Actin in spindles of Haemanthus Katherinae endosperm : II. distribution of actin in chromosomal spindle fibres, determined by analysis of serial sections

In this paper we report on the arrangements of actin-containing filaments in chromosomal spindle fibres of Haemanthus endosperm cells; these arrangements were determined by 3-dimensional reconstructions deduced from micrographs of consecutive longitudinal sections. In the preceding paper we described the effects on these cells of various glycerination procedures, and we argued that spindle structure is preserved during the glycerination procedures: chromosome positions do not change during glycerination, spindle fibre birefringence is maintained during glycerination, and, electron-microscopically, spindle microtubules seem qualitatively normal (Forer & Jackson, 1979). In considering whether actin-containing filaments in spindles might have a functional role, we argued that if spindle actin functions in producing force for chromosome movement, then the spindle actin must be associated with each and every chromosomal spindle fibre, because the chromosomal spindle fibres transmit forces to the associated chromosomes. If, on the other hand, actin identified in the spindle was just ‘trapped’ in the spindle as the spindle formed, or just ‘relocated’ into the spindle during the preparatory procedures, as various other workers have speculated (review in Forer, 1978 a), then actin-containing filaments would be present in the spindle but would be arranged irregularly and would not necessarily be associated with chromosomal spindle fibres. One criterion, then, for assuming a possible functional role of actin-containing filaments in the spindle is to see if actin-containing filaments are associated with each chromosomal spindle fibre. As reported herein, actin-containing filaments are present in each chromosomal spindle fibre so far analysed. Preliminary description of these results has been given elsewhere (Forer, 1978a; Forer & Jackson, 1978).

Haemanthus endosperm cells were glycerinated and treated with heavy meromyosin (HMM) or HMM subfragment 1 (S1), as described previously (Forer & Jackson, 1979). We studied 2 cells, one in metaphase and one in anaphase: the metaphase cell was glycerinated with 50% glycerol, then centrifuged and resuspended in HMM in standard salts solution, and then pelleted and fixed with glutaraldehyde. The anaphase cell was treated with only one solution before fixation, namely, 25% glycerol/o.25% Triton-X-100/HMM, as described previously (Forer & Jackson, 1979). All procedures were carried out as in Forer & Jackson (1979).

Serial sections were collected on single-slot grids and stored in LKB grid boxes or in Philips grid holders. We took pictures of 50 consecutive sections of the metaphase cell, and each section was photographed in 2 ways. Each section was photographed in overview, at about × 1000 on the negative, and this gave the entire cell in each picture. Each section was also photographed at higher power, for we could identify microtubules and HMM-decorated actin-containing filaments only at higher power: for this, each section was photographed at a primary magnification of about 7000 (on the negative), and it took about 30 such micrographs to cover the entire cell in any given section. Similarly for 25 sections of the anaphase cell (1 section was missing in the middle of this series, though). We studied the micrographs in positive prints which were printed from the negatives at an enlargement of about 7 times. For analysis we first used the overview pictures: we traced the chromosomes and identified kinetochores. (In this way we eliminated the possible bias of choosing and photographing ‘better kinetochores’ using the electron microscope and ignoring ‘worse kinetochores’.) Each kinetochore was then followed at higher magnification. We made tracings from each higher-power picture which included a kinetochore region (these pictures were at final magnifications of around 50000), as follows. A transparent cellulose acetate sheet was placed on top of each photograph, and the microtubules, the actin-containing filaments, and the chromosome outlines were traced onto it; we used a different colour for each component, namely, microtubules were red lines, actin-containing filaments were blue lines, and the outlines of the chromosomes were black lines. From each section, then, we had one sheet of transparent cellulose acetate which had a red-blue-black tracing for each kinetochore in that section. We obtained a 3-dimensional image of the arrangements of microtubules and of actin-containing filaments in any given kinetochore region by superposing, in sequence, the 12—20 tracings corresponding to the 12–20 sections through that particular kinetochore region; this also let us see the shape of the kinetochore region. In obtaining this 3-dimensional image, we placed transparent 3.2-mm-thick methyl methacrylate (Plexiglass, or Perspex) sheets between the tracings: at the final magnification at which we worked, namely 50000, this thickness of Plexiglass corresponds to about 64 nm, which is near the estimated average thickness of each section.

The descriptions in the text are based on observation of the tracings as placed between the plastic sheets and as viewed through a light box. When these tracings were viewed at various angles one indeed could see the 3-dimensional arrangement of microtubules, actin-containing filaments, and chromosomes; our descriptions in the text are based on such observations of 7 kinetochores in the anaphase cell and 6 kinetochores in the metaphase cell.

In addition to verbal description of these observations, we have represented the 3-dimensional images in 2 dimensions (and have included them in this article). This was done in 2 ways, namely, photographically and by drawings.

The tracings for each kinetochore were superposed, and the resultant ‘montages ‘were photographed as viewed from above and as illuminated from below. We took 2 black-and-white photographs for each kinetochore. (a) One photograph was with no filter; in this case all the lines, black (chromosomes), red (microtubules), and blue (actin-containing filaments) were imaged, (b) The other was through a red filter; in this case the red lines (microtubules) were not imaged, and in these photographs one sees, therefore, only chromosomes and actin-containing filaments. We have included paired photographs of this kind for 4 of the kinetochores we have studied (Figs. 3—6); in each pair the A photograph was taken without filter, and shows chromosomes, microtubules, and actin-containing filaments, whilst the B photograph was taken through a filter, and shows only chromosomes and actin-containing filaments. (In these photographs the dashed lines indicate linear elements that we could not identify with certainty: we thought that they were actin-containing filaments, but we were not sure. The arrows, such as in Fig. 3 or 6, represent filaments in which the arrowhead directions on the HMM-decorated actin-containing filaments could be unambiguously identified, giving the arrowhead directions as indicated.)

Drawings were of 2 kinds: one illustrates the arrangement of actin-containing filaments and of microtubules within individual bundles of kinetochore microtubules (Figs. 7–9), and the other illustrates the shapes and ‘intermingling’ of various microtubule bundles (Figs, 1, 2); both kinds were drawn from the arrays of superposed tracings, by Vai Page (whom we thank for this). The first kind of drawing (e.g. Fig. 7, illustrating the arrangements of microtubules and of actin-containing filaments in the bundle) was done as follows. First the tracings were superposed. Then we chose 3 positions along the length of the bundle of kinetochore microtubules at which we would indicate the relative arrangements of microtubules and actin-containing filaments: these 3 positions included the kinetochore (indicated by k in Figs. 7–9) and 2 positions along the length of the bundle. Three circles were drawn, of the diameter of the kinetochore-microtubule bundle at each of the 3 positions, and each circle represented the cross-section of the kinetochore-microtubule bundle at that position. Each circle was then divided into compartments by equally spaced horizontal lines; if there were n electron-microscopic sections which included that kinetochore, then the circles were divided by horizontal lines into n compartments, so that each tracing corresponded to one of the compartments in the circle. The positions of the microtubules and of the actin-containing filaments in each tracing were then transferred to corresponding positions in the compartments in the circles; once these were all done, the results were then drawn to scale, as in Figs. 7–9, in which the original 3 circles are represented by disks (k is the disk representing the surface of the kinetochore) and in which the microtubules are represented by open circles and the actin-containing filaments are represented by closed circles (or dots). The second kind of drawing (e.g. Fig. 1, illustrating the 3-dimensional arrangements of the various kinds of microtubule bundles) was done from direct observation of the superposed tracings. The cross-sectional shapes of the bundles and the paths of the bundles were determined by direct observation, and scale drawings were made (Figs, 1, 2). In these drawings, the cross-sectional shapes of the various bundles of microtubules are indicated by the shapes of the figures at the places where the bundles intersect the surfaces of the rectangular parallelograms. (These parallelograms have no cytological meaning; they are added solely to aid in seeing the arrangements of the bundles and in seeing the crosssectional shapes of the bundles.)

To help the reader assess the accuracy of the drawings, we have included photographs of the tracings of the same bundles of kinetochore microtubules which are illustrated in the drawings, and we have included micrographs of 8 consecutive sections through one of the kinetochores.

The accuracy of our identifications of microtubules and of actin-containing filaments was assessed in 2 ways, (a) The same person made tracings from a series of 6 sections of one kinetochore, and then after 2 weeks traced the same 6 again; there were only minor differences in the results, (b) Two different persons made tracings from the same series of sections: here, too, there were only minor differences. Hence we believe that the assessments of actin and microtubule positions are quite accurate.

We have argued in the preceding paper (Forer & Jackson, 1979) that spindle structure is preserved during the glycerination procedures. Further verification that the glycerination procedures preserve spindle structure comes from study of spindle microtubules in glycerinated cells : as deduced by reconstruction from serial sections, the arrangements of microtubules in glycerinated Haemanthus endosperm cells are exactly the same as those which have been described by others in non-glycerinated glutaraldehyde-fixed Haemanthus endosperm cells, as follows. (The results from untreated cells are those of Bajer, 1968 a, b;Bajer & Molé-Bajer, 1971 ; Harris & Bajer, 1965; and Jensen & Bajer, 1973.)

In both untreated and glycerinated cells, kinetochore regions consist of lighter ‘balls’ inserted in the darker chromosomal ‘cups’ (or ‘sockets’). The kinetochore balls are themselves composed of lighter and darker components, and the microtubules terminate in the lighter component. The kinetochores are either stretched out from the body of the chromosome (the sockets) or remain in the chromosome body and do not extend out past the usual chromosome contour. The kinetochores in glycerinated cells are about the same size as those in un-treated cells (Table 1).

In both untreated and glycerinated cells, the bundle of kinetochore microtubules diverges from the kinetochore to the pole; the divergence is to a greater or lesser extent in different bundles (see drawings in Figs, 1, 2). Whereas the bundles of kinetochore microtubules are circular in cross-section, giving the appearance of a megaphone as the microtubules diverge from the kinetochore towards the pole, the non-kinetochore bundles are in the form of ‘sheets,’ or bands, which give an elongate elliptical appearance in cross-section (Figs. 1, 2) just as in untreated cells.

In glycerinated cells, the non-kinetochore microtubules intermingle with the kinetochore microtubules (i.e. those from the non-kinetochore microtubule sheet join up with and enter the ‘megaphone’ of kinetochore microtubules, as illustrated in Figs, 1, 2), and this occurs at varying distances from the kinetochore, from 1.5 to 4.5 μm. Sometimes the non-kinetochore microtubules pass right through the chromosome alongside the kinetochores and then join up with the kinetochore microtubules within a few microns of the kinetochore, or sometimes they pass the chromosomes further away, and join up with the kinetochore microtubules further from the kinetochore. Again, this is just as in untreated cells.

In glycerinated cells, as in untreated cells, the non-kinetochore microtubules form sheet-like bundles (i.e. with elliptical rather than circular profiles), and, as in untreated cells, the bundles of non-kinetochore microtubules often branch : the bundles bifurcate, and the resulting 2 bundles go in different directions (e.g. Figs. 1, 2). The ‘secondary’ bundles can themselves branch.

Finally, ‘wavy’ microtubules are seen in glycerinated cells (illustrated in Forer & Jackson, 1979), in just those places where wavy microtubules are found in untreated cells (Jensen & Bajer, 1969).

We conclude, then, that the arrangement of microtubules in glycerinated cells is identical to that in untreated cells, and this, together with the evidence presented in the previous paper (Forer & Jackson, 1979), suggests strongly that spindle structure is preserved during glycerination.

On one point only do the results on glycerinated spindles differ from those on untreated spindles, and that is in the numbers of microtubules associated with each kinetochore: the glycerinated cells have fewer microtubules per kinetochore than do untreated cells (Table 1). This discrepancy might be due in part to the fact that Jensen & Bajer (1973) and Bajer & Molé-Bajer (1971) determined the number of microtubules per kinetochore from cross-sections whilst we determined the number from longitudinal sections. The former is more accurate: because of the great depth of field of the electron microscope, several microtubules could appear as one in longitudinal section, but they would be separate in cross-section. Until other experiments are done, however, we cannot decide how much of the difference in number is due to technique, or to biological variation, or to glycerination indeed causing loss of microtubules.

In summary, the microtubule distribution in glycerinated cells is exactly that seen by others in untreated cells, except that there seems to be loss of some microtubules. Where are the actin-containing filaments?

Actin-containing filaments are seen in every one of the 13 kinetochore microtubule bundles studied : the actin-containing filaments are in close association with the microtubules. As described before (Forer & Jackson, 1979), the filaments are so closely associated with the microtubules that in longitudinal section they are often obscured by the microtubules; thus, we are unable to follow with certainty any given filament from section to section, and sometimes even within a section, and thus we cannot state how long the filaments are. Nonetheless, actin-containing filaments are seen in every bundle of kinetochore microtubules, in amongst the microtubules in the bundle. Actin-containing filaments seem to terminate at each kinetochore and to extend from the kinetochore for several microns polewards (or at least until the kinetochore microtubules intermingle with non-kinetochore microtubules). Actin-containing filaments are also found in amongst the microtubules in the non-kinetochore microtubule bundles, and some actin-containing filaments are not associated with any microtubules. As illustration of these statements, black and white images of microtubules or actin-containing filaments in their relationship to kinetochores are given in Figs. –6, and 3 of the same kinetochore bundles are illustrated in 3-dimensional perspective in Figs. 7–9. Electron micrographs of a series of 8 consecutive sections through the kinetochore illustrated in Figs. 6 and 7 are given in Figs. 10 and 11 ; these micrographs illustrate clearly that the actin-containing filaments are found throughout the bundles of kinetochore microtubules.

The actin-containing filaments are in amongst the microtubules in the bundles of kinetochore microtubules, and it is difficult to judge how many of these are continuous, or how far individual filaments extend. Nor can we judge if there is a consistent arrangement of actin-containing filaments in each bundle. We have, however, tried to estimate the number of actin-containing filaments that are present per bundle of kinetochore microtubules. To do this, we counted the number of individual stretches of actin-containing filaments we identified in each kinetochore microtubule bundle, and we compared these numbers with the numbers of microtubules which terminated at the kinetochore in each bundle. This comparison indicates that we were able to identify roughly 1 actin-containing filament for each microtubule in the bundle of kinetochore microtubules (Table 2).

Actin-containing filaments seem to extend into and terminate at the kinetochores: electron micrographs illustrating this are given in Figs. 10-12 and one can see this in the tracings (Figs. 3–9). Our earlier statement (Forer & Jackson, 1976), that we could not see actin-containing filaments end at kinetochores, must therefore be amended : in careful study of serial sections, actin-containing filaments are seen to end at each of the 13 kinetochores studied.

In addition to the positions of actin-containing filaments, one would like to know the polarities of the actin-containing filaments, to see, for example, if the kinetochore region might be a surrogate Z-line, or, alternatively, a surrogate myosin (see discussion in Forer, 1978 a) ; one would expect consistent polarities if the actin-containing filaments are functional, but one would not expect consistent polarities if they are trapped in the spindle, or move randomly into the spindle. Polarities are difficult to see in sections with certainty with the techniques used prior to 1978 (review in Forer, 1978b); we have identified polarities of only 5% of the actin-containing filaments identified as associated with the bundle of kinetochore microtubules, and some of these are illustrated in Fig. 13 and in figures in Forer & Jackson (1979). Nevertheless, for actin-containing filaments associated with kinetochore microtubules, all polarities were of arrows pointing towards the kinetochore (Table 3) whereas for filaments associated with non-kinetochore microtubules, polarities were in both directions (Table 3). Thus, actin-containing filaments associated with kinetochore microtubules seem to be of consistent polarities; clearly, though, one would like to identify polarities in a higher percentage of filaments before accepting as proved that all filaments associated with kinetochore microtubules have polarities such that arrows point towards the kinetochores.

Spindle structure is well preserved during glycerination : the 3-dimensional arrangements of kinetochore microtubules and of non-kinetochore microtubules in spindles in glycerinated Haemanthus endosperm cells are exactly as seen in untreated, glutaraldehyde-fixed cells (Bajer, 1968a, b; Bajer & Molé-Bajer, 1971; Jensen & Bajer, 1969, 1973). Evidence that microtubule arrangements are preserved, together with the evidence given in the preceding paper (Forer & Jackson, 1979,) shows that spindle structure is not disrupted during the glycerination procedure, and gives confidence that we are not studying artifactual rearrangements brought about by disruptive procedures.

Actin-containing filaments are found in each bundle of kinetochore microtubules - at least in each of the 13 we so far have studied in Haemanthus endosperm. Actin-containing filaments must be present in each bundle of kinetochore microtubules if actin-containing filaments function to move chromosomes: they are present in each bundle, and thus, this criterion seems to be satisfied. If actin were trapped in the spindle, or if actin moved into the spindle during glycerination, one would expect that some bundles of kinetochore microtubules would be without actin, and this is not observed. Hence our data support the proposition that actin-containing filaments are functional components of each chromosomal spindle fibre.

It is possible that if a large amount of actin were trapped in the spindle, then all chromosomal fibres might be expected to have actin-containing filaments associated with them. One way of distinguishing between functional and trapped filaments might be to see if there are consistent numbers of filaments associated with the kinetochore microtubules (functional) or if there is great variation (trapped). As discussed above, however, the actin-containing filaments seem to be obscured by microtubules and therefore we are unable to decide with confidence the numbers or arrangements of actin-containing filaments in each bundle of kinetochore microtubules. Indeed, we anticipate that there are in fact more actin-containing filaments present than we identified. Thus, while each cell seems to have a consistent number of actin-containing filaments in each kinetochore bundle relative to the number of microtubules in that bundle (Table 2, B), we will not be able to describe the exact numbers and arrangements of actin-containing filaments in any given bundle of kinetochore microtubules until our techniques are improved.

Fluorescence microscopy data (using fluorescent antibodies or fluorescently labelled HMM) also suggest that actin is present in spindle fibres extending between chromosomes and poles (review in Forer, 1978a). These experiments are ambiguous, however, because some necessary control experiments have not been done (review in Forer, 1978a) ; furthermore, the cells studied had so many chromosomes (and chromosomal spindle fibres) and such small (in diameter) chromosomal spindle fibres that one could not tell if each chromosomal spindle fibre stained, or even if the stain was of the chromosomal spindle fibre per se or rather just associated material. An experiment published too recently for inclusion in the review by Forer (1978a) also used fluorescently labelled HMM to argue that actin is present in mitotic spindles (Herman & Pollard, 1978). These experiments by and large repeat those presented earlier by Schloss, Milsted & Goldman (1977), and the discussion in Forer (1978a) of the Schloss et al. (1977) experiments also pertains to the data of Herman & Pollard (1978). As one example of why these results are not compelling, a control was ambiguous: staining was blocked when 10 mM pyrophosphate was added simultaneously with the fluorescent HMM (Herman & Pollard, 1978), and from this result it was argued that the staining was ‘specific’. It is possible, however, that, as argued previously (Forer, 1978a), the pyrophosphate acts as a solvent of ‘non-specific’ components rather than as a specific blocking agent of an actin-myosin interaction. A different control is needed to avoid this ambiguity in interpretation, and that is to add fluorescent-HMM again after the pyrophosphate is washed away, to demonstrate that the pyrophosphate did not just extract some component. But this control was not done by Herman & Pollard (1978). Nor was this control reported by Sanger (1977). Hence, for this reason and for other reasons discussed in Forer (1978a), the fluorescence microscopy data are not compelling and our electron microscopy data remain for the moment the only unambiguous identifications of actin associated with individual chromosomal spindle fibres.

In addition to finding actin-containing filaments in each chromosomal spindle fibre studied, we found that actin-containing filaments interact with spindle microtubules (Forer & Jackson, 1979); evidence for this is that microfilaments and microtubules are often closely parallel, even though sharply curved. Since actin is found in each bundle of kinetochore microtubules, and since the actin-containing filaments seem to interact with the microtubules, one could imagine that microtubules in spindles act analogously to myosin in myofibrils, and that force is produced when actin-containing filaments slide relative to kinetochore microtubules (analogous to force being produced in myofibrils when actin-containing filaments slide relative to myosin-containing filaments). Furthermore, the kinetochore microtubules are attached to the chromosomes, which move polewards, and therefore if microtubules and actin-containing filaments do move in opposite directions, then one presumes that the relative motions are of microtubules moving towards the poles and of actin-containing filaments moving towards the chromosomes. (If this motion indeed occurs, the filaments eventually would reach the kinetochore and they would presumably be depolymerized there, much as it is thought by some (e.g. Inoué & Ritter, 1975) that microtubules are depolymerized at the poles.) The observed polarities of actin-containing filaments associated with kinetochore microtubules are consistent with this postulated force-producing interaction between microtubules and actin, as follows.

When HMM is added to actin-containing filaments in myofibrils, the arrowheads on the actin-containing filaments point in the direction in which the actin-containing filaments move (Huxley, 1963), i.e. away from the Z-line and towards the myosin. Similarly, in other systems the arrowheads point in the direction that the actin-containing filaments move relative to the second component (e.g. Kersey, Hepler, Palevitz & Wessells, 1976; or Mooseker & Tilney, 1975). Hence, if actin-containing filaments interact with microtubules and slide towards the kinetochore, then when such filaments are reacted with HMM the arrowheads should point in the direction in which the filaments move, namely towards the chromosome: indeed, this is what we found (Table 3). We conclude, then, that the polarities of actin-containing filaments that we found are quite consistent with actin-containing filaments interacting with microtubules or with some other component to move chromosomes towards the poles.

The idea that actin-containing filaments might interact with microtubules to produce force is not original to this report. That such interactions might produce force during mitosis has been considered by Forer (1974), Nicklas (1977), and Oakley & Heath (1978), for example. In addition, Forer & Behnke (1972), Forer (1974), Gawadi (1974), Oakley & Heath (1978), Griffith & Pollard (1978), and Forer & Jackson (1979) discuss the possibility of microfilament-microtubule interactions and cite evidences for such interactions. Our data suggest strongly that spindle microtubules do interact with actin-containing filaments in Haemanthus endosperm (Forer & Jackson, 1979), and, as we have discussed above, our data are consistent with such interactions producing force during mitosis. Possible mechanisms for actin-containing filaments producing force during mitosis other than the one we considered above (e.g. the one proposed by Gawadi, 1974) are not ruled out by our data, however.

It is relevant to point out that whilst our data are consistent with actin having an active role in chromosome movement, the data do not prove this and counter interpretations are possible. Consider the following possible counter argument, for example. Actin filaments polymerize in the direction opposite to that in which the arrows point (Hayashi& Ip, 1974,1976; Woodrum, Rich & Pollard, 1975; Kondo & Ishiwata, 1976). One might argue then, that when cells are glycerinated kinetochores act as ‘organizing centres’ for polymerization of actin filaments. Hence the observed polarity. The interactions with microtubules would occur because actin is a ‘sticky’ molecule, and interacts with various other proteins, including DNase and thrombin (see Forer, 1978 b), aldolase (Clarke & Morton, 1976; Morton, Clarke & Masters, 1977), and fibrin (Mui & Ganguly, 1977). This counter-interpretation of our data is possible, as are others, but we nonetheless feel that actin-containing filaments in spindles are most likely involved in producing force for chromosome movement (Forer & Jackson, 1976, 1978; Forer, 1974, 1978 a). In support of our view we emphasize that spindle structure is well preserved by the procedures, as evidenced by polarization microscopy observation (Forer & Jackson, 1979), by measurements of chromosome positions before and after glycerination (Forer & Jackson, 1979), and by comparison of microtubule arrangements with those in non-treated cells (this report); further, aside from speculation, there is no evidence/or extra-spindle actin in Haemanthus endosperm cells, or for actin polymerizing or relocating during the treatment, or for actin being trapped in the spindle as the spindle forms. Hence, we feel that actin-containing filaments in the spindle are probably involved in producing force for chromosome movement.

Physiological experiments have been used to argue against actin and myosin having a role in causing chromosome movements (review in Forer, 1978a). These experiments have been critically reviewed in Forer (1978a), and one concludes from review of the data that none of the experiments purported to negate the role of actin is at all conclusive. Kiehart, Inoué & Mabuchi (1977) report in abstract form experiments which might seem to rule out some of the objections raised by Forer (1978 a), and while one cannot really evaluate these experiments until details are published, even from the data presented in the abstract one can raise objections. For example, the basic assumptions in the experiments (Kiehart et al. 1977) are that antibodies to myosin will block all myosin-requiring systems and that the antibodies block cleavage by directly acting on myosin in the contractile ring. But exactly the same results are achieved in another cell type by treating cells with a ‘membrane mobility agent’ : cytokinesis is blocked but mitosis continues and multinucleate cells are produced (Lustig, Kosower, Pluznik & Kosower, 1977). Without evidence regarding the site of action of the antibodies against myosin, therefore, it is possible that the blocking of cleavage is indirect (via an effect on membrane mobility or other system) and is not due to myosin antibodies acting directly on the cleavage furrow (contractile ring). We emphasize that no experiments to our knowledge rule out a role for actin in producing force for chromosome movement.

We acknowledge the skilful technical help of Barbara G. Doyle. The work was supported by grants from the National Research Council of Canada (the Natural Sciences and Engineering Research Council of Canada) to A.F., and from the U.S. National Science Foundation to W.T.J. Dr Brad Amos gave helpful advice on the drawings.