Mitotic spindles isolated from the diatom Stephano-pyxis furris will elongate in vitro in the presence of ATP with a concurrent decrease in the width of the zone of microtubule overlap. A spindle-associated phosphoprotein that co-localizes with the zone of microtubule overlap in isolated spindles serves as a convenient marker for midzone-associated proteins other than microtubules. We have used a monoclonal antibody that labels this protein when it is artificially thiophosphorylated and studied its redistribution during spindle reactivation in vitro. As the spindle elongates midzone label accumulates in a successively narrower and brighter ring at the spindle midpoint with increasing time in ATP. Biotinylated bovine microtubule segments polymerized onto the ends of the diatom microtubules increase the overall width of the zone of microtubule overlap and serve as a marker for the boundary of the original diatom overlap zone. During elongation in ATP, the biotinylated segments move into the area marked by the monoclonal antibody, which does not decrease in width until the spindle has elongated to the point at which the zone of microtubule overlap delineated by the newly polymerized microtubules is smaller than the original overlap zone. We use these results to develop a model to explain the behaviour of nonmicrotubule midzone-associated proteins during spindle elongation.
The mitotic spindle is responsible for the equidistribution of genetic material to daughter cells at each cell division. One approach toward understanding how the mitotic spindle functions is to identify parts of the mitotic apparatus and analyze how these parts move relative to one another. Microtubules comprise the predominant structural element in the mitotic spindle and they fall into two major categories: kinetochore microtubules that extend from the chromosomes to the poles, and polar microtubules that extend from the poles to interdigitate in the center of the spindle (McDonald et al. 1977). In addition to microtubules, indirect immunofluorescence studies have identified a variety of proteins associated with discrete regions of the mitotic spindle, such as the spindle poles, midzone or kinetochores (Vandre & Borisy, 1986; Cooke et al. 1987; Kingwell et al. 1987). We believe that the diatom spindle is well suited for analyzing the rearrangements of spindle components that occur during anaphase because the components that comprise the mitotic spindle of the diatom are positioned with a high degree of structural regularity relative to the mitotic spindles of mammalian and higher plant cells (Pickett-Heaps & Tippit, 1978; McDonald et al. 1979).
We have used spindles isolated from the diatom Stephanopyxis turris to analyze the mechanism of spindle elongation (anaphase B). In S’, turris the kinetochore and non-kinetochore (central spindle) microtubules are spatially separated and the closely packed central spindle microtubules of each half-spindle are of relatively uniform length so that the zone of microtubule overlap of the spindle midzone is well defined. With the addition of ATP, the spindles elongate concurrently with a decrease in the extent of the zone of microtubule overlap. Electron microscopic evidence suggests that the ultrastructural organization of the two interdigitating half-spindles remains relatively unchanged during spindle elongation and that the observed increase in length results from the sliding apart of the two half-spindles (Cande & McDonald, 1986). We have found that spindle elongation also occurs in DNase I-digested spindles in which there are no other cytoplasmic elements in association with the isolated central spindle (Baskin & Cande, 1988). In view of these results, we believe that the zone of microtubule overlap is the most likely location of force generation for half-spindle sliding. Studies by Saxton & Macintosh (1987), in which the distance between bars bleached perpendicularly across the spindle midzone increased during anaphase, and earlier work by Leslie & Pickett-Heaps (1983) on spindles cut in various locations by a laser microbeam, furnish additional evidence that microtubule sliding and force generation take place in the zone of microtubule overlap. Since, in S. turns, spindle elongation in vitro can occur in the absence of exogenous tubulin and in the presence of drugs that promote microtubule depolymerization, we have eliminated models that suggest that microtubule polymerization is a requirement for force generation during anaphase B. Furthermore, isolated spindles in which purified bovine neurotubulin has been polymerized onto the ends of the original diatom microtubules at the edge of the zone of microtubule overlap will elongate in ATP, resulting in spindles in which the overlap zone is composed entirely of neurotubulin (Masuda et al. 1988). In order for this to occur, we predict that some midzone components, including the mechanochemical enzymes responsible for anaphase B, must remain stationary in the midzone even as the original half-spindle microtubules slide away from the center of the spindle.
To study the behaviour of proteins in the zone of microtubule overlap during spindle reactivation we use a marker for overlap zone proteins and for microtubule ends. Previously we demonstrated that isolated spindles must be highly phosphorylated to sustain spindle elongation in vitro (Wordeman & Cande, 1987). A spindle-associated kinase(s) can use ATPyS as a substrate to thiophosphorylate spindle-associated proteins///vitro. We have used an antibody against thiophosphorylated proteins (anti-thio-P) to visualize the distribution of these proteins in the overlap zones, poles and kinetochores. In this paper we describe the behavior of the thiophosphorylated proteins in the zone of microtubule overlap during spindle elongation in vitro. The extent of microtubule overlap is determined by polymerizing biotinylated microtubule segments onto the edges of the original zone of microtubule overlap prior to reactivation in ATP (Masuda et al. 1988). We show that as the two halfspindles slide apart in ATP the midzone-associated proteins remain localized specifically in the zone of microtubule overlap, even when the microtubules are composed entirely of exogenous bovine tubulin that has been polymerized onto the ends of the diatom microtubules prior to ATP addition. The result is used to develop a theoretical model to explain the behavior of spindle midzone-associated proteins during anaphase B.
MATERIALS AND METHODS
Thiophosphorylation and reactivation of isolated mitotic spindles
S. turris cells were synchronized and harvested as described by Cande & McDonald (1986) and Wordeman et al. (1986). Nuclei were isolated in MEM buffer (50mM-Mes, pH 6·5, 10mM-EGTA, 5 mM-MgSO4, 5 mM-dithiothreitol, 1/200 proteolytic inhibitor cocktail (Cande et al. 1983), 0·5 mM-phenylmethylsul-fonyl fluoride and 10^M-taxol) and 30% glycerol, 0·2% Brij 58 and SOftM-ATPyS. Spindles were reactivated in MEM buffer containing 1 mM-ATP, then fixed for 20 min in 0·1% glutaraldehyde, 0 ·05% paraformaldehyde. Fixed spindles were reduced in lmgml-1 sodium borohydride in Tris-saline, pH 7 4, rinsed in and then stained with a monoclonal antibody to thiophosphorylated proteins (Gerhart et al. 1985; Wordeman & Cande, 1987). Fluorescently labeled mitotic spindles were photographed with Kodak Technical Pan SO-115 film using a Zeiss Photoscope HI. Spindle lengths were measured from a TV monitor using a DAGE-MTI low light level TV camera.
Preparation and reactivation of spindles in the presence of exogenous biotinyla ted tubulin
Spindles were prepared as described above except that, prior to incubation in 1 mM-ATP, the spindles were incubated for 5 min in MEM buffer plus 20βM-biotinylated bovine neurotubulin (Masuda & Cande, 1987; Mitchison & Kirschner, 1985). Fixed spindles were labeled with a monoclonal antibody against thiophosphorylated proteins, rhodamine-conjugated avidin (Vector Labs) and fluorescein-conjugated goat anti-mouse (Cappel Labs).
Isolated spindles incubated with ATPγS and then labeled with an antibody to thiophosphorylated proteins, exhibit a distinct staining pattern. Before the addition of ATP, the thiophosphorylated proteins are located in the spindle midzone, in kinetochores and in a portion of the pole complex (Fig. 1A). Measurements of the extent of the midzone label versus the total length of a population of isolated spindles are shown in Fig. 2. Before the addition of ATP, the extent of the anti-thio-P midzone label is approximately 25–30% of the total spindle length. This agrees with ultrastructural measurements of the extent of microtubule overlap in vivo and in isolated spindles (McDonald et al. 1986) and with the extent of the midzone as defined by the zone of increased birefringence observed with polarization optics in unfixed, isolated spindles (Cande & McDonald, 1985; Baskin & Cande, 1988). It is likely that the midzone distribution of thiophosphorylated proteins co-localizes with the zone of microtubule overlap in isolated spindles.
Thiophosphorylated spindles that are reactivated in 1 mM-ATP, in the absence of exogenous tubulin, exhibit a striking alteration in the midzone anti-thio-P label (Fig. IB and C). The width of the central spindle labeling decreases considerably and becomes brighter in intensity, while the distribution of label at the poles remains unaltered. After several minutes in ATP, the midzone anti-thio-P label consists of a narrow band that, in many spindles, appears to bulge outward from the sides of the spindle. In spindles that have lost structural integrity after reactivation and are bent in half, the anti-thio-P label is located at the junction between the bent half-spindles (Fig. 1C and F). The increased intensity of the midzone anti-thio-P label argues against the loss of thiophosphorylated proteins into the surrounding milieu during spindle elongation. Furthermore, the behavior of phase-dense material in the overlap zone observed on video during spindle reactivation in vitro is identical to the behavior of the thiophosphorylated material in these experiments (Baskin & Cande, 1988).
When biotinylated tubulin is polymerized onto the plus ends of the diatom microtubules the extent of the zone of microtubule overlap is increased, and the extent of subsequent spindle elongation is also increased. Furthermore, the position of the added tubulin marks the plus ends of the original diatom interzone microtubules (Masuda et al. 1988). Ultrastructural studies show that, following tubulin addition, there are few microtubule fragments in the spindle midzone, and that the length of the half-spindle microtubules has increased (Masuda et al. 1988). Biotinylated microtubules also polymerize extensively off the spindle poles. These two areas overlap in unreactivated spindles such as the one in Fig. 3A and E. Slight bleed-through from the rhodamine-labeled polar biotinylated microtubules into the fluorescein channel results in the poles appearing more heavily labeled by the anti-thio-P antibody in Fig. 3 than in Fig. 1. During reactivation the biotinylated regions of the microtubules were observed to move into the midzone (Fig. 3A,B,E,F) and eventually to colocalize with the anti-thio-P label (Fig. 3C,D,G,H). This result was confirmed by quantifying the distribution of anti-thio-P label and newly added biotinylated microtubules for populations of isolated spindles before and after the addition of 1 mM-ATP (Fig. 4 and Table 1). Although the midzone anti-thio-P label eventually decreases in width in these spindles in the presence of ATP, this decrease becomes pronounced only after the spindles have elongated to the extent that the edges of the thio-P midzone label and the ends of the added biotinylated segments are coincident (Fig. 4 and Table 1). Moreover, no spreading outward of the anti-thio-P label was observed, either before ATP addition when extensive biotinylated tubulin incorporation had doubled the size of the overlap zone, or at early times after the addition of ATP, before the extent of overlap had decreased.
Our studies suggest that some spindle midzone components, including perhaps the anaphase B motor, remain stationary in the midzone even as overlapping microtubules move relative to each other. This suggests that another discrete structure, equivalent perhaps to the osmophilic matrix seen in many electron micrographs, exists in the zone of microtubule overlap and may play a major role in spindle elongation. Furthermore, this implies that the microtubules of one half-spindle do not necessarily push off against the microtubules of the other half-spindle. Rather, force may be transmitted via a third structural component, such as a multimeric motor complex residing between the antiparallel microtubules.
The sliding model for spindle elongation, proposed by McIntosh et al. (1969), posits the existence of dynein-like crossbridges that are permanently attached to one microtubule and transiently to another (Fig. 5A and E). However, this model is inconsistent with our results and also with studies of insect spermatocytes (Nicklas & Koch, 1972) on the displacement of granules positioned on anaphase spindles. Granules placed in the spindle interzone do not move, confirming earlier observations of Bajer & Mole-Bajer (1972) on Haemanthus, that proximal-poleward forces are present only within the bordering half-spindles (i.e. between the chromosomes and the poles). Our results describing the behavior of the midzone protein(s) support these observations. The causal relationship between the persistent interzone localization of matrix material and the balanced poleward movement of closely opposed antiparallel microtubules in opposite directions is not known. However, it is clear, on the basis of our results, that it is possible for material to remain stationary in the spindle interzone during spindle elongation in vitro. The behavior of the anti-thio-P labeling of midzone components is not consistent with any model of spindle organization in which force-generating components are anchored permanently to microtubules and become transported out of the overlap zone as the half-spindles slide apart. Although it is possible that, in vivo, disassembled motors as well as tubulin subunits could be in equilibrium with the overlap zone, in vitro this is not possible because a soluble pool of motors molecules is not present. Finally, we have not eliminated the possibility that the labeled midzone material that is not in contact with antiparallel microtubules simply disassociates from the mitotic spindle (Fig. 5B). However, we believe that this is unlikely because of the increase in staining intensity of the labeled midzone material that we observe during spindle elongation (Fig. IB).
We have thought of two explanations consistent with our results (Fig. 5C and D). In one model, a matrix is present in the midzone and is anchored to free microtubule ends at the edges of the zone of microtubule overlap. This matrix would become passively aggregated as half-spindles slide and the ends approach each other (Fig. 5C). Alternatively, a midzone protein complex could move toward the plus ends of microtubules during spindle elongation in order to remain in association with domains of antiparallel microtubules (Fig. 5D). This protein complex would slide along microtubules in an ATP-dependent manner, which is the behavior originally suggested by Margolis et al. (1986) for STOP proteins.
We can distinguish between these two alternatives by considering the behavior of the midzone-associated proteins when biotinylated tubulin is polymerized onto the ends of diatom microtubules, increasing the size of the overlap zone. Thiophosphorylated components of the overlap zone do not move onto the newly polymerized microtubules; thus microtubule polymerization does not simply extend an elastic matrix that is anchored to the ends of the original diatom microtubules in the overlap zone. Furthermore, the movement of biotinylated microtubule segments into the labeled midzone region suggests that the labeled midzone material is not a matrix that maintains an association with diatom microtubule ends even after foreign tubulin is polymerized onto these ends (Fig. 4). We believe that these results argue against the passive aggregation of midzone material as the spindle elongates.
The diagrams in Fig. 5E and F are extensions of the two models presented earlier for permanently anchored midzone proteins (Fig. 5A and E) versus movable midzone protein (Fig. 5D and F) with the addition of biotinylated microtubule segments. One can see that the results shown in Fig. 4 force us to reject the model shown diagrammatically in Fig. 5E in favor of that presented in Fig. 5F. Although we do not know how many proteins actually constitute the midzone material, we favor the idea that at least one of these components can move along microtubules in an ATP-dependent manner and therefore maintain a fixed position in the midzone relative to the sliding microtubules.
Recently, using indirect immunofluorescence, several researchers have identified proteins from diverse organisms that remain localized in the spindle midzone and eventually in midbodies during spindle elongation in vivo (Sellitto & Kuriyama, 1988; Cooke et al. 1987; Kingwell et al. 1987). However, the function of these proteins are unknown. Nevertheless, these studies and previous ultrastructural studies suggest that a variety of spindle-associated proteins accumulate in the spindle midzone during mitosis. These proteins may include structural components, regulatory proteins and mechanochemical transducers of force production, all of which could play roles in controlling spindle elongation, and would have an association with microtubules similar to that postulated above.
We have previously demonstrated that the thiophosphorylated proteins labeled in the spindle midzone most likely correspond to only one major species of 205K (K=103Mr). Furthermore, we have shown that the phosphorylated state of this 205K protein is positively correlated with reactivation of spindle elongation in vitro. This protein is released from the nucleus when spindle microtubules are depolymerized, suggesting that it (and perhaps other midzone proteins) is not part of an independent structural nuclear matrix but is dependent on the integrity of the spindle microtubules to remain associated with the mitotic apparatus (Wordeman & Cande, 1987). Microtubule-associated proteins (MAPs) in the range of 200–215K that co-localize with interphase, spindle and midbody microtubules have been described in a variety of cell systems (Bulinski & Borisy, 1980; Parysek et al. 1984; Goldstein et al. 1986). Some of these 200–215K MAPs are phosphoproteins (Gard & Kirschner, 1987; Vandre & Borisy, 1986). Izant et al. (1983) described a 200K MAP that is unique to spindle microtubules in PtK cells. A 210K MAP isolated from Xenopus eggs that promotes plus-end addition of tubulin subunits into microtubules has been identified (Gard & Kirschner, 1987). Although the cytoplasmic distribution of this MAP is not known, such a MAP could play a role in regulating tubulin polymerization in the spindle midzone during spindle elongation.
In summary, we have shown that a component of the spindle midzone remains stationary relative to sliding antiparallel microtubules during spindle elongation in vitro. Previously, using fluorography, indirect immunofluorescence and immunoblots, we have demonstrated that the major protein thiophosphorylated in the spindle midzone is a 205K peptide whose phosphorylated state is correlated with enhanced spindle elongation in vitro. On the basis of these results we propose that midzone-associated proteins important for spindle elongation, including putative force-generating molecule(s), remain in the spindle midzone while newly polymerized antiparallel microtubules belonging to each half-spindle slide past each other.