to indicate the relative position of DNA. Throughout mitosis, unphospho-KLP61F is diffuse and cytoplasmic, displaying no discernible association with MT arrays, nuclei, or spindles.

Embryos stained with anti-phospho-KLP61F show a markedly different staining pattern (Fig. 6). During interphase (Fig. 6 A), phospho-KLP61F staining is punctate and concentrated in the nucleus. This same pattern is observed in prophase (Fig. 6 B), after the centrosomes have separated but before nuclear envelope breakdown. However, on occasion, a very faint but noticeable dot just next to the nucleus can be seen. Double labeling with anti-tubulin antibodies indicate that these dots colocalize with the centrosome (not shown). Little or no colocalization with the microtubule bundles rimming the nuclear envelope or between the centrosomes is observed (not shown). By metaphase, after the fenestration of the nuclear envelope (Fig. 6 C), the phospho-KLP61F staining localizes to the spindle and appears filamentous. Phospho-KLP61 F is particularly concentrated on the half spindles and densely stained bridges between the half spindles in the midzone (Fig. 6 C, arrow). A slightly less concentrated staining is also visible at the poles and a grainy cytoplasmic staining is present. As the chromosomes begin to separate during anaphase A (Fig. 6 D), phospho-KLP61F staining resembles metaphase, with the most intense staining on the half spindles and filamentous bridges between the half spindles in the midzone. Interestingly, very little or no phospho-KLP61F staining can be seen on the astral MTs that begin to emanate from the poles at the onset of anaphase (see Fig. 1 C). By anaphase B (Fig. 6 E; see interzone morphology in Fig. 1 D) the staining pattern of the motor changes dramatically, as nearly all of the phospho-KLP61F is present in the spindle between the separating nuclei. Again, filaments of phospho-KLP61f staining cross the midzone while little or no staining is associated with the spindle poles and astral MTs or within the daughter nuclei. In telophase (Fig. 6 F), as the nuclear envelopes reform around daughter nuclei, phospho-KLP61F is most pronounced on two adjacent patches on the remnants of the central spindle. Close examination of the phospho-KLP61F staining pattern in the daughter nuclei also reveals that bright punctae begin to form within them. Based on these data, we conclude that, similar to its Eg5 homologues from frog and human (Sawin and Mitchison, 1996; Blangy et al., 1996), KLP61F associates with the mitotic spindle in a phosphorylation dependent fashion. Moreover, phospho-KLP61F motors associate with “filaments” crossing the spindle midzone that we believe correspond to the interpolar bundles as determined by EM (below).

The localization of phospho-KLP61F to distinct spindle regions may provide information about the subsets of MTs upon which the motor acts. Conversely, it may simply mirror differences in tubulin concentration within the spindle. To address this, Fig. 7 shows confocal immunofluorescence overlaying anti-phospho-KLP61F staining with anti-β-tubulin. Instead of determining the concentration of phospho-KLP61F throughout the spindle in absolute terms, these images show the relative fluorescence intensity of phospho-KLP61F when compared with MTs. During metaphase (Fig. 7 A), the region of the half spindles between the poles generally shows relatively equal intensity for tubulin and phospho-KLP61F fluorescence. In contrast, the poles and astral MTs display disproportionately low phospho-KLP61F staining whereas the interpolar MT bundles crossing the midzone display disproportionately high phospho-KLP61F fluorescence. Somewhat similar results are obtained as the spindle elongates during anaphase B (Fig. 7 B), as the poles and astral MTs display disproportionately low phospho-KLP61F and bundles of MTs traversing the midzone display disproportionately high phospho-KLP61F staining. However, in metaphase some phospho-KLP61F staining is visible at the poles while, during anaphase B, polar localization of phospho-KLP61F is nearly undetectable. These data support the hypothesis that phospho-KLP61F is associated with “filaments” corresponding to interpolar MT bundles and becomes concentrated in the region of these bundles containing antiparallel MTs.

Although our immunofluorescence analyses indicate that phospho-KLP61F is present in the spindle, they cannot resolve whether the motor associates directly with individual spindle MTs. We performed immunoEM analyses using our anti-phospho-KLP61F antibody in order to examine the spatial relationship between the bipolar kinesin and MTs. Because the antibody is specific for a phospho-epitope in the tail of KLP61F these analyses allow us to examine not only whether phospho-KLP61F associates with spindle MTs but also how it is positioned relative to them by determining the position of the tail.

Fig. 8 shows the results of anti-phospho-KLP61F immunostaining performed on thin sections from embryos undergoing either metaphase (Fig. 8, A–D) or anaphase B (Fig. 8, E and F) labeled with secondary antibodies conjugated to 10- or 5-nm gold particles. The images shown in Fig. 8, A–C and E are from embryos embedded in the standard immunoEM resin, LR-White. At low magnification (Fig. 8, A and E), the staining patterns observed are consistent with the immunofluorescence shown above. Dense staining of the metaphase half spindles (Fig. 8 A), the anaphase interzone (E), and filaments bridging the midzone in both spindles (A and E) is observed. Gold labeling is also present on the poles but does not appear as concentrated as in the half spindles. The gold particle density in the spindle region is extremely high in relation to the cytoplasm (∼253 vs. 7.3 particles/μm² during metaphase and ∼237 vs. 54 particles/μm² during anaphase B in spindles vs. cytoplasm, respectively). Higher magnification reveals that gold labeling is clearly concentrated on or near interpolar MTs in the metaphase midzone (Fig. 8, B–D) and anaphase interzone (F). Note that immunolabeling only occurs at the surface of the section (Stierhof et al., 1991) and this accounts for the patchy labeling of MTs. The close juxtapositioning of gold particles to MTs was best seen in sections from embryos embedded in Eponate 12/Araldite resin to optimize spatial resolution (Fig. 8, D and F). Fig. 8 D shows the staining pattern along a metaphase interpolar MT bundle and Fig. 8 F shows the labeling of a bundle from an anaphase spindle. In both images it is apparent that the gold labeling is located extremely close to the surface of MTs. Measurements of the distance between gold particles and MTs indicates that, on average, gold particles lie ∼9 and 7 nm from the surface of MTs during metaphase and anaphase, respectively (see Table II).

Our structural model posits that KLP61F motors cross-link interpolar MTs as bipolar homotetramers. A prediction of this model is that our anti-phospho-KLP61F antibody should lie at spacings ∼60 nm apart at opposite ends of MT crossbridges (Fig. 9 A). In support of this hypothesis, we observed gold-labeled MT cross-links when the images shown in Fig. 8 were digitally enhanced to increase contrast. Fig. 9 B (top panel) shows a region of the anti-phospho-KLP61F-labeled interpolar MT bundle indicated by the inset in Fig. 8 A. In this image, a portion of the gold particles can be seen clearly associated with electron-dense filaments running between adjacent MTs (red arrows point down the long the axis of these filaments). In many cases, such as the one shown, these filaments are extremely numerous, producing a formation similar to rungs on a latter. Similar crossbridges were observed in different regions of multiple spindles and all showed nearly identical morphologies (Fig. 9 B, bottom panels). We identified 115 of these filaments that were decorated with gold particles on both ends. The spacing between gold particles on the filaments ranged from ∼30–90 nm but was most commonly between 60–65 nm (Fig. 9 C).

The purpose of this study was to visualize KLP61F within mitotic spindles of Drosophila syncytial blastoderms, in order to determine whether KLP61F acts as a MT-MT sliding motor involved in spindle pole separation or if it cross-links MTs to a spindle matrix and organizes the poles. To this end, we immunolocalized KLP61F in mitotic spindles using immunofluorescence and immunoEM and we performed a detailed structural analysis of the mitotic spindle components with which KLP61F putatively associates. The results of these analyses support the hypothesis that KLP61F participates in the formation and function of bipolar mitotic spindles by cross-linking and sliding MTs in relation to one another.

In order for a MT-MT sliding mechanism to play a role in spindle pole positioning, the half spindles must be in contact through overlapping antiparallel MTs, most likely at the spindle midzone. Yet, tubulin immunofluorescence shows little or no staining of the spindle midzone in Drosophila embryos. Strikingly, however, EM reveals the presence robust interpolar MT bundles running through this region to bridge together the half spindles. Reconstruction of a representative interpolar bundle from serial cross-sections indicates that they consist of two MT arrays emanating from opposite poles that interdigitate in the midzone region. The region of antiparallel MT interdigitation extends ∼1 μm into each half spindle and, thus, motor proteins that drive MT-MT sliding within these interpolar MT bundles could play a role in spindle pole positioning. For example, plus-end-directed motors could cross-link and slide apart antiparallel MTs in the region of interdigitation resulting in a repulsion of the poles (outward force), while minus-end-directed motors could pull the poles together (inward force).

The morphological characteristics of Drosophila embryonic spindles are consistent with the exertion of such counterbalancing forces. Since MTs will grow as straight tubes in the absence of an external force, the curvature of the interpolar MT bundles on the perimeter of the fusiform metaphase and anaphase A spindles (Fig. 1, B and C) suggests that they are being subjected to compressive forces resulting from a counterbalance of antagonistic forces pulling the poles inward and pushing them outward. An abrupt change in this counterbalance appears to coincide with the onset of anaphase B, which is characterized by straightening of interpolar MTs (Fig. 1 D). In such a scenario, a release of the inward force would relax the compressive forces acting on MTs allowing the poles to be pushed apart. A similar counterbalance of forces involving bipolar kinesins has been proposed for S. cerevisiae (Saunders et al., 1997b), S. pombe (Pidoux et al., 1996), and A. nidulans (O'Connell et al., 1993).

The hypothesis that KLP61F functions by cross-linking and sliding spindle MTs apart is supported by the observations that a severe loss-of-function mutation in the gene encoding KLP61F results in a late larval lethal phenotype, hallmarked by monoastral spindles in proliferating cells (Heck et al., 1993) and that the native KLP61F holoenzyme is a bipolar homotetramer, capable of cross-linking adjacent MTs and sliding them in relation to one another (Cole et al., 1994; Kashina et al., 1996a,b). To test this model, it is essential to determine the structural basis of the interaction between KLP61F motors and the mitotic spindle. The immunolocalization of KLP61F provides data relevant to this issue by revealing that KLP61F associates with spindle MTs in a phosphorylation dependent fashion, that it associates directly with interpolar MT bundles, and, strikingly, that bipolar KLP61F motors cross-link MTs within these bundles.

Work on the bipolar Eg5 kinesin in humans and frogs has shown that the phosphorylation of bipolar Eg5 kinesin motors by cdk1/cyclinB at a specific site in the BCB targets these motors to spindles. Similarly, our results show that KLP61F associates with the spindle when phosphorylated in the BCB. This contrasts with the results from studies on the bipolar kinesins from the budding yeast S. cerevisiae that lack BCBs (Heck et al., 1993) and fission yeast S. pombe where the phosphorylation of this site does not affect the spindle association of the motor (Drummond et al., 1998). The significance of this difference is unclear but may reflect the fact that in both types of yeast, mitosis occurs completely within the confines of the nuclear envelope so that mitotic motors can interact with spindle MTs but not cytoplasmic MTs. However, in most other systems the spindle forms in the cytoplasm and precocious contact between mitotic motors and cytoplasmic MTs might perturb interphase MT arrays.

In Drosophila early embryos, the nuclear envelope never completely breaks down but instead fenestrates only at the points where spindle MTs enter the nuclear region. Our results showing that phospho-KLP61F associates with nuclei and spindles while nonphosphorylated KLP61F is excluded from these compartments, suggest that phosphorylation serves to sequester an active pool of the motor in a compartment where it can contact spindle MTs but not cytoplasmic MTs. How phosphorylation regulates the localization of KLP61F is unknown. It does not appear to do so by inducing the formation of bipolar tetramers, as our hydrodynamic analyses show that both phospho- and unphospho-KLP61F motors are predominantly tetrameric (Table I). It is also possible that phosphorylation increases the processivity of KLP61F motor activity that would, in turn, increase the time the motor spends associated with MTs. However, data regarding this are lacking. Clearly, the mechanism of the phospho-regulation of KLP61F localization and function requires additional investigation and will not be discussed further here.

Our data show a clear association of KLP61F with interpolar MT bundles, visible as thin filaments of phospho-KLP61F bridging the midzone during both metaphase and anaphase by immunofluorescence (Fig. 6) that colocalize with interpolar MT bundles (Fig. 7). Interestingly, double label immunofluorescence images also show that the concentration of KLP61F is disproportionately high on MT bundles in the midzone and disproportionately low at the poles and on astral MTs when compared with tubulin staining. (Whole mount immunofluorescence provides a superior method for comparing the concentrations of motors and microtubules than does immunoEM of ultrathin sections, so the relative enrichment for KLP61F in these bundles is not apparent in Fig. 8.) This pattern shows a striking correspondence with the region of half spindle interdigitation within interpolar MT bundles suggesting that phospho-KLP6F has a particular affinity for antiparallel MTs (conversely, KLP61F could have an affinity specifically for the plus-ends of spindle MTs). However, it should be noted that other factors, such as the masking of both tubulin and phospho-KLP61F antigenic sites within the midzone might also affect the relative intensity of their fluorescence; for example, if the masking of tubulin and KLP61F antigenic sites in this region is nonequivalent, the ratio of their fluorescence intensities would not mirror their relative concentrations. We note, however, that the region of the spindle showing disproportionately high phospho-KLP61F fluorescence intensity extends ∼1–2 μm into each half spindle, well beyond the zone of potential antigenic masking, indicating that the concentration of KLP61F in regions of MT interdigitation is truly disproportionately high.

Whereas immunofluorescence reveals a general colocalization of KLP61F with interpolar MT bundles, immunoEM shows a close juxtaposition of the motor with the MTs within these bundles. At low magnification the immunolabeling pattern is obviously linear, similar to that expected from a microtubule stain. At higher magnifications, the gold particles are clearly aligned along MTs, and closely juxtaposed to the MT surfaces (Table II). Since our phospho-KLP61F antibody recognizes an epitope in the tail domain of the motor, this close juxtaposition of gold particles and MTs strongly suggests that the tail is positioned adjacent to a motor domain (see Fig. 9 A). The average distance between gold particles and the MT surface is ∼10 nm, approximately the diameter of a KLP61F motor domain, supporting this suggestion. Furthermore, the analysis of the deduced sequence of KLP61F together with rotary shadowing EM analysis of the native holoenzyme, place the BCBs at the opposite end of an α-helical rod spaced ∼60 nm apart, consistent with the model shown in Fig. 9 A.

Direct evidence in support of MT-MT cross-linking by bipolar KLP61F motors was obtained by visualizing gold decorated spindles after digital enhancement of contrast. In the images shown in Fig. 9 B, for example, gold-labeled phospho-BCBs clearly associate with electron-dense crossbridges at a spacing of ∼60 nm, in strong support of our previously proposed structural model for KLP61F. The average distance of 10 nm between the gold-labeled BCB and the MT surface (discussed above) summed with the ∼60 nm spacing between the gold particle clusters on individual crossbridges, suggests a spacing between cross-linked MTs of ∼80 nm, consistent with the relatively wide spacing between antiparallel MTs within interpolar MT bundles (Fig. 3 C). However, KLP61F is also clearly present in regions of the interpolar MT bundles that contain only parallel MTs, in which cross-links of a similar length (∼60 nm) are also observed. Therefore, these data are most consistent with the idea that bipolar KLP61F motors cross-link both parallel and antiparallel MTs but other factors (MAPs or motors) must contribute to the generally closer packing of nearest neighbor MTs observed within parallel MT bundles (Fig. 3 B). Moreover, the close spacing of parallel MTs near the poles suggests that in this region of the spindle, KLP61F motors must cross-link parallel MTs that are not nearest neighbors. Further testing of this hypothesis requires a more thorough understanding of the organization of parallel vs. antiparallel MTs within interpolar MT bundles.

Based on the localization of KLP61F shown here, its transport properties, and the phenotype of severe loss of function mutants, we propose the following expanded model for KLP61F function (shown schematically in Fig. 10). During centrosome migration, KLP61F is active and homotetrameric but sequestered away from MTs within the intact nuclear envelope. As the nuclear envelope fenestrates during prometaphase and MTs enter the nuclear region, KLP61F is free to bind and cross-link parallel and antiparallel spindle MTs, alike, riding slowly toward their plus-ends. The ability of KLP61F to cross-link MTs of both polarities assumes that theKLP61F motor domain is capable of swiveling 360° as has been shown for conventional kinesin (Hunt and Howard, 1993). Because of the plus-end–directed motility of KLP61F, a steady state is established in which KLP61F is present throughout the spindle but concentrated in the midzone, where the plus-ends of MTs from both half spindle interdigitate. The interaction of KLP61F with parallel MTs results in the formation of MT bundles capable of capturing kinetochores or interacting with MT bundles emanating from the opposite pole but, since the force vector on both MTs is the same, no net MT-MT sliding occurs. Conversely, when KLP61F interacts with antiparallel MTs, cross-linked MTs slide apart, generating an outward force on the poles. During metaphase and anaphase A, this force produces isometric tension counterbalancing an inward force to hold the poles apart. As the antagonistic inward force is released during anaphase B, the outward force serves to drive the poles apart. In other systems, where anaphase B involves strong pulling forces on the poles (Kronebusch and Borisy, 1982; Aist et al., 1993), the MT-MT interactions mediated by KLP61F might also serve as a brake, regulating the rate of spindle elongation.

While our data clearly show KLP61F cross-linking interpolar MTs, it does not directly address several important issues. First, implicit in this model is that centrosome migration during prophase does not involve KLP61F, since this occurs before nuclear envelope fenestration when the motor is sequestered in the nucleus (Fig. 1 A). While we cannot exclude the possibility that the low level of KLP61F present in the cytoplasm is sufficient to drive prophase centrosome migration, it is also conceivable that spindle formation proceeds normally before nuclear envelope fenestration in the absence of KLP61F function and that the monoastral spindles seen in KLP61F mutants result from the collapse of a preformed bipolar spindle. It should be noted that such a spindle collapse has been observed in S. cerevisiae mutants carrying null alleles for the bipolar kinesins, cin8 and kip1 (Saunders et al., 1997). Moreover, although the monoastral spindles observed in bipolar kinesin mutants in A. nidulans (Enos and Morris, 1990), and S. pombe (Hagan and Yanagida, 1990) have been interpreted as resulting from defects in the initial separation of the spindle poles during spindle assembly, one cannot rule out the possibility that these structures result from the collapse of already assembled spindles. Direct visualization of spindle formation in KLP61F mutants should help resolve precisely when KLP61F function is needed. Second, although we have not observed KLP61F concentrated on kinetochore MT bundles or associated with a spindle matrix we cannot rule out the possibility that such interactions can occur at least in some systems.

The bipolar kinesin, KLP61F is critically important for mitosis. The data reported here, in concert with the motility properties of the motor and the phenotype of severe loss of function mutants, strongly suggest that KLP61F functions in a sliding filament mechanism to hold the spindle poles apart during metaphase and anaphase A and to drive spindle elongation during anaphase B. Given the strikingly similar molecular architecture of myosin II and KLP61F, it is appealing to speculate that the bipolar configuration of force generating enzymes involved in the sliding of cytoskeletal filaments is highly conserved and a fundamental feature of cytoskeletal activity.

This work was supported by National Institutes of Health (NIH) grant number GM55507 to J.M. Scholey, and D.J. Sharp was supported by NIH postdoctoral fellowship number GM19262.

BCB

unphosphorylated BimC box

cdk1

cyclin-dependent kinase 1

CSF

cytostatic factor

HPF/FS

high-pressure freezing/freeze substitution

HSS

high speed supernatant

immunoEM

immunoelectron microscopy

LB

Luria broth

MAP

microtubule-associated protein

MT

microtubule

p-BCB

phosphorylated BCB

PI

protease inhibitor

Rs

Stokes radius

TEM

transmission EM