Optogenetic reconstitution reveals that Dynein-Dynactin-NuMA clusters generate cortical spindle-pulling forces as a multi-arm ensemble

To position the mitotic spindle within the cell, dynamic plus ends of astral microtubules are pulled by membrane-associated cortical force-generating machinery. However, in contrast to the chromosome-bound kinetochore structure, how the diffusion-prone cortical machinery is organized to generate large spindle-pulling forces remains poorly understood. Here, we develop a light-induced reconstitution system in human cells. We find that induced cortical targeting of NuMA, but not dynein, is sufficient for spindle pulling. This spindle-pulling activity requires dynein-dynactin recruitment/activation by NuMA’s N-terminal long arm, and NuMA’s direct microtubule-binding activities to achieve a multiplicity of microtubule interactions. Importantly, we demonstrate that cortical NuMA assembles specialized focal structures that cluster multiple force-generating modules to generate cooperative spindle-pulling forces. This clustering activity of NuMA is required for spindle positioning, but not for spindle-pole focusing. We propose that cortical Dynein-Dynactin-NuMA (DDN) clusters act as the core force-generating machinery that organizes a multi-arm ensemble reminiscent of the kinetochore.


Introduction
Forces generated at dynamic plus-ends of microtubules drive directional movement of chromosomes and the mitotic spindle to achieve successful cell division (Inoue & Salmon, 1995). During animal mitosis, dynamic plus-ends of microtubules emanating from the spindle interact with two macro-molecular complexes; kinetochores and the cortical force-generating machinery. Kinetochores consist of more than 100 different proteins assembled on centromeric DNA and surround dynamic microtubule plus-ends using multiple fibril-like microtubule-binding proteins and/or ring-like couplers to harness the energy of microtubule depolymerization for chromosome segregation (Cheeseman, 2014;Dimitrova, Jenni, Valverde, Khin, & Harrison, 2016;McIntosh et al., 2008). In contrast, the cortical force-generating machinery assembles on the plasma membrane and pulls on the dynamic plus-ends of astral microtubules to define spindle position and orientation (Gonczy, 2008;Grill & Hyman, 2005). Spindle positioning determines daughter cell fate by controlling the distribution of polarized cell fate determinants and daughter cell size during both symmetric and asymmetric cell division (di Pietro, Echard, & Morin, 2016;Kiyomitsu, 2015;Morin & Bellaiche, 2011;Williams & Fuchs, 2013). In metaphase human cells, the cortical machinery consists of evolutionary conserved protein complexes, including the cytoplasmic dynein motor, its binding partner dynactin, and the cortically-anchored NuMA-LGN-Gαi complex (Fig. 1A) (Kiyomitsu & Cheeseman, 2012). Prior work has conceptualized that the cortical complex is distributed along the cell cortex and individually pulls on astral microtubules using dynein-based motility and/or by controlling microtubule dynamics (Kiyomitsu & Cheeseman, 2012;Kotak & Gonczy, 2013;Laan et al., 2012). However, compared to 4 the focal kinetochore structure, how this diffusion-prone membrane-associated complex efficiently captures and pulls on dynamic plus-ends of astral microtubules remains poorly understood. Here, we sought to understand the mechanisms of cortical pullingforce generation by reconstituting a minimal functional unit of the cortical forcegenerating complex in human cells using a light-induced membrane tethering. We found that cortical targeting of NuMA is sufficient to control spindle position, and that NuMA makes multiple, distinct contributions for spindle pulling through its N-terminal dynein recruitment/activation domain, central long coiled-coil, and C-terminal microtubulebinding domains. In addition, we demonstrate that NuMA assembles focal clusters at the mitotic cell cortex that coordinate multiple dynein-based forces with NuMA's microtubule binding activities. We propose that the cortical Dynein-Dynactin-NuMA clusters (hereafter referred to as the cortical DDN clusters) act as the core spindlepulling machinery that efficiently captures astral microtubules and generates cooperative pulling forces to position the mitotic spindle.

Optogenetic targeting of NuMA to the mitotic cell cortex is sufficient for dyneindynactin recruitment and spindle pulling
To understand the molecular mechanisms that underlie cortical force generation, we sought to reconstitute a minimal functional unit of the cortical force-generating machinery in human cells using a light-induced hetero-dimerization system (iLID) (Guntas et al., 2015). In this system, cytoplasmic RFP-Nano fusion proteins can be targeted to a locally illuminated region of the mitotic cell cortex by interacting with membrane-bound iLID (Fig. 1A, Fig. S1A-B, and Movie S1). Because the N-terminal fragment of NuMA is sufficient to recruit dynein-dynactin to the cell cortex (Kotak, Busso, & Gonczy, 2012), we first sought to manipulate endogenous NuMA. We established triple knock-in cell lines that stably express membrane-targeted BFP-iLID Importantly, following asymmetric NuMA-RFP-Nano targeting, the mitotic spindle was gradually displaced toward the light-illuminated region in 82.4% of cells (n=17, Fig.   1B, D-E, and Movie S2), whereas spindle displacement and cortical dynein recruitment 7 right) (Rankin & Wordeman, 2010). Importantly, even in the presence of 10 µM taxol, the spindle was gradually displaced toward the light-illuminated region :00), although the velocity was slow than that observed in control cells . In these experiments, we visualized spindle microtubules with 50 nM SiR-tubulin (Lukinavicius et al., 2014), a fluorescent docetaxel derivative, and confirmed the presence of 10 µM taxol by the decrease of SiR-tubulin intensity ( Fig. 2G t=0:00), likely due to competition for the same microtubule-binding site. These results suggest that the light-induced cortical DDN complex generates cortical pulling forces by using dyneinbased motility on astral microtubules even if microtubule depolymerization is inhibited.

Light-induced cortical targeting of dynein is not sufficient to pull on the spindle in human cells
A dimerized version of the yeast dynein motor domain is sufficient to position microtubule asters in microfabricated chambers (Laan et al., 2012). To understand the sufficiency of cortical dynein for generating spindle-pulling forces within a human cell, we next directly targeted dynein to the cell cortex (Fig. 3A). Similar to the NuMA-RFP-Nano fusion, endogenously tagged Nano-mCherry-DHC asymmetrically accumulated at the light-illuminated region within several minutes ( Fig. 3B and Fig. S3A), and subsequently recruited SNAP-tagged endogenous p150/dynactin to this cortical region . These results suggest that cortical dynein targeting is not sufficient for generating cortical pulling forces in human cells, consistent with recent studies demonstrating that human dynein is auto-inhibited (Torisawa et al., 2014;Zhang et al., 2017) and dynactin and cargo adaptors are required to activate dynein motility (McKenney, Huynh, Tanenbaum, Bhabha, & Vale, 2014;Schlager, Hoang, Urnavicius, Bullock, & Carter, 2014;Zhang et al., 2017).

A Spindly-like motif in NuMA is required for cortical dynein recruitment, but not sufficient for spindle pulling
The above results suggest that NuMA is required to activate dynein at the cell cortex.
Thus, we next sought to define the minimal functional region of NuMA as a dynein adaptor (Fig. 4A). Importantly, our truncation analyses revealed that the NuMA Nterminal region contains a Spindly-like motif sequence ( Fig. 4B-E, and Fig. S4A-G) that was recently identified as a conserved binding motif for the pointed-end complex of dynactin in dynein cargo adaptors (Gama et al., 2017). We found that NuMA wild type (WT) fragment (1-705), but not 4A mutant containing alanine mutation in the Spindlymotif (Fig. 4D), recruited dynein to the light-illuminated cortical region ( Fig. 4F and Fig.   S4H). However, the NuMA (1-705) WT and longer NuMA (1-1700) fragments were unable to fully displace the spindle despite the presence of substantial levels of cortical dynein ( Fig. 4B-C, H-I, Fig. S4I-K), whereas ectopically expressed full length NuMA (1-2115 ΔNLS) was able to displace the spindle in ~40% of cells ( Fig. 4G; NLS was deleted to eliminate dimerization with endogenous NuMA.). These results suggest that NuMA recruits dynein-dynactin via its N-terminal Spindly motif, likely to activate dynein's 9 motility at the mitotic cell cortex similarly to other dynein cargo adaptors (Gama et al., 2017;McKenney et al., 2014;Schlager et al., 2014). However, despite this activation, additional NuMA domains are required to produce cortical spindle-pulling forces.

NuMA's central coiled-coil is required for pulling on the spindle
The work described above defines two important molecular features for cortical force generation: dynein recruitment/activation through the Spindly-like motif and a distinct direct microtubule-binding activity by NuMA. To test whether these features are sufficient to generate cortical pulling forces, we next expressed a fusion construct, NuMA (N+C ΔNLS), that contains both its dynein-recruiting N-terminal and microtubulebinding C-terminal domains, but lacks a ~1000 aa region of its central coiled-coil ( However, the NuMA (N+C ΔNLS) fusion was unable to fully displace the spindle (Fig.   5G). These results indicate that NuMA's 200 nm long, central coiled-coil (Harborth, Wang, Gueth-Hallonet, Weber, & Osborn, 1999) also functions with its N-terminal and C-terminal domains to efficiently capture and pull on astral microtubules.

Identification of a clustering domain on NuMA's C-terminal region
Our results reveal that NuMA has multiple functional modules for force generation.
However, considering the sophisticated kinetochore structure that surrounds a plus-end of microtubule with multiple microtubule-binding proteins (Cheeseman, 2014;Dimitrova et al., 2016), we next sought to define the architecture of the cortical attachment site that is required to efficiently capture and pull on dynamic plus-ends of astral microtubules. Intriguingly, we found that NuMA constructs containing its C-terminal region displayed punctate signals at the cell cortex (e.g. Fig. 5H-I), suggesting that NuMA forms oligomeric structures at the mitotic cell cortex as observed in vitro 11 (Harborth et al., 1999). To understand mechanisms of the NuMA's C-terminal oligomerization/clustering at the mitotic cell cortex, we took advantages of a NuMA-C 3A fragment, which eliminates CDK phosphorylation sites (Compton & Luo, 1995) allowing NuMA to localize to the metaphase cell cortex independently of LGN (Kiyomitsu & Cheeseman, 2013). Similar to the NuMA-C-RFP-Nano (Fig. 5I), GFP-NuMA-C 3A displayed punctate cortical signals ( Fig. 6A-B #C1), which was distinct from that of its potential cortical interacting partners -phospholipids, 4.1 proteins and actin cytoskeleton (Kiyomitsu & Cheeseman, 2013;Kotak, Busso, & Gonczy, 2014;Mattagajasingh, Huang, & Benz, 2009;Zheng, Wan, Meixiong, & Du, 2014) -that localize homogenously to the cell cortex ( Fig. S6A-C). This suggests that the NuMA Cterminal fragment self-assembles on the membrane.
By analyzing different truncations and mutants, we found that a 100 aa region (aa: 1700-1801) of NuMA adjacent to its 4.1 binding domain is required for the formation of punctate foci ( Fig. 6A-B, compare #C1 to #C2), and further that a highly conserved 10 amino acid region, E1768-P1777 (Fig. 6C), is necessary for cluster formation ( Clustering by NuMA is required for spindle pulling and positioning, but not for spindle-pole focusing 12 Above we identified NuMA mutants (5A-2, 5A-3) that are unable to form clusters at the mitotic cell cortex ( Fig. 6B-C). To test the functional importance of the novel clustering behavior of NuMA, we next analyzed cortical force generation by full length NuMA wildtype (WT) compared to the 5A-3 mutant using Nano fusions. In cells expressing NuMA (1-2115 ΔNLS)-RFP-Nano (WT), NuMA and DHC-SNAP became gradually detectable as punctate foci (Fig. 6D, 4:30 and 11:00), and the spindle was displaced towards the light-illuminated region (Fig. 6D, 13:00). In contrast, when the NuMA 5A-3 mutant was targeted to the cell cortex, both NuMA 5A-3 mutant and DHC failed to form punctate foci To further probe functional importance of the NuMA's clustering activity, we next replaced endogenous NuMA with either NuMA WT or the 5A-3 mutant using the auxininduced degron (AID) system ( Fig. 7A) (Natsume, Kiyomitsu, Saga, & Kanemaki, 2016). with the NuMA KO phenotypes in human hTERT-RPE1 cells (Hueschen, Kenny, Xu, & Dumont, 2017). However, both NuMA WT and the 5A-3 mutant were able to rescue these abnormal spindle phenotypes (Fig. 7B #3 and #4), suggesting that clustering of NuMA is dispensable for microtubule focusing at the spindle poles. In contrast, when endogenous NuMA was replaced with NuMA 5A-3 mutant, the metaphase spindle was tilted and randomly oriented on the x-z plane (26.8 ± 20.7°, n=37, Fig. 7C and Fig. S7E) whereas the spindle in NuMA WT cells was oriented parallel to the substrate (10.7 ± 9.6°, n=34, Fig. 7C) as observed in control metaphase cells (11.5 ± 11.8°, n=41, Fig.   13 7C). These results suggest that NuMA's C-terminal clustering is required for proper spindle orientation. We note that the 5A-3 mutation site contains Y1774 (Fig. 6C), which is phosphorylated by ABL1 kinase and contributes to proper spindle orientation (Matsumura et al., 2012). However, treatment with the ABL1 kinase inhibitor Imatinib caused only a mild spindle orientation phenotype (12.3 ± 14.7°, n=27, Fig. 7C), suggesting that the spindle mis-orientation phenotype observed in the 5A-3 mutant is largely attributable to defects in NuMA clustering. Taken together, these results indicate that clustering activity of NuMA is required at the mitotic cell cortex, but not at the spindle poles, for generating cortical pulling forces. Thus, NuMA has a locationdependent structural function that clusters multiple DDN complexes to efficiently capture and pull on dynamic plus ends of astral microtubules.

Discussion
The cortical DDN complex acts as a core functional unit of the cortical force-

generating machinery
Here, we applied a light-induced targeting system, iLID (Guntas et al., 2015), for in cell reconstitution of the cortical force-generating machinery (e.g. Fig. 1A-B). Our work demonstrates that light-induced targeting of NuMA, but not dynein, is sufficient to control spindle position and orientation in human cells. This is consistent with recent findings that mammalian dynein requires cargo adaptors to activate its motility in vitro (McKenney et al., 2014;Schlager et al., 2014;Zhang et al., 2017). In addition, our findings suggest that LGN/Gαi are dispensable for force generation, and instead act as receptors that specify the position of NuMA at the cell membrane. Consistent with this model, LGN-independent pathways that target NuMA to the cell cortex have been reported, such as Dishevelled (Segalen et al., 2010) and phospho-lipids (Zheng et al., 2014). Thus, we propose that the Dynein-Dynactin-NuMA (DDN) complex is a universal core unit that constitutes the cortical force-generating machinery, whereas LGN and other receptors specify the targeting of the DDN complex to the membrane.

NuMA acts as a force-amplifying platform at the mitotic cell cortex
Our work demonstrates 4 distinct functions for NuMA at the mitotic cell cortex. First, NuMA recruits dynein-dynactin through its N-terminal region. We found that the conserved Spindly-like motif in NuMA is required for dynein recruitment (Fig. 4D-F).
NuMA may directly interact with the dynactin pointed-end complex through this Spindlylike motif similarly to other dynein cargo adaptors (Gama et al., 2017), and activate 15 dynein motility at the mitotic cell cortex. Second, the central long coiled-coil of NuMA is required for spindle pulling (Fig. 5G). Purified NuMA displays a long (~200 nm) rodshaped structure that shows flexibility with a main flexible-linker region near the middle of central coiled coil (Harborth et al., 1999). Longer flexible arms of NuMA may increase the efficiency of astral microtubule capture by the dynein-dynactin complex, similarly to fibril-like Ndc80 complexes and CENP-E motors at kinetochores (Kim, Heuser, Waterman, & Cleveland, 2008;McIntosh et al., 2008). Third, NuMA contributes to cortical force generation with its own C-terminal microtubule-binding domains (MTBDs) and remains associated with stalled and/or deploymerizing microtubules (Seldin et al., 2016). By using its two microtubule-binding domains, NuMA may harness the energy of microtubule depolymerization for pulling on astral microtubules similar to the human Ska1 complex or yeast Dam1 ring complex at kinetochores, both of which track with depolymerizing microtubules (Schmidt et al., 2012;Westermann et al., 2006).
Finally, we demonstrate that NuMA generates large pulling forces by clustering the DDN complexes through its C-terminal clustering domain ( Fig. 6C-E), similar to lipid microdomains on phagosomes that achieve cooperative force generation of dynein (Rai et al., 2016). Previous studies demonstrated that the 1700-2003 region of NuMA is required for oligomerization in vitro (Harborth et al., 1999). We found that 1700-1895 of 16 NuMA is necessary for NuMA fragments to form clusters at the mitotic cell cortex ( Fig.   4H and Fig.5C). Because this 1700-1895 fragment itself localizes to the cytoplasm, and showed no punctate signals (Fig. S6D, S6G #C6), the clustering activity of this region may be enhanced by its recruitment and concentration at membranes, as observed for CRY2 clusters (Che, Duan, Zhang, & Cui, 2015). Consistently, NuMA's clustering function is required for spindle pulling at the cell cortex ( Fig. 6E and Fig. 7C), but not for microtubule focusing at spindle poles (Fig. 7B).

Mechanisms of astral-microtubule capture and pulling by the cortical DDN clusters
Our live-cell imaging revealed that DDN clusters gradually assemble at the cell cortex and then displace the spindle (Fig. 5H, and 6D). Based on the results obtained in this study, we propose a multiple-arm capture model of astral microtubules by the DDN clusters ( Fig. 7D and Fig. S7F). Following nuclear envelope break down, cytoplasmic NuMA and DDN complexes are recruited to the mitotic cell cortex by binding to the LGN/Gαi complex, and then assemble DDN clusters on the cell cortex via the NuMA Cterminal domain. In vitro, up to 10-12 NuMA dimers self-assemble and form ring-like structures with an average diameter of 48 ± 8 nm (Harborth et al., 1999) (Fig. S7F), which are similar to those of the central hub of yeast kinetochores (37 ± 3 nm) (Gonen et al., 2012), and of the Dam1 ring complex (about 50 nm) which encircles a single kinetochore microtubule (Miranda, De Wulf, Sorger, & Harrison, 2005;Westermann et al., 2005). Given that the NuMA MTBD interacts with depolymerizing microtubules (Seldin et al., 2016), and dynein-dynactin moves along the lattice of microtubules, it is tempting to speculate that the DDN cluster encircles the plus tip of a single astral microtubule with NuMA's MTBDs, and holds the lateral wall of the astral microtubule with multiple dynein/dynactin-containing arms ( Fig. 7D and Fig. S7F). This multiple-arm capture by the DDN cluster leads to larger cooperative pulling forces by increasing the number of both dynein-dynactin containing modules and NuMA's microtubule binding per an astral microtubule. Additionally, this clustering may contribute to force generation by increasing both the stability of the DDN complex at the membrane, and the frequency for dynein-dynactin to capture or re-bind to astral microtubules. To produce pulling forces at dynamic plus-ends of microtubules, the cortical force-generating machinery appears to develop multiple molecular and structural features analogous to the kinetochore (Cheeseman, 2014;Dimitrova et al., 2016).
In conclusion, our optogenetic reconstitution and AID-mediated replacement reveal that the cortical DDN cluster acts as a core spindle-pulling machinery in human cells. Analyzing the structure and regulation of the DDN cluster will provide further information to understand the basis of spindle positioning in both symmetric and asymmetric cell division, and the general principles for microtubule plus-end capture and pulling.   Pole-to-cortex distance for control (black, n=5), and nocodazole-treated cells (blue or light-blue). Blue and light-blue graphs indicate immobile (n=5/9) and partially mobile pools (n=4/9), respectively. Error bars indicate SEM. (F and G) Live fluorescent images of NuMA-RFP-Nano (upper), and SiR-Tubulin and SiR-DNA (Lukinavicius et al., 2015) (lower), treated with DMSO (F) or taxol (G). DMSO and Taxol were added at -1:00, and light illumination began at 0:00, when SiR-tubulin images were selectively abolished by taxol treatment. (H) Kymographs obtained from image sequences in (F) and (G) at 30sec intervals. In taxol-treated cells, the spindle did not attach to the cell cortex as indicated with an asterisk, likely due to stabilized astral microtubules. (I) Pole-to-cortex distance for control (black, n=4), and taxol-treated cells (red, n=5). Error bars indicate SEM. Scale bars = 10 µm.       (Natsume et al., 2016). To construct CRISPR/Cas9 vectors, pX330-U6-Chimeric_BB-CBh-hSpCas9 (#42230, Addgene, Cambridge, MA) was used (Ran et al., 2013). PAM and 20-bp single guide RNA sequences were selected by the optimized CRISPR design tool (http://crispr.mit.edu) (Table S2). To construct donor plasmids containing homology arms for NuMA (~500bp homology arms), p150 (~200-bp arms) and DHC (N-terminal, ~500-bp arms), a gene synthesis service (Genewiz, South Plainsfield, NJ) was used. To construct the donor plasmid for DHC (C-terminal), a ~2,000-bp sequence was amplified by PCR from genomic DNA and then cloned into the pCR2.1-TOPO vector. A BamHI site was introduced at the center of the 2,000-bp fragment to facilitate the subsequent introduction of cassettes encoding tag and selection marker genes. To express Mem-BFP-iLID from the AAVS1 locus, membrane-targeted BFP2 ("Mem" from Neuromodulin; Clontech, Mountain View, CA) was fused to the N-terminus of iLID (#60411, Addgene) with a 53-amino acid (aa) linker derived from pIC194 (Kiyomitsu & Cheeseman, 2012) (#44433, Addgene), and the resulting fusion construct was introduced between the AfeI and HindIII sites in pMK231 (AAVS1 CMV-MCS-Puro, #105924, Addgene). Note that the Venus-iLID-caax construct (#60411, Addgene) was able to recruit RFP-Nano, but not NuMA-RFP-Nano to the membrane. To construct the RFP-Nano-NeoR cassette, a tagRFPt-Nano fragment (#60415, Addgene) was introduced between the SacI and MfeI sites in pMK277 (#72793, Addgene). The RFP-Nano-NeoR cassette was excised by BamHI and cloned into the BamHI site in the donor plasmid containing NuMA's homology arms. A 24-aa linker sequence containing 4× GGGS was introduced between the last codon of NuMA and the first codon of RFP. To construct the Nano-mCherry cassette, the Nano coding sequence was fused to the N-terminal region of mCherry from pIC194 with a 2× GGGS linker. To express Nano-mCherry-DHC, the BSDR sequence from pIC242 (Kiyomitsu & Cheeseman, 2012) (#44432, Addgene) was linked to the Nano-mCherry sequence with a P2A sequence, and the resulting BSDR-P2A-Nano-mCherry cassette, which contained a BamHI site at each end, was inserted into the BamHI site of the donor plasmid for DHC (N-terminal). A 47-aa linker sequence derived from pIC 194 was introduced between the last codon of mCherry and the start codon of DHC. To generate the SNAP-HygroR cassette, the mCherry coding sequence in pMK281 (#72797, Addgene) was replaced with the SNAPf coding sequence (N9186, New England BioLabs, Ipswich, MA) using In-Fusion ® cloning (Takara Bio, Ōtsu, Japan). The SNAP-HygroR cassette was excised by BamHI and cloned into the BamHI site of the donor plasmids. To make the DHC donor plasmid containing a SNAP-BSDR cassette, HygroR of the SNAP-HygroR cassette was replaced with BSDR from pIC242 using In-Fusion ® cloning. To conditionally express NuMA-RFP-Nano constructs from the Rosa 26 locus, a fragment containing Tet-On 3G, the TRE3GS promoter, and a multiple cloning site (MCS) derived from pMK240 (Tet-On-AAVS1-MCS-PuroR, #105925, Addgene) was introduced into pMK247 (Rosa26-CMV-MCS-HygroR, #105926, Addgene), which contains homology arms for the Rosa 26 locus. An RFP-Nano coding sequence was integrated between MluI and AgeI in the MCS, and NuMA fragments were subsequently inserted into the MluI site. NuMA truncation fragments and mutants were generated by PCR using NuMA cDNA (Compton & Luo, 1995;Kiyomitsu & Cheeseman, 2012) as a template, and the sequences were confirmed by DNA sequencing. These NuMA fragments encode isoform 2 (aa: 1-2101), which lacks a 14-aa region (aa: 1539-1552) in the longer isoform 1. However, the human NuMA constructs presented in the present study conform to isoform 1 (aa: 1-2115; NP_006176) to avoid confusion.
To construct mAID-mClover-3×FLAG-NeoR, a 3× FLAG sequence with a GGGS linker was introduced at the C-terminus of mClover of pMK289 (#72827, Addgene) by PCR. To conditionally express mCherry-NuMA WT or the 5A-3 construct from Rosa 26 locus, a fragment containing the TRE3GS promoter and the MCS derived from pMK240 was introduced into pMK247. The mCherry coding sequence derived from pIC 194 was integrated between the MluI and AgeI sites in the MCS, and the NuMA fragments were subsequently inserted Between the SalI and AgeI site.
・ Cell Culture and Cell Line Generation HCT116 and HeLa cells were cultured as described previously (Kiyomitsu & Cheeseman, 2012;Natsume et al., 2016;Tungadi, Ito, Kiyomitsu, & Goshima, 2017). Knock-in cell lines were generated according to the procedures described in Natsume et al., (Natsume et al., 2016) with minor modifications. CRISPR/Cas9 and donor plasmids were transfected into the cell lines using Effectene ® (Qiagen, Venlo, Netherlands). For drug selection, 1 µg/mL puromycin (Wako Pure Chemical Industries, Osaka, Japan), 800 µg/mL G418 (Roche, Basel, Switzerland), 200 µg/mL hygromycin B (Wako Pure Chemical Industries), and 8 µg/mL blasticidin S hydrochloride (Funakoshi Biotech, Tokyo, Japan) were used. Selection medium was replaced with fresh selection medium 4-5 days after starting selection. After 10-14 days, colonies grown on a 10-cm culture dish were washed once with PBS, picked up with a pipette tip under a microscope (EVOS ® XL, Thermo Fisher Scientific, Waltham, MA) located on a clean bench, and subsequently transferred to a 96-well plate containing 50 µL of trypsin-EDTA. After a few minutes, these trypsinized cells were transferred to a 24-well plate containing 500 µL of the selection medium, and then further transferred to a 96-well plate (200 µL per well) for the preparation of genomic DNA. The remaining cells in the 24-well plate were grown and frozen using Bambanker™ Direct (Nippon Genetics, Tokyo, Japan). For the preparation of genomic DNA, cells in the 96-well plate were washed once with PBS and then mixed with 60 µL of DirectPCR ® lysis solution (Viagen Biotech, Los Angeles, CA) containing 0.5 mg/mL proteinase K (Wako Pure Chemical Industries). The 96-well plate was sealed with an aluminum plate seal and incubated first at 56 °C for 5-6 h, then at 80 °C for 2-3 h in a water bath. To confirm the genomic insertion, PCR was performed using 1-2 µL of the genomic DNA solution and Tks Gflex™ DNA polymerase (Takara Bio). The cell lines and primers used in this study are listed in Tables S1 and S3, respectively.
・ Microscope System Imaging was performed using spinning-disc confocal microscopy with a 60× 1.40 numerical aperture objective lens (Plan Apo λ, Nikon, Tokyo, Japan). A CSU-W1 confocal unit (Yokogawa Electric Corporation, Tokyo, Japan) with three lasers (488, 561, and 640 nm, Coherent, Santa Clara, CA) and an ORCA-Flash4.0 digital CMOS camera (Hamamatsu Photonics, Hamamatsu City, Japan) were attached to an ECLIPSE Ti-E inverted microscope (Nikon) with a perfect focus system. A stage-top incubator (Tokai Hit, Fujinomiya, Japan) was used to maintain the same conditions used for cell culture (37 °C and 5% CO2). For light illumination, a Mosaic-3 digital mirror device (Andor Technology, Belfast, UK) and a 488-nm laser (Coherent) were used. The microscope and attached devices were controlled using Metamorph (Molecular Devices, Sunnyvale, CA).
To express NuMA-RFP-Nano constructs from the Rosa 26 locus in LGNdepleted cells, cells were treated with LGN siRNA (Kiyomitsu & Cheeseman, 2012) and Dox at 24 h and 48 h, respectively, according to the procedure described in Fig  S4B. RO-3306 was added at 48 h to cells that were then synchronized at G2 at 68 h. The NuMA-RFP-Nano fusion protein was expressed in most cells, but its expression frequency was reduced in cells that expressed longer NuMA fragments.
To compare the intensities of cortically targeted NuMA-Nano fusions, images of NuMA-Nano fusions and DHC-SNAP were acquired using the same parameters (Exposure time: NuMA, 1000 msec; DHC, 500 msec), except for Fig. 1B (NuMA, 1500 msec; DHC, 500 msec). To optimize image brightness, same linear adjustments were applied using Fiji and Photoshop. Supplemental movie files were generated using Metamorph and Fiji.
To activate the auxin-inducible degradation of NuMA-mAID-mClover-3FLAG (mACF), cells were treated with 2 µg/mL Dox and 500 µM indoleacetic acid (IAA) for 20-24 h. Cells with undetectable mClover signals were analyzed. A small population of cells showed mClover signals even after being treated with Dox and IAA. For replacement experiments, either mCherry-NuMA WT or the 5A-3 mutant was expressed from the Rosa 26 locus following Dox treatment. This caused the cells to simultaneously express OsTIR1 from the AAVS1 locus to initiate the auxin-inducible degradation of endogenous NuMA-mACF.

・ Light-inducible Targeting
Except for Figure S1B, HCT116 cells expressing Mem-BFP-iLID and NuMA-Nano fusion proteins were treated with RO-3306 and MG-132 according to the procedure described in Figure S4B to increase the proportion of metaphase-arrested cells.
To target Nano fusion proteins at the metaphase cell cortex, cells were illuminated using a Mosaic-3 digital mirror device (Andor Technology) at the indicated regions (circles with a diameter of 3.03 µm) with a 488-nm laser pulse (500 msec exposure, 25 mW). To manually control the frequency of the light pulse and the position of the illuminated region during time-lapse experiments, a custom macro was developed using Metamorph. Using this macro, indicated regions were illuminated ~10 times with the light pulse during time intervals (30 sec) between image acquisitions. The illuminated position was adjusted to precisely illuminate the cortical region of each cell. In response to the expression level of the Nano fusion proteins, the frequency of the light pulse was reduced to prevent the targeting of Nano fusion proteins throughout the cell cortex.
To reposition NuMA-RFP-Nano at the mitotic cell cortex in Fig. 1F, a cortical region adjacent to the spindle axis was illuminated. The light-illuminated region was changed once the spindle started to move but before the spindle was completely attached to the cell cortex. Spindles that rotated by approximately 90° within 15 min were counted.

・ Quantification of Cortical Fluorescent Signals and Spindle Displacement
Cortical and cytoplasmic fluorescence intensities were determined using Fiji by calculating the mean pixel intensity along three different straight lines (length 3 µm, width 3 pixels) drawn along the cell cortex showing Nano signals or the cytoplasm near the cell cortex but without any aggregations. The background intensity was subtracted from each measurement. The distance from the pole to the cell cortex was measured using Metamorph or Fiji. Line scans for cortical fluorescence intensity were generated using Fiji by calculating the mean pixel intensity along the segmented line (width 3 pixels) drawn along the cell cortex. Kymographs were generated using Photoshop (Adobe Systems, San Jose, CA). Spindle displacement was judged by the definition given in Figure S4I. In addition, cells that satisfied the following conditions were analyzed; 1. NuMA-RFP-Nano fusion proteins were asymmetrically recruited at the light-illuminated region, but not distributed to a whole cell cortex. 2. The cortical intensities of NuMA-Nano fusion proteins were higher than that of NuMA Δex24-RFP-Nano (Fig. 5F). 3. DHC-SNAP was detectable at the light-illuminated region except for the case of the cortical targeting of NuMA-C (#13). 4. The spindle was monitored for >10 min, and not vertically rotated. 5. The bipolar spindle was properly formed without severe membrane blebbing.

NuMA
(1-705+1700-2115) Dynein recruitment NuMA fragments Punctate signals Spinlde displacement NuMA-C (1700NuMA-C ( -2115 0.0% (0/6) (7/7) (0/7) + - NuMA G E P A S P I S Q R L P P K V E S L E S L Y F T P I P A R S G E P A S P I S Q R L P P K V E S L E S L Y F T P I P A R G G E P A S P I S Q R L P P K V E S L E S L Y F T P T P A R G G G T E P R A S Q R L P R K V E S L E S L Y F T P I A S R T S S P D S L I S G Q L P K K V E S L E S L Y F T P I P T R A R S S E R L A A K R R A L G A E S L E T L Y F T P M G K Q G LGN RNAi + MG132 LGN RNAi + MG132   LGN RNAi + MG132 WT    Figure S4 A B LGN RNAi + MG132  (Mattagajasingh et al., 2009) LGN BD: 1900-1926(Zhu et al.,2011 CDK phosphorylation sites: T2015, T2055, S2087 (Compton and Luo, 1995) MTBD (1): 1914-1985(Du et al.,2002 MTBD (2): 2002-2115 (Gallini et al.,2016) PIPs BD (1): 1699-1876 (Kotak et al., 2014) PIPs BD (2): 1996-2074 (Zheng et al., 2014) Region required for oligomerization  Oligomerization: 1700-2003 (Harborth et al., 1999) NuMA's C-terminal domains