Kinetochore fibers (K-fibers) of the mitotic spindle are force-generating units that power chromosome movement during mitosis. K-fibers are composed of many microtubules that are held together throughout their length. Here, we show, using 3D electron microscopy, that K-fiber microtubules (MTs) are connected by a network of MT connectors. We term this network ‘the mesh’. The K-fiber mesh is made of linked multipolar connectors. Each connector has up to four struts, so that a single connector can link up to four MTs. Molecular manipulation of the mesh by overexpression of TACC3 causes disorganization of the K-fiber MTs. Optimal stabilization of K-fibers by the mesh is required for normal progression through mitosis. We propose that the mesh stabilizes K-fibers by pulling MTs together and thereby maintaining the integrity of the fiber. Our work thus identifies the K-fiber meshwork of linked multipolar connectors as a key integrator and determinant of K-fiber structure and function.https://doi.org/10.7554/eLife.07635.001
Before a cell divides, its genetic material must be copied and then equally distributed between the newly formed daughter cells. In the cells of plants, animals, and fungi, a structure known as the spindle pulls the two copies of the chromosomes apart. The spindle is made up of a network of long, protein filaments called microtubules, and the bundles of microtubules that attach to the chromosomes are referred to as ‘K-fibers’.
K-fibers are organized in a way that provides strength. These bundles of microtubules are held together throughout their entire length and, in 2011, it was suggested that a group of proteins including one called TACC3 could cross-link adjacent microtubules within K-fibers. However, it remained unclear how these proteins achieved this.
Now, Nixon et al.—including several of the researchers involved in the 2011 work—have used a technique called 3D electron tomography to analyze what holds the K-fibers together in human cells. This analysis revealed struts or connectors that hold together adjacent microtubules within K-fibers. These connectors can vary in size and a single connector can link up to four microtubules. This means that, in a three-dimensional view, the connectors appear as a ‘mesh’ between the microtubules in the bundle.
Nixon et al. then increased the levels of the TACC3 protein and found that the K-fibers became disorganized. The spacing of the microtubules with the K-fibers was reduced so that they were more tightly packed than normal. These observations suggest that ‘the mesh’ influences the microtubule spacing within a K-fiber.
Nixon et al. analyzed how disorganized K-fibers affected dividing cells and found that it took longer for the chromosomes to move to the newly forming daughter cells. This suggests that cells must maintain optimal levels of TACC3 to ensure that the K-fibers can effectively separate the chromosomes. Further work is needed to identify the other proteins and molecules that make up the mesh.https://doi.org/10.7554/eLife.07635.002
Accurate mitosis is essential to eukaryotic life. It requires the correct assembly of a bipolar array of microtubules (MTs) into a mitotic spindle which, in concert with hundreds of different motors and non-motor proteins, segregates the duplicated sister chromatids to the two daughter cells. Many of the chromosome movements in mitosis are governed by the kinetochore fibers (K-fibers) of the spindle apparatus. In human cells, K-fibers are bundles of 20–40 parallel MTs that typically run from the kinetochore to the spindle pole (McDonald et al., 1992; Mastronarde et al., 1993; McEwen et al., 1997; Booth et al., 2011; Sikirzhytski et al., 2014). K-fibers can be thought of as coherent units: their constituent MTs are held together throughout their length, as well as being focused at either end (Rieder, 1981; Spurck et al., 1997). The coherence of the K-fiber is thought to be crucial for accurate chromosome segregation in mitosis. However, the ultrastructural and molecular basis of K-fiber coherence is not well understood.
Lateral MT connectors are important for the function of MT arrays, such as K-fibers, in cells (Brangwynne et al., 2006; Ward et al., 2014). Classic electron microscopy (EM) studies uncovered the presence of electron density between MTs of the K-fibers (Wilson, 1969; Hepler et al., 1970; Witt et al., 1981; Bastmeyer and Fuge, 1986). These inter-MT bridges appear as bipolar struts laterally connecting two MTs in 2D electron micrographs. The bridges are typically ∼5 nm thick and range from 6–20 nm in length. The morphology of bridges is heterogeneous, and they are likely composed of a variety of proteins (Booth et al., 2011). The mitotic spindle is an ensemble of hundreds of MT-associated proteins (Sauer et al., 2005), some of which are candidates for inter-MT bridges (Manning and Compton, 2008). The number of inter-MT bridges scales with the number of MTs and also with the number of paired MTs in a bundle (Bastmeyer and Fuge, 1986), and it was proposed that it is the inter-MT bridges that hold K-fiber MTs together (Witt et al., 1981). The connection between the kinetochore and K-fiber MTs has been studied by 3D-EM (Dong et al., 2007; McIntosh et al., 2008). However, a similar 3D investigation of inter-MT connections away from the kinetochore has not been reported.
One molecular candidate for inter-MT bridges in K-fibers is TACC3–ch-TOG–clathrin. The assembly of this complex is regulated by phosphorylation of TACC3 at serine 558 by Aurora-A kinase (Fu et al., 2010; Hubner et al., 2010; Lin et al., 2010; Booth et al., 2011). This allows two domains from TACC3 and clathrin to come together in space, making a single, composite MT interaction surface (Booth et al., 2011; Cheeseman et al., 2011; Hood et al., 2013). Previous work showed that this complex is necessary for K-fiber stabilization (Royle et al., 2005; Booth et al., 2011; Cheeseman et al., 2013) and that the complex forms a distinct class of short inter-MT bridges (Booth et al., 2011).
In this study, we set out to examine the 3D ultrastructure of K-fiber inter-MT bridges. Surprisingly, we found that inter-MT bridges are not simply bipolar connections between two MTs but are a network of interconnected struts that contact multiple MTs. We term this structure ‘the mesh’. This novel subcellular structure positions MTs within the K-fiber and is required for normal mitosis.
Inter-MT bridges were first described in 2D electron micrographs as faintly stained fine threads projecting from the surface of the tubules (Hepler et al., 1970). For example, 2D views of inter-MT bridges connecting adjacent MTs within a K-fiber are shown in Figure 1A. The morphology of inter-MT bridges is heterogeneous (Booth et al., 2011), as noted in the original report (Hepler et al., 1970). In order to examine the morphology of bridges in more detail, we used electron tomography of sections taken from mitotic HeLa cells fixed by high-pressure freezing/freeze substitution (HPF/FS). This gave a 3D view of the K-fiber MTs and the material that connected them (Figure 1B, Video 1). Careful examination of the tomograms revealed that inter-MT bridges are not just simple bipolar struts connecting two adjacent MTs: they are interconnected and can contact multiple MTs within the K-fiber. This network of inter-MT connectors is best illustrated by rendering the densities seen in the tomogram and visualizing the resultant computer model (Figure 1B). We term the interconnecting material ‘the mesh’.
To obtain an unbiased view of the mesh, we developed a semi-automated segmentation method for 3D model building (Figure 1B, Video 1). This method was used for all subsequent quantification. We found that mesh preservation was superior using HPF/FS compared with chemical fixation (Figure 1—figure supplement 1). Moreover, the mesh was associated with K-fiber MTs. In contrast, segmentation of non-K-fiber areas resulted in distinct, particulate density and translating MTs to a non-K-fiber area meant that globular density, which passed through the MTs, was detected (Figure 1—figure supplement 2).
Having established a detection method, we next explored the anatomy of this novel subcellular structure. Three types of connectors within the mesh could be distinguished in these 3D models. Bipolar, tripolar, or quadrupolar connectors could be selected using the criterion of uninterrupted density that connects two, three, or four MTs, respectively (Figure 2A). These connectors were highly heterogeneous in size. Where the volumes of individual connectors could be determined easily, we found that bipolar, tripolar, and quadrupolar connectors had mean volumes of 6,689, 17,370, and 32,376 nm3, respectively (Figure 2B). Therefore, there is not a linear relationship between the number of struts in a connector and the connector volume. The volume of mesh within each fiber was, on average, equivalent to the volume of MTs that it encapsulates and larger than the volume of MT walls alone (Figure 2C). This observation establishes that the mesh is a major component of every K-fiber.
We next examined the contacts made between the mesh and MTs to determine if there were preferred locations on the MT for mesh attachment, for example, on the seams of MTs. We found that mesh-MT contacts were highly heterogeneous (Figure 2D). For example, a crosslink between two MTs could be a simple bipolar strut with a small footprint on both connected MTs, or at the other extreme, the crosslink could be a network of multiple-linked struts, with large mesh contact areas on both MT surfaces. These larger, composite footprints extended for some distance along both linked MTs, but these were not confined to a co-axial line on the MT wall. Moreover, several mesh connectors could contact the same MT at the same axial position (e.g., Figure 1B), which further argues that the mesh has no preference for the MT seam.
Previous work indicated that one component of the mesh is a complex of TACC3–ch-TOG–clathrin whose assembly is regulated by Aurora-A kinase (Booth et al., 2011; Hood et al., 2013). Moreover, the amount of this complex on K-fibers can be increased simply by overexpressing TACC3 (Booth et al., 2011). In order to experimentally manipulate the mesh, we therefore made a stable inducible HeLa cell line where GFP-TACC3 could be overexpressed in a controlled manner (Figure 3—figure supplement 1). The most obvious effect of this manipulation was to alter MT organization. K-fibers in cells expressing GFP-TACC3 had more MTs per fiber, and the cross-sectional area that those MTs occupied was larger, compared to fibers in control uninduced HeLa cells (Figure 3A). The MT density was similar between the two groups suggesting that the fiber area scaled with the number of MTs (Figure 3A).
To look in more detail at MT packing within a K-fiber, we analyzed the distance from each MT to its nearest neighbor, and the number of neighboring MTs found within 80 nm. Both analyses showed that local MT packing density within the fiber had increased substantially in TACC3 overexpressing cells compared to uninduced controls (Figure 3B), although the fibers themselves were larger overall. A simple manifestation of this tighter local packing was the increased frequency of doublet and triplet MTs within TACC3 overexpressing K-fibers (Figure 3B). The median distance to the nearest neighboring MT had decreased from 56.1 to 48.1 nm (Figure 3C), a change in edge-to-edge proximity from 31.1 to 23.1 nm. This means that in TACC3 overexpressing cells, the average nearest neighboring MT is less than the width of one MT away. Because the overexpression of TACC3 alters the MT packing within the K-fiber, these experiments suggested to us that the mesh might influence MT spacing within the K-fiber. We return to the hypothesis that the mesh has an important role in MT spacing below.
Are the additional MTs in TACC3 overexpressing K-fibers stably attached to the kinetochore? To address this question, we used a 3D confocal microscopy assay of tubulin staining in the vicinity of kinetochores (Cheeseman et al., 2013). In agreement with the EM analysis, we detected a higher tubulin signal in cells expressing GFP-TACC3 compared to those expressing GFP alone. Following cold treatment, the tubulin intensity in the vicinity of kinetochores was reduced to comparable levels in both conditions, suggesting that the additional MTs in TACC3 overexpressing K-fibers are attached by mesh to the rest of the K-fiber but were not stably attached to the kinetochore (Figure 3D).
Overexpression of TACC3 also increased the volume of mesh between K-fiber MTs in a tomogram to 5.2 ± 1.0 × 106 nm3 (mean ± s.e.m.). This corresponds to 9.1 ± 0.01% of the fiber volume in the tomogram, whereas control mesh was 5.7 ± 0.01%. This change is somewhat difficult to interpret because of the significant increase in the number of MTs per fiber and the tighter local packing. More MTs per fiber might push up the volume of mesh, but the closer proximity of MTs limits the space available for mesh to be present.
One defining characteristic of the mesh is that it connects multiple MTs within K-fibers. This interconnectivity means that a MT that is contacted by the mesh is connected to one or more MTs and each of these, in turn, may be connected to one or more MTs and so on. We defined these connected MTs as ‘chains’. In uninduced cells, chain sizes were small, containing at most 6 MTs (Figure 4A). By contrast, cells overexpressing TACC3 had chains containing up to 12 MTs (Figure 4A), suggesting that the MTs were more interconnected as a result of TACC3 overexpression. Although MTs in TACC3 overexpressing K-fibers were more interconnected, the constitution of the connectors within the mesh was not noticeably altered (Figure 4B). In both conditions, the mesh was composed of a predominance of bipolar connectors and similar proportions of tripolar and quadrupolar connectors (Figure 4B).
We next wondered if the larger chains in TACC3 overexpressing cells were the result of the tighter local MT packing. Accordingly, we constructed 2D MT maps, where each MT's chain membership was displayed, and we compared these to the heat maps of local packing as previously described (Figure 4C). The entire dataset was analyzed computationally in order to test the idea that chain membership depended on MT proximity. For each MT, we calculated the number of neighboring MTs within a given search radius and compared these values for single (‘unchained’) MTs and those that were part of a chain (Figure 4D). This analysis showed that in controls, the single MTs and chained MTs had similar numbers of neighbors, that is, meshed MTs were not more likely to be in tightly packed regions of the fiber. However, in the TACC3 overexpressing fibers, the MTs that were part of a chain had significantly more neighbors than single MTs (Figure 4D), this pattern was seen for search radii >50 nm and was not seen if the dataset was randomized (Figure 4D, inset, see ‘Materials and methods’). The closer proximity of chained MTs vs single MTs indicates that the mesh influences MT positioning, effectively pulling them closer together. An alternative possibility is that the mesh is passive and only encapsulates MTs if they are in close proximity. However, this possibility is less likely because MTs in control fibers exhibit no difference in their proximity relative to their chain status.
The two possibilities for mesh stabilization of K-fibers are shown in (Figure 5A). To test the possibility that the mesh can influence MT positioning in K-fibers, we turned to an in vitro assay. Fluorescently labeled MTs assembled in vitro and stabilized with taxol were incubated with proteins, and any effect on MT bundling was observed by light or EM. To reconstitute the mesh component containing TACC3, a protein mixture comprising clathrin, TACC3, ch-TOG, and GTSE1 was prepared and phosphorylated by Aurora-A kinase (see ‘Materials and methods’). As a positive control, we used PRC1, which is known to bundle MTs (Mollinari et al., 2002). As a negative control, we used an equivalent amount of GST and MBP-His6 protein as in the test condition, phosphorylated by Aurora-A kinase. Thick bundles of MTs could be seen by fluorescence microscopy for the TACC3-containing complex and for PRC1 but not for the negative control (Figure 5B). These images indicated specific bundling activity for the TACC3 complex.
To look in more detail at the bundled MTs, we analyzed the control, PRC1, and TACC3 complex conditions by EM. In the PRC1 and TACC3 complex conditions, pairs of MTs were interconnected by short electron dense connectors (Figure 5C). In control conditions, MTs were randomly oriented, but on occasions when they were in close proximity, no density was seen.
These MT-bundling experiments were complicated by the purification of several large proteins (Figure 5—figure supplement 1). As an alternative source, we immunoprecipitated protein complexes, which contained clathrin and TACC3 directly from mitotic spindle of HeLa cells at metaphase (Figure 5—figure supplement 2A,B). We observed bundling of Taxol-stabilized MTs in vitro, using this complex specifically (Figure 5—figure supplement 2C).
Together these experiments indicate that the TACC3 complex can crosslink MTs and drive the recruitment of stabilized MTs into bundles. They are not compatible with a passive role for the mesh in responding to MT positioning. In the context of the K-fiber, this suggests a role for the mesh in maintaining fiber integrity by tethering MTs together and influencing MT positioning.
To look further for any evidence of a role for the mesh in MT positioning, we examined the trajectories of MTs within each fiber. K-fibers are bundles of parallel MTs, and we sought to characterize how parallel the MTs are and to test if this is altered by TACC3 overexpression. Deviations from parallel would suggest that the mesh tethers MTs closer together and interferes with their trajectory. Qualitatively, we could see that MTs in TACC3 overexpressing K-fibers were more disorganized (Figure 6A, Video 2 and Video 3). To measure parallelism more rigorously and to allow comparison of multiple K-fibers, we used several computational approaches (described in ‘Materials and methods’), which allowed us to normalize the 3D trajectory of each fiber and then examine the deviation of MT trajectories from this vector. K-fibers in cells overexpressing TACC3 had a higher proportion of MTs that deviate from parallel. They frequently had significantly larger polar angles compared to MTs in K-fibers from control cells (Figure 6B,C). The deviation of MTs from a completely parallel condition is easier to visualize, by examining the Cartesian intersection coordinates of each MT vector with an x-y plane at a given distance (z) from a common origin (see ‘Materials and methods’). These scatter plots show that MTs in TACC3 overexpressing cells deviate further from parallel than those in control cells (Figure 6D). The fraction of MT vectors intersecting this plane at radial distances of <10 nm from the center is 0.39 and 0.18 in control and TACC3 overexpressing cells, respectively. In other words, there are half as many parallel MTs after TACC3 overexpression.
Where are these deviant MTs? If they resided towards the periphery of the fiber, they may represent ‘extra MTs’, perhaps non-kinetochore MTs, that became enmeshed in the fiber when TACC3 was overexpressed. Alternatively, if the deviant MTs are throughout the fiber, this could be evidence for a role of the mesh in influencing MT spacing and packing within the fiber. To assess this, we plotted the polar and azimuthal angles as a function of distance from the K-fiber center (Figure 6E). These plots show that deviant MTs in TACC3 overexpressing cells are distributed at all distances from the center of the fiber. In controls, there was a weak tendency for deviant MTs to be at the fiber periphery, but the overall extent of trajectory deviancy was lower than in TACC3 overexpressing cells (Figure 6C,E). The lack of relation between distance from the fiber center and the deviancy from the fiber trajectory of MTs in fibers from TACC3 overexpressing cells indicates that the mesh plays an influential role in organizing MTs within the fiber. Together our results show that the change in composition of the mesh, caused by TACC3 overexpression, results in tighter local packing of MTs, more interconnectivity and disruption of the parallel organization of MTs within a K-fiber.
What are the functional consequences of the ultrastructural changes in MT organization within K-fibers? To address this question, we used live-cell imaging of our stable inducible cells expressing H2B-mCherry, growing asynchronously and compared mitotic progression in cells, where GFP-TACC3 expression was induced vs not induced. Figure 7 shows that TACC3 overexpression increases the time taken for cells to congress all chromosomes to the metaphase plate (prometaphase-to-metaphase). In addition, the time from last chromosome alignment to the onset of anaphase (metaphase-to-anaphase) is also longer (Figure 7). No change was seen in the duration of anaphase (anaphase-to-telophase). Interestingly, these changes in mitotic progression are similar to those seen following depletion of TACC3 (Lin et al., 2010; Cheeseman et al., 2013) (Figure 7—figure supplement 1). We conclude from these experiments that mitosis is sensitive to the levels of TACC3 and that this sensitivity correlates to changes in MT organization within the K-fiber.
In this paper, we described the mesh: a network of MT connectors in K-fibers. We showed that the mesh is comprised of different connector types and is a major component of the K-fiber. Overexpression of TACC3 alters the mesh and changes the trajectory of MTs within the fiber, indicating that the mesh is a determinant of K-fiber structural integrity. This manipulation causes defects in mitosis and establishes that TACC3 levels at the spindle must be regulated for optimal mitosis.
Since the first EM studies of mitotic spindles, it has been clear that K-fiber MTs are held together throughout their length, presumably to function as an integrated unit (Hepler et al., 1970; Rieder, 1981; Witt et al., 1981). The importance of inter-MT connections along the length of the K-fiber was demonstrated by severing K-fibers at their midpoint (Spurck et al., 1997). This resulted in two K-fiber stubs attached to the pole and kinetochore. The stubs do not immediately collapse but instead maintain their structural integrity. Our work identifying the mesh suggests that it is the structure that holds MTs together. This makes sense structurally, but there is a possibility that the mesh is passive and simply connects MTs in the fiber, wherever they may be. However, our work points to a role for the mesh in influencing MT position within the K-fiber. The evidence for this is: first, in K-fibers overexpressing TACC3, interconnected MTs are closer together than those not contacted by mesh. Second, a complex containing TACC3 was able to bundle MTs in vitro. Third, mesh in TACC3 overexpressing fibers pulls MTs out of alignment. In this way, the mesh plays an active role in MT positioning in K-fibers by favoring inter-MT interactions of defined distances. An extension of this is that the mesh can actively pull MTs together, in an energy-consuming process. We have no evidence for this intriguing possibility, which is compatible with our findings.
It seems counterintuitive that overexpression of a mesh component results in skewed MTs; shouldn't it make the MTs in the fiber ‘more parallel’? However, it is likely that the mesh is made of other proteins, besides the TACC3–ch-TOG–clathrin complex, for example, CHD4, HURP/DLGAP5, and HSET/KIFC1 (Mountain et al., 1999; Sharp et al., 1999; Koffa et al., 2006; Yokoyama et al., 2013). We favor the idea that because we have increased only one component of the mesh experimentally and not others, we do not see uniform changes in spacing and packing. This seems plausible as previous analysis determined TACC3–ch-TOG–clathrin was the shortest crosslinks in K-fibers (Booth et al., 2011), and so an imbalance of connector sizes in the mesh is predicted to skew MT trajectories.
The spindle matrix was originally proposed as a microtrabecular lattice that assists the motors and MTs of the mitotic spindle during chromosome movement (Pickett-Heaps et al., 1982). The existence of the matrix has been elusive, although several large macromolecular complexes have been proposed as matrix components (Pickett-Heaps et al., 1984; Chang et al., 2004; Ma et al., 2009; Johansen et al., 2011; Schweizer et al., 2014). Although superficially similar, we do not think that the K-fiber mesh is the ultrastructural correlate of the spindle matrix. The mesh that we have observed exists between MTs within K-fibers (rather than between K-fibers) and is most likely composed of MT-associated proteins that function on K-fiber MTs. Most models of the spindle matrix envisage an elastic milieu that encompasses the entire spindle, yet is independent of the MTs (Scholey et al., 2001; Yao et al., 2012). Despite this, micromanipulation experiments have left open the possibility that an anisotropic matrix could exist along K-fibers (Gatlin et al., 2010), and this would be compatible with our observations. The K-fiber mesh is in agreement with theoretical and experimental work showing that lateral MT connectors are important for MT arrays to sustain physical forces in cells (Brangwynne et al., 2006).
We used overexpression of TACC3 as our primary tool to manipulate the mesh. Progression through mitosis is slower under these conditions due to problems in chromosome congression and satisfying the spindle assembly checkpoint. Interestingly, a similar phenotype is observed when TACC3 is depleted from the spindle (Lin et al., 2010; Booth et al., 2011; Cheeseman et al., 2013). This indicates that mitosis is sensitive to TACC3 levels. We propose that this sensitivity reflects the robustness of the K-fibers. Too little TACC3–ch-TOG–clathrin and the mesh cannot strengthen the fibers adequately. Too much, and the MTs within the fibers become less parallel and oversupported. Previous work has suggested that hyperstabilization of K-fibers is counter-productive, leading to errors in mitosis (Bakhoum et al., 2009). TACC3 levels are altered in several human cancers, for example, ovarian, non-small cell lung cancer, and bladder (Schmidt et al., 2010), and fusions between TACC3 and FGFR3 are reported in glioblastoma and bladder cancer (Singh et al., 2012; Williams et al., 2013). Moreover, the amount of TACC3–ch-TOG–clathrin at the spindle is limited by the availability of phosphorylated TACC3 (LeRoy et al., 2007; Hood et al., 2013), and since Aurora-A kinase is amplified or upregulated in a wide range of cancers (Nikonova et al., 2013), it is possible that hyperstabilization of K-fibers occurs in these cells. Under any of these conditions, erroneous mitosis leading to aneuploidy may contribute to cancer initiation or progression. While we favor the idea that normal mitosis requires optimal K-fiber stabilization by the mesh, it must be noted that TACC3 has at least one additional function in mitosis (Gutiérrez-Caballero et al., 2015), and future work will seek to delineate the role of TACC3 in the mesh vs at the MT plus-ends.
Identifying and locating proteins within the mesh is the next challenge. Potentially, 3D views of K-fibers in situ will allow us to dock molecular structures of mesh components as has been achieved for other cellular systems (Lucic et al., 2013). Methods to unambiguously assign specific electron density to a particular protein will enable us to achieve this, and efforts are currently underway in our lab to do this.
HeLa cells (HPA/ECACC #93021013) were maintained in Dulbecco's Modified Eagle's Medium (DMEM) plus 10% fetal bovine serum (FBS) and 100 U/ml penicillin/streptomycin, in a humidified incubator at 37°C and 5% CO2.
For TACC3 overexpression, HeLa Tet-On cells (Clontech, Takara, Saint-Germain-en-Laye, France) were used to inducibly express GFP-TACC3. HeLa Tet-On cells were maintained in full DMEM with 300 μg/ml G418, and HeLa Tet-On cells with stably integrated pTRE2hyg-GFP-TACC3(KDP) plasmid were maintained with 300 μg/ml G418 and 200 μg/ml Hygromycin B. Expression of GFP-TACC3 was induced with 0.5 μg/ml doxycycline, 24 hr before analysis (EM analysis or mitotic progression experiments). For TACC3 knockdown, cells were transfected with plasmids to co-express GFP and shRNA using GeneJuice according to the manufacturer's instructions 2 days before imaging.
For immunofluorescence, HeLa cells on coverslips transfected to express GFP or GFP-TACC3 were either kept at 37°C or cold-treated for 5 min then fixed with ice-cold methanol for 10 min. Cells were then blocked (PBS with 5% BSA and 5% goat serum) before immunostaining with rabbit anti-alpha-tubulin (Thermo, PA5-19489, 1:500) and mouse anti-CENPA (Abcam, ab13939, 1:500) and Alexa568/Alexa633-conjugated secondary antibodies.
For EM analysis, cells were synchronized using 2 mM thymidine block for 16 hr, release for 6 hr, followed by 9 μM RO-3306 for 16 hr (see Figure 3—figure supplement 1). After 30–40 min release, mitotic cells were collected by shake-off and centrifugation at 300×g for 2 min at 37°C and resuspended in DMEM containing 20% FBS. Cell synchronization was necessary to harvest sufficient numbers of cells for our analysis. Similar results were obtained with cells synchronized using nocodazole, and with no synchronization, using a correlative light-EM approach, where a single cell was targeted for freezing.
For inducible expression, GFP-TACC3KDP (Booth et al., 2011) was inserted into pTRE2hyg vector at NheI and NotI sites. Plasmids for expression of GST-TACC3-His6 and GST-TPX2(1–43) were available from previous work (Hood et al., 2013). To make MBP-ch-TOG-His6, the coding sequence of full-length human ch-TOG was amplified by PCR to add BamHI and EagI sites before subcloning into pMALPreHis vector. For MBP-GTSE1-His6, the coding sequence of human G-2 and S-phase expressed 1 was amplified from an IMAGE clone (4138532) to add EcoRI and BamHI sites. For His6-PRC1, the coding sequence of human Protein Regulator of Cytokinesis 1 was amplified from an IMAGE clone (2958690) and inserted into pRSETB at NheI and EcoRI sites. Plasmids to co-express GFP with shRNA targeting GL2 or TACC3 (pBrain-GFP-shGL2 and pBrain-GFP-shTACC3) were available from previous work (Booth et al., 2011; Lopez-Murcia et al., 2014).
Clathrin was purified from rat liver as described previously (Hood et al., 2013; Lopez-Murcia et al., 2014). Recombinant human Aurora-A kinase was purchased from EMD Millipore (Watford, UK). Reagents for in vitro MT assembly were from Cytoskeleton Inc (Bioquote, York, UK). All proteins were purified from BL21pLysS bacteria as described previously (Hood et al., 2013).
For in vitro MT bundling experiments, the following conditions were used: Control (210 nM MBP-His6, 100 nM GST), PRC1 (2 µM His6-PRC1), and Mesh Complex (10 nM MBP-ch-TOG-His6, 100 nM clathrin, 100 nM GST-TACC3-His6, 100 nM MBP-GTSE1-His6), and visualized by fluorescence microscopy and EM. Concentrations refer to the final concentration with MTs. Proteins were incubated at 30°C for 90 min to allow for phosphorylation by TPX2(1–43)/Aurora-A, 2 µg/ml each. MTs were polymerized from 100 µM tubulin (1:10 labeled to unlabeled) in general tubulin buffer (80 mM K-Pipes, 2 mM MgCl2, 0.5 mM EGTA, pH 7.0) with 1 mM GTP and 6% glycerol for 15 min at 37°C. Diluted 1:20 in general tubulin buffer supplemented with 10 μM taxol (paclitaxel, Sigma) and 1 mM GTP, incubated for a further 5 min at 37°C. To 92.4 µl of protein mixture, 69.9 µl of MTs were added and incubated at RT for 10 min. Then 7.25 µl of this was fixed with glutaraldehyde in BRB80 (1.45 µl; 0.1% final), then diluted with 10 µl 70% glycerol/BRB80. 6 µl was spotted onto glass slides, overlaid with a cover slip, and sealed with nail varnish. The remainder was split into two and spun for 90 min RT 21,000×g and the pellets fixed in 1% glutaraldehyde for 2 hr RT, followed by post-fixing with 3% glutaraldehyde, 15 min RT. Following three washes in PBS, samples were osmicated in 1% osmium tetroxide, 1 hr RT, washed in PBS (3 × 20 min), then dH2O (2 × 20 min), and into 30% ethanol for 20 min. Samples were incubated in 0.5% uranyl acetate in 30% ethanol for 40 min. Samples were dehydrated and infused with resin and cured at 60°C for 2 days. Sections were cut and visualized by transmission EM. The MT bundling assay by fluorescence microscopy alone was carried out more than five times. The experiment where the same samples were visualized by fluorescence microscopy and EM was performed twice using different protein preparations.
Immunopurification of the mesh complex was as described previously (Booth et al., 2011). Beads from this procedure were used directly for MT-bundling experiments as described above.
For the mitotic progression assays, 6.8 × 104 cells/ml were seeded in 12 well plates and transfected with H2B-mCherry. GFP-TACC3 expression was induced the day after transfection. 24 hr later, the cells were imaged on a Nikon Ti epifluorescence microscope with 40× ELWD objective (0.6 NA) and a Coolsnap Myo camera (Photometrics (Tucson, AZ)) for 14 hr using NIS Elements AR software; H2B-mCherry was imaged once every 3 min, and GFP-TACC3 monitored once every 30 min. Cells were kept at 37°C, in supplemented CO2-independent medium containing doxycycline (0.5 μg/ml) for cells with GFP-TACC3 induction. Light intensity and exposure were minimized to avoid light-induced cell damage. Analysis of mitotic staging was by manual inspection of videos, with automated time look-up. The same microscope was used to assay the bundling of fluorescent MTs.
For analysis of MTs in the proximity of kinetochores, a similar method to that previously described was used (Cheeseman et al., 2013). Briefly, confocal imaging was performed on a spinning disc confocal system (Ultraview Vox, Perkin Elmer) using a 100× ∼1.4 NA oil immersion objective lens with a Hamamatsu C10600-10B ORCA-R2 camera. Fixed cells expressing GFP or GFP-TACC3, stained as described above, were excited at 488 nm, 561 nm, 405 nm, and 640 nm, and z-sections were taken every 0.25 μm. OME-TIFF image stacks were transferred to IMARIS (Bitplane) and subjected to spot detection in the CREST/CENP-A channel with a minimal spot size of 0.58 μm to detect kinetochores. Spots that were not properly assigned (∼5%) were corrected manually. The mean pixel densities in this sphere were background subtracted, and the median intensity per cell was found using IgorPro.
HPF was performed using either EM PACT2 or EM HPM100 (Leica Microsystems, Milton Keynes, UK). Mitotic cells were transferred to 100 μm-depth membrane carriers or Type A 100 μm-depth carriers for the EM PACT2 or EM HPM100, respectively. Once frozen, carriers were transferred in liquid nitrogen to the EM AFS2 (Leica Microsystems) FS machine, which was precooled to −90°C. Carriers were immersed in a FS medium of 1% osmium tetroxide, 0.5% uranyl acetate, and 5% H2O in acetone. Cells were brought to RT over 67 hr: 27 hr at −90°C; ramp to −60°C over 15 hr; 8 hr at −60°C; ramp to −30°C over 15 hr; 1 hr at −30°C; ramp to 4°C over 1 hr; cells were taken to the fume hood to reach RT. Cells were embedded in epoxy resin (TAAB) and polymerized at 60°C for 48 hr. Sections (70 nm) were cut using an EM UC6 (Leica Microsystems) and post-stained using uranyl acetate and Reynold's lead citrate. Sections were imaged using the FEI Tecnai G2 Spirit BioTWIN microscope at 100 kV. Putative K-fibers were identified as bundles of more than 10 MTs in sections taken orthogonal to the spindle axis. K-fibers were imaged 1 µm away from the kinetochore, so as not to include any kinetochore-MT linkages (Dong et al., 2007). Each tilt series was taken using TIA software (TEM imaging and Analysis, FEI) from +50° to −50° in 1° steps. The IMOD etomo package was used to generate tomograms from these tilt series (Mastronarde, 1997). Tomogram thickness ranged from 28.8 to 66.4 nm.
For visualization, 3D rendering of tomograms was done in Amira 5.6.0 (Visualization Sciences Group, FEI, Eindhoven, Netherlands) using a combination of manual and automated detection. Each MT was rendered individually by hand, and then any 3D density connecting the MTs was automatically detected throughout the volume of the tomogram, using the average gray value of the MTs as an initial threshold for segmentation. The space between MTs in a K-fiber is less granular than the surrounding cytoplasm, which improves mesh detection. This unbiased method superseded our initial attempts at hand-rendering the mesh and allowed for unbiased quantification and visualization of the mesh. Note that measurements are taken from the unsmoothed segmentation maps. The segmentation method detects unbroken density connected to MTs. However, after smoothing for visualization, breaks in the mesh appear. In Amira, MTs were smoothed using unconstrained smoothing, and the mesh was rendered with smooth surface (50 iterations, 0.6 lambda). Hand-rendered mesh in Figure 1B was smoothed with constrained smoothing and then smooth surface generated (20 iterations, 0.6 lambda).
In longitudinal sections, K-fibers were defined as bundles of MTs contacting both the kinetochore and the pole (McDonald et al., 1992; Booth et al., 2013). In orthogonal sections, K-fiber bundles were defined as collections of 10 or more MTs, using an 80 nm boundary around each MT (105 nm from the MT center). This distance corresponds to the longest inter-MT bridges visualized previously by 2D EM (Booth et al., 2011).
For analysis of MT packing, coordinates of each MT at the midpoint of the tomogram were collected, and the area of a convex hull that enclosed the coordinates was found. The distance to the nearest neighboring MT and the number of MTs within 80 nm (105 nm from the MT center) were calculated. Heat maps were generated by Voronoi interpolation of 3D coordinate sets that comprised the x-y coordinates together with the number of MTs within 80 nm in the z dimension.
To analyze the mesh connections and MT chaining, individual connections in rendered tomograms were classified and recorded together with the chain sizes for each K-fiber. Coordinate sets were supplemented with chain membership data. This was used for an automated comparison of the number of neighboring MTs within a given search radius, comparing chained vs single MTs. The analysis was performed from 20 to 120 nm, and non-parametric Wilcoxon–Mann–Whitney two-sample rank test used to test the hypothesis that there was no difference in the number of neighboring MTs. Chain membership data were then randomized and the analysis repeated. Note that chain membership is probabilistic and is a function of tomogram thickness and mesh density/MT interconnectivity. ‘Single MTs’ are those that are not detected to be attached to another but are likely to be attached to other MTs in a subsequent slab of K-fiber.
For analysis of MT trajectories, 3D coordinates for each MT at the bottom and top of the tomogram were logged in ImageJ and fed into IgorPro. Custom-written procedures ran through the following steps. First, coordinates were compiled into individual MT matrices. Second, the center of the bundle was found using farthest-point clustering, and then the distance of each MT from this point was calculated. Third, the fiber direction was ‘normalized’, that is, rotated through 3D space so that most MTs pointed towards the zenith. This was necessary because of the variability in the axes of different K-fibers relative to sectioning and the resultant tomograms. To do normalization, all MTs were multiplied by rotation matrices
for (in radians) 0 ≤ α ≤ 2π, 0 ≤ β ≤ π/2, γ = 0 in 180/π increments. The Cartesian distance of the orthogonal projection of each MT on the reference plane after rotation was summed, to find the rotation that produced the minimum value. The spherical coordinates, given by
of the MTs set at optimal rotation, were used for plotting. Fourth, a 2D plot to visualize the degree of MT deviancy was generated. To do this, the point at which each 3D MT vector starting at the origin intersects an x-y plane that was set arbitrarily at z = 100 nm was calculated and plotted. Note that the even radial dispersal in Figure 6D and the even spread of azimuthal angles in Figure 6E show that capture of tilt series was randomized, and the data set has no bias towards a particular trajectory. Several other strategies were explored to analyze deviations in trajectory vs the fiber axis. These were: examining the variance in trajectory angles, pairwise comparison of all MTs in the bundle, and comparison to a reference MT that represented the fiber axis, using spherical rotation and rotating by an average value. These produced similar results, however, the one described here was the most robust and represents our best method for this kind of spatial statistical analysis. The computer code used in the main analysis pipeline can be found as a Source code 1.
Images were cropped in Photoshop, and figures were assembled in Illustrator CS5.1. IgorPro 6.36 (Wavemetrics, Portland, OR) was used for all analysis and plotting.
Deviant kinetochore microtubule dynamics underlie chromosomal instabilityCurrent Biology 19:1937–1942.https://doi.org/10.1016/j.cub.2009.09.055
Studying kinetochore-fiber ultrastructure using correlative light-electron microscopyMethods in Cell Biology 115:327–342.https://doi.org/10.1016/B978-0-12-407757-7.00020-7
Microtubules can bear enhanced compressive loads in living cells because of lateral reinforcementThe Journal of Cell Biology 173:733–741.https://doi.org/10.1083/jcb.200601060
Aurora A kinase activity is required for localization of TACC3/ch-TOG/clathrin inter-microtubule bridgesCommunicative & Integrative Biology 4:409–412.https://doi.org/10.4161/cib.4.4.15250
Specific removal of TACC3-ch-TOG-clathrin at metaphase deregulates kinetochore fiber tensionJournal of Cell Science 126:2102–2113.https://doi.org/10.1242/jcs.124834
Clathrin recruits phosphorylated TACC3 to spindle poles for bipolar spindle assembly and chromosome alignmentJournal of Cell Science 123:3645–3651.https://doi.org/10.1242/jcs.075911
Directly probing the mechanical properties of the spindle and its matrixThe Journal of Cell Biology 188:481–489.https://doi.org/10.1083/jcb.200907110
Intermicrotubule bridges in mitotic spindle apparatusThe Journal of Cell Biology 45:438–444.https://doi.org/10.1083/jcb.45.2.438
Coordination of adjacent domains mediates TACC3-ch-TOG-clathrin assembly and mitotic spindle bindingThe Journal of Cell Biology 202:463–478.https://doi.org/10.1083/jcb.201211127
Quantitative proteomics combined with BAC TransgeneOmics reveals in vivo protein interactionsThe Journal of Cell Biology 189:739–754.https://doi.org/10.1083/jcb.200911091
Clathrin heavy chain mediates TACC3 targeting to mitotic spindles to ensure spindle stabilityThe Journal of Cell Biology 189:1097–1105.https://doi.org/10.1083/jcb.200911120
Presynaptic clathrin levels are a limiting factor for synaptic transmissionThe Journal of Neuroscience 34:8618–8629.https://doi.org/10.1523/JNEUROSCI.5081-13.2014
Cryo-electron tomography: the challenge of doing structural biology in situThe Journal of Cell Biology 202:407–419.https://doi.org/10.1083/jcb.201304193
Requirement for Nudel and dynein for assembly of the lamin B spindle matrixNature Cell Biology 11:247–256.https://doi.org/10.1038/ncb1832
Structural and regulatory roles of nonmotor spindle proteinsCurrent Opinion in Cell Biology 20:101–106.https://doi.org/10.1016/j.ceb.2007.11.004
Dual-axis tomography: an approach with alignment methods that preserve resolutionJournal of Structural Biology 120:343–352.https://doi.org/10.1006/jsbi.1997.3919
Kinetochore fiber maturation in PtK1 cells and its implications for the mechanisms of chromosome congression and anaphase onsetThe Journal of Cell Biology 137:1567–1580.https://doi.org/10.1083/jcb.137.7.1567
PRC1 is a microtubule binding and bundling protein essential to maintain the mitotic spindle midzoneThe Journal of Cell Biology 157:1175–1186.https://doi.org/10.1083/jcb.200111052
Aurora A kinase (AURKA) in normal and pathological cell divisionCellular and Molecular Life Sciences 70:661–687.https://doi.org/10.1007/s00018-012-1073-7
Direct kinetochore-spindle pole connections are not required for chromosome segregationThe Journal of Cell Biology 206:231–243.https://doi.org/10.1083/jcb.201401090
Ultraviolet microbeam irradiations of epithelial and spermatocyte spindles suggest that forces act on the kinetochore fibre and are not generated by its disassemblyCell Motility and the Cytoskeleton 36:136–148.https://doi.org/10.1002/(SICI)1097-0169(1997)36:2<136::AID-CM4>3.0.CO;2-7
Arms and bridges on microtubules in the mitotic apparatusThe Journal of Cell Biology 40:854–859.https://doi.org/10.1083/jcb.40.3.854
A nuclear-derived proteinaceous matrix embeds the microtubule spindle apparatus during mitosisMolecular Biology of the Cell 23:3532–3541.https://doi.org/10.1091/mbc.E12-06-0429
Anna AkhmanovaReviewing Editor; Utrecht University, Netherlands
eLife posts the editorial decision letter and author response on a selection of the published articles (subject to the approval of the authors). An edited version of the letter sent to the authors after peer review is shown, indicating the substantive concerns or comments; minor concerns are not usually shown. Reviewers have the opportunity to discuss the decision before the letter is sent (see review process). Similarly, the author response typically shows only responses to the major concerns raised by the reviewers.
Thank you for sending your work entitled “The mesh is a network of microtubule contacts that stabilizes individual kinetochore fibers of the mitotic spindle” for consideration at eLife. Your article has been favorably evaluated by Vivek Malhotra (Senior editor), a Reviewing editor, and three reviewers.
The following individuals responsible for the peer review of your submission have agreed to reveal their identity: J Richard McIntosh and Helder Maiato. A further reviewer remains anonymous.
The Reviewing editor and the reviewers discussed their comments before we reached this decision, and the Reviewing editor has assembled the following comments to help you prepare a revised submission.
This is an interesting and one can even say “courageous” paper. Most cell biologists like to work on defined proteins or periodic structures that allow experimental rigor by established criteria. This paper describes work on a component of the mitotic spindle that certainly exists and some of whose components are known, but whose structure is ill-defined and therefore very hard to study. This group has been pursuing aspects of the inter-microtubule “mesh” seen in kinetochore fibers for some time, and this paper takes that work a big step along. However, before the paper can be published, some additional analyses and clarifications are required.
1) While the 3D rendering of the mesh suggests interaction with up to 4 microtubules, careful inspection of the electron densities in the tomograms corresponding to bridges involving more than 2 microtubules is not very convincing from the images provided. For instance, in Figure 1A, quadrupolar and tripolar bridges are often interrupted, but are shown as continuous electron densities in the 3D rendered images. Because the criterion of uninterrupted density was chosen for this classification, the authors need to explain why this is not visible in the tripolar and quadrupolar connections and clarify if any threshold was used for 3D-rendering.
2) The authors should show how the mesh compares to an area without microtubules in the spindle and, ideally, to non-kinetochore microtubules. This is important to understand whether the features revealed are specific for K-fibers, as they claim.
3) The authors should characterize K-fiber stability (even if by just a crude cold assay) after TACC3 overexpression. If possible, it would also be nice to compare the results with TACC3 depletion.
4) The decision of the authors to use populations of cells synchronized by thymidine and then RO-3306 (both for 16 hrs) in the analysis of control K-fibers is somewhat puzzling. 16 hrs of late G2 arrest is not benign. For example, centrosome organization becomes abnormal. This in turn affects the organization of the mitotic spindle. Whether the fine structure of K-fibers is affected is not known but there is a danger that it is. The procedure used and its justification should be clearly explained in the main text of the paper.
5) The in vitro reconstitution experiments are performed with a rather ill-defined mixture of four components. The authors should at least illustrate the quality of the individual proteins by Coomassie-stained gels. They should also explain in the main text why they think that all four components need to be added to observe microtubule bundling, whether the HIS-tags present on the individual proteins can contribute to microtubule bundling and also explain the rationale of adding ch-TOG in sub-stoichiometric amounts compared to the other components. Do the authors have any indications that the four components purified by them indeed form a (defined) complex in vitro?
1) We couldn't find information about the mitotic stage in which the EM analysis was performed and how K-fibers were identified.
2) In the Introduction, the authors mention that: “chromosome movements are governed by the kinetochore fibers (K-fibers) of the mitotic spindle”. Not all chromosome movements are governed by K-fibers. This is the case for example of chromosome congression mediated by motor proteins, which actually do not need K-fibers to move chromosomes to the spindle equator (work by the Walczak and Khodjakov labs). This should be corrected.
3) In the section “Ultrastructural morphology of the mesh”, the authors state: “The density that we observe as mesh is likely made from many different proteins and the lack of scaling in these measurements suggest that connectors are not multimeric assemblies of a single subunit type”. This sentence is highly speculative and does not really add anything. The lack of scaling could simply reflect conformational changes in highly flexible structures that form inter-microtubule bridges.
4) The analogy with the spindle matrix is unavoidable and the authors disclose this in a nice way in the Discussion, but it does raise the question of whether the mesh is resistant to microtubule depolymerization (related to point 2 in the major concerns, above).
5) The argument against the preferential binding to the microtubule seam is totally dispensable, since one microtubule is able to bind more than 2 connectors.
6) How much is TACC3 being overexpressed and what is the effect over the other proteins of the complex? This important control is missing.https://doi.org/10.7554/eLife.07635.021
- Cristina Gutiérrez-Caballero
- Stephen J Royle
- Faye M Nixon
- Stephen J Royle
The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.
We thank Tom Honnor and Julia Brettschneider, members of the Royle lab and Anne Straube for critical discussion. We also thank Alison Beckett, Sam Williams, and Liam Cheeseman for technical help and Rob Cross for commenting on the manuscript. This work was supported by a project grant from North West Cancer Research (CR928) and by a Senior Cancer Research Fellowship from Cancer Research UK (C25425/A15182) to SJR.
- Anna Akhmanova, Reviewing Editor, Utrecht University, Netherlands
© 2015, Nixon et al.
This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.