The mitotic protein NuMA plays a spindle-independent role in nuclear formation and mechanics

Most eukaryotic cells enclose their genome within a single, round nucleus. When chromosomes are not properly incorporated into the nucleus, they can form micronuclei, which can be vulnerable to DNA damage and chromothripsis, leading to genomic instability and disease (Andreassi et al., 2011; Crasta et al., 2012; Hatch et al., 2013; Ly et al., 2019; Zhang et al., 2015). Furthermore, while different cell types have nuclei of different shapes, within a given cell type, nuclear shape is regulated and stereotypical, and nuclear shape defects can correlate with altered gene expression, disease, and aging (Le Berre et al., 2012; Dahl et al., 2008; Skinner and Johnson, 2017; Thomas et al., 2002). While some of the mechanisms required for nuclear formation are known, how mammalian cells pack many chromosomes into a single nucleus with a simple rounded shape is not well understood. More broadly, understanding how nanometer-scale molecules give rise to higher-order micrometer-scale structures of a given number and shape remains a frontier.

Several mechanisms are known to promote the formation of a single nucleus and to give rise to its shape. For example, bringing anaphase chromosomes into a tight cluster at mitosis promotes the formation of a single nucleus. Indeed, erroneous chromosome-to-spindle attachments (Cimini et al., 2004; Liu et al., 2018; Liu and Pellman, 2020; Thompson and Compton, 2011), misregulated centrosome amplification (Ganem et al., 2009), and defective metaphase chromosome alignment and anaphase compaction (Fonseca et al., 2019) can all lead to micronucleation. At late mitosis, the protein Barrier-to-autointegration factor (BAF) builds a dense chromatin layer around the anaphase chromosome mass, limiting membrane penetration to prevent the formation of multiple nuclei (Samwer et al., 2017). Similarly, diverse factors are thought to influence nuclear shape. These range from the shape of the anaphase chromosome mass (Fonseca et al., 2019; Mora-Bermúdez et al., 2007) and chromatin organization and compaction (Furusawa et al., 2015; Stephens et al., 2017) to the mechanics of the nuclear lamina, cytoskeleton, and extracellular environment (Buxboim et al., 2010; Chu et al., 2017; Dahl et al., 2008; Lammerding et al., 2005).

Despite what we know about nuclear formation, understanding when, at mitosis versus at nuclear formation, and how directly certain molecules play a role in this process remains challenging. The protein nuclear mitotic apparatus (NuMA; Lydersen and Pettijohn, 1980), for example, is essential to spindle assembly (Compton and Luo, 1995; Kallajoki et al., 1991; Yang and Snyder, 1992) and chromosome segregation (Silk et al., 2009; Haren et al., 2009), as well as to the formation of a single nucleus (Compton and Cleveland, 1993; Kallajoki et al., 1991, 1993). NuMA contains two globular domains, its N and C termini, connected through seven coiled-coil domains (Compton et al., 1992; Yang et al., 1992). At mitosis, NuMA localizes to the spindle and cell cortex, where it drives spindle structural stability and orientation, respectively (Compton et al., 1992; Kallajoki et al., 1991; Kiyomitsu and Cheeseman, 2012; Kotak et al., 2013). At interphase, NuMA localizes to the nucleus. It has been proposed to have a postmitotic role in nuclear formation (Compton and Cleveland, 1993; Kallajoki et al., 1993; Van Ness and Pettijohn, 1983), which could be a direct or indirect role (Cleveland, 1995; Kallajoki et al., 1991, 1993; Merdes and Cleveland, 1998), and to promote DNA repair and chromatin organization in the nucleus (Abad et al., 2007; Kivinen et al., 2010; Vidi et al., 2014). Furthermore, NuMA has been reported to self-assemble into a filamentous network in the nucleus (Gueth-Hallonet et al., 1998; Zeng et al., 1994), and to directly bind DNA in vitro (Ludérus et al., 1994). However, NuMA’s essential function at mitosis has made it difficult to determine whether it also acts at nuclear formation independently of its spindle function. If it does, we do not know what this nuclear formation function is, or how NuMA’s structure supports it and regulates it over the cell cycle.

Here, we show that NuMA is essential for the formation of a single, round nucleus in human cells independent of its spindle role. We find that in the absence of NuMA, nuclear defects appear at the moment of nuclear envelope reassembly, even without prior spindle defects. We also show that in cells lacking NuMA, the chromosome mass expands faster at mitotic exit, and that nuclei are mechanically compromised. We demonstrate that NuMA’s coiled-coil is essential to forming a single nucleus and to promoting the formation of stable nuclear NuMA assemblies. In addition, we find that NuMA’s coiled-coil prevents the premature binding of NuMA’s C terminus, which we show directly binds DNA in vitro and to chromosomes at mitosis. Thus, NuMA’s coiled-coil serves as a central element that controls both its localization and its function over the cell cycle. Together, these findings establish a spindle-independent role for NuMA in nuclear formation and mechanics, and suggest that NuMA’s long coiled-coil helps it organize chromosomes at interphase. More broadly, they highlight NuMA’s key structural role in building and maintaining two of the largest cellular structures throughout the cell cycle, the spindle and nucleus.

To investigate whether NuMA contributes to the formation of the mammalian nucleus independently of its spindle function, we used inducible CRISPR/Cas9 human retinal pigment epithelial (RPE1) cells to knock out (KO) NuMA (Hueschen et al., 2017), and looked at nuclear formation after a spindle-less mitosis. To do so, we treated cells with nocodazole to depolymerize spindle microtubules, and with reversine to bypass the spindle assembly checkpoint. With this approach, mitotic cells do not assemble a spindle, exit mitosis and form a single nucleus of unsegregated chromosomes (Samwer et al., 2017). We induced NuMA KO cells at time (t) = 0 with doxycycline (DOX) to deplete endogenous NuMA (Fig. S1 A), treated them with nocodazole and reversine after 72 h, and fixed and imaged them 18–24 h after drug addition (termed “96 h” onward for simplicity; Fig. 1 A). As expected, control (uninduced) drug-treated cells did not assemble a spindle (Fig. S1 B), and yet still formed a single, round (termed “normal”) nucleus after mitosis (Fig. 1, B and C; Samwer et al., 2017), as did Cas9-induced cells without a single guide RNA (sgRNA) for NuMA (Fig. S1 C).

In turn, NuMA KO cells with a defective spindle (untreated with drugs; Fig. S1 B) presented nuclei with several fragments (Fig. 1, B and C), as previously observed (Compton and Cleveland, 1993; Kallajoki et al., 1991, 1993). Additionally, these cells also presented micronuclei (one to three small nuclei in addition to the primary nucleus) and atypically shaped nuclei (Fig. 1, B and C). In spindle-less NuMA-KO cells (i.e., treated with nocodazole and reversine drugs), the fraction of cells with fragmented nuclei was significantly reduced, while the fraction of cells with micronuclei and atypically shaped nuclei remained similar (Fig. 1, B and C). This suggests that fragmented nuclei result at least partly from spindle-related failures, and that micronucleation and atypically shaped nuclei result from spindle-independent failures. To quantitatively compare nuclear roundness, we measured the solidity of primary nuclei. Solidity significantly decreased in spindle-less NuMA KO cells compared with control cells (Fig. 1, D and E), and decreased as NuMA depletion extent increased (Fig. S1, D and E). Finally, because NuMA depletion occurred gradually over 4 d (Fig. S1 A), we sought to determine whether nuclear defects appeared when spindles were present (before drug addition) or absent (after drug addition). To do so, we repeated the same assay (Fig. 1 A) and compared cells fixed at 72 h, just before drug addition, and at 96 h, after drug treatment. The fraction of NuMA KO cells with fragmented nuclei, micronuclei, and atypically shaped nuclei all increased significantly in the 18–24 h spindle-less period (Fig. 1 F). Together, these findings indicate that NuMA plays a role in the formation of a single and round nucleus, independent of its spindle function.

To gain insight into NuMA’s nuclear function, we sought to determine whether it is required before, after, or at the time of nuclear formation. To do so, we performed live cell imaging of inducible NuMA KO RPE1 cells treated with nocodazole and reversine and stably expressing the histone marker mCherry-H2B and the nuclear envelope marker EGFP-Lap2β. As for fixed cells (Fig. 1), live imaging revealed a significant increase in the fraction of spindle-less NuMA KO cells with fragmented nuclei, micronuclei, and atypically shaped nuclei at 96 h compared with control cells (Fig. 2, A and B). To define when NuMA is needed, we followed the trajectories of individual cells that entered mitosis with a single, round nucleus (“mother cells”) until they exited spindle-less mitosis, formed a nucleus, and beyond (“daughter cells”; Fig. 2 C). From mother cells with normal nuclei, NuMA KO daughter cells exited mitosis with more fragmented nuclei, micronuclei, and atypically shaped nuclei than control cells (Fig. 2, D and E; and Videos 1 and 2). In these cell trajectories, we observed morphology defects appear as chromosomes decondensed and the nuclear envelope formed. We did not observe defects appear before or soon after nuclear formation. To test whether NuMA could help keep the decondensing chromosome mass together (either directly or indirectly) as the nuclear envelope formed, we measured the expansion of this mass with and without NuMA. We found that the chromosome mass expanded faster, and further, at nuclear formation in NUMA KO cells than in control cells (Fig. 2 F). Together, these findings indicate that NuMA’s spindle-independent role in building the nucleus takes place at the time of nuclear envelope formation, rather than before or after, and that NuMA keeps the chromosome mass compact during this time.

Electron microscopy and overexpression of NuMA coiled-coil truncations suggest that the coiled-coil helps form nuclear NuMA filaments, which could in principle contribute to nuclear stability (Gueth-Hallonet et al., 1998; Harborth et al., 1999; Zeng et al., 1994). To determine if NuMA’s coiled-coil is required for nuclear formation, we tested the ability of a truncation of NuMA that lacks most of the coiled-coil (NuMA-Bonsai; Fig. 3 A) to rescue the KO of endogenous NuMA, when expressed at near endogenous levels (Fig. 3, B and C; and Fig. S2 A). As a control, overexpression of similar levels of NuMA–full-length (FL)–EGFP (Fig. S2 A) in NuMA KO cells mostly rescued the nuclear defects observed in the absence of endogenous NuMA, even though some cells still presented micronuclei (Fig. 3, B and C). In turn, NuMA-Bonsai-EGFP, which contains only a short portion of the coiled-coil, failed to significantly rescue micronucleation (Fig. 3, B and C). This is consistent with NuMA-Bonsai being sufficient for NuMA’s role in the spindle body (Forth et al., 2014; Hueschen et al., 2017) but not at nuclear formation, with nuclear fragmentation coming largely from spindle defects, and micronucleation coming from spindle-independent defects, being revealed without spindle defects (Fig. 1 C). Consistently, inducible NuMA KO cells stably expressing NuMA-Bonsai-EGFP (at higher levels than endogenous NuMA; Fig. S2 B) had micronuclei (Fig. S2 C) but no detectable chromosome segregation errors (Fig. S2 D). Importantly, this indicates that NuMA’s nuclear formation role is essential not only after a drug-induced spindle-less mitosis but also after normal mitosis. Together, these findings indicate that either all or part of NuMA’s coiled-coil is necessary for formation of a single nucleus.

Given the role of NuMA’s coiled-coil in nuclear formation (Fig. 3, B and C; and Fig. S2 C), and that NuMA can assemble into higher-order filament-like structures (Zeng et al., 1994) whose size depends on the coiled-coil’s length (Gueth-Hallonet et al., 1998; Harborth et al., 1999), we hypothesized that the coiled-coil contributes to NuMA’s self-assembly in the nucleus. To test this, we measured the turnover of NuMA-FL-GFP, NuMA-Bonsai-GFP, and NuMA-NC-GFP, a construct that has only N and C termini and lacks the entire coiled-coil, in the nucleus using FRAP (Fig. 3, D–F). NuMA-FL-GFP and NuMA-Bonsai-GFP displayed similar recovery kinetics, while NuMA-NC-GFP recovered faster. Using a double exponential model, the fast half-life was shorter for NuMA-NC-GFP than NuMA-FL-GFP and NuMA-Bonsai-GFP, while the slow half-life was indistinguishable for all three constructs, independent of GFP-tagged protein expression levels within the analyzed range (Fig. S2, E and F). Furthermore, we observed that cells expressing the highest NuMA-FL-EGFP levels, excluded from the above analysis, can form stable cable-like nuclear structures that show much slower recovery after photobleaching (Fig. 3 G and Video 3). This supports the notion that NuMA can self-assemble in the nucleus when highly overexpressed, though it may or may not do so when endogenously expressed. In turn, we never observed such cable-like structures in NuMA-NC-EGFP cells, even at high expression levels. Together, our findings indicate that NuMA’s coiled-coil is necessary for nuclear formation, and suggest that it modulates NuMA’s mobility in the nucleus, by modulating its interaction either with itself or with other players.

NuMA is only detected on chromosomes after initiation of nuclear envelope reassembly (Fig. S3), when it is imported into the nucleus (Compton et al., 1992). What mechanism ensures that NuMA only associates with chromosomes in interphase, when chromosomes are kept together, and not in mitosis, when chromosomes must move apart? To begin answering this question, we asked how NuMA binds chromosomes in interphase, testing whether NuMA’s globular N- and C-terminal domains associate with chromosomes once in the nucleus. NuMA’s N and C termini were suggested to bind DNA (Radulescu and Cleveland, 2010), though no specific and direct comparisons were made. Using a light-controlled dissociation system, we sought to disconnect NuMA’s N and C termini from each other once they were both imported together into the nucleus. We engineered an optocontrollable NuMA-Bonsai protein (Opto-NuMA) using the LOV2 trap and release of protein (LOVTRAP) system (van Haren et al., 2018; Wang et al., 2016; Fig. 4 A). In the dark, when LOV2 and Zdk1 bind, NuMA-N(1–705)-PhusionRed-LOV2 and EGFP-Zdk1-NuMA-C localized to the nucleus, as expected since NuMA’s C terminus contains a NLS (Tang et al., 1994). Upon blue light illumination, the nuclear localization of NuMA-C (imaged with the same wavelength that activates LOV2 and induces dissociation) remained unchanged. Furthermore, NuMA-C’s nuclear localization pattern resembles that of NuMA-N before and after blue light illumination, but not during illumination (Fig. S4 A), suggesting that NuMA-C drives NuMA-N localization. Meanwhile, NuMA-N became diffuse (Fig. 4, B and C; and Video 4), likely remaining in the nucleus due to its high molecular weight (123 kD). Thus, whether chromatin binding is direct or indirect, NuMA's C terminus has a higher affinity for chromatin than its N terminus.

Previous work suggested that FL NuMA can directly bind DNA (Ludérus et al., 1994), but whether binding occurs through the C or N terminus or both, and the relative affinity of each for DNA, are not known. To address these questions and understand how NuMA binds DNA, we purified different SNAP-tagged NuMA truncation constructs (Fig. S4 B) and tested their ability to bind a DNA fragment in vitro using an electrophoretic mobility shift assay (EMSA) under stringent binding conditions (Fig. 4 D). We observed that the construct that lacks the coiled-coil (NuMA-NC) bound DNA with the highest affinity, followed by NuMA-Bonsai (NuMA-NC with an additional 493–amino acid–long coiled-coil) and finally NuMA-C with the lowest detectable affinity. NuMA-N did not have detectable DNA affinity in our conditions, and negative control tubulin had no detectable affinity. Thus, NuMA directly binds DNA primarily through its C terminus, and this binding could in principle underlie NuMA-C’s chromatin binding in vivo (Fig. 4, A–C). Furthermore, these in vitro observations suggest that NuMA’s DNA binding may be intramolecularly regulated, as NuMA’s N terminus increases its DNA binding, and the coiled-coil appears to decrease it.

Given that NuMA-C binds chromatin in interphase (Fig. 4, A–C) and DNA in vitro (Fig. 4 D), we asked what prevents NuMA-C from binding chromosomes in mitosis. To answer this question, we mapped the localization of NuMA-FL, NuMA-Bonsai, NuMA-NC, and NuMA-C through mitosis (Fig. 4 E and Video 5), testing whether other parts of NuMA regulate NuMA-C’s chromatin binding, as suggested in vitro (Fig. 4 D). Similar to endogenous NuMA (Fig. S3) and NuMA-FL, NuMA-Bonsai did not localize to chromosomes in mitosis. In contrast, NuMA-NC and NuMA-C were recruited to mitotic chromosomes as early as anaphase onset. Therefore, NuMA’s coiled-coil actively regulates NuMA’s localization during mitosis by preventing the premature binding of NuMA-C to chromosomes. NuMA is heavily phosphorylated by the mitotic kinase CDK1 at mitotic entry (Compton and Luo, 1995; Kotak et al., 2013; Yang et al., 1992). CDK1 activity negatively controls NuMA levels at the cell cortex during metaphase, and NuMA’s dephosphorylation at anaphase promotes the protein’s translocation to the cortex (Kiyomitsu and Cheeseman, 2013; Kotak et al., 2013). Premature RO3366-mediated CDK1 inhibition at metaphase did not change NuMA-Bonsai’s localization, while NuMA-C and NuMA-NC still bound to chromosomes (Fig. S5). This indicates that the coiled-coil regulates NuMA localization during mitosis, apparently doing so independently of CDK1 activity, and may depend on other regulatory processes. Together, these findings indicate that the coiled-coil of NuMA acts as a central element that drives not only the protein’s nuclear specific function and the dynamics of its assemblies (Fig. 3), but also when NuMA is allowed to bind chromosomes and which structures it acts on over the cell cycle.

So far, we have shown that NuMA is essential for forming a single, round nucleus (Figs. 1 and 2), keeps the chromosome mass compact at mitotic exit (Fig. 2 F), can form higher-order nuclear assemblies (Fig. 3 G; Zeng et al., 1994), and binds chromatin in a cell cycle–regulated manner (Fig. 4). Together, these findings indicate that NuMA plays a structural role in forming the nucleus, and suggest how it may do so. This led us to hypothesize that NuMA also plays a structural role in maintaining the nucleus. To test for such a role, we compared nuclear morphologies and then mechanically perturbed the nuclei of cells stably expressing EGFP-Lap2β and mCherry-H2B with and without NuMA. If nuclei were more deformable without NuMA, this would indicate that NuMA plays a structural role in nuclear stability. Imaging in 3D, and at higher resolution than above experiments, revealed invaginations in the nuclear surface of NuMA KO cells that were undetectable in control cells (Fig. 5 A). While reduced cell spreading can cause a reduced nuclear area (Buxboim et al., 2017) and nuclear wrinkling (Li et al., 2015), this is unlikely the reason NuMA KO nuclei have wrinkles as NuMA KO cells spread more (Fig. 5 B) and their nuclear area is larger (Fig. 5 C) than control cells. The presence of wrinkles suggested that nuclei are more deformable in NuMA KO cells. Furthermore, nuclei had a larger volume in NuMA KO than control cells (Fig. 5 D), suggesting that NuMA may keep chromosomes together not only at nuclear formation (Fig. 2 F) but also during nuclear maintenance. To compare how nuclear shape responds to mechanical force with and without NuMA, we confined and thereby compressed cells using a polydimethylsiloxane (PDMS) device (Guild et al., 2017; Le Berre et al., 2012) with 3-µm-high pillars (Fig. 5 E), and measured resulting changes in nuclear morphology. Under confinement, the nuclei of NuMA KO cells lost their nuclear invaginations as these “unfolded” into a smooth nuclear surface similar to control cells (Fig. 5 F). In addition, nuclei without NuMA were more deformable under external mechanical force than control nuclei: while nuclei were thicker (i.e., higher in the confinement axis) before confinement in NuMA KO cells, and had a larger volume in NuMA KO cells (Fig. 5 C), they became thinner than in control cells under confinement (Fig. 5, G–I). Thus, NuMA either directly or indirectly provides mechanical stability to the nucleus, which could help maintain nuclear structure and function. Altogether, our work indicates that NuMA plays a spindle-independent, structural role in both nuclear formation and mechanics in human epithelial cells.

Although many of the molecules required for nuclear formation are known, we still do not understand how they together give rise to a single, round, and robust mammalian nucleus. Because mitotic errors can lead to abnormal nuclei, one challenge is to define when, at mitosis or nuclear formation, a given molecule acts, and what roles it plays at different times. A long-standing example has been NuMA, an ∼200-nm-long coiled-coil protein (Harborth et al., 1995, 1999; Yang et al., 1992) essential for both mitosis (Compton and Luo, 1995; Kallajoki et al., 1991; Yang and Snyder, 1992) and nuclear integrity (Compton and Cleveland, 1993; Kallajoki et al., 1991, 1993). Here, we uncouple NuMA’s mitotic and nuclear functions. We show that NuMA has a spindle-independent role in building a single, round nucleus (Fig. 1), keeping the decondensing chromosome mass compact at nuclear formation (Fig. 2) and promoting the nucleus’ mechanical robustness (Fig. 5). Thus, NuMA plays a structural role in both building and maintaining the nucleus, as it does in the spindle (Compton and Luo, 1995; Hueschen et al., 2019; Silk et al., 2009; Yang and Snyder, 1992). We find that NuMA’s coiled-coil acts as a central element: it is required for nuclear function, and regulates NuMA’s mobility in the nucleus (Fig. 3) and its binding to chromosomes throughout the cell cycle (Fig. 4). As such, the coiled-coil compartmentalizes NuMA’s function to where and when it is needed: on chromosomes at interphase (when these are kept together) and off chromosomes at mitosis (when these are pulled apart). We propose a model whereby NuMA drives cellular organization throughout the cell cycle, building and maintaining the nucleus at interphase and the spindle at mitosis (Fig. 6).

Our current understanding of NuMA’s spindle function and its underlying basis much surpasses that of its nuclear function. Here, our work raises the question of how NuMA’s structure can give rise to its function in nuclear formation (Figs. 1 and 2) and mechanics (Fig. 5). While most of NuMA’s coiled-coil is dispensable for building the spindle (Forth et al., 2014; Hueschen et al., 2017), it is essential for building a single, round nucleus (Fig. 3, B and C; and Fig. S2 C). Additionally, the coiled-coil can stabilize nuclear NuMA (Fig. 3, D–F), and nonphysiological overexpression of FL NuMA, but not coiled-coil truncations, induces the formation of stable filament-like nuclear structures (Fig. 3 G). Furthermore, we know that NuMA can, when ectopically expressed in the cytoplasm, assemble into an organized network (Saredi et al., 1996) and, when overexpressed in the nucleus, assemble into a 3D network whose lattice size is determined by its coiled-coil’s length (Gueth-Hallonet et al., 1998; Harborth et al., 1999). We also know that NuMA self-assembles at the cell cortex to position the spindle using its coiled-coil and C terminus (Okumura et al., 2018). Together, these observations are consistent with a model where NuMA assembles into a network that shapes nuclear morphology and mechanics (Figs. 1, 2, and 5), with its coiled-coiled key to function and possibly assembly.

NuMA’s mobility in the nucleus (Fig. 3, D–G) suggests that it assembles into a dynamic structure. We do not know whether and how NuMA’s mobility on chromosomes changes over nuclear formation and maintenance, and how NuMA performs its functions given its mobility. One possibility is that NuMA’s dynamic structure enables its diverse nuclear functions. Indeed, NuMA regulates the localization and mobility of proteins involved in DNA repair (Salvador Moreno et al., 2019; Vidi et al., 2014) and in the nucleolar stress response (Jayaraman et al., 2017). Since NuMA-C binds DNA and chromosomes (Fig. 4) and altered NuMA expression impacts higher chromatin organization (Abad et al., 2007), such a network could in principle help cross-link DNA within and between chromosomes. Alternatively, NuMA may also establish connections with other nuclear networks that determine higher-order nuclear organization, such as RNA and ribonucleoproteins, as previously suggested (Barboro et al., 2002, 2003). Independently of the mechanism by which NuMA acts in the nucleus, its long structure and its self-assembling and DNA-binding properties make it well-suited to bridge large distances and impact higher-order nuclear organization. Determining to what extent nuclear NuMA self-assembles, and the structure of this assembly, will be essential to understanding how NuMA’s structure gives rise to its dynamics (Fig. 3, D–F) and architectural (Figs. 1, 2, and 5) and regulatory functions in the nucleus. Uncovering NuMA’s nuclear functions may in turn help understand why its localization varies depending on cell phenotype (Knowles et al., 2006; Vega et al., 2017). Similarly, it will shed light on why NuMA’s expression level and impact on nuclear architecture vary dramatically between epithelial cells and highly specialized or terminally differentiated cells (Merdes and Cleveland, 1998).

While NuMA-C binds DNA in vitro (Fig. 4 D) and appears to bind chromosomes in the nucleus (Fig. 4, A and B), FL NuMA does not bind chromosomes at mitosis. As such, its binding must be regulated. Here, we show that a small portion of NuMA’s coiled-coil is sufficient to prevent NuMA-C from binding chromosomes at mitosis (Fig. 4 E), and sufficient to decrease NuMA’s DNA binding affinity in vitro (Fig. 4 D). Preventing NuMA from binding chromosomes at mitosis could prevent it from keeping chromosomes together when they must segregate or, since NuMA recruits dynein (Hueschen et al., 2017), from recruiting dynein-based forces there. However, we find that prematurely localizing NuMA-NC on chromosomes at mitosis does not lead to detectable mitotic defects (Fig. 4 E). This could be because NuMA-NC lacks the coiled-coil needed for NuMA to keep chromosomes together (Fig. 3, B and C; and Fig. S2 C) and that is likely necessary to activate dynein motility (Hueschen et al., 2017; Reck-Peterson et al., 2018), or because its function on chromosomes is cell cycle–regulated. We propose NuMA’s coiled-coil plays a central role throughout the cell cycle, acting as a central structural (Fig. 3, B and C; and Fig. S2 C) and regulatory (Fig. 3, D–F; and Fig. 4 E) element for NuMA’s function on chromosomes. How the cell regulates this hub, such that NuMA binds chromosomes at mitotic exit and dissociates at mitotic entry, remains an open question. Answering it will require defining how NuMA’s C terminus binds DNA and chromosomes and will be key to understanding how the same protein performs such different physical roles in interphase and mitosis. Notably, NuMA’s C terminus also targets NuMA in the spindle, specifically to microtubule minus-ends (Hueschen et al., 2017).

What role NuMA plays in nuclear formation (Figs. 1 and 2) and mechanics (Fig. 5) is unclear. One model is that NuMA directly contributes to the organization and mechanics of the chromosome mass at mitotic exit (Fig. 6, model A). Indeed, NuMA-C binds DNA in vitro (Fig. 4 D) and chromosomes in vivo (Fig. 4, A–C), NuMA keeps the mass of decondensing chromosomes compact at mitotic exit (Fig. 2 F), and it localizes throughout the entire nucleus (Compton et al., 1992; Yang et al., 1992; Fig. S3), where it can interact with chromosomes. RNA-dependent NuMA localization and tethering to lamins has been suggested (Barboro et al., 2002), and NuMA influences higher-order chromatin organization (Abad et al., 2007; Kivinen et al., 2010) and binds chromatin remodelers (Vidi et al., 2014). Furthermore, we know that chromatin compaction impacts the shape and robustness of the nucleus (Stephens et al., 2019). If NuMA cross-linked DNA and chromosomes, or established connections with other nuclear structures, its import into the nucleus as it is forming and chromatin starts decondensing (Schooley et al., 2012) could keep the chromosome mass more compact at nuclear formation (Fig. 2 F) and maintenance (Fig. 5 D), and make it more rigid (Fig. 5). This could make it harder for the assembling nuclear envelope to penetrate into the chromosome mass, thus decreasing multinucleation and abnormally shaped nuclei (Figs. 1 and 2), and provide mechanical robustness to the nucleus (Fig. 5). In this model, NuMA plays a nuclear formation role with similarities to BAF’s, with BAF localizing at the chromosome mass’s rim (Samwer et al., 2017) and NuMA through the entire volume of the nucleus. Finally, we note that NuMA could also organize chromosomes at mitotic exit by keeping them in a tight focus around spindle poles (Cleveland, 1995; Compton and Cleveland, 1993), given NuMA’s pole-focusing role. However, such a role cannot explain our observations that NuMA plays a role in nuclear formation after a spindle-less mitosis (Figs. 1 and 2).

Alternatively, and not mutually exclusively, NuMA’s binding to chromosomes during nuclear formation could promote assembly and maturation of the nuclear envelope and its components (LaJoie and Ullman, 2017; Ungricht and Kutay, 2017) or mediate the tethering of chromatin to the nuclear envelope (Schreiner et al., 2015; Fig. 6, model B). These functions, when compromised, could lead to defects in nuclear morphology (Figs. 1 and 2) and mechanics (Fig. 5; Davidson and Lammerding, 2014; Webster et al., 2009). Incidentally, NuMA-C–coated beads mixed with Xenopus extracts recruit membranes, suggesting that NuMA could contribute to nuclear envelope assembly (Lu et al., 2012). Though NuMA is not enriched at the nuclear rim (Figs. 1 A and S3), we cannot exclude that it regulates the nuclear envelope and as such indirectly impacts chromosome organization. While in the two models we discuss that NuMA plays a direct role in nuclear formation and mechanics (Fig. 6), we cannot exclude the possibility that it plays an indirect role. For example, as a nuclear protein, it could in principle change the expression of genes that contribute to nuclear formation or to nuclear deformation under force, for example by changing the mechanics of nonnuclear elements such as the cytoplasm or cell cortex.

Looking forward, testing these and other models will be critical to understanding NuMA’s physical role in building and maintaining the nucleus. Toward this goal, it will be important to quantitatively measure NuMA’s physical contributions and to determine whether NuMA can cross-link chromatin in vitro and in vivo, and whether and how it influences the nuclear envelope and regulates gene expression. Tools to acutely inhibit NuMA will be essential to separating its roles in nuclear formation versus maintenance. Defining NuMA’s physical role in the nucleus, and with which molecular players it acts, will be key to understanding its relationship to other mechanisms regulating nuclear formation and maintenance, and to understanding its role in other, nonepithelial cell types and contexts. More broadly, these efforts will provide insight into how NuMA is repurposed and regulated across the cell cycle to build and maintain both the nucleus and the spindle.

Tet-on inducible CRISPR/Cas9 NuMA KO RPE1 cells (hereafter “NuMA KO RPE1 cells”; Hueschen et al., 2017) were cultured in DMEM/F12 with GlutaMAX (10565018; Thermo Fisher Scientific) supplemented with 10% tetracycline-screened fetal bovine serum (SH30070.03T; Hyclone Labs). Cells were maintained at 37°C and 5% CO₂. NuMA was depleted by inducing spCas9 expression with 1 µg/ml DOX hyclate in Tet-on inducible CRISPR/Cas9 RPE1 sgNuMA cells for 96 h before each experiment, unless other induction times are indicated. For transient expression experiments (Fig. 3, B and D–G, NuMA-FL-EGFP only; Fig. 4 E, NuMA-FL-EGFP only; Fig. S2 A; Fig. S2, E and F, NuMA-FL-EGFP only), inducible NuMA KO RPE1 cells were transfected with DNA using ViaFect (E4981; Promega) according to the manufacturer’s instructions 48 h before imaging. For rescue experiments (Fig. 3 B; and Fig. S2, A), cells were transfected 72 h before imaging. For long-term live imaging in Fig. 2, cells were treated with nocodazole and reversine at 72 h after KO induction, and imaged for 16–18 h. NuMA depletion at the end of the live imaging has occurred for 88–90 h, but for simplicity, we refer to it as “96 h.” At 88–90 h of induction, NuMA is expected to be nearly fully depleted (Fig. S1 A). Cell lines were not short-tandem repeat (STR)-profiled for authentication. All cell lines tested negative for mycoplasma.

To generate a cell line stably expressing mCherry-H2B and EGFP-Lap2β, NuMA KO RPE1 cells were infected with mCherry-H2B and EGFP-Lap2β lentiviruses (Figs. 2 and 5). Viruses were made in HEK293T cells from a pLenti6/V5-DEST plasmid (Invitrogen) containing mCherry-H2B (gift from T. Wittmann, University of California, San Francisco, San Francisco, CA) or EGFP-Lap2β and lentiviral packaging plasmids using calcium phosphate transfection. The medium of transfected HEK293T cells was replaced after 24 h, and virus-containing medium was collected 48 h after transfection. NuMA KO RPE1 cells were infected with virus-containing medium supplemented with 10 µg/ml polybrene and selected with 2.5 µg/ml blasticidin. A polyclonal stable line with low expression of mCherry-H2B and EGFP-Lap2β was further selected by FACS sorting and used herein.

To generate the Opto-NuMA cell line (Fig. 4, B and C), NuMA KO RPE1 cells were infected with NuMA-N-PhusionRed-LOV2 and EGFP-Zdk1-NuMA-C using retroviruses made in HEK293T-GP cells expressing the retroviral packaging genes gag and pol and transfected with calcium phosphate, as described above. A polyclonal stable line was established after FACS sorting.

To generate cell lines stably expressing NuMA truncations, the above protocol was used to infect NuMAKO RPE1 cells. We made the following cell lines: monoclonal NuMA-Bonsai-EGFP (Fig. 3, D–F; Fig. 4 E; Fig. S2, B–D; and Fig. S5), EGFP-Zdk1-NuMA-C (Figs. 4 E and S5), and NuMA-NC-EGFP (Fig. 3, D–F; Fig. 4 E; Fig. S2, E and F; and Fig. S5). In the NuMA-Bonsai-EGFP line, exogenous NuMA is expressed at higher levels than endogenous NuMA (Fig. S2 B).

Cells were treated with 664 nM nocodazole (M1404; Sigma-Aldrich) to depolymerize spindle microtubules and with 320 nM reversine (R3904; Sigma-Aldrich) to inhibit MPS1 and silence the spindle assembly checkpoint (Fig. 1; and Fig. S1, A–C). To avoid the confounding negative effect of a too-short mitosis on successful nuclear formation, we adjusted reversine concentrations in control cells (1 µM) so that NuMA KO cells spent at least as much time in spindle-less mitosis as control cells (Fig. 2; and Fig. S1, D and E). Nocodazole and reversine were added 10–30 min before live imaging. Cells were treated with 9 µM RO-3306 to acutely inhibit CDK1. To visualize DNA by live imaging in cells not expressing mCherry-H2B (Figs. 4 E and S5), 500 nM silicon rhodamine (SiR)-DNA dye (CY-SC007; Cytoskeleton) was added 30 min before imaging.

NuMA-FL-EGFP (FL human NuMA; NM_006185.3), used for transient transfection, is Cas9-resistant and was made by creating silent mutations (5′-GTG​TCA​GAG​AGA​CTG​GAC​TTT-3′ mutated to 5′-GTT​AGT​GAA​CGC​TTG​GAT​TTT-3′, preserving amino acids 57–62 of NP_006176.2 [“VSERLD”]). The modified sequence was subsequently cloned into a pEGFP-N1 vector (Clonetech, Takara Bio). These silent mutations made the sequence resistant to the sgRNA with the sequence 5′-AAG​TCC​AGT​CTC​TCT​GAC​AC-3′ (Hueschen et al., 2017). NuMA-Bonsai-EGFP used for transient transfection was made by amplifying NuMA^{1-705+1701-2115} from NuMA-FL-EGFP (Hueschen et al., 2017); for reference, the NuMA-Bonsai-EGFP herein is called NuMA-N-C-EGFP in Hueschen et al. (2017).

To make the mCherry-H2B and EGFP-Lap2β cell line, the EGFP-Lap2β coding sequence was amplified from a EGFP-Lap2β lentiviral plasmid (gift from D. Gerlich, Institute of Molecular Biotechnology of the Austrian Academy of Sciences [IMBA)] Vienna, Austria; plasmid laboratory ID: 1282; Samwer et al., 2017), and inserted into the lentiviral vector pLenti6/V5-DEST plasmid (Invitrogen, gift from T. Wittman). The lentiviral mCherry-H2B plasmid was subcloned from pH2B_mCherry_IRES_neo (Addgene plasmid 21044, gift from T. Wittman; Pemble et al., 2017).

To generate the Opto-NuMA cell line, NuMA-N-PhusionRed-LOV2 was cloned into a human cytomegalovirus (CMV) promoter plasmid as follows: NuMA-N^1-705 was amplified from NuMA-Bonsai^{1-705+1701-2115}-EGFP (Hueschen et al., 2017), and PhusionRed was amplified from pmKate2.7C-N1 (the Michael Davidson Fluorescent Protein Collection, University of California, San Francisco, Nikon Imaging Center). These were subsequently cloned into a CMV promoter plasmid containing the fast dissociation LOV2 domain (EB1-N-LZ-LOV2 described in van Haren et al., 2018) by replacing the EB1-N-LZ sequence (LZ stands for GCN4 leucine zipper) with the NuMA-N-PhusionRed sequence. EGFP-Zdk1-NuMA-C was cloned into the CMV promoter plasmid mCherry-Zdk1-EB1-C (van Haren et al., 2018) as follows: mCherry was replaced by EGFP amplified from pEGFP-C1 (Clonetech, Takara Bio), and EB1-C was replaced by NuMA-C^1701-2115, which was amplified from NuMA-Bonsai-EGFP (Hueschen et al., 2017). As a final step, NuMA-N-PhusionRed-LOV2 and EGFP-Zdk1-NuMA-C were subcloned into a Moloney murine leukemia retroviral plasmid (Addgene plasmid 51403, gift from I. Cheeseman, Whitehead Institute for Biomedical Research, Cambridge, MA; Welburn et al., 2009). For this, EGFP-TEV-S-Ska1 (contaning a tobacco etch virus [TEV] protease cleavage site) was removed from the Addgene plasmid 51403, and NuMA-N-PhusionRed-LOV2 and EGFP-Zdk1-NuMA-C were amplified and inserted in that site.

To generate cell lines stably expressing NuMA truncations, NuMA-Bonsai^{1-705+1701-2115}-EGFP was amplified from NuMA-Bonsai^{1-705+1701-2115}-EGFP (Hueschen et al., 2017), EGFP-Zdk1-NuMA-C was made as described above, and NuMA-NC^{1-212+1701-2115}-EGFP was made from assembling both NuMA-N^1-212 and NuMA-C^1701-2115 from NuMA-Bonsai^{1-705+1701-2115}-EGFP (Hueschen et al., 2017). Then these were each inserted into the lentiviral vector pLenti6/V5-DEST plasmid (Invitrogen, gift from T. Wittmann). Notably, NuMA-FL-EGFP, NuMA-Bonsai-EGFP, and NuMA-NC-EGFP plasmids used in this study are resistant to the NuMA sgRNA, as described above and in Hueschen et al. (2017).

To express and purify proteins for EMSA, the SNAP-tagged NuMA proteins were cloned as follows: NuMA-N^1-505, NuMA-C^1701-2115, NuMA-Bonsai^{1-705+1701-2115}, and NuMA-NC^{1-212+1701-2115} were amplified from NuMA-FL-EGFP and inserted into an Sf9 expression vector (pACEBac1). αβTubulin was purified from porcine brains (Hyman et al., 1991).

Cells were plated in six-well plates (08–772-1B; Thermo Fisher Scientific) in parallel with setting up an imaging experiment. The medium was removed, and cells were collected at the indicated time after induction of NuMA KO. Cells were subsequently lysed in cold radioimmunoprecipitation assay (RIPA) buffer, supplemented with protease inhibitor cocktail (11836153001; Roche), and incubated on ice for 30 min, and protein extracts were collected after 30 min of centrifugation at 4°C. Protein concentrations were measured using a Bradford assay kit (Bio-Rad) based on absorbance (A595 nm), and samples were prepared at equal concentrations in NuPage Sample buffer and NuPage Sample reduction agent (Invitrogen). Proteins were denatured for 10 min at 95°C, separated on a 3–8% Tris-Acetate gel (Invitrogen) by SDS-PAGE, and transferred to a nitrocellulose membrane by Western blotting using a Bio-Rad electrophoresis system. Membranes were blocked with 4% milk in TBS-Triton for 1 h at room temperature, probed with rabbit anti-NuMA (1:1,000; ab97585; Abcam; Fig. S2 B), rabbit anti-NuMA (1:300; NB500-174; Novus Biologicals; Fig. S1 A), and mouse anti-α-tubulin (1:5,000; T6199; Sigma-Aldrich) primary antibodies in 1% milk in TBS-T at 4°C overnight, washed with TBS-T, and incubated with anti-rabbit and anti-mouse HRP-conjugated secondary antibodies (0.01 µg/ml; sc-2357 and sc-2005; Santa Cruz Biotechnologies) for 1 h at room temperature. Proteins were detected with the ImageQuant LAS4000 using ECL (Thermo Fisher Scientific)–based chemiluminescence.

Cells were plated on 25-mm glass coverslips coated with 1 mg/ml poly-L-lysine. Cells were fixed with 95% (vol/vol) methanol containing 5 mM EGTA for 2 min at −20°C, washed with 0.05% Triton X-100 in TBS-T, and blocked with 2% BSA in TBS-T for 1 h at room temperature. Primary and secondary antibodies were diluted in 2% BSA in TBS-T and incubated with cells for 1 h at room temperature. DNA was labeled with Hoechst 33342 (B2261; Sigma-Aldrich). Cells were mounted in ProLongGold Antifade (P36934; Thermo Fisher Scientific). Cells were imaged using a spinning disk confocal inverted or an epifluorescence microscope as described in the following section. The following antibodies were used: rabbit anti-NuMA (1:300; NB500-174; Novus Biologicals; Fig. 1; Fig. S1, A–E; and Fig. S3), rabbit anti-NuMA (1:130; ab97585; Abcam; Fig. 3, B and C; and Fig. S2, A and C), anti-GFP conjugated to Atto488 (1:100; ChromoTek gba-488; Fig. 3, B and C; and Fig. S2, A and C), anti-mouse secondary antibodies conjugated to Alexa Fluor 488 (5 µg/ml; A11001; Invitrogen), and anti-rabbit secondary antibodies conjugated to Alexa Fluor 647 (5 µg/ml; A21244; Invitrogen).

NuMA KO RPE1 cells were plated on no. 1.5 glass-bottom 35-mm MatTek dishes coated with poly-D-lysine (MatTek Corporation) and imaged in DMEM/F12 with GlutaMAX (10565018; Thermo Fisher Scientific). Imaging was performed on a spinning disk confocal (CSU-X1; Yokogawa) inverted microscope (Eclipse Ti-E, Nikon) with a perfect focus system (Nikon). Cells were imaged in a stage-top incubation chamber (Tokai Hit) and maintained at 37°C with 5% CO₂. Experiments were performed with a Di01-T405/488/568/647 head dichroic (Semrock), along with 405-nm (100 mW), 488-nm (150 mW), 561-nm (150 mW), and 642-nm (100 mW) diode lasers, and emission filters ET455/50M, ET525/50M, ET630/75M, and ET705/72 (Chroma). Cells were imaged with a Zyla 4.2 sCMOS camera (Andor Technology) or an iXon3 camera (Andor Technology) operated by MetaMorph 7.7.8.0 (Molecular Devices; 50–100 ms exposures) with a 60× 1.4 Ph3 oil objective or a 100× 1.45 Ph3 oil objective (Fig. 4 B) in a single plane every 1–60 s. For the confinement experiments (Fig. 5), images were acquired with a 100× 1.45 Ph3 oil objective over 32 z planes spaced 0.25 µm apart. For long-term live imaging (Fig. 2), a 20× 0.5 Ph1 air objective was used, and images were acquired every 4 min over 11 z planes spaced 2 µm apart to account for focal plane changes over long periods. For fixed cell imaging (Figs. 1, 3, and S1), images were acquired over 30 z planes spaced 0.3 µm apart using the same microscope and objectives setting. Fixed cell imaging in Fig. S3 was performed on a Zeiss AxioPlan2 epifluorescence microscope. FRAP experiments were performed on a OMX-SR microscope (GE Health Care), equipped with a PCO Edge 5.5 cMOS camera (PCO), a four-line laser launch, 405/488/568/640 nm (Toptica), and a live cell chamber (GE Health Care) at 37°C with 5% CO₂. All images were acquired through a Plan ApoN 60× 1.42 oil objective (Olympus) with Laser liquid 1.518 (Cargile). GFP was imaged with the 488-nm laser, with a 528/48M emission filter in place. Photobleaching was performed on a 2-µm-diameter circular region in the nucleus. After defining the region of interest, a 405-nm laser was used to bleach the region. Images were acquired at 250-ms intervals for a total of 39 s; five baseline images were captured 1 s before bleaching and a time-lapse of 39 s after the bleaching event. Cells were plated on glass-bottom 35-mm dishes coated with poly-D-lysine (MatTek Corporation) for all live imaging experiments or on 25 mm no. 1.5 glass coverslips acid-cleaned and coated with poly-L-lysine (0117650; Marienfeld) for all immunofluorescence experiments and imaged in DMEM/F12 with GlutaMAX (10565018; Thermo Fisher Scientific) supplemented as for NuMA KO RPE1 cell culture as described above.

NuMA-N^1-505-SNAP, NuMA-Bonsai-SNAP, and NuMA-C^1701-2115-SNAP were purified from baculovirus-infected Sf9 cells. Briefly, cell pellets were resuspended in lysis buffer (50 mM Hepes, pH 7.4, 300 mM NaCl, 10% glycerol, 1 mM DTT, and 2 mM PMSF) supplemented with 1× protease inhibitor cocktail (Roche cOmplete Protease Inhibitor Cocktail) and lysed using a Dounce homogenizer (20 strokes with a loose plunger followed by 20 strokes with a tight plunger). The lysate was clarified by centrifugation (65,000 rpm for 45 min) and incubated with IgG Sepharose (GE Healthcare) for 2 h at 12°C, applied to a gravity flow column, and washed extensively with wash buffer (same as lysis buffer without PMSF or protease inhibitor cocktail). Protein labeling was performed by concentrating bead slurry to 5 ml, followed by the addition of 5 nmol of BG-LD555 dye (Lumidyne) and then incubation (1 h at 12°C). The slurry was readded to a gravity flow column and washed extensively in wash buffer. The protein–bead complexes were then treated with TEV protease at 12°C overnight. The mixture was then centrifuged (4,000 rpm for 5 min), and the supernatant was retrieved. The mixture was then concentrated using an Amicon Ultra-0.5 ml 50k spin column (EMD Millipore). The protein concentration was determined using the Nanodrop 1000.

Binding reactions were performed by mixing NuMA truncation proteins with 60 ng H1-H3 spacer DNA (Ludérus et al., 1994) followed by incubation on ice for 1 h. Binding buffer consisted of 20 mM Hepes, pH 7.9, 100 mM NaCl, 50 nM KCl, 1 mM MgCl₂, 1 mM DTT, 5% glycerol, 0.1 mg/ml BSA, and 0.1% NP-40. After incubation, binding reactions were then loaded onto a 2% Tris-glycine agarose gel, which was run for 90 min at 90 V. The gel was then poststained with 1× Sybr-Gold (S11494; Thermo Fisher Scientific) for 20 min followed by UV imaging to visualize DNA.

NuMA KO RPE1 cells transiently overexpressing NuMA-FL-EGFP, or stably expressing NuMA-Bonsai-EGFP, NuMA-NC-EGFP, or EGFP-Zdk1-NuMA-C, were plated on no. 1.5 glass-bottom 35-mm MatTek dishes coated with poly-L-lysine (MatTek Corporation) for 48 h. Cells were subsequently imaged, and FRAP experiments were performed as described above in the Microscopy section.

NuMA KO RPE1 cells, noninduced and stably expressing NuMA-N-PhusionRed-LOV2 and EGFP-Zdk1-NuMA-C, were imaged with a 561-nm laser (1 s interval) to visualize NuMA-N-PhusionRed-LOV2. The 488-nm laser was subsequently turned on (1 s interval) to simultaneously activate the LOV2 domain and visualize EGFP-Zdk1-NuMA-C. The 488-nm laser was turned off after 30 s to end activation of the LOV2 domain, while the 561-nm laser was used to continue imaging the behavior of NuMA-N-PhusionRed-LOV2.

Cells were plated on no. 1.5 glass-bottom 35-mm MatTek dishes coated with poly-D-lysine (MatTek Corporation). A PDMS-based suction cup device that contains a 10-mm-diameter glass coverslip with PDMS-based pillar structures (3-µm height; 200-µm diameter; spacing 700 µm to center) was generated (Guild et al., 2017; Le Berre et al., 2012). The device was attached to a 1-ml syringe, placed on top of a glass-bottom 35-mm dish, and sealed by pulling the syringe to apply negative pressure. After imaging 0.3-µm-spaced z planes of unconfined cells and initiating live acquisition, additional negative pressure was created to lower the pillared coverslips onto cells. While the pillars are 3 µm high, the “roof” of the device may be <3 µm from the glass bottom as it can deform under negative pressure (Guild et al., 2017). The negative pressure was maintained while z-stacks of compressed cells were acquired. Confinement was applied gradually over a period of ∼1 min and sustained for 5 min.

Observed nuclear phenotypes were qualitatively categorized into (1) single, round (shaped approximately like a sphere, without clear lobules or deep invaginations) nucleus, (2) nucleus with one to three micronuclei, (3) fragmented nucleus (one main nucleus and more than three micronuclei or two or more larger nuclei), and (4) atypical shape nucleus (cells with a single, abnormally shaped nucleus) based on DNA signal (Hoechst or mCherry-H2B). To determine percentages of these phenotypes before and after spindle-less mitosis (Fig. 2 B), the phenotype was analyzed based on mCherry-H2B signal for all cells at t = 0 (∼10 min after nocodazole and reversine addition) and at about t = 16 h.

A threshold of Hoechst or mCherry-H2B intensity was calculated and used to extract the outline of the nucleus using a particle analysis plugin from ImageJ. Based on this outline, the area and convex hull area were extracted. Solidity was calculated as the ratio of the area of the nucleus to the area of a convex hull (Fig. 1, D and E; and Fig. S1, D and E). NuMA and EGFP-tagged NuMA proteins mean intensities were measured in the nucleus based on a threshold of mCherry-H2B signal, as described above. The mean background intensity from three independent regions was subtracted. For each experiment, all values were normalized to the mean value of control (noninduced) cells (Fig. S1, A and E).

Individual cells were followed, and only cells that entered and exited spindle-less mitosis during image acquisition were included in the analysis. Phenotypes were determined before mitotic entry (mother cells) and after mitotic exit (daughter cells) for all cells. Only cells that had a single and round nucleus at mitotic entry were analyzed.

Control and NuMA KO cells stably expressing mCherry-H2B and EGFP-Lap2β treated with nocodazole and reversine were randomly selected for analysis. Nuclear envelope reformation (NER; t = 0) was determined by the first appearance of EGFP-Lap2β around condensed mitotic chromosomes, and the area of the chromosome mass was calculated for the subsequent 132 min. Area was calculated based on a threshold of mCherry-H2B intensity using a particle analysis plugin from ImageJ. All thresholds were checked and manually corrected when the automated threshold was not accurate before areas were calculated.

The mean fluorescence intensity value in the bleached and unbleached area, as well as background, were measured using a custom macro in ImageJ written by Arthur Molines. The signal was corrected for background and photobleaching and normalized 0 to 1, with 0 corresponding to the lowest signal and 1 to the highest signal after photobleaching. Recovery measurements were quantified by fitting normalized fluorescence intensities of bleached areas to a double exponential for EGFP-tagged NuMA constructs using GraphPad Prism. Cells with an average intensity 3.5 SDs above the NuMA-FL-EGFP mean intensity were excluded. The recovery halftimes were calculated from double exponential fits (FL, R²= 0.87; Bonsai, R² = 0.91; NC, R² = 0.87).

NuMA-N intensity in the nucleus was measured on a 2.2-µm-diameter circular region before, during, and after illumination with a 488-nm laser. The signal was corrected for background and normalized to average intensity.

The colocalization Pearson’s correlation coefficient was calculated by comparing images of nuclear regions with an area of 33.4 µm² using the ImageJ plugin Coloc 2.

The area of the cell was estimated by calculating the cell area in the plane where it was largest, based on the transmitted light signal. Because the data in the rest of Fig. 5 did not allow us to view all cells in their entirety, we acquired a separate dataset of unconfined control and NuMA KO cells to image the entire cells, now all fully in the field of view.

To calculate the area of the nucleus, a sum projection from a z-stack was generated and the area of the projected nucleus was extracted, based on H2B signal and EGFP-Lap2β signal at the nuclear rim.

3D volumes were measured using ImageJ and the 3D image segmentation plugin LimeSeg (Machado et al., 2019), selecting a region of interest corresponding to the nucleus. The segmentation plugin was used on z-stack images of EGFP-Lap2β with the following parameters: D_0 = 3.0, F_pressure = 0.025, Z_Scale = 3.8, Range_in_D0_units = 2.0, NumberOfIntegrationStep = −1, and RealXYsize = 0.066. The smallest feature size the segmentation could detect with these parameters was 3 pixels large. After each 3D analysis converged, we used the 2D result viewer to verify whether or not the contour found by LimeSeg overlapped with the EGFP-Lap2β outline. When there was an occasional discrepancy, we performed other analyses with multiple circular regions of interest.

To determine nuclear height before and after compression, orthogonal xz and yz views were generated from z-stacks of live cells (0.25-µm step size; 8-µm range), a line was drawn from top to bottom of the nucleus (based on EGFP-Lap2β signal), and its length was determined in ImageJ.

All data are expressed as average ± SD or SEM, as indicated in the figures. All statistical analyses were performed using Prism 8 software (GraphPad). Data were tested for normality using the D’Agostino–Pearson normality test when n was not too small. P values were calculated using unpaired t tests and Mann–Whitney U tests as indicated in the figure legends. For the contingency tables of Fig. 1, C, E, and F; Fig. 2, B and C; and Fig. 3 C, the Fisher's exact test was applied. All P values were two-tailed, and we used P < 0.05 as the threshold for statistical significance. Quoted n’s are described in detail in the figure legends, and in general refer to individual measurements (individual cells, nuclei).

Time-lapse images show single planes from spinning disk confocal imaging, except Fig. 2, A and D, which show maximum intensity projections (20 µm total z distance) of spinning disk confocal z-stacks. Immunofluorescence images of nuclei and mitotic spindles (Fig. 1 B; Fig. 3 B; and Fig. S1 B and D) show maximum intensity projections (20–30 µm total z distance) of spinning disk confocal z-stacks or single planes (Fig. S3). All images and videos were formatted using ImageJ.

Videos 1 and 2 show a maximum intensity projection from a spinning disk confocal z-stack. Video 3 shows a single z-slice from an OMX-SR microscope. Videos 4 and 5 show a single z-slice from a spinning disk confocal microscope. Videos were prepared for publication using ImageJ and set to play at 10 (Videos 1, 2, 3, and 5) or 20 frames per second (Video 4).

Fig. S1 shows the validation of the RPE1-inducible Cas9 NuMA KO cell line and of the nocodazole/reversine treatment (related to Fig. 1). Fig. S2 shows the validation of the RPE1 NuMA KO line stably expressing NuMA-Bonsai-EGFP and the estimate of NuMA-EGFP intensity in cells used in the FRAP analysis (related to Fig. 3). Fig. S3 shows that NuMA is only detectable on chromosomes after initial Lap2β recruitment (related to Fig. 4). Fig. S4 shows NuMA-N and -C responses in Opto-NuMA in vivo experiments, and purified NuMA truncation proteins used for in vitro EMSAs (related to Fig. 4). Fig. S5 shows that CDK1 inhibition does not affect NuMA binding to chromosomes during mitosis (related to Fig. 4). Video 1 shows that control RPE1 cells enter and exit mitosis with a single and round nucleus (related to Fig. 2). Video 2 shows that nuclear defects arise at nuclear envelope formation as NuMA KO cells exit mitosis (related to Fig. 2). Video 3 shows that overexpressed NuMA-FL-EGFP can form stable cable-like structures in the nucleus (related to Fig. 3). Video 4 shows that NuMA-C has a higher affinity than NuMA-N for chromosomes in the nucleus (related to Fig. 4). Video 5 shows that NuMA’s coiled-coil prevents its C terminus from binding chromosomes at mitosis (related to Fig. 4).

We thank Christina Hueschen (Stanford University, Stanford, CA) for inducible RPE1 NuMA KO cells; Daniel Gerlich (Institute of Molecular Biotechnology of the Austrian Academy of Sciences [IMBA], Vienna, Austria) for reagents; Jeffrey van Haren (Erasmus University Medical Center Rotterdam, Netherlands) and Torsten Wittmann (University of California, San Francisco, San Francisco, CA) for LOVTRAP plasmids and discussions; Joël Lemière for help with FRAP and nuclear volume analyses and discussions; Arthur Molines for help with FRAP analysis and discussions; Amanda Jack for advice on DNA binding experiments; Delaine Larson and Kari Herrington at the University of California, San Francisco (UCSF), Nikon Imaging Center for assistance with FRAP experiments; Abby Buchwalter, Duane Compton, Daniel Gerlich, and Megan King for discussions; Abby Buchwalter and Christina Hueschen for critical reading of the manuscript; the Fred Chang laboratory for discussions; and the Dumont laboratory for discussions and critical reading of the manuscript. We thank Sachin Kotak for generously sharing findings prior to publication and for insightful discussions.

The FRAP data for this study were acquired at the UCSF Nikon Imaging Center with instruments obtained by using funding from the National Institutes of Health (5R35GM118119); the UCSF Program for Breakthrough Biomedical Research, funded in part by the Sandler Foundation; the UCSF Research Resource Fund Award; and the Howard Hughes Medical Institute. We acknowledge the UCSF Parnassus Flow Cytometry Core supported in part by National Institutes of Health grant P30 DK063720 and National Institutes of Health S10 Instrumentation Grant S10 1S10OD021822-01 for assistance. This work was funded by National Institutes of Health grants DP2GM119177, R01GM134132 and R35GM136420 (to S. Dumont), the Rita Allen Foundation (to S. Dumont), the National Science Foundation Center for Cellular Construction grant 1548297 (to S. Dumont), National Science Foundation CAREER grant 1554139 (to S. Dumont), National Science Foundation grant MCB-1617028 (to A. Yildiz), and a Human Frontier Science Program Long Term Postdoctoral Fellowship LT001061/2016-L (to A. Serra-Marques).

The authors declare no competing financial interests.

Author contributions: A. Serra-Marques, Conceptualization, Methodology, Software, Validation, Formal Analysis, Investigation, Resources, Data Curation, Writing – Original Draft Preparation, Writing – Review & Editing, Visualization, Supervision, and Funding Acquisition; R. Houtekamer, Conceptualization, Methodology, Validation, Formal Analysis, Investigation, Data Curation, Writing – Review & Editing, and Visualization; D. Hintzen, Conceptualization, Methodology, Validation, Investigation, and Writing – Review & Editing; J.T. Canty, Methodology, Validation, Investigation, Data Curation, Writing – Review & Editing, and Visualization; A. Yildiz, Methodology, Writing – Review & Editing, Supervision, and Funding Acquisition; and S. Dumont, Conceptualization, Methodology, Resources, Writing – Review & Editing, Supervision, and Funding Acquisition.