Kinesin-5 Mediated Chromosome Congression in Insect Spindles

Abstract

Introduction

The microtubule motor protein kinesin-5 is well known to establish the bipolar spindle by outward sliding of antiparallel interpolar microtubules. In yeast, kinesin-5 also facilitates chromosome alignment “congression” at the spindle equator by preferentially depolymerizing long kinetochore microtubules (kMTs). The motor protein kinesin-8 has also been linked to chromosome congression. Therefore, we sought to determine whether kinesin-5 or kinesin-8 facilitates chromosome congression in insect spindles.

Methods

RNAi of the kinesin-5 Klp61F and kinesin-8 Klp67A were performed separately in Drosophila melanogaster S2 cells to test for inhibited chromosome congression. Klp61F RNAi, Klp67A RNAi, and control metaphase mitotic spindles expressing fluorescent tubulin and fluorescent Cid were imaged, and their fluorescence distributions were compared.

Results

RNAi of Klp61F with a weak Klp61F knockdown resulted in longer kMTs and less congressed kinetochores compared to control over a range of conditions, consistent with kinesin-5 length-dependent depolymerase activity. RNAi of the kinesin-8 Klp67A revealed that kMTs relative to the spindle lengths were not longer compared to control, but rather that the spindles were longer, indicating that Klp67A acts preferentially as a length-dependent depolymerase on interpolar microtubules without significantly affecting kMT length and chromosome congression.

Conclusions

This study demonstrates that in addition to establishing the bipolar spindle, kinesin-5 regulates kMT length to facilitate chromosome congression in insect spindles. It expands on previous yeast studies, and it expands the role of kinesin-5 to include kMT assembly regulation in eukaryotic mitosis.

Electronic supplementary material

The online version of this article (doi:10.1007/s12195-017-0500-0) contains supplementary material, which is available to authorized users.

Introduction

During mitosis, duplicated chromosomes attach to the mitotic spindle machinery through their chromosome-associated kinetochores, which attach to the plus ends of dynamic kinetochore microtubules (kMTs) that have their minus ends anchored near the spindle poles.24 Biorientation of chromosomes requires that the kinetochores attach to kMTs from opposite spindle poles. A key step in mitotic progression is chromosome congression, the alignment of the bioriented chromosomes at the mitotic spindle equator during metaphase. Although chromosome congression is a well-documented phenomenon in mitosis, the mechanism by which it is achieved is still a subject of debate.1

A polar ejection force (PEF) has been proposed in which pushing forces act on chromosome arms to bias movement of chromosomes away from the poles and toward the spindle equator to promote congression. The pushing force is attributed to: (1) plus-end directed chromokinesins (kinesin-4/10) that reside on chromosome arms and walk along spindle microtubules, (2) microtubule “pushing” by polymerizing microtubules at spindle poles, or (3) a dense array of microtubules inhibiting movement into the spindle poles.3,6,13,25,36,50 However, several lines of evidence argue against a PEF-based mechanism for congression. First, cells injected with antibodies to chromokinesins show only a modest defect in chromosome congression.26 Second, mitotic cells with unreplicated genomes, or “MUGS,” which essentially lack chromosome arms, still have congressed kinetochores.2,33,34,55 Finally, experiments in which chromosome arms were removed using lasers did not produce conclusive changes in the oscillations of the resultant chromosomes sufficient to reject the null hypothesis that the PEF is not involved in chromosome oscillations (see expanded discussion in Materials and Methods). Altogether, these results argue that the PEF is not critical for achieving chromosome congression.

Because the polar ejection force does not adequately explain how chromosomes congress, another mechanism must exist to promote congression. Kinetochores attach to the plus ends of dynamic kMTs that shorten or lengthen. The progressive alignment of chromosomes at the spindle equator requires that the plus-end dynamics of the kMTs are spatially regulated, so that they grow well from the spindle poles and poorly near the equator.14,15,47 The quality of congression can be quantitatively characterized by a signal-to-noise ratio, SNR = 〈L〉/σ, where the mean kMT length, 〈L〉, is normalized to the standard deviation of lengths, σ, or, equivalently, by the coefficient of variation, CV = 1/SNR, where CV = 0 means perfect congression and CV = 1 means poor congression.30 Previous studies have reported that mitotic microtubule motor proteins other than chromokinesins mediate chromosome congression by regulating the plus-end dynamics of kMTs. Specifically, kinesin-8 is a dimeric, plus-end directed motor protein that promotes disassembly of long microtubules in vitro, mediates spindle length, and dampens chromosome oscillations in vivo 29,41,42,48–50,52,53 Kinesin-5 is a homotetrameric, plus-end directed motor protein best known for cross-linking and sliding apart antiparallel microtubules to separate spindle poles during bipolar spindle assembly.11,19,21,23,39,43,45,46 Kinesin-5 also facilitates spindle elongation in anaphase B.4,5,48 In yeast, kinesin-5 has demonstrated activity in bi-directional motility (switching from plus-end to minus-end directed activity) and as a polymerase.8,12,16,40 Consistent with polymerase activity, the yeast kinesin-5 Kip1p at the midzone stabilizes iMTs during late anaphase.12 In both the fungi Saccharomyces cerevisiae and Candida albicans, the kinesin-5 mutants cin8Δ (Cin8p and Kip1p are the only two kinesin-5 genes in S. cerevisiae) and KIP1/kip1Δ (Kip1p is the only kinesin-5 in diploid C. albicans) were still able to form bipolar mitotic spindles; however, in these mutants the kMTs were longer and more variable in length, thus impairing chromosome congression.15,30 In contrast, kinesin-8 mutants showed no evidence of longer kMTs or impaired chromosome congression, but did exhibit an increase in spindle lengths.15,30 Gardner et al. established a model where tetrameric kinesin-5 crosslinks microtubules, which naturally guides it toward kMT plus ends where it promotes kinetochore microtubule disassembly and mediates chromosome congression.15 By contrast, dimeric kinesin-8, unable to crosslink between MTs, tends to remain persistent on the longer interpolar microtubules (iMTs); therefore, kinesin-8 is guided toward iMT plus ends where it promotes interpolar microtubule disassembly and controls spindle length.15 Significantly, kinesin-5 is only known to mediate chromosome congression in fungi, raising the question of whether this motor-based congression mechanism preceded fungal divergence from other eukaryotes during evolution, or emerged as a later elaboration after divergence.

To address whether kinesin-5 drives chromosome congression in insect spindles, we examined the function of kinesin-5 in Drosophila melanogaster S2 cells. The Drosophila kinesin-5, Klp61F, has been reported to mediate bipolar spindle assembly, promote poleward flux, and drive spindle elongation during anaphase B.4,18–20 We chose the Drosophila system because it is very amenable to protein knockdown with RNAi via dsRNA-induced gene silencing.9,18,19,28 In previous experiments with near-complete protein reduction, Klp61F RNAi led to monopolar spindles.18–20 In our experiments, the partial reduction of kinesin-5 allowed us to study bioriented spindles assembled in insect cells, to ask whether kinesin-5 also regulates kMT assembly in a length-dependent manner to facilitate chromosome congression.

Results and Discussion

Klp61F RNAi Cells Undergo Delayed Mitosis with Asymmetric Chromosome Segregation in Anaphase

Mitotic delay is a typical consequence of impairment of microtubule dynamics, e.g. via mitotic poisons used for cancer treatment, such as paclitaxel, nocodazole, and kinesin-5 inhibitors.35,37 Outcomes are mitotic slippage (completion of mitosis without satisfaction of the spindle assembly checkpoint), aneuploidy, and eventual apoptosis.35,37 To determine the longer-term mitotic effects of Klp61F reduction, we performed RNAi of Klp61F (5 µg dsRNA/10⁶ cells) on days 0 and 3 in S2 cells, indicating 10% protein reduction after 0 and 2 days, 50% protein reduction after 3 days, and 90% protein reduction after 5 days, normalized to control Klp61F for each day (Fig. 1a).9,31 This was completed in S2 cells stably expressing mCherry-α-Tubulin and GFP-Cid (Centromere identifier, CENP-A histone, a core kinetochore component).22,51 We observed cells via time-lapse fluorescence microscopy 2.5 days after the initial transfection (estimated Klp61F reduction 30%) and found that most control spindles were bipolar and Klp61F RNAi spindles were monopolar with approximately half transitioning to bioriented spindles (Figs. 1b and 1c). In bipolar spindles, microtubules clustered at two poles on opposite sides of the spindle, while in bioriented spindles, microtubules clustered at one pole and were parallel at the other pole, suggesting that the kinetochores were bioriented, but that only one true microtubule organizing center existed. When the width across unclustered microtubules at a spindle pole was equal to or wider than the width across the kinetochores, spindles were categorized as “bioriented.” The Klp61F RNAi bioriented phenotype (Fig. 1c middle row) has been observed previously in S2 cells, and called “monastral bipolars;” it was found that this pathway required Klp61F, indicating that sufficient Klp61F remained to form bioriented spindles.19,20 Anaphase was identified as Cid fluorescence localized at spindle poles. For monopolar spindles, the Cid mass was highly asymmetric (Monopolar, Figs. 1c and 1e).

Figure 1.

Klp61F RNAi spindles have delayed mitoses with unequal segregation of chromosomes relative to control spindles. MCherry-α-tubulin, GFP-Cid expressing cells were imaged with 10 × 10 array tilescans at 10 min intervals with 5 z-planes for 10–12 h. Movie experiment repeated three times. (a) Immunoblot of tubulin and Klp61F in control and Klp61F RNAi cells shows reduction of Klp61F throughout time following RNAi Klp61F (C = Control, R = Klp61F RNAi; 10% reduction after 0 and 2 days, 50% reduction after 3 days, and 90% reduction after 5 days, normalized to control Klp61F for each day). Antibodies used were DM1α (mouse-anti-α-tubulin) (1:10,000), goat anti-mouse (1:10,000), rat-anti-Klp61F (1:5,000), and goat anti-rat (1:10,000). (b) Spindles were bipolar for control and initially monopolar for Klp61F RNAi, although over half of the monopolar spindles eventually transitioned to bioriented spindles (Control, n = 102; Klp61F RNAi, n = 112). (c) Representative images of mitosis in control and Klp61F RNAi cells obtained from time-lapse fluorescence imaging. (d) Klp61F RNAi results in a mitotic delay. Times are not from the same distribution (p = 10⁻¹⁴, n = 96, 36,10). Statistically significant times indicated (α = 0.05 by Tukey–Kramer test). Bars are median ± SEM. (e) Klp61F RNAi results in improper kinetochore segregation given by Cid fluorescence in anaphase. Bars represent the ratio (low/high) of the mean fluorescence for the background-subtracted Cid signals in anaphase. Ratios are not from the same distribution (p = 10⁻¹⁴, n = 98,62,40). Bars are mean ± SEM.

To determine if the cells underwent mitotic delay with reduced Klp61F levels, we measured the time in mitosis, defined as the difference between anaphase onset and nuclear envelope breakdown, observed in 97% of control cells, 64% of Klp61F RNAi bioriented cells, and 20% of Klp61F RNAi monopolar cells (Fig. 1d). Overall, time in mitosis correlated with severity of phenotype, with Klp61F RNAi spindles that never bioriented spending the most time in mitosis (Fig. 1d). Cid distribution in anaphase indicated chromosome segregation and was quantified by taking the ratio of the lower side’s fluorescence to the higher side’s fluorescence, where a ratio of 1 means perfect symmetry (i.e. high chromosome segregation fidelity) and zero means complete asymmetry (i.e. massive aneuploidy). Again, the trend was increased asymmetry with severity of phenotype, indicating that chromosome segregation was more impaired in Klp61F RNAi spindles than control (Fig. 1e).

Our results demonstrate that a weak knockdown of Klp61F causes mitotic delay followed by mitotic slippage and asymmetric segregation of chromosomes in S2 cells, similar to aberrant mitoses for cancer cell lines treated with kinesin-5 inhibitors.35

Klp61F RNAi Bioriented and Bipolar Spindles Have Longer kMTs and Less Congressed Chromosomes Than Control Bipolar Spindles

In addition to collecting time-lapse images, we also collected single time point images of control and Klp61F RNAi spindles with varying amounts of dsRNA Klp61F (0,1,10 µg dsRNA/10⁶ cells) transfected and various times following transfection. These experiments were performed in both mCherry-α-Tubulin/GFP-Cid S2 cells and GFP-α-Tubulin S2 cells.19,22,51 Overall, we found that increased time following transfection of dsRNA led to greater monopolarity, more so than increased dsRNA (Fig. 2b). This was not surprising because transfecting cells with dsRNA initiates RNAi to inhibit protein expression, but it does not directly target the existing protein level. The low level of monopolarity one day after transfection indicates that the Klp61F protein level was still high in most cells, but two and three days after transfection the high level of monopolarity indicates that the remaining protein was degraded and not replaced. As in the movie data (Fig. 1), we again observed bipolar and bioriented spindles, which allowed us to investigate congression with reduced kinesin-5. The reason to image the spindles at single time points was because we could use more aggressive imaging (higher laser power, longer pixel dwell) which causes the undesirable effect of photobleaching in time-lapse experiments.

Figure 2.

Klp61F RNAi bipolar and bioriented spindles have impaired chromosome congression and shorter spindle lengths relative to control spindles. MCherry-α-tubulin, GFP-Cid expressing cells and GFP-α-tubulin expressing cells were imaged with 1–5 z-planes at single time points (10 experiments involving ranges of dsRNA concentrations and time following dsRNA transfection). (a) Representative single time point images of control and Klp61F RNAi spindles. (b) Spindles were primarily bipolar for control (0 µg dsRNA) and bipolarity decreased more with longer incubation following transfection of dsRNA rather than more dsRNA transfected (n = 54,66,179,185,97). (c) Tubulin fluorescence distribution of bipolar and bioriented spindles indicates that kMTs were longer and more variable in length in Klp61F RNAi spindles (n = 161) compared to control (n = 314). Bars are mean ± SEM. (d) Cid fluorescence distribution with aligned peaks indicates that chromosomes were more broadly distributed in Klp61F RNAi bipolar and bioriented spindles (n = 73) compared to control (n = 180). Bars are mean ± SEM. (e) Spindles lengths were shorter for Klp61F RNAi bipolar and bioriented spindles (n = 161) compared to control (n = 314). Bars are mean ± SEM.

We measured the tubulin fluorescence distribution, Cid fluorescence distribution, and spindle length in bipolar and bioriented spindles in time-lapse and single time point data across different dsRNA concentrations, times following transfection, and fluorescence cell lines. In all cases, we found that the Klp61F RNAi tubulin fluorescence distribution was statistically different (p < 0.05) from the control tubulin distribution. Therefore, we pooled all control data and all Klp61F RNAi data across the experiments (for results of time-lapse experiments, see Supplemental Fig. 1 and for results of single time point experiments see Supplemental Fig. 2). Cells were defined as being in metaphase when their spindle lengths were within two standard deviations of the mean of control cells’ spindle lengths and this criterion was applied to limit our investigation to metaphase spindles.15,30 A bilobed tubulin fluorescence distribution, in which fluorescence rises from the spindle poles and then decreases at the spindle equator indicates that the mean location of kMT plus ends is where the fluorescence signal decays most rapidly in space, which effectively positions sister kinetochores on either side of the equator.14,47 This fluorescence distribution is observed experimentally in wild-type yeast.15,30 The loss of the bilobed fluorescence distribution, characterized by a relatively flat “bar” of fluorescence across the spindle, indicates that the kMTs are longer and more variable in length and the chromosomes are not properly congressed.14,15,30 This fluorescence distribution is observed experimentally in S. cerevisiae cin8Δ and C. albicans KIP1/kip1. 15,30 We used this same assay in S2 cells where the majority of minus ends are located at the spindle poles.17 Whereas a bilobed tubulin distribution was observed in control bipolar spindles, a “flattened” bilobed tubulin distribution was observed in Klp61F RNAi bipolar and bioriented spindles, indicating that the kMTs in Klp61F RNAi bipolar and bioriented spindles were longer and more variable in length than kMTs in control spindles (Fig. 2c). Additionally, the aligned fluorescent peaks of the Cid distributions for control and Klp61F RNAi bipolar and bioriented spindles revealed that the Klp61F RNAi Cid fluorescence distribution was broader than control, consistent with a loss of congression (Fig. 2d). Furthermore, the CV metric was applied to quantify the level of congression.30 The mean kMT lengths were estimated as the location where the normalized tubulin fluorescence was steepest multiplied by the respective mean spindle lengths, (0.75)(8.2 µm) = 6.2 µm for control and (0.92)(7.5 µm) = 6.9 µm for Klp61F RNAi.30 The standard deviations of kMT lengths were estimated by the product of the normalized Cid Gaussian fit standard deviations and spindle lengths, (0.17)(8.2 µm) = 1.4 µm for control and (0.24)(7.5 µm) = 1.8 µm for Klp61F RNAi. The CV was calculated by the ratio of the standard deviations of the spindle lengths to the mean kMT lengths, thus, the spindle lengths dropped out of the equation. For control spindles, the CV was

$$\frac{\left(0.17\right)\left(8.2\right)}{\left(0.75\right)\left(8.2\right)}=0.22$$ 1

For Klp61F RNAi spindles, the CV was

$$\frac{\left(0.24\right)\left(7.5\right)}{\left(0.92\right)\left(7.5\right)}=0.26$$ 2

indicating that congression was reduced in Klp61F RNAi spindles compared to control. Lastly, with fewer kinesin-5 motors to force the poles apart, the spindle length is expected to be shorter, and we found that indeed the spindle lengths were 8.5% shorter in Klp61F RNAi bipolar and bioriented spindles compared to control bipolar spindles (Fig. 2e).

Our results show Klp61F RNAi bipolar and bioriented spindles have longer kMTs, less concentrated Cid fluorescence at the spindle equator, and shorter spindles. Taken together with the result that Klp61F RNAi bioriented spindles segregate their chromosomes more asymmetrically than control, this demonstrates that a lack of chromosome congression correlates with poor chromosome segregation and that kinesin-5 has a role in facilitating both proper chromosome congression and segregation. Additionally, Brust-Mascher et al. found that in Drosophila embryos injected with anti-Klp61F antibodies that had only partial dissociation of kinesin-5 from the spindles (avoiding monopolar spindle collapse), the kinetochores were less concentrated at the spindle equator compared to control cells, although they did not propose a mechanism for this observation.5 Here, we find similar results in Drosophila S2 cells using RNAi, namely that kMTs are longer and the Cid distribution is broader than control, indicating that chromosome congression is impaired, verified by a larger CV. Furthermore, because these cells still formed bipolar or bioriented spindles, there was very likely a residual effect of the remaining Klp61F that allowed the spindles to remain somewhat congressed. Thus this means that Klp61F may be a mediator of congression.

Klp67A RNAi Bipolar Spindles Have Control-Like kMT Organization But Spindles are Longer

Kinesin-8 has shown both length-dependent microtubule depolymerase activity in vitro and chromosome congression activity in vivo, and kinesin-8 mutant spindles are abnormally long.15,29,30,41,48–50,52 In fungi kinesin-8 mutant spindles were longer than control but chromosome congression was not impaired, consistent with kinesin-8 acting preferentially as a length-dependent microtubule depolymerase on iMTs instead of kMTs.15,30 To test the role of kinesin-8 in S2 cells, we performed RNAi of the kinesin-8 Klp67A using different amounts of dsRNA and took single time point images one and three days following transfection (Fig. 3a), following the same protocol as single time point measurements for Klp61F RNAi. We found that nearly all spindles were bipolar (Fig. 3b). We found that the Klp67A RNAi tubulin fluorescence distributions were not statistically different (p < 0.05) from the control tubulin distribution regardless of amount of dsRNA or time following transfection. Therefore, we pooled all control spindles and all Klp67A RNAi bipolar spindles. With the spindle length criterion, almost 30% of Klp67A RNAi bipolar spindles would be eliminated, a percentage much higher than what would be expected for outliers. Therefore, we did not apply our spindle length criterion and instead evaluated all spindles. If kinesin-8 affected chromosome congression in S2 cells, we would expect that, like the Klp61F RNAi bipolar and bioriented spindles, the tubulin fluorescence distribution would be flatter for Klp67A RNAi spindles indicating longer kMTs; however, the tubulin fluorescence distributions were visibly similar and did not show a statistically significant difference (p = 0.77, Fig. 3c), demonstrating that the kMTs in Klp67A RNAi spindles were on average the same length, relative to the spindle, as control. We also found that the spindles in Klp67A RNAi spindles were significantly longer than control (Fig. 3d), consistent with previous studies of kinsesin-8 mutants.15,29,30,41,48–50 Furthermore, the same results were obtained when we did apply the spindle length criterion (Figure S3). These results are consistent with Klp67A acting as a plus-end directed, length-dependent disassembly promoter that acts primarily on iMTs instead of kMTs, thus affecting the overall spindle length but not affecting the kMT length and subsequent chromosome congression.15,30 However, our simple interpretation of the fluorescence data does not exclude alternative interpretations for Klp67A.

Figure 3.

Klp67A RNAi bipolar and bioriented spindles have normal chromosome congression and longer spindle lengths relative to control spindles. GFP-α-tubulin expressing cells were imaged with 1–5 z-planes at single time points. Experiment repeated two times. (a) Representative single time point images of control and Klp67A RNAi spindles. (b) Spindles were predominantly bipolar for both control and Klp67A RNAi cells (n = 37, 60, 56, 33, 26, 17). (c) Tubulin fluorescence distribution indicates that kMTs in Klp67A RNAi bipolar and bioriented spindles were the same length (n = 158), relative to the spindle length, as kMTs in control spindles (n = 68). Bars are mean ± SEM. (d) Spindles lengths were longer for Klp67A RNAi bipolar and bioriented spindles (n = 158) compared to control (n = 68). Bars are mean ± SEM. (e) Model of kinesin-5 mediated length control of kMTs and kinesin-8 mediated length control of iMTs. In control bipolar spindles, tetrameric kinesin-5 crosslinks parallel MTs, promoting catastrophe at the plus ends thus facilitating congression of kinetochores, while dimeric kinesin-8, unable to crosslink, remains persistent on iMTs, promoting catastrophe at the plus ends thus facilitating maintenance of spindle length. In Klp61F RNAi bioriented spindles, reduced kinesin-5 leads to both lack of kinetochore congression because of lack of control of kMT lengths and lack of formation of complete bipolar spindle because of lack of antiparallel MT force generation. In Klp67A RNAi bipolar spindles, reduced kinesin-8 leads to maintenance of congressed kinetochores because of presence of kinesin-5 but longer spindle length because of lack of control of iMT lengths.

There are many differences between the budding yeast spindle and Drosophila S2 spindle including number of microtubules/kinetochore (1 MT/KT for budding yeast, 11 ± 2 MTs/KT for S2), kinetochore separation (0.5–0.6 µm in budding yeast, >50 µm for S2), and spindle length (1.3–1.6 µm for budding yeast, 8.2 ± 1.6 for S2).7,27,54 However, the algorithm to measure tubulin and Cid fluorescence across the spindle provided an estimated location of mean kMT plus ends, Cid distribution, and spindle length. The algorithm was independent of the differences between the two different organisms and thus could be used to make measurements regardless of the organism.

In cells with modestly reduced levels of kinesin-5 (i.e. enough Klp61F remains to form bipolar or bioriented spindles), kMTs were longer and more variable in length, compromising chromosome congression, and spindle lengths were shorter presumably because of decreased outward pushing force. This result supports a model in which chromosome congression is driven by motor-mediated disassembly of long kMTs by tetrameric kinesin-5 motors that act primarily on kMTs because of their cross-linking ability (Fig. 3e). In cells with reduced levels of kinesin-8, kMTs were the same length relative to the spindle length as control, maintaining proper chromosome congression, yet the overall spindle lengths were longer. This result supports a model in which spindle length is driven by motor-mediated disassembly of iMTs by dimeric kinesin-8 motors that act primarily on iMTs because of their inability to crosslink (Fig. 3e).

Mitotic mechanisms have focused primarily on chromokinesin (kinesin-4/10) mediated PEF to promote chromosome congression, yet kinetochores still align at the mitotic spindle equator despite the loss of both chromokinesins and the chromosome arms on which the chromokinesins act, demonstrating that neither is indispensible for proper congression.2,25,26,33,34,55 On the other hand, three separate studies now point to kinesin-5 mediated chromosome congression spanning yeast to insect spindles, an alternative model focusing on length-dependent control of kMT dynamics, instead of a poorly-supported PEF.15,30 Kinesin-5 mediated disassembly of kMTs contrasts kinesin-5’s role in yeast as a polymerase in vitro and stabilizer of iMT plus ends during budding yeast anaphase.8,12 However, unlike the current work, the study in which kinesin-5 demonstrated polymerase activity involved an engineered (hybrid Xenopus-Drosophila), dimeric kinesin-5 and the polymerase activity occurred under in vitro conditions.8 In addition, the study in which kinesin-5 stabilized the plus ends of iMTs in anaphase found that different kinesin-5 proteins, Cin8p and Kip1p exhibited different activity, namely Cin8p facilitated spindle pole body separation and Kipl1 stabilized iMT plus ends in anaphase. Thus, not only did that study acknowledge different activities for kinesin-5, but it also focused on anaphase while the current work focuses on metaphase.12 Altogether, none of the studies definitively refute the results of the others. Overall, we conclude that kinesin-5 consistently acts as length-dependent kinetochore microtubule depolymerase in budding yeast, Candida albicans, and now, in the present study, Drosophila S2 cells to mediate chromosome congression in metaphase. These results show the conservation of the kinesin-5 depolymerase activity in metaphase across phylogeny from fungi to animals.

Future work is needed to test for kinesin-5 mediated chromosome congression in other eukaryotes, including human cells, and this work must overcome the technical challenge of reducing kinesin-5 while avoiding monopolar spindle collapse, but if it is found that kinesin-5 promotes chromosome congression in higher eukaryotes this will fundamentally change mitotic spindle assembly models, by both replacing the polar ejection force with motor-mediated kMT length regulation and by providing an additional role for kinesin-5, separate from (but not inconsistent with) its role in bipolar spindle assembly and maintenance.

Materials and Methods

Cell Culture and RNA Interference

Drosophila melanogaster Schneider (S2) cells and plasmids (Addgene) were obtained from Professor Ronald Vale’s lab.19,22,51 Cells were maintained in T25 flasks at room temperature and passed weekly into normal media consisting of Shields and Sang M3 Media (Sigma-Adrich #S3652), Insect Medium Supplement (Sigma-Aldrich #I7267), and Schneider’s Drosophila Media (Sigma-Aldrich #S0146) supplemented with 2% fetal bovine serum and penicillin/streptomycin.31

For RNAi, dsRNAs were prepared following in vitro transcription using the Megascript T7 kit (Ambion).31 PCR was performed using genomic Drosophila DNA as template and with the following primers Klp61F: forward 5′-t7-GCT CTG AGT CAC CTT TTC GAT-3′, reverse 5′-t7-CAC-GAT-ATG-GAA-CGT-GAG-GAaG-3′ Klp67A: forward 5′-t7-AAC-GAA-CAT-GTG-ATG-AAT-CTG-3′ reverse 5′-t7-TCT-TCA-CAT-AGA-ACT-CGG-TTG-3′, which were designed using published results of the Klp61F sequence from flybase.org (Klp61F) or from published results (Klp67A).19 PCR products were purified (Qiagen kit) and used as transcription templates in the MEGAScript T7 kit to generate RNA. The RNA was then heated at 65 °C for 30 min to promote annealing and the resulting dsRNA was cooled to room temperature over 4–5 h, aliquoted, and stored at −20 °C until use.

RNAi was performed following the methods of Clemens et al.9 Prior to transfection with dsRNA, 4.5–6 million cells were exchanged from normal media to 1× M3 media supplemented with 1× IMS and placed in a well of a 6-well plate to obtain a concentration of 2 million cells/mL. DsRNA was added directly to the well (amounts varied according to experiment) and the plate was placed on a rotator at 50 rpm. After one hour an equal volume (to the existing media) of 1× Schneider’s Drosophila media supplemented with 10% FBS was added to the well to obtain a concentration of 1 million cells/mL, and the cells remained in the well until imaging.9 DsRNA was transfected again in the cells three days after the initial transfection in the first experiment to verify protein knockdown (Fig. 1) but only single transfections were performed in all remaining experiments.

Immunoblotting

Antibodies used for immunoblotting were DM1α (mouse-anti-α-tubulin) (1:10,000, Sigma-Aldrich #T6199) with goat anti-mouse IgG/IgM Secondary Antibody, AP conjugate (1:10,000, Thermo Fisher Scientific #T2192) and rat-anti-Klp61F (1:5,000, a gift from Dr. Lawrence Goldstein) with goat anti-rat IgG Secondary Antibody, AP conjugate(1:10,000, Thermo Fisher Scientific #31350) antibodies.

Microscopy

Cells were plated on 35-mm dishes with round, 14-mm no. 1.5 cover glass bottoms (P35G-1.5-14C; MatTek) coated with Concanavalin A (Sigma-Aldrich) and allowed to adhere to the dishes for 20 min before imaging.38 All cells were imaged at room temperature on a Zeiss LSM 7 Live swept-field confocal microscope using a ×100 1.45 NA plan-fluor oil immersion objective with 488 and 561 nm lasers and 495–555 nm and 575–615 nm plus longpass 655 nm filter sets for imaging the mCherry-Tubulin, GFP-Cid cells, and longpass 495 nm filter set for imaging the GFP-tubulin cells. Images were collected on a 1 × 512 line scanning 12-bit CCD camera with 60% quantum efficiency scanning 512 lines to obtain 512 × 512 images of each confocal plane. Time-lapse images of a 10 × 10 array of 512 × 512 fields tiled together, with 5 z-planes per tile (3.75 µm sections), were collected at 10 min intervals for 10–12 h with a zoom of 0.5, which yields a pixel size of 263 nm. Single time-point images with variable numbers of z-planes (typically 1–5 planes) in order to collect images of complete spindles were collected with a zoom of 1.3, which yields a pixel size of 101 nm.

Image Analysis

Identification of spindle poles and estimation of the fluorescence distribution for tubulin and Cid across the spindle axis were performed using a semi-automated MATLAB (Mathworks) algorithm as previously described.30 Briefly, spindles were rotated to lie along the horizontal axis and the tubulin and Cid fluorescence intensities were collected for each vertical section in a background-subtracted region containing the spindle and along the axis of the spindle. The tubulin fluorescence decrease near the ends of the spindle were fit to Gaussian survival functions for each half-spindle, and the spindle pole locations were identified as the mean locations of the survival curves.10,30,44 The intensities between the centrosomes were then binned with a constant number of bins (24) to allow comparison between different spindle lengths. The left and right halves were averaged to obtain a half spindle fluorescence intensity distribution, and finally the distributions were normalized by dividing each bin position’s fluorescence intensity by the sum of all of the bin positions’ fluorescence intensities.30 For each spindle, the background-subtracted, binned Cid fluorescence distributions were fit to Gaussian functions centered at the Cid fluorescence peaks with constrained maxima and minima given by the maximum peak fluorescence intensity and minimum tail fluorescence intensity for each Cid fluorescence distribution and unconstrained standard deviations. The standard deviations were then obtained by minimizing the sums of the least-squared fits to the data. A single Cid fluorescence distribution was removed from the data because its peak occurred in the last bin, making a Gaussian fit impossible. The Cid fluorescence distributions were aligned with their peaks at the center to avoid misinterpretation of broad Cid fluorescence due to off-center peaks. Cid distribution in anaphase was collected from movie data by rotating anaphase spindles to lie along the horizontal spindle axis, generating a rectangle sufficiently large to contain the segregated Cid fluorescence, dividing the rectangle in equal thirds, and subtracting the mean signal of the middle third from both the left and right thirds (background subtraction).

Acceptance Criteria for Evaluating Metaphase Cells

To avoid evaluating prometaphase or anaphase spindles, a spindle length criterion was developed where only spindles with spindle lengths within two standard deviations of the mean of the control spindle length were analyzed for tubulin fluorescence distribution, Cid fluorescence at the spindle equator, and spindle length for each set of experiments.30

Bootstrapping Method for Comparing Tubulin Fluorescence Distributions

Tubulin fluorescence distributions were compared using a bootstrapping method implemented with a MATLAB script as previously described.30 Briefly, for N cells of condition A, the mean fluorescence intensity distribution (mean_A) was obtained by taking the mean across all N cells at each bin location (12 bins total). The same procedure was followed for obtaining the mean fluorescence intensity distribution of M cells of condition B (mean_B). At each bin position, the difference of the means between the two distributions (mean_A and mean_B) was squared and the sum of the squared differences gave the value of the experimental sum of squares difference between the distributions. A matrix of size N + M rows by 12 bins columns was then created. The first N rows were populated by the fluorescence distributions of the N cells of condition A and then last M rows were populated by the fluorescence distributions of the M cells of condition B. The rows were then shuffled. The first N rows of the shuffled set became the simulated cells of condition A and the last M rows became the simulated cells of condition B. As before, the mean fluorescence distributions were obtained but now for the simulated cells of each condition. The sum of squares difference for the mean simulated distributions was obtained and stored. The process of preparing simulated distributions and comparing them was repeated 10,000 times to obtain 10,000 simulated sum of squares differences, and the values were sorted. The experimental sum of squares difference was then ranked within the list of simulated sum of squares and the p value was calculated by dividing the difference between total number of simulated sum of squares differences and the rank, R, by the total number of simulated sum of squares differences,

$$p=\frac{10,000-R}{10,000}.$$

This boot-strapping method is sensitive to outliers when the sizes of the groups to be compared are very different and thus the fluorescence distributions of outliers must be removed prior to making the comparisons. This was found when comparing the tubulin fluorescence distributions of control (n = 83) and Klp61F RNAi bipolar and bioriented spindles (n = 16) in single time point measurements of mCherry-α-Tubulin/GFP-Cid cells (Figure S2) that appeared different but did not give a significant p-value. A single fluorescence distribution in the control spindles where the normalized fluorescence intensity values in 11 out of 12 bins had outlier values (deviating 1.5 times the interquartile range above the third quartile or below the first quartile) was identified, the fluorescence distribution for that spindle was removed, the comparison repeated, and a significant p-value was obtained.32 This was the only case when a tubulin fluorescence distribution was removed from the data set.

Statistical Comparisons

With the exception of comparing tubulin fluorescence distributions, all groups were compared using the Kruskal–Wallis test, which does not require normality of data, and multiple comparisons (Tukey–Kramer with α = 0.05) were conducted if the Kruskal–Wallis test revealed that the groups did not all have the same mean ranks. Heights on all bar graphs represent mean values unless stated otherwise. All error bars represent standard error of the mean (SEM).

Calculation of p-Value in Laser-Cutting Chromosome Arm Experiments

If one assumes a null hypothesis that PEF does not play a major role in congression, then each of the 16 laser cutting experiments would be equally probable to exhibit larger oscillations as smaller oscillations. In this case, the probability of larger oscillations in each chromosome cutting experiment is 0.5 (i.e. random chance), and the expected number of chromosomes that would increase oscillations after cutting by chance alone is 8 out of 16. Using a Bernoulli trial calculation with n = 16 trials and p = 0.5 of success in each trial we obtain the probability of having 11 or more (or 5 or fewer) spindles with increased oscillations by chance alone is 0.21. In other words, the null hypothesis that PEF does not play an important role in congression cannot be rejected from these laser cutting experiments (p = 0.21).

Electronic Supplementary Material

Below is the link to the electronic supplementary material.

Abbreviations

kMT

Kinetochore microtubule

iMT

Interpolar microtubule

RNAi

RNA interference

dsRNA

Double stranded RNA

NEB

Nuclear envelope breakdown

AO

Anaphase onset

PEF

Polar ejection force