NuMA assemblies organize microtubule asters to establish spindle bipolarity in acentrosomal human cells

Abstract

In most animal cells, mitotic spindle formation is mediated by coordination of centrosomal and acentrosomal pathways. At the onset of mitosis, centrosomes promote spindle bipolarization. However, the mechanism through which the acentrosomal pathways facilitate the establishment of spindle bipolarity in early mitosis is not completely understood. In this study, we show the critical roles of nuclear mitotic apparatus protein (NuMA) in the generation of spindle bipolarity in acentrosomal human cells. In acentrosomal human cells, we found that small microtubule asters containing NuMA formed at the time of nuclear envelope breakdown. In addition, these asters were assembled by dynein and the clustering activity of NuMA. Subsequently, NuMA organized the radial array of microtubules, which incorporates Eg5, and thus facilitated spindle bipolarization. Importantly, in cells with centrosomes, we also found that NuMA promoted the initial step of spindle bipolarization in early mitosis. Overall, these data suggest that canonical centrosomal and NuMA‐mediated acentrosomal pathways redundantly promote spindle bipolarity in human cells.

Studies on acentrosomal human cells suggest that canonical centrosomal and NuMA/dynein‐mediated mechanisms may redundantly promote establishment of bipolar mitotic spindles after nuclear envelope breakdown.

Introduction

The mitotic spindle, a large cytoskeletal structure, ensures proper chromosomal segregation (Prosser & Pelletier, 2017). In somatic animal cells, the mitotic spindle poles consist of centrosomes and several factors involved in their organization. The centrosomes consist of a pair of centrioles surrounded by pericentriolar material (PCM; Bornens, 2002), which nucleates and anchors microtubules. Centrosomes act as the major microtubule‐organizing centers and organize the radial array of microtubules during early mitosis (Kaseda et al, 2012). At the G2/M transition, centrosomes move away from each other through the activity of centrosome separation pathways (Rattner & Berns, 1976; Waters et al, 1993; Whitehead & Rattner, 1998; Toso et al, 2009; Raaijmakers et al, 2012) in order to assemble a bipolar mitotic spindle. At the onset of centrosome separation, the evolutionary conserved kinesin Eg5 (Enos & Morris, 1990; Hagan & Yanagida, 1990; Heck et al, 1993; Blangy et al, 1995) is loaded onto centrosomes. Therefore, the centrosomal microtubules promote the initial steps establishing mitotic spindle bipolarity (Megraw et al, 2001; Bertran et al, 2011; Smith et al, 2011; Kaseda et al, 2012; Hata et al, 2019).

In addition to centrosomes, spindle pole organization is regulated by several proteins, such as nuclear mitotic apparatus protein (NuMA) and the dynein–dynactin complex. The functions of these proteins for spindle pole organization and spindle positioning have been well studied in meiotic and mitotic spindles. It has been shown that NuMA is recruited to the minus‐ends of microtubules dependent on (Merdes et al, 2000) or independently of dynein (Hueschen et al, 2017). During prophase, NuMA begins to form aggregates in the nucleus (Kisurina‐Evgenieva et al, 2004). These small aggregates of NuMA are incorporated around centrosomes within a few minutes after nuclear envelope breakdown (NEBD; Kisurina‐Evgenieva et al, 2004). Specific inhibition of NuMA or dynein disrupts spindle pole formation (Merdes et al, 1996, 2000). Thus, the spindle pole is organized by NuMA and the dynein–dynactin complex (Merdes et al, 1996, 2000; Hueschen et al, 2017). In addition to their functions in spindle pole organization, NuMA and the dynein–dynactin complex cooperate to regulate spindle positioning by forming dynein–dynactin–NuMA clusters at the mitotic cell cortex through the clustering activity of NuMA (Okumura et al, 2018).

In contrast, oocytes do not have centrosomes; instead, they exhibit acentrosomal spindle formation. In mouse oocytes, the meiotic spindle poles are organized by NuMA, and the microtubule binding activity of NuMA is required for female fertility (Kolano et al, 2012). It has been shown that NuMA localizes at the meiotic spindle poles during meiosis I and II in human oocytes (Xu et al, 2011). In addition to meiotic spindles, previous studies have shown that vertebrate and fly somatic cells can divide continuously, even after the removal of centrosomes (Bonaccorsi et al, 2000; Khodjakov et al, 2000; Basto et al, 2006; Hornick et al, 2011; Sir et al, 2013) and, in green monkey fibroblasts, NuMA accumulates at the acentrosomal spindle poles (Khodjakov et al, 2000). Although these studies suggest that meiotic and mitotic spindle poles are organized by NuMA, the exact functions of NuMA in the initial steps of spindle bipolarization remain to be elucidated.

The microtubule organization activity of NuMA has been studied in detail (Nachury et al, 2001; Wiese et al, 2001). It has been shown that NuMA possesses clustering properties through its C‐terminus (Harborth et al, 1999) and forms microtubule‐independent insoluble complexes that are resistant to nocodazole treatment (Dionne et al, 1999; Merdes et al, 2000; Kisurina‐Evgenieva et al, 2004; Hueschen et al, 2017). It has also been reported that the C‐terminus of NuMA interacts with microtubules (Du et al, 2002; Haren & Merdes, 2002), presumably mediating centrosome‐independent aster formation (Gaglio et al, 1995; Du et al, 2002; Haren & Merdes, 2002). The properties of NuMA, which assemble and organize an aster‐like microtubule array, are similar to those of centrosomes in early mitosis. However, the role of NuMA in the initial step of spindle bipolarization, rather than spindle pole organization or spindle positioning, is not fully understood yet.

In this study, using live‐cell imaging and auxin‐inducible degradation (AID) experiments, we revealed that acentrosomal human somatic cells organized bipolar mitotic spindles in a NuMA‐dependent manner. In acentrosomal cells, NuMA started to assemble and formed microtubule asters shortly after NEBD. Depletion of NuMA induced umbrella‐like monopolar spindles in acentrosomal cells, suggesting that NuMA organizes the microtubule array and Eg5 localization to promote proper separation of the poles. In addition, we found that the depletion of NuMA delayed spindle bipolarization in cells with centrosomes. Collectively, based on the present data, we propose the following model: (i) Canonical centrosomal and NuMA‐mediated acentrosomal pathways redundantly promote spindle bipolarization in early mitosis of cells with centrosomes; and (ii) NuMA can compensate for the function of centrosomes in the initial step of bipolar spindle formation in acentrosomal human cells.

Results

NuMA comprises a structure specific to acentrosomal human cells in the process of bipolar spindle assembly

We initially induced the formation of acentrosomal spindles in human cells to understand the mechanism through which human somatic cells can establish spindle bipolarity without centrosomes. For this purpose, we treated HeLa cells with centrinone, a specific PLK4 inhibitor (Wong et al, 2015), to remove the centrosomes. We subsequently examined the mechanism through which the two separate spindle poles were established in these acentrosomal cells. Immunofluorescence was used to analyze the localization patterns of proteins known to accumulate at the mitotic spindle poles (Fig 1A–F, Appendix Figs S1 and S2). We confirmed that, in cells with two centrosomes, the spindle poles consist of NuMA/p150^Glued and ASPM–katanin complexes (Fig 1A–C). Interestingly, the acentrosomal cells assembled a bipolar spindle with a dense NuMA structure between the spindle poles (Fig 1A and B). In addition, p150^Glued, a dynactin subunit that forms a complex with NuMA, also showed this dense structure (Fig 1C). In contrast, in these spindles, the ASPM–katanin complex (Jiang et al, 2017) was distributed only near the spindle poles (Fig 1B and C). We subsequently observed the localization pattern of the PCM components in acentrosomal cells. Several PCM components (e.g., pericentrin, CDK5RAP2, and Cep192) were not detected at most acentrosomal spindle poles in HeLa cells (Appendix Fig S1A and B). Furthermore, pericentrin was consistently undetectable at the acentrosomal spindle poles of various human cell lines (Appendix Fig S1C and D). Collectively, these data indicate that the organization of NuMA/p150^Glued and ASPM–katanin complexes at acentrosomal spindle poles differs.

Figure 1. Acentrosomal human cells form a specific structure of NuMA in the process of bipolar spindle assembly.

• A–C
  Distribution of spindle pole factors in centrosomal and acentrosomal spindles. DMSO‐treated control mitotic spindles (2‐centrosomes) and centrinone‐treated acentrosomal spindles (0‐centrosome) of HeLa cells. (A) Red, gray, green, and blue represent NuMA, α‐tubulin, Cep152, and DNA, respectively. Z‐projections of eight sections, every 0.3 μm. Scale bar, 5 μm. (B) Red, gray, and blue represent NuMA, katanin p60, and DNA, respectively. Z‐projections of five sections, every 0.3 μm. Scale bar, 5 μm. (C) Red, gray, and blue represent p150^Glued, ASPM, and DNA, respectively. Z‐projections of 10 (2‐centrosomes) or five (0‐centrosome) sections, every 0.3 μm. Scale bar, 5 μm.
• D
  The structure of NuMA and microtubules in centrinone‐treated acentrosomal spindles of HeLa cells. Two‐way arrows indicate the bipolarity of elongated NuMA structures. Gray, red, green, and blue represent α‐tubulin, NuMA, Cep152, and DNA, respectively. Scale bar, 5 μm.
• E
  The localization of p150^Glued in centrinone‐treated acentrosomal spindles of HeLa cells. Red, gray, green, and blue represent NuMA, p150^Glued, Cep152, and DNA, respectively. Scale bar, 5 μm.
• F
  NuMA distribution in 3D. Images in (D) of NuMA (Red) and chromosomes (blue) were reconstructed in 3D and surface‐rendered.
• G
  Representative images of NuMA (arrowheads) in centrosomal and acentrosomal cells before and after photobleaching. DMSO‐ or centrinone‐treated HCT116 TetOsTIR1 NuMA–mAID–mClover–FLAG cells were observed. Time after photobleaching (s) is indicated. Scale bar, 5 μm.
• H
  Normalized fluorescence intensities of NuMA–mClover at centrosomal poles (control, 2‐centrosomes) or at acentrosomal spindle poles (0‐centrosome) during the FRAP time course (N = 20 cells for each condition). The first three time points are before photobleaching. Error bars show standard deviation (SD).
• I
  Mobile fractions from the normalized recovery curves of NuMA–mClover at centrosomal spindle poles (control, 2‐centrosomes) or at acentrosomal spindle poles (0‐centrosome). Mobile fractions (roughly corresponds to normalized fluorescence intensities at the last time point of (H)) were determined by single exponential fitting. Line and error bars represent the median and interquartile range (N = 20 cells for each condition). The Mann–Whitney U‐test (two‐tailed) was used to obtain a P value.

Subsequently, we observed the localization patterns of NuMA, α‐tubulin, and chromosomes during mitosis to further analyze the establishment of spindle bipolarity in acentrosomal cells. Acentrosomal cells formed a unique pattern of NuMA and p150^Glued in the center of the condensed chromosome ring (monopolar‐like; Fig 1D–F). These cells organized the radial array of microtubules around the NuMA structure (Fig 1D). In acentrosomal cells, NuMA also showed a bundle‐like pattern, which colocalized with microtubules (bundle‐like; Fig 1D and F). It has been shown that microtubule bundles form within the meiotic spindle through the assembly of central spindle components (So et al, 2019). However, the bundle‐like structure in acentrosomal human cells did not contain the central spindle components, such as PRC1 (Mollinari et al, 2002) and KIF4 (Kurasawa et al, 2004; Appendix Fig S2A and B). The structure of NuMA in acentrosomal cells was similarly formed in various cell lines derived from different human tissues, including nontransformed and hTERT‐immortalized human retinal pigment epithelium cells (RPE1; Appendix Fig S2C and D). It has been reported that PLK4 contributes to microtubule nucleation in oocytes (Bury et al, 2017). Thus, we examined whether the observed structure of NuMA could be attributed to the inhibition of PLK4 or to the low microtubule density induced by the removal of centrosomes. Monopolar‐ and bundle‐like NuMA structures were similarly observed in centrosome‐eliminated cells after Sas6 depletion using the AID system (Appendix Fig S2E; Yoshiba et al, 2019), suggesting that the aforementioned phenotype was not the result of PLK4 inhibition. We further confirmed that the microtubule density in the acentrosomal spindle was higher than that in the centrosomal spindle (Appendix Fig S2F). This result suggests that the NuMA structure in acentrosomal cells is not formed due to low microtubule density.

In addition, we performed fluorescence recovery after photobleaching (FRAP) analysis using HCT116 cells, in which endogenous NuMA was tagged with mClover to further characterize the properties of NuMA in the acentrosomal spindle pole. Prior to photobleaching, the fluorescence intensity of NuMA in acentrosomal spindle poles was significantly higher than that measured in centrosomal poles (Fig 1G, Appendix Fig S3A, and Movies EV1 and EV2). In control cells, the fluorescence of NuMA recovered to 54% on average within 3 min (Fig 1G–I, Movie EV1). In contrast, the fluorescence of NuMA in acentrosomal cells recovered to only 29% on average (Fig 1G–I, Movie EV2), suggesting that a large proportion of NuMA is more static in the acentrosomal spindle poles than in the centrosomal poles. However, the half‐time for the recovery and absolute value of the mobile fraction in acentrosomal spindle poles were comparable with those determined in centrosomal poles (Appendix Fig S3B–D). Thus, these data suggest that the acentrosomal spindle pole consists of two populations of NuMA: a dynamic population, which is also present in centrosomal poles; and a static fraction, which is specific to acentrosomal spindle poles.

Collectively, these results suggest that a bipolar spindle in acentrosomal human cells is established in two steps: (i) organization of the radial array of microtubules around the dense NuMA structure and (ii) separation of the two spindle poles. This system may be conserved for the formation of bipolar mitotic spindles in acentrosomal human cells.

NuMA organizes small asters during NEBD and assembles a small spindle prior to spindle elongation and chromosome congression

Next, we further investigated the time course of mitotic events in acentrosomal cells using live‐cell imaging. We used time‐lapse fluorescence microscopy to track the localization and dynamics of endogenous NuMA tagged with mCherry or mClover in HeLa cells. In control cells with centrosomes, the centrosomes are dissociated through degradation of the centrosomal linker before mitosis (Wang et al, 2014). In these cells, NuMA was rapidly recruited to the two separated spindle poles during NEBD (Appendix Fig S4A and B) as previously described (Kisurina‐Evgenieva et al, 2004). In contrast, acentrosomal cells formed several aggregates of NuMA after NEBD (4 min: Fig 2A; 0 min: Fig 2B, Movies EV3 and EV4). Subsequently, the bipolarity of the acentrosomal mitotic spindle, which was judged based on two separate NuMA structures, was established on average within 11.8 min after NEBD (Fig 2B and C). After the establishment of bipolarity of the acentrosomal mitotic spindle, the chromosomes started to be aligned on the metaphase plate (Fig 2B and C). This observation was also confirmed through immunofluorescence analysis with fixed cells, which showed that chromosomes and kinetochores localized around the elongating bundle‐like structure of NuMA (Fig 2D). These acentrosomal cells showed a delay in chromosome alignment after NEBD and in the transition from chromosome alignment to the onset of anaphase versus control cells with centrosomes (Appendix Fig S4C and D), suggesting that both processes of spindle bipolarization and spindle assembly checkpoint satisfaction were delayed in acentrosomal spindle formation. In addition, a high frequency of chromosome segregation errors was detectable in acentrosomal cells as previously described (Sir et al, 2013; Appendix Fig S4E). Collectively, these findings indicate that bipolarity within the acentrosomal mitotic spindle is established after NEBD and that the process precedes chromosomal alignment.

Figure 2. NuMA organizes small asters at the time of NEBD and assembles a small spindle before spindle elongation and chromosome congression.

• A
  Time‐lapse observation of the establishment of bipolarity in acentrosomal cells. HeLa cells expressing EGFP–centrin1 and mCherry–NuMA were observed with a 63× objective. Magenta and green represent mCherry–NuMA and EGFP–centrin1, respectively. Arrowheads indicate the assembling NuMA after NEBD. Two‐way arrows indicate the bipolarity. Z‐projections of 20 sections, every 1.2 μm. Scale bar, 10 μm. Time zero corresponds to NEBD.
• B
  Time‐lapse observation of the dynamics of NuMA and chromosomes. Centrinone‐treated HeLa cells expressing mCover–NuMA were observed with a 63× objective. Magenta and green represent SiR‐DNA and mClover–NuMA, respectively. Arrowheads and two‐way arrows indicate the assembling NuMA after NEBD and bipolarity, respectively. Z‐projections of 20 sections, every 1.2 μm. Scale bar, 10 μm. Time zero corresponds to NEBD.
• C
  The timing of bipolarity establishment in the acentrosomal pole, chromosome alignment, and anaphase onset in (B). Each plot shows the cumulative percentage of each event at individual time points (N = 49 cells from two independent experiments).
• D
  The distribution of kinetochores during acentrosomal spindle pole separation was confirmed in fixed cells. Two‐way arrow indicates the bipolarity of elongated NuMA structures. Red, gray, green, and blue represent NuMA, CENP‐C, GT335, and DNA, respectively. Scale bar, 5 μm.
• E, F
  Time‐lapse observation of the structure of NuMA and microtubules upon centrosome removal. Centrinone‐treated HCT116 TetOsTIR1 NuMA–mAID–mClover–FLAG cells were observed with a 60× objective. Red and green represent NuMA and SiR‐tubulin, respectively. Z‐projections of 17 sections, every 1 μm. Scale bar, 5 μm. Time zero corresponds to NEBD. (F) An example of NuMA formed several asters (arrowheads) at the time of NEBD.

We subsequently examined the mechanism through which separation of the two spindle poles is achieved in acentrosomal cells, with a particular focus on the structure of NuMA and microtubules. We used a spinning disk confocal microscopy SpinSR10 system with a sCMOS camera to track the structure of NuMA and microtubules. Acentrosomal HCT116 NuMA–mAID–mClover cells showed several patterns of NuMA structures (Fig 2E and F, Movies EV5 and EV6). In one example, NuMA started to form two aggregates after NEBD (10 min, Cell 1, Fig 2E). These aggregates subsequently assembled into one structure (30 min, Cell 1, Fig 2E). After the assembly of the monopolar‐like spindle (30 min, Cell 1, Fig 2E), the spindle elongated with a bundle‐like structure of NuMA (60–70 min Cell 1, Fig 2E). This NuMA structure was eventually separated into two poles (80 min, Cell 1, Fig 2E). In other examples, NuMA formed several asters during NEBD (Fig 2F) or after NEBD (10 min, Cell 2, Fig 2E). These asters then assembled into two aggregates of NuMA to form a small bipolar spindle (40 min, Cell 2, Fig 2E), and this small spindle subsequently started to elongate (40, 70 min, Cell 2, Fig 2E). These observations suggest that NuMA organizes small asters during NEBD and assembles into two separate aggregates that eventually function as the two spindle poles.

Microtubules and dynein are required for NuMA assembly to form two acentrosomal spindle poles

We next addressed the mechanisms underlying NuMA assembly for acentrosomal spindle pole formation. It has been established that NuMA is transported toward the spindle poles in a microtubule‐based motor dynein‐dependent manner (Merdes et al, 1996, 2000; Hueschen et al, 2017). Therefore, we validated the requirement of microtubules and dynein for NuMA assembly at acentrosomal spindle poles. Nocodazole, a microtubule polymerization inhibitor, induced multiple asters with NuMA, suggesting that NuMA was assembled into two acentrosomal spindle poles in a microtubule‐dependent manner (Fig 3A).

Figure 3. Microtubule, dynein, and clustering activity of NuMA are required for the assembly of NuMA aggregates in acentrosomal human cells.

1. NuMA structure upon microtubule depolymerization. Nocodazole‐treated HCT116 CMVOsTIR1 CEP152–mClover–mAID cells were treated with 100 ng/ml nocodazole. Gray, red, green, and blue represent α‐tubulin, NuMA, CEP152 (mClover), and DNA, respectively. Scale bar, 5 μm.
2. Immunostaining of DHC1 and NuMA in centrinone‐treated HCT116 TetOsTIR1 DHC1–3X–mAID–mClover cells. Gray, green, red, and blue represent DHC1 (mClover), GT335, NuMA, and DNA, respectively. Scale bar, 5 μm.
3. Schematic illustration of target protein depletion using the AID system to analyze the requirement of dynein for acentrosomal spindle pole formation.
4. NuMA structure upon DHC1 depletion. Centrinone‐treated HCT116 TetOsTIR1 DHC1–3X–mAID–mClover cells were treated with 1 μg/ml doxycycline (Dox) and 500 μM indole‐3‐acetic acid (IAA). Arrowhead indicates the spindle pole. Green, red, gray, and blue represent DHC1 (mClover), NuMA, GT335, and DNA, respectively. Scale bar, 5 μm.
5. Frequency of acentrosomal pole patterns upon DHC1 depletion in (D). Values are mean percentages ± SD from three independent experiments (N ≧ 24 spindles in each experiment).
6. The distribution of NuMA upon DHC1 depletion. Arrowheads indicate the presence of NuMA on microtubules. Green, red, and blue represent α‐tubulin, NuMA, and DNA, respectively. Scale bars, 5 μm.
7. Structure of acentrosomal spindle poles upon replacement of endogenous NuMA–mAID–mClover–FLAG with either mCherry–NuMA WT or 5A‐3. Arrowheads indicate the assembled NuMA structure. Green, red, gray, and blue represent endogenous NuMA (mClover), expressed NuMA (mCherry), GT335, and DNA, respectively. Z‐projections of 21 sections, every 1 μm. Scale bar, 5 μm.
8. The number of NuMA structure in (G). Values are mean percentages ± SD from four independent experiments (N ≧ 15 spindles in each experiment). One‐way ANOVA with Tukey's multiple comparisons test was used to obtain a P value.

Next, time‐lapse fluorescence microscopy revealed localization patterns of dynein during mitosis by tracking endogenous dynein heavy chain 1 (DHC1) tagged with 3X–mAID–mClover. In control cells with centrosomes, DHC1 initially showed a kinetochore‐like distribution and was rapidly recruited into the spindle poles (Appendix Fig S5A). On the other hand, in acentrosomal cells, DHC1 showed a kinetochore‐like distribution before being redistributed into the proximity of NuMA structures in the center of the condensed chromosome ring (Appendix Fig S5B). As the separation of two spindle poles proceeded, DHC1 localized around the NuMA structure at the spindle poles or at the periphery of the poles in acentrosomal cells (Fig 3B). In contrast, in cells with centrosomes, these proteins colocalized at the spindle poles (Appendix Fig S5C).

We then depleted DHC1 protein using the AID system and tested whether this depletion affected the formation of the NuMA structure (Fig 3C–F). The AID system enabled us to efficiently reduce the expression levels of the target protein (Nishimura et al, 2009; Natsume et al, 2016; Okumura et al, 2018) and specifically test the requirement of DHC1 for the assembly of NuMA. Interestingly, DHC1 depletion disrupted NuMA assembly, resulting in small dispersed aggregates of NuMA (Fig 3C–E and Appendix Fig S5D). We also noted that NuMA continued to localize at the microtubules (Fig 3F). This finding suggests that, under these conditions, dynein is required for the assembly of NuMA proteins in acentrosomal spindle pole formation. Taken together, these results suggest that the microtubule and the dynein motors facilitate NuMA assembly in acentrosomal cells.

The clustering activity of NuMA partially contributes to NuMA assembly in acentrosomal spindle poles

It has been shown that NuMA forms oligomeric structures in vitro (Harborth et al, 1999), and its clustering activity is important for the generation of spindle‐pulling forces at the cortex (Okumura et al, 2018). These observations raised the question of whether the clustering activity of NuMA is also required for acentrosomal spindle pole formation. To address this, we replaced endogenous NuMA with either exogenous NuMA wild‐type (WT) or a clustering‐defective mutant (5A‐3) by combining the AID system with overexpression. Specifically, we used the AID system to deplete endogenously tagged NuMA–mAID–mClover in cells expressing exogenous mCherry–NuMA WT or 5A‐3 mutant (Appendix Fig S6A). The expression levels of mCherry–NuMA WT and 5A‐3 were comparable upon simultaneous depletion of centrioles and endogenous NuMA (Appendix Fig S6B). Upon replacement of the endogenous NuMA–mAID–mClover with either mCherry–NuMA WT or mCherry–NuMA 5A‐3, we checked whether the cells could assemble two NuMA structures for bipolar spindle formation. In the absence of indole‐3‐acetic acid (IAA), cells expressing both NuMA–mAID–mClover and mCherry–NuMA WT or 5A‐3 assembled 2.2 aggregates of NuMA on average (Fig 3G and H). In the presence of IAA, cells in which endogenous NuMA was replaced with mCherry–NuMA WT assembled 2.8 aggregates of mCherry–NuMA (Fig 3G and H). In contrast, the cells in which endogenous NuMA was replaced with mCherry–NuMA 5A‐3 showed defects in NuMA assembly; these cells assembled 3.4 aggregates of mCherry–NuMA 5A‐3 (Fig 3G and H). Such defects in the assembly of mCherry–NuMA were not observed in centrosomal spindle pole formation (2.0 aggregates of NuMA) (Appendix Fig S6C and D). Collectively, these results suggest that the clustering activity of NuMA is required for proper formation of the two acentrosomal spindle poles.

NuMA plays a critical role in promoting mitotic spindle bipolarity in acentrosomal cells

We next analyzed the function of NuMA in microtubule organization and subsequent bipolar spindle formation using the AID system (Okumura et al, 2018). We induced the degradation of NuMA (Appendix Fig S7A and B) and observed the structure of mitotic spindles in acentrosomal cells (Fig 4A–C). Immunofluorescence analyses revealed that 47.8% of control acentrosomal cells showed bipolar spindles (Fig 4A–C), whereas the rest of the cells exhibited monopolar spindles, mostly with a radial array of microtubules (Pattern 1 in Fig 4B and C). Interestingly, 88.9% of acentrosomal cells in which NuMA was depleted using the AID system showed umbrella‐like monopolar spindles (Pattern 2 in Fig 4B and C). However, these were not observed in control acentrosomal cells (Fig 4A–C). In these cells, such umbrella‐like monopolar spindles were well focused, suggesting that NuMA is required for organizing the radial array of microtubules rather than focusing of spindle poles in acentrosomal cells in this experimental condition. The structure of the acentrosomal mitotic spindles was not affected by treatment with doxycycline and IAA in the negative control cells, in which endogenous NuMA was not tagged with mAID–mClover–FLAG. This result confirmed that the phenotype was not the result of nonspecific effects caused by the toxicity of doxycycline and/or IAA (Appendix Fig S7E and F). Although the NuMA‐depleted acentrosomal cells finally managed to segregate the chromosomes after the umbrella‐like monopolar state, the duration of mitosis in these cells was significantly prolonged (Appendix Fig S7C and D). In addition, we confirmed that, in the absence of NuMA, p150^Glued was strongly displaced from the acentrosomal spindle poles (Appendix Fig S7H). Thus, it is likely that NuMA is essential for the recruitment of p150^Glued at the acentrosomal spindle pole, as was observed at centrosomal pole (Hueschen et al, 2017; Appendix Fig S7G). Collectively, these results suggest that NuMA is important for placing the monopolar spindle in the center of acentrosomal cells, in prior to formation of bipolar spindles.

Figure 4. NuMA promotes the establishment of spindle bipolarity in acentrosomal human cells.

• A, B
  Structure of the spindle upon NuMA depletion. Centrinone‐treated HCT116 TetOsTIR1 NuMA–mAID–mClover–FLAG cells were treated with 1 μg/ml doxycycline (Dox) and 500 μM IAA. Two‐way arrow indicates the bipolarity. Gray, green, red, and blue represent α‐tubulin, NuMA (mClover), GT335, and DNA, respectively. Z‐projections of 10 sections, every 0.3 μm. Scale bar, 5 μm.
• C
  Frequency of spindle structure patterns upon NuMA depletion in (A, B). Values are mean percentages ± SD from three independent experiments (N = 30 spindles in each experiment).
• D
  The localization of Eg5 in centrinone‐treated acentrosomal HeLa cells. Red, gray, green, and blue represent NuMA, Eg5, GT335, and DNA, respectively. Scale bar, 5 μm.
• E
  The localization of Eg5 in the spindle upon NuMA depletion. Centrinone‐treated HCT116 TetOsTIR1 NuMA–mAID–mClover–FLAG cells were treated with Dox and 500 μM IAA. Gray, green, red, and blue represent Eg5, NuMA (mClover), CP110, and DNA, respectively. Z‐projections of five sections, every 0.3 μm. Scale bar, 5 μm.
• F
  Structure of the spindle upon Eg5 inhibition. Centrinone‐treated HCT116 TetOsTIR1 NuMA–mAID–mClover–FLAG cells were treated with 30 μM monastrol. Gray, green, red, and blue represent α‐tubulin, NuMA (mClover), GT335, and DNA, respectively. Z‐projections of five sections, every 0.3 μm. Scale bar, 5 μm.
• G
  Frequency of spindle structure patterns upon Eg5 inhibition in (F). Values are mean percentages ± SD from three independent experiments (N ≧ 23 spindles in each experiment).

The establishment of bipolarity in spindle assembly depends on the tetrameric kinesin motor protein Eg5 (Sawin et al, 1992; Heck et al, 1993; Gaglio et al, 1996; Walczak et al, 1998; Kapitein et al, 2005; Schuh & Ellenberg, 2007). To understand the system of NuMA‐dependent spindle bipolarization, we first observed the localization of Eg5 in acentrosomal spindles. In the initial process where NuMA structures accumulated in the center of the cell to form a monopolar‐like spindle, Eg5 localized around the NuMA structures and colocalized on the radial array of microtubules in both HeLa and HCT116 NuMA‐mAID–mClover–FLAG cells (Fig 4D and E). As the NuMA structure was stretched in a bundle‐like pattern and spindle bipolarization progressed, Eg5 appeared to localize on a small bipolar spindle (Fig 4D). We confirmed that monastrol, an Eg5 inhibitor, arrested acentrosomal cells with monopolar spindles in a dose‐dependent manner (Fig 4F and G). This finding indicates that Eg5 on microtubules around acentrosomal spindle poles is functionally required for spindle bipolarization. In NuMA‐depleted acentrosomal cells, Eg5 localized on the umbrella‐like monopolar spindle (Fig 4E), suggesting that NuMA is not required for Eg5 loading onto microtubules. Taken together, these observations imply that, in acentrosomal cells, the aggregation of NuMA gathers microtubule asters in one place and properly arranges Eg5 to organize the radial array of microtubules for the subsequent bipolar spindle formation.

Eg5 motor activity and stable kinetochore–microtubule attachment promote spindle pole separation in acentrosomal cells

In addition to Eg5, spindle pole separation depends on kinetochore–microtubule (KT–MT) attachment (Maiato et al, 2002; McAinsh et al, 2006; Toso et al, 2009; Mottier‐Pavie et al, 2011; Mchedlishvili et al, 2012; Gayek & Ohi, 2014). To confirm that these systems are also involved in spindle pole separation in acentrosomal human cells, we observed the localization and dynamics of NuMA upon inhibition of Eg5 or KT–MT attachment, using time‐lapse fluorescence microscopy.

Cells treated with monastrol, an Eg5 inhibitor, showed two types of defects in the generation of two spindle poles in acentrosomal cells (Fig 5A and B, Movies [Link], [Link], [Link]): Bipolarity within the acentrosomal spindle poles was only transiently established and subsequently the poles returned to a monopolar state (Pattern 1, Movie EV8), and bipolarity was never established within the poles (Pattern 2, Movie EV9). In contrast, bipolarity of the poles was rapidly established in control acentrosomal cells treated with dimethyl sulfoxide (DMSO; Fig 5A, Movie EV7). These observations indicate that Eg5 motor activity promotes both the initial establishment of bipolarity within the acentrosomal spindle poles, which is consistent with the results shown in Fig 4, and the subsequent separation of the poles.

Figure 5. Contribution of Eg5 and stable KT–MT attachment to the initial establishment of bipolarity and subsequent separation of acentrosomal spindle poles.

1. Time‐lapse observation of the establishment of bipolarity within the acentrosomal spindle pole upon inhibition of Eg5. HeLa cells expressing EGFP–centrin1 and mCherry–NuMA were observed in the presence of 50 μM of monastrol with a 63× objective. Magenta and green represent mCherry–NuMA and EGFP–centrin1, respectively. Arrowheads and two‐way arrows indicate the monopolar and bipolar states, respectively. Z‐projections of 20 sections, every 1.2 μm. Scale bar, 10 μm. Time zero corresponds to NEBD.
2. The time of bipolarity establishment and being stuck in the monopolar‐like state of acentrosomal spindle poles in (A). Each plot shows the cumulative percentage of each event at each time point (N = total 75 cells from two independent experiments).
3. Time‐lapse observation of the dynamics of mClover–NuMA and chromosomes. HeLa cells expressing mClover–NuMA were observed in the presence of SiR‐DNA with a 63× objective. Magenta and green represent SiR‐DNA and mClover–NuMA, respectively. Arrowheads indicate initial bipolarity in the acentrosomal spindle pole (8 min), dispersed chromosomes (328 min), and the monopolar‐like state of the acentrosomal spindle pole (332 min), respectively. Z‐projections of 20 sections, every 1.2 μm. Scale bar, 10 μm. Time zero corresponds to NEBD.
4. The time of bipolarity establishment and being stuck in the monopolar‐like state of acentrosomal spindle poles in (C). Each plot shows the cumulative percentage of each event at each time point (N = total 35 cells from three independent experiments).

KT–MT attachment has been shown to accelerate centrosome separation in prometaphase (Maiato et al, 2002; McAinsh et al, 2006; Toso et al, 2009; Mottier‐Pavie et al, 2011; Mchedlishvili et al, 2012; Gayek & Ohi, 2014). We attenuated KT–MT attachment via knockdown of NDC80, one of the essential kinetochore proteins involved in the KT–MT interface (Cheeseman et al, 2006), to verify whether KT–MT attachment is also involved in pole separation in the absence of centrosomes. Depletion of NDC80 disrupted spindle bipolarity in acentrosomal cells, but not in control centrosomal cells despite the abnormal spindle structure (Appendix Fig S8A–E). Upon depletion of NDC80, the acentrosomal cells initially exhibited bipolarity within the poles (Fig 5C and D, Movie EV10). However, pole separation did not proceed thereafter, and the poles shrank back to monopolar state (Fig 5C and D, Movie EV10).

Together, these results suggest that, in acentrosomal human cells, the motor activity of Eg5 plays important roles in both the initial establishment of bipolarity and the separation of the poles, while the requirement of stable KT–MT attachment is limited to the separation of the poles.

NuMA promotes the initial step of spindle bipolarization in cells with centrosomes

It has been shown that NuMA promotes centrosome separation after monastrol washout in prometaphase‐arrested mouse embryo fibroblasts (Silk et al, 2009). Therefore, to complement the revelation of the function of NuMA in acentrosomal bipolar spindle formation, we also characterized the requirement of NuMA in the initial step of spindle bipolarization in asynchronous human cells with centrosomes.

Most of the cells in which NuMA was depleted using the AID system established bipolar spindles (72.2%) (Fig 6A and B). However, interestingly, 18.8% of the cells showed umbrella‐like monopolar spindles (Pattern 2, Fig 6A and B), which were also observed in acentrosomal cells (Fig 4A–C). The population of NuMA‐depleted mitotic cells with umbrella‐like monopolar spindles was increased by the addition of monastrol, suggesting that both NuMA and Eg5 are important for the organization of mitotic spindle poles (Fig 6A and B). Upon NuMA depletion in cells with centrosomes, Eg5 localized to the umbrella‐like monopolar spindles (Fig 6C).

Figure 6. NuMA promotes the initial step of spindle bipolarization in cells with centrosomes.

1. Structure of the spindle upon NuMA and Eg5 inhibition. HCT116 TetOsTIR1 NuMA–mAID–mClover–FLAG cells with 2‐centrosomes were treated with 1 μg/ml doxycycline (Dox), 500 μM IAA, and monastrol. Green, red, and blue represent NuMA (mClover), α‐tubulin, and DNA, respectively. Z‐projections of 10 sections, every 0.3 μm. Scale bar, 5 μm.
2. Frequency of spindle structure patterns upon NuMA depletion and Eg5 inhibition in (A). Values are mean percentages ± SD from three independent experiments (N ≧ 22 spindles in each experiment).
3. The localization of Eg5 in the spindle upon NuMA depletion. HCT116 TetOsTIR1 NuMA–mAID–mClover–FLAG cells with 2‐centrosomes were treated with 1 μg/ml Dox and 500 μM IAA. Gray, green, red, and blue represent α‐tubulin, NuMA (mClover), Eg5, and DNA, respectively. Z‐projections of five sections, every 0.3 μm. Scale bar, 5 μm.
4. Time‐lapse observation of the establishment of spindle bipolarity upon NuMA depletion. HCT116 TetOsTIR1 NuMA–mAID–mClover–FLAG cells with 2‐centrosomes were treated with 1 μg/ml Dox and 500 μM IAA and observed with a 40× objective. Red and green represent SiR‐tubulin and NuMA, respectively. Z‐projections of 14 sections, every 2 μm. Scale bar, 10 μm. Time zero corresponds to NEBD.
5. The time required for the initial establishment of spindle bipolarity in (D). Line and error bars represent the mean and SD (N ≧ 60 cells from two independent experiments). The Mann–Whitney U‐test (two‐tailed) was used to obtain a P value.

We further analyzed the timing of the establishment of spindle bipolarization in NuMA‐depleted cells. In control cells, separation of two spindle poles was detectable within 13.4 min after NEBD, on average (Fig 6D and E, Movie EV11). Depletion of NuMA delayed the spindle pole separation to 40.8 min after NEBD, on average (Fig 6D and E, Movie EV12). Collectively, these results suggest that NuMA promotes the separation of the two spindle poles to establish spindle bipolarity in asynchronous human cells with centrosomes.

Discussion

Based on the findings of this study, we propose a model for the NuMA‐mediated establishment of spindle bipolarity in acentrosomal human cells (Fig 7). In this model, acentrosomal cells can establish spindle bipolarity in the following steps: (i) NuMA organizes microtubule asters; (ii) the clustering activity of NuMA and dynein promotes the assembly of those asters; (iii) NuMA, together with Eg5, organizes a radial array of microtubules to promote spindle elongation; and (iv) subsequently, the motor activity of Eg5 and KT–MT attachment further promote spindle pole separation. Eventually, chromosome congression occurs, and a short bipolar spindle is established.

Figure 7. A model for NuMA‐dependent establishment of spindle bipolarity independently of centrosomes.

Schematic illustration of acentrosomal spindle assembly in human cells. For details, see Discussion.

Acentrosomal human cells assemble spindle poles by accumulating NuMA in the center of the chromosome ring during early mitosis. Interestingly, the assembly of NuMA in acentrosomal spindle formation is presumably promoted by the coordination of dynein‐dependent microtubule minus‐end focusing and the clustering activity of NuMA. Under physiological conditions in human cells, the clustering activity of NuMA is important for the generation of spindle‐pulling forces at the cell cortex and regulation of spindle orientation, but not for spindle pole focusing (Okumura et al, 2018). In the presence of centrosomes, the minus‐end of spindle microtubules and K‐fibers are sorted into the spindle poles (Khodjakov et al, 2003; Maiato et al, 2004; Goshima et al, 2005), where centrosomes predominantly provide the site for microtubule minus‐end focusing. As previously described (Okumura et al, 2018), in this situation, the clustering of NuMA is not absolutely required for spindle pole focusing. This finding was also confirmed in the present study (Appendix Fig S6C and D). However, in the absence of centrosomes, the role of NuMA becomes more dominant as a platform for spindle pole organization by accumulating contractile motors and microtubules (Fig 3 and Appendix Fig S7H). Thus, NuMA may have a redundant function that complements the centrosome function in spindle pole organization in acentrosomal cells.

Our data showed that NuMA organizes a radial array of microtubules onto which Eg5 is loaded (Fig 4). These results suggest that NuMA assembles to accumulate the asters and Eg5 subsequently cross‐links the microtubules in an antiparallel fashion. This process may promote the spontaneous establishment of bipolarity within the acentrosomal spindle. Similar process is known to regulate centrosome separation in prometaphase (Kaseda et al, 2012). Hence, NuMA may play a central role in the initial establishment of bipolarity during acentrosomal spindle assembly in human cells.

Importantly, NuMA is also partially required for efficient spindle bipolarization in cells with centrosomes (Fig 6). However, the relationship between the centrosomal and NuMA‐mediated mechanisms remains unclear. It is possible that NuMA regulates the focusing of K‐fibers at spindle poles and promotes spindle pole separation, as suggested in previous studies (Khodjakov et al, 2003; Silk et al, 2009). Alternatively, even in the presence of centrosomes, NuMA and centrosomes may redundantly organize microtubules to activate the Eg5 motor for the initial bipolarization within the mitotic spindle, as observed in the acentrosomal situation in this study. Understanding this mechanism will be an interesting objective of future studies.

Several studies have suggested the compensation mechanisms in mitosis (Prosser & Pelletier, 2017). It has been shown that somatic cells can divide in the absence of centrosomes (Bonaccorsi et al, 2000; Khodjakov et al, 2000; Basto et al, 2006; Hornick et al, 2011; Sir et al, 2013). Overexpression of Kif15 could drive bipolar spindle formation in the absence of Eg5 activity (Tanenbaum et al, 2009; Vanneste et al, 2009). Furthermore, it is suggested that the machinery of centrosome separation at the onset of mitosis is coordinated by redundant mechanisms (Hata et al, 2019). The results of our study suggest that human cells possess the redundant mechanism of canonical centrosomal and NuMA‐mediated acentrosomal pathways to promote spindle bipolarity. The present findings further suggest that an additional redundant machinery may function in human cells in the absence of centrosomes and NuMA. Interestingly, as shown in Appendix Fig S7C and D, NuMA‐depleted acentrosomal cells manage to segregate chromosomes after prolonged mitosis in an unknown manner. Understanding these compensation mechanisms in mitosis is important for the development of anticancer drugs targeting the formation of the mitotic spindle (Jackson et al, 2007).

Oocytes do not have centrosomes; instead, they exhibit acentrosomal spindle formation (Bennabi et al, 2016). It has been proposed that meiotic spindle formation in meiosis I is archived by an inside‐out mode of assembly (Dumont et al, 2007; Schuh & Ellenberg, 2007; Breuer et al, 2010; Bennabi et al, 2018; Letort et al, 2019). In mouse, meiotic spindle formation is mediated by microtubule nucleation from chromatin and multiple MTOCs, which progressively establishes the spindle bipolarity (Schuh & Ellenberg, 2007; Ma & Viveiros, 2014; Baumann et al, 2017). Furthermore, in the mouse oocyte, proper spindle assembly requires the sorting and clustering of multiple small asters, which contain PCM components (Schuh & Ellenberg, 2007; Breuer et al, 2010; Manil‐Ségalen et al, 2018). These two properties of oocytes, spindle assembly with inside‐out mode and the clustering of asters, are similar to our observations in this study.

It has been suggested that centrosomes and Ran‐dependent microtubule nucleation pathways have been suggested to work together in somatic cells (Zheng, 2004; Kaláb et al, 2006). NuMA is a nuclear protein in somatic cells and is controlled by the Ran pathway (Nachury et al, 2001; Wiese et al, 2001). In addition to NuMA, other spindle assembly factors (SAFs) such as TPX2 and HURP are also controlled by the Ran pathway in somatic cells and oocytes (Forbes et al, 2015). Therefore, the central function of NuMA that was revealed in our study is probably controlled by the Ran pathway, and NuMA may organize spindle bipolarization cooperatively with other SAFs. In addition, immunostaining approaches have shown that NuMA forms a large aggregate in the processes of pre‐meiosis I and subsequently localizes at the spindle poles of meiosis I and II in acentrosomal meiotic spindle of human oocytes (Xu et al, 2011). Therefore, it is important to determine whether the mechanism identified in this study (i.e., NuMA‐dependent spindle bipolarization, rather than spindle pole maintenance), is also involved in the formation of the meiotic spindle in the oocytes of humans and other species.

Materials and Methods

Cell culture and transfection

HeLa, RPE‐1, and U2OS cells were obtained from the European Collection of Authenticated Cell Cultures (ECACC). These cell lines were authenticated by the ECACC via short tandem repeat (STR) profiling. HeLa cells expressing NuMA endogenously tagged with mCherry or mClover were generated using the CRISPR/Cas9 system, as previously described, with slight modifications (Natsume et al, 2016). GuideRNA oligos (NuMA_gRNA_F: CACCGAGATGACACTCCACGCCACC and NuMA_gRNA_R: AAACGGTGGCGTGGAGTGTCATCTC) were hybridized and cloned into the BbsI site of pX330 (Addgene). To construct the donor plasmid for homology‐directed repair, homology arms of the NuMA locus (chr11:72035555‐72036331) were amplified (pBS2_NuMA N‐ter_InsF: GGTATCGATAAGCTTTCAGCATCCAGACCCAGGTC and pBS2_NuMA N‐ter_InsR: CGCTCTAGAACTAGTCCGCCACCACAGCCAGCTAA) from the genomic DNA of HeLa cells and cloned into pBluescript using an Infusion Cloning Kit (Takara). The P2A‐mCherry/mClover cassette, containing a hygromycin‐resistant gene, was inserted immediately after the transcription start site in the middle of the homology arms. The plasmids were introduced into a HeLa cell line stably expressing EGFP–centrin1 (Tsuchiya et al, 2016), and selected using hygromycin. A549, DU145, GI‐1, MCF7, and PANC‐1 cells were obtained from the RIKEN BRC (RIKEN BioResource Research Center). These cell lines were authenticated by the RIKEN BRC via STR profiling. The SKOV‐3 cell line was provided by Dr. Yoko Nagumo. HCT116 TetOsTIR1 and HCT116 TetOsTIR1 DHC1–3X–mAID–mClover cells were provided by Dr. Masato Kanemaki and Dr. Toyoaki Natsume (Natsume et al, 2016). The HCT116 TetOsTIR1 NuMA–mAID–mClover–3X–FLAG, HCT116 TetOsTIR1 NuMA–mAID–mClover–3X–FLAG DHC1–SNAP mCherry–NuMA WT, and HCT116 TetOsTIR1 NuMA–mAID–mClover–3X–FLAG DHC1–SNAP mCherry–NuMA 5A‐3 were provided by Dr. Tomomi Kiyomitsu (Okumura et al, 2018). HCT116 CMVOsTIR1 HsSAS6–AID have been described previously (Yoshiba et al, 2019). These HCT116 cells were cultured in McCoy's 5A medium (Thermo Fisher Scientific) supplemented with 10% fetal bovine serum, 2 mM l‐glutamine, 100 U/ml penicillin, and 100 μg/ml streptomycin. The HeLa, U2OS, A549, and GI‐1 cells were cultured in Dulbecco's modified Eagle's medium, containing 10% fetal bovine serum, 100 U/ml penicillin, and 100 μg/ml streptomycin, at 37°C in a 5% CO₂ atmosphere. The DU145, PANC‐1, and SKOV‐3 cells were cultured in RPMI 1640 medium, containing 10% fetal bovine serum, 100 U/ml penicillin, and 100 μg/ml streptomycin, at 37°C in a 5% CO₂ atmosphere. MCF‐7 cells were cultured in MEM, containing 10% fetal bovine serum, 1 mM sodium pyruvate, 100 U/ml penicillin, and 100 μg/ml streptomycin, at 37°C in a 5% CO₂ atmosphere. The RPE‐1 cells were cultured in DMEM/F‐12 medium, containing 10% FBS, 100 U/ml penicillin, and 100 μg/ml streptomycin, at 37°C in a 5% CO₂ atmosphere. Transfection of siRNA constructs into HeLa cells was conducted using Lipofectamine RNAiMAX (Life Technologies).

RNA interference

The following siRNAs were used: Silencer Select siRNA (Life Technologies) against NDC80 #1 (s20350), NDC80 #2 (s20351), and negative control #1 (4390843).

Chemicals

The following chemicals were used in this study: centrinone (PLK4 inhibitor, MedChem Express, HY‐18682) and centrinone B (PLK4 inhibitor, donated by Drs. Shiau and Oegema), monastrol (Eg5 inhibitor, Toronto Research Chemicals, M505750), SiR‐Tubulin (microtubule probe, SPIROCHROME, CY‐SC002), and SiR‐DNA (DNA probe, SPIROCHROME, CY‐SC007).

Antibodies

The following primary antibodies were used in this study: rabbit polyclonal antibodies against ASPM (Novus Biologicals, NB100‐2278, immunofluorescence (IF) 1:250), CDK5RAP2 (Bethyl Laboratories, IHC00063, IF 1:500), Cep192 (Bethyl laboratories, A302–324A, IF 1:1,000), Cep152 (Bethyl laboratories, A302–480A, IF 1:1,000), CP110 (proteintech, 12780‐1‐AP, IF 1:1,000), GFP (MBL, 598, IF 1:1,000, Western blotting (WB) 1:2,500), α‐tubulin (MBL, PM054, IF1:300), NuMA (Cell Signaling Technology, 3888S, IF 1:1,000, abcam, ab109262, IF 1:500), and pericentrin (Abcam, ab4448, IF 1:1,000), RFP (MBL, PM005, IF 1:1,000, WB 1:2,500); mouse monoclonal antibodies against Eg5 (BD Biosciences, BD611186, IF 1:1,000), HSP90 (BD Biosciences, BD610419, WB 1:2,500), katanin p60 AL1 (A‐10; Santa Cruz Biotechnology, sc‐373814, 1:100), KIF4 (E‐8, Santa Cruz Biotechnology, sc‐365144, IF 1:100), NDC80 (C‐11; Santa Cruz Biotechnology, sc‐515550, 1:50), Polyglutamylation Modification (GT335, mAb; AdipoGen, AG‐20B‐0020‐C100, IF 1:1,000), PRC1 (C‐1, Santa Cruz Biotechnology, sc‐376983, IF 1:100), p150^Glued (BD Biosciences, BD610473, IF 1:1,000), and α‐tubulin (Sigma‐Aldrich, T5168, IF1:1,000, WB 1:5,000); and guinea pig polyclonal antibodies against CENP‐C (MBL, PD030, IF 1:500). Alexa Fluor 647‐labeled anti‐Eg5 was purchased from Abcam (ab202571, IF 1:100). FITC‐labeled anti‐GFP was purchased from Abcam (ab6662, IF: 1:250 or 1:500). Alexa 488‐labeled anti‐Cep152 antibody (Bethyl laboratories, A302–480A, IF 1:250) was generated using an Alexa Fluor Labeling Kit (Life Technologies). The following secondary antibodies were used: Alexa Fluor 488 goat anti‐mouse IgG (H+L; Molecular probes, A‐11001, 1:500), Alexa Fluor 555 goat anti‐rabbit IgG (H+L; Molecular probes, A‐21428, 1:500), Alexa Fluor 647 goat anti‐mouse IgG (Abcam, ab150115, 1:500), Alexa Fluor 647 goat anti‐guinea pig IgG (Abcam, ab150187, 1:500), and goat polyclonal antibodies horseradish peroxidase against mouse IgG (Promega, W402B, 1:10,000) and rabbit IgG (Promega, W401B, 1:10,000) for WB.

Sample preparation for immunostaining

All cells were treated with 100 nM of centrinone for 3–5 days to induce acentrosomal cells. For the Sas‐6 depletion experiments using the AID system (Appendix Fig S2E), the cells were incubated 50 μM of indole‐3‐acetic acid (IAA) for 3 days. To examine the effect of siNDC80 on acentrosomal cells (Appendix Fig S8C–E), after 2 days of treatment with 500 nM of centrinone B, 20 nM of siNDC80 was added for an additional 2 days. For the DHC1 depletion experiments using the AID system (Fig 3D and E, Appendix Fig S5D), after 4 days of treatment with 0.1% DMSO or 100 nM of centrinone, the cells were incubated with 1 μg/ml doxycycline (Dox) and 500 μM of IAA for an additional 24 h. In Fig 3F, after 2 days of treatment with 0.1% DMSO or 100 nM of centrinone, the cells were incubated with 1 μg/ml Dox. After 24 h of incubation, 100 μM of IAA was added for an additional 3 h. For the NuMA replacement experiments using the AID system (Fig 3G and H, Appendix Fig S6C and D), the cells were incubated with 100 nM of centrinone, 1 μg/ml Dox, and 500 μM of IAA for 3 days. For the NuMA depletion experiments using the AID system (Fig 4A–C and E, Appendix Fig S7A–D), after 3 days of treatment with 0.1% DMSO or 100 nM of centrinone, the cells were incubated with 1 μg/ml Dox and 500 μM of IAA for an additional 24 h, as previously described, with slight modifications (Okumura et al, 2018). For the NuMA depletion and Eg5 inhibition experiments (Fig 6A and B), after treatment with 1 μg/ml Dox and 500 μM of IAA for 24 h, the cells with centrosomes were incubated with monastrol for an additional 6 h. To perform the chemical perturbation experiments (Figs 3A, and 4F and G), after 3 days of treatment with 0.1% DMSO or 100 nM of centrinone, cells were treated with nocodazole or monastrol for an additional 6 h.

Western blotting

For preparation of total cell lysates, cells were washed with PBS, lysed in 1× SDS sample buffer. SDS–PAGE was performed using 6 or 10% polyacrylamide gels, followed by transfer on Immobilon‐P membrane (Millipore Corporation). The membrane was probed with the primary antibodies, washed with 0.02% Tween‐20 in PBS, and incubated with HRP‐conjugated secondary antibodies. The signals were detected with ECL Prime/Select reagents (GE Healthcare) or Chemi‐Lumi One (Nacalai Tesque) via the ChemiDoc XRSþ system (Bio‐Rad).

Sample preparation for live imaging

For live imaging, HeLa cells expressing mCherry–NuMA and EGFP–centrin1, and HeLa cells expressing mClover–NuMA, HCT116 TetOsTIR1 NuMA–mAID–mClover–3X–FLAG, HCT116 CMVOsTIR1 CEP152–mClover–mAID, or HCT116 Tet–OsTIR1 DHC1–3X–mAID–mClover were cultured in 35‐mm glass‐bottom dishes (Greiner‐bio‐one, #627870) or 24‐well SENSOPLATE (Greiner‐bio‐one, #662892) at 37°C in a 5% CO₂ atmosphere.

To explore the NuMA dynamics (Fig 2A–C, Movies EV3 and EV4), after 2 or 3 days of treatment with 100 nM of centrinone, HeLa cells expressing mCherry–NuMA and EGFP‐centrin1 or HeLa cells expressing mClover–NuMA were arrested in the S phase through treatment with 2 mM thymidine for 16–17 h. The thymidine was then removed by washing the cells with medium three times. Following 7–8 h of incubation, we performed a time‐lapse analysis. To analyze chromosome segregation error upon centrosome removal (Appendix Fig S4C–E), after 4 days of treatment with 100 nM of centrinone, HCT116 CMVOsTIR1 CEP152–mClover–mAID cells were observed for 24 h. To observe the structures of NuMA and microtubules (Fig 2E and F, Movies EV5 and EV6), after 4 days of treatment with 100 nM of centrinone, HCT116 TetOsTIR1 NuMA–mAID–mClover–3X–FLAG cells were observed for 12 h. We also used a time‐lapse analysis to observe the DHC1–mClover dynamics (Appendix Fig S5A and B), after 4 days of treatment with 0.1% DMSO or 100 nM of centrinone. To perform the NDC80‐knockdown experiment (Fig 5C and D, Movie EV10), after 3 days of treatment with 100 nM of centrinone, HeLa cells expressing mClover–NuMA were treated with 20 nM of siNDC80. The cells were then arrested in the S phase through treatment with 2 mM thymidine for 16 h. The thymidine was then removed by washing the cells with medium three times. After incubation for 8 h, we performed a time‐lapse analysis. To perform the chemical perturbation experiments (Fig 5A and B, Movies [Link], [Link], [Link]), after 3 days of treatment with 0.1% DMSO or 100 nM of centrinone, HeLa cells expressing mCherry–NuMA and EGFP–centrin1 were arrested in the S phase through treatment with 2 mM thymidine for 15–17 h. The thymidine was then removed by washing the cells with medium three times. After treatment with 50 μM monastrol for 8 h, we performed a time‐lapse analysis. To observe the initial establishment of spindle bipolarity in cells with centrosomes upon NuMA depletion (Fig 6D and E, Movies EV11 and EV12), after 24 h of treatment with 1 μg/ml Dox and 500 μM of IAA, HCT116 TetOsTIR1 NuMA–mAID–mClover–3X–FLAG cells were observed. Before imaging, the cells were incubated with 100 (HCT116) or 200 (HeLa) nM of SiR‐DNA or 100 nM of SiR‐tubulin for 2–3 h to visualize the DNA or microtubules, respectively.

Microscopy for immunofluorescence analyses

For the immunofluorescence analyses, cells cultured on coverslips [Matsunami: No 1] were fixed using −20°C methanol for 7 min and washed with phosphate‐buffered saline (PBS). After fixation, using PBS with 0.05% Triton X‐100 (PBSX) for 5 min, the cells were permeabilized. The cells were blocked in 1% bovine serum albumin in PBSX for 30 min at room temperature. The cells were then incubated with primary antibodies for 7–24 h at 4°C, washed with PBSX three times, and incubated with secondary antibodies and 0.2 μg/ml Hoechst 33258 (DOJINDO) for 45–60 min at room temperature. The cells were washed three times with PBSX and mounted onto glass slides.

We counted the number of spindle patterns using a DeltaVision Personal DV‐SoftWoRx system (Applied Precision) equipped with a CoolSNAP CH350 CCD camera, or an Axioplan2 fluorescence microscope (Carl Zeiss) equipped with a CoolSNAP CH350 CCD camera.

Confocal images were captured using a Leica TCS SP8 system equipped with a Leica HCX PL APO 63×/1.4 oil CS2 objective. The diameter of the pinhole was adjusted to one airy unit, and the scan speed was set to 200 Hz with 3× line averaging. The images were collected at 300‐nm intervals in Z with 1,024 × 500 pixels in the XY plane. For deconvolution, we used the Huygens Essential software (SVI; Scientific Volume Imaging).

Maximum‐intensity Z‐projections of representative images were generated using FIJI distribution in the ImageJ (NIH) software. Optical slice numbers and step sizes of Z‐stacks are described in the figure legends. We used the Imaris software (Bitplane) for three‐dimensional surface rendering (Fig 1F).

Microscopy for live imaging

A confocal scanner box, the Cell Voyager CV1000 (Yokogawa Electric Corp) equipped with a 63× oil‐immersion objective and a back‐illuminated EMCCD camera, or CQ1 Benchtop High‐Content Analysis System equipped with a 40× objective and a sCMOS camera (Yokogawa Electric Corp) were used for live‐cell imaging (Figs 2A, B, 5A, C, and 6D, Appendix Figs S4, S5A, B, and S7C). Maximum‐intensity Z‐projections of representative images were generated using FIJI distribution in the ImageJ (NIH) software. Optical slice numbers and step sizes of Z‐stacks are described in the figure legends.

We observed the structures of NuMA and microtubules (Fig 2E and F) using spinning disk confocal microscope optics, IXplore Spin, equipped with a 60× oil‐immersion objective (NA1.5) and a stage incubator for a 35‐mm dish. The images were captured using an sCMOS camera (ORCA Flash4.0 V3).

FRAP experiments

FRAP experiments were performed using HCT116 TetOsTIR1 NuMA–mAID–mClover–3X–FLAG cells after 4 days of treatment with DMSO or 100 nM centrinone. The confocal system involved a Leica TCS SP8 equipped with a Leica HCX PL APO 63×/1.4 oil CS2 objective. A stage‐top incubator (Tokai Hit) was used to maintain the cells at 37°C and 5% CO₂. Three sequential images were captured, and then, the fluorescence of NuMA‐mClover at a centrosomal pole or the acentrosomal spindle pole was photobleached with the lowest laser power required for photobleaching. Fluorescence images were acquired every 5 s over a 3‐min period. Fluorescence intensities were measured using the ImageJ software (NIH). A square region of interest (8 × 8 pixels) covering the entire centrosomal spindle pole or acentrosomal spindle pole was used to measure the fluorescence intensities. After background subtraction, the measured fluorescence intensities were normalized against the averaged intensities of the three images captured before photobleaching. Recovery data were fitted with a single exponential curve to determine the mobile fraction and half‐recovery time, using our original code written in Mathematica (Wolfram).

Statistical analysis

Statistical analyses were performed using the GraphPad Prism 7 software.

Author contributions

TC and DK designed the study; TC, SY, and DK designed experiments; TC, YT, KW, KK, KH, and DT performed experiments; TC, YT, KK, and DT analyzed data; TC and DK wrote the manuscript. All authors reviewed and approved the manuscript.

Conflict of interest

The authors declare that they have no conflict of interest.

Supporting information

Appendix

Movie EV1

Movie EV2

Movie EV3

Movie EV4

Movie EV5

Movie EV6

Movie EV7

Movie EV8

Movie EV9

Movie EV10

Movie EV11

Movie EV12

Review Process File

The EMBO Journal (2020) 39: e102378