p37/UBXN2B regulates spindle orientation by limiting cortical NuMA recruitment via PP1/Repo-Man

Mitotic spindle orientation determines the axis of cell division and plays a key role in cell fate determination in tissues (Panousopoulou and Green, 2014). Spindle orientation is controlled by forces exerted by cortical dynein–dynactin motor complexes on the astral microtubules emanating from the spindle poles (di Pietro et al., 2016). The strength of these forces is proportional to the abundance of motor complexes at the cortex (Du and Macara, 2004; Kotak et al., 2012). In metaphase, dynein–dynactin is recruited via the conserved Gαi–leucine-glycine-asparagine (LGN)–nuclear and mitotic apparatus (NuMA) complex: Gαi, a G protein α subunit, anchors the complex at the plasma membrane, LGN bridges the GDP-bound form of Gαi and the C terminus of NuMA, and NuMA recruits the dynein–dynactin complex to the cortex via its N terminus (di Pietro et al., 2016). The NuMA–dynein–dynactin complex is also present at spindle poles, where it physically tethers kinetochore fibers to focus the poles (Merdes et al., 1996; Gordon et al., 2001). In anaphase, additional Gαi/LGN-independent platforms recruit NuMA to the cortex, including the actin-binding protein 4.1R/G and phosphoinositides (Kiyomitsu and Cheeseman, 2013; Seldin et al., 2013; Kotak et al., 2014; Zheng et al., 2014).

NuMA recruitment to the cortex must be tightly controlled, as both too little and too much cortical NuMA impairs spindle orientation (Du and Macara, 2004; Kotak et al., 2012). In metaphase, NuMA phosphorylation by Cdk1 displaces it from the cortex, directing it to spindle poles. When CDK1 activity drops at anaphase onset, the protein phosphatase PP2A dephosphorylates NuMA, resulting in cortical enrichment (Kotak et al., 2013; Zheng et al., 2014). Conversely, Aurora A phosphorylation directs NuMA to the cortex (Gallini et al., 2016; Kotak et al., 2016). Finally, the Plk1 kinase displaces LGN and dynein–dynactin when centrosomes or unaligned chromosomes come too close to the cortex (Kiyomitsu and Cheeseman, 2012; Tame et al., 2016). This regulation ensures appropriate levels of cortical dynein to orient the spindle in metaphase and to elongate it in anaphase.

Our recent work identified p37, a cofactor of the p97^CDC48 AAA ATPase, as a regulator of spindle orientation (Kress et al., 2013). p97^CDC48 regulates multiple processes both in interphase and mitosis. It hydrolyzes ATP to segregate modified substrates from cellular structures, multiprotein complexes, and chromatin, and targets them either to degradation or recycling (Yamanaka et al., 2012). Functional specificity is given by p97 adapters such as p37. How p37 controls spindle orientation is, however, unknown. In this study, we find that p37 ensures proper spindle orientation by preventing the excessive recruitment of NuMA to the cortex in metaphase. Epistasis experiments indicate that p37 acts in a Gαi/LGN-independent manner via the protein phosphatase PP1 and its regulatory subunit Repo-Man, which promote NuMA recruitment to the cortex.

In tissue culture cells with an intact spindle orientation control, the mitotic spindle is oriented parallel to the growth surface, whereas spindle orientation defects result in a higher median angle between the spindle and the growth surface (called from here on “spindle angle”; Figs. 1 A and S1 A; Toyoshima and Nishida, 2007). As we previously showed, p37 depletion in HeLa cells increased the spindle angle when compared with control treatment (Fig. S1, A–D; Kress et al., 2013). This effect is rescued by exogenous p37 expression, indicating that this is not a result of an off-target effect (Kress et al., 2013). To understand how p37 controls spindle orientation, we depleted it in HeLa cells, labeled the spindle with SiR-tubulin, a live microtubule marker (Lukinavičius et al., 2014), and monitored it by time-lapse imaging. In siControl cells, the mitotic spindle remained parallel to the growth substratum and oscillated along the spindle axis (Fig. 1, A–C). In contrast, in 73% of sip37 cells, the mitotic spindle exhibited excessive oscillations in all axes, with a mean spindle rotation of 20.5° every 3 min (called spindle rotations from now on; Fig. 1, A–C; and Fig. S1 B), confirming our previous study (Kress et al., 2013). These orientation defects appeared immediately after mitotic entry, implying that they were not caused by a prolonged mitotic arrest, in contrast with other protein depletions that only deregulate spindle orientation several hours into a mitotic arrest (Tame et al., 2016).

These excessive spindle rotations were similar to the ones observed in cells overexpressing Gαi, LGN, or NuMA (Fig. 1 D; Du and Macara, 2004; Kotak et al., 2012). We therefore immunostained siControl and sip37 cells against Gαi, LGN, NuMA, and p150^glued (a dynein–dynactin subunit) and quantified their levels at the cortex. Both p150^glued and NuMA levels were increased by 50% in sip37 cells, whereas Gαi and LGN levels were unchanged (Fig. 1, E and F). NuMA has a complex localization pattern, as it forms large crescents at spindle poles and is present in the cytoplasm and the cortex (Fig. 1 E). At the cortex, NuMA is enriched proximal to spindle poles and is displaced from regions close to the chromatin via the Ran-GTP gradient (Fig. 1 E; Kiyomitsu and Cheeseman, 2012; Bird et al., 2013). We used line profiles in siControl and sip37 cells to quantify cortical and cytoplasmic NuMA levels and a threshold function to measure the volume it occupies at spindle poles (Fig. S3, D–F). p37 depletion increased cortical NuMA levels by 48% and in the cytoplasm by 22% (Fig. 1 G) but decreased the volume of NuMA spindle pole crescents by 41% (Fig. 1 H). The increase in cortical levels did not reflect a disruption of local regulation, as NuMA was still enriched proximal to the spindle poles (Fig. 1 I). Immunoblotting of siControl and sip37 cell extracts revealed constant total cellular NuMA levels (Fig. 1, J and K), indicating that p37 specifically regulates NuMA localization. The increase in cortical NuMA levels was responsible for the spindle orientation defects, as partial NuMA codepletion restored normal spindle angles (Fig. 1 L) and suppressed the excessive spindle rotations (Fig. 1 M) in sip37 cells; it did not, however, change the spindle angle in siControl cells. This rescue was not caused by an RNAi competition effect because p37 was depleted to the same extent whether NuMA was codepleted or not (Fig. 1 N). We conclude that p37 controls spindle orientation by limiting the amount of cortical NuMA in metaphase.

p37 depletion did not affect cortical Gαi and LGN levels, but it increased cortical NuMA levels, suggesting a Gαi-independent recruitment. To test this hypothesis, we codepleted Gαi and p37. Gαi depletion abolished cortical NuMA in control metaphase cells as previously shown (Fig. 2, A–C; Woodard et al., 2010), but it left a cortical NuMA pool in sip37 cells (Fig. 2, A and B). Equivalent results were observed after LGN depletion (Fig. S1, E–H), implying that a pool of cortical NuMA in sip37 cells is recruited in a Gαi/LGN-independent manner.

Aurora A phosphorylates NuMA in metaphase to promote its cortical enrichment (Fig. 2 D; Gallini et al., 2016; Kotak et al., 2016). Because p37 depletion increases the centrosomal levels of Aurora A in prophase (Kress et al., 2013), we tested whether the increase in cortical NuMA levels depended on Aurora A. We inhibited Aurora A activity in siControl and sip37 cells using 20 nM MLN8237 for 24 h and then quantified cortical NuMA levels. This treatment impaired Aurora A activity, as assessed by phospho-T288 antibodies recognizing active Aurora A, and severely reduced cortical NuMA levels (Fig. 2, E–H). Nonetheless, p37 depletion still increased cortical NuMA in MLN8237-treated cells, indicating that p37 can regulate NuMA independently of Aurora A (Fig. 2, G and H).

Phosphorylation by Cdk1 negatively regulates NuMA cortical localization, whereas dephosphorylation by PP2A promotes it (Fig. 2 D; Kotak et al., 2013; Zheng et al., 2014). High cortical levels of NuMA in sip37 are unlikely to be a result of low Cdk1 activity because sip37 cells enter mitosis with no significant delay (unpublished data) and show no changes in the levels of epitopes stained by MPM2, a phosphoantibody that recognizes Cdk1 phosphorylation targets (Fig. S1, I and J; Westendorf et al., 1994). To test the involvement of a phosphatase, we treated sip37 cells with the phosphatase inhibitor calyculin A (50 nM for 10 min). As previously shown, this treatment prevented cortical NuMA recruitment in siControl cells (Kotak et al., 2013). It also abolished the increase in cortical NuMA levels seen after p37 siRNA (Fig. 2, I and J) but did not alter the cortical Gαi and LGN levels (Fig. S1, K and L). We conclude that p37 limits cortical NuMA levels by preventing the action of a phosphatase.

We next asked whether p37 regulates cortical NuMA levels via PP2A. As shown previously, PP2A depletion abolished cortical NuMA localization in siControl cells (Kotak et al., 2013); it did not, however, suppress the increase in cortical NuMA levels in sip37 cells (Fig. 3, A–C). This suggests that PP2A is not the main phosphatase through which p37 regulates cortical NuMA levels. Calyculin A inhibits both PP2A and PP1 (Ishihara et al., 1989). Because exogenous p37 efficiently pulls down the PP1 isoform PP1α (Raman et al., 2015), we tested whether codepletion of the catalytic PP1α subunit impairs the excessive cortical recruitment of NuMA. PP1α depletion left cortical NuMA levels unchanged in control cells but abolished the excessive NuMA recruitment in sip37 cells (Fig. 3, D–F) and restored normal spindle angles (median of 15.6° in siControl/sip37-treated cells versus 8.0° in sipp1α/sip37 cells; Fig. 3 G). We conclude that p37 regulates cortical NuMA abundance via PP1.

Cdk1 and PP2A regulate NuMA in metaphase via threonine 2055 at the C terminus (Kotak et al., 2013; Seldin et al., 2013; Zheng et al., 2014). To test whether p37 regulates NuMA via the same site, we asked whether its depletion affects the localization of a phosphomimetic T2055E NuMA mutant. As previously shown, cortical GFP-NuMA-T2055E levels were reduced compared with GFP-NuMA in control depleted cells (Kotak et al., 2013), but in sip37 cells, they were comparable, indicating that p37 regulates NuMA via a pathway that is separate from the one controlled by Cdk1 and PP2A (Fig. 3, H and I).

Our results infer that the excessive cortical NuMA levels observed in sip37 cells are caused by high PP1 activity. PP1 activity could be generally higher in sip37 cells, or p37 could spatially regulate PP1 activity. In metaphase, PP1α is enriched at kinetochores, where it stabilizes kinetochore–microtubule attachments and silences the spindle assembly checkpoint by dephosphorylating Aurora B kinase and its substrates (Murnion et al., 2001; Pinsky et al., 2009; Vanoosthuyse and Hardwick, 2009; Liu et al., 2010). When we compared PP1α levels at kinetochores in HeLa cells stably expressing GFP-PP1α, we found a significant reduction in sip37 cells (Fig. 3, J and K). We also observed higher phospho–Aurora B (T232) levels on metaphase chromosomes, whereas Aurora B levels were unchanged (Fig. S2, A–D), consistent with lower PP1 activity. We conclude that p37 depletion does not lead to a generally higher PP1 activity in metaphase. We instead postulate that it leads to a lower PP1 activity at kinetochores and a higher cytoplasmic activity that promotes cortical NuMA localization.

PP1 phosphatases are composed by a catalytic and a regulatory subunit. In human cells, there are ∼200 regulatory subunits each influencing PP1 localization, substrate specificity, and activity (Bollen et al., 2010). The two best-characterized mitotic regulatory PP1 subunits are Sds22 and Repo-Man (Trinkle-Mulcahy et al., 2006; Vagnarelli et al., 2006; Posch et al., 2010; Wurzenberger et al., 2012; Eiteneuer et al., 2014; Rodrigues et al., 2015). We codepleted either of them in siControl or sip37 cells and measured cortical NuMA levels. Whereas Sds-22 depletion had no effect, depletion of Repo-Man abolished cortical NuMA enrichment both in siControl and sip37 metaphase cells (Fig. 4, A–D; and Fig. S2, E and F), whereas it left Gαi and LGN levels unchanged (Fig. S2, G–J). This effect was not a result of an off-target effect, as a second validated Repo-Man siRNA also impaired cortical NuMA recruitment (Fig. S2 K; Wurzenberger et al., 2012). To validate that PP1/Repo-Man positively regulates NuMA localization at the cortex, we transiently overexpressed eGFP–Repo-Man or a mutant that cannot bind PP1, eGFP–Repo-Man^RAXA (Trinkle-Mulcahy et al., 2006), and monitored cortical NuMA levels and spindle orientation. eGFP–Repo-Man expression led to higher cortical NuMA levels than eGFP expression, whereas eGFP–Repo-Man^RAXA expression led to lower levels (Fig. 4, E and F). 63% of the cells expressing eGFP–Repo-Man but only 27% of the cells expressing eGFP–Repo-Man^RAXA and 13% of untransfected cells also showed excessive spindle rotations (Figs. 1 and 4, G and H). These excessive spindle rotations depended on NuMA because partial NuMA depletion suppressed this percentage to 29% in eGFP–Repo-Man–expressing cells (Fig. 4, I and J). We conclude that PP1/Repo-Man activity is essential for cortical recruitment of NuMA in metaphase and that p37 must restrain this activity to prevent spindle orientation defects caused by excessive cortical recruitment of NuMA.

Cortical NuMA levels are temporally regulated as new Gαi-independent populations of NuMA are recruited to the cortex once cells enter anaphase (Kiyomitsu and Cheeseman, 2013; Seldin et al., 2013; Kotak et al., 2014). We thus investigated whether Repo-Man or p37 depletion also affect NuMA localization in anaphase (Fig. 5, A and B). Repo-Man depletion decreased cortical NuMA levels in anaphase. In contrast, sip37 did not affect NuMA levels in anaphase. This indicates that PP1/Repo-Man activity positively regulates cortical NuMA in metaphase and anaphase and that p37 limits this activity specifically in metaphase.

Our results identify a novel regulatory pathway controlling the abundance of the NuMA–dynein motor complex at the cortex (Fig. 5 C): PP1/Repo-Man promote NuMA localization at the cortex but are kept in check by p37 until anaphase. A failure to do so leads to severe spindle orientation defects in metaphase as a result of excessive dynein-dependent forces. This argues that the PP1/Repo-Man complex has a more general role than just acting at kinetochores and/or chromatin, unlike what is currently thought. We further postulate that in anaphase, PP1 promotes recruitment of NuMA at the cortex in conjunction with PP2A. Future investigations will uncover how PP1/Repo-Man controls NuMA. NuMA has a PP1 binding motif (RVxF; Egloff et al., 1997), and mass spectroscopy analysis of PP1 immunoprecipitates identified NuMA (Moorhead et al., 2008), suggesting a direct interaction. In all our experiments, Repo-Man depletion had a stronger effect on NuMA than PP1α depletion, suggesting that Repo-Man might regulate NuMA in combination with several PP1 isoforms. A second important aspect of future investigation is the relationship between p37 and PP1/Repo-Man. Our data point to a local regulation as we find lower PP1 activity at kinetochores and higher NuMA levels at the cortex. We speculate that p37 mobilizes PP1/Repo-Man from cytoplasmic complexes in metaphase, allowing it to bind to kinetochores and chromatin. When p37 levels are low, high cytoplasmic PP1/Repo-Man activity dephosphorylates NuMA, which leads to higher cytoplasmic and cortical NuMA levels. This model is compatible with proteomic studies, which reported interactions between several PP1 subunits and p37 (Huttlin et al., 2015, 2017; Raman et al., 2015). Moreover, Shp1, the Saccharomyces cerevisiae orthologue of p37, positively regulates nuclear accumulation of yeast PP1 (called Glc7) to promote Glc7 activity at kinetochores (Cheng and Chen, 2010; Böhm and Buchberger, 2013), suggesting that the overall principle of the pathway is conserved.

HeLa Kyoto cells and HeLa dynein heavy chain–GFP cells (gift from I. Cheeseman, Massachusetts Institute of Technology, Cambridge, MA) were grown in DMEM containing 10% FCS, 100 U/ml penicillin, and 100 mg/ml streptomycin at 37°C with 5% CO₂ in a humidified incubator. HeLa eGFP-centrin1/eGFP–CENP-A cells were supplemented with 0.2 mg/ml of G418 and 0.5 µg/ml of puromycin. HeLa GFP-PP1α cells (GA4.3; gift from L. Trinkle-Mulcahy; Trinkle-Mulcahy et al., 2006) were supplemented with 0.4 mg/ml of G418. Live-cell imaging experiments were performed at 37°C in Lab-TekII (Thermo Fisher Scientific) and Ibidi chambers in Leibovitz’s medium (L15) supplemented with 10% FCS. For drug treatments, Aurora A was inhibited for 24 h with 20 nM MLN8237 (Selleckchem) and PP1/PP2A for 10 min with 50 µM calyculin A (Thermo Fisher Scientific; Ishihara et al., 1989). To visualize microtubules during live-cell imaging, 50 nM SiR-tubulin (Spirochrome AG) was added to the cells 3 h before acquisition. To synchronize HeLa cells in mitosis, cells were treated for 14 h with 20 µM of the Eg5 inhibitor (+)-S-Trityl-l-cysteine (Sigma-Aldrich).

The siRNA oligonucleotides used for protein depletions (Thermo Fisher Scientific, GE Healthcare, or QIAGEN) were all already validated by other laboratories (references cited for each siRNA) and are as follows: control (scrambled, 5′-GGACCUGGAGGUCUGCUGU-3′; Mchedlishvili et al., 2012), p37 (5′-CUCCAGAAGAGGAGGAUAA-3′; Uchiyama et al., 2006), NuMA (5′-GGCGUGGCAGGAGAG-3′; Kiyomitsu and Cheeseman, 2013), Gαi_1/3 (5′-CCGAAUGCAUGAAAGCAUG-3′; Kiyomitsu and Cheeseman, 2013), Gαi₂ (5′-CUUGAGCGCCUAUGACUUG-3′; Kiyomitsu and Cheeseman, 2013), LGN (SMARTpool, 5′-GAACUAACAGCACGACUUA-3′, 5′-CUUCAGGGAUGCAGUUAUA-3′, 5′-ACAGUGAAAUUCUUGCUAA-3′, and 5′-UGAAGGGUUCUUUGACUUA-3′; Kiyomitsu and Cheeseman, 2012), PP1αc (5′-CCGCAATTCCGCCAAAGCCAA-3′; Wurzenberger et al., 2012), PP2Ac (5′-ATGGAACTTGACGATACTCTA-3′; Kotak et al., 2013), Repo-Man (5′-ATGGGACTCATCCGAGCTTAA-3′ in all figures except Fig. S2 and 5′-AGCAAATACTCCATTGCGTAA-3′ in Fig. S2; Wurzenberger et al., 2012), and Sds22 (5′-AGAGTTCTGGATGAACGACAA-3′; Wurzenberger et al., 2012). Cells were transfected with siRNAs for 48 h unless otherwise indicated at final concentrations of 50 nM using Lipofectamine RNAiMAX (Thermo Fisher Scientific). For codepletion experiments, selected siRNA oligonucleotides were transfected at a 1:1 ratio with a total concentration of 50 nM. For gene expression, the following plasmids were used: pEGFP-C1 (Invitrogen), pEGFP-(C1)–Repo-Man, and pEGFP-(C1)–Repo-Man/RAXA (gift from A. Lamond; plasmids 44212 and 44213; Addgene Trinkle-Mulcahy et al., 2006), GFP-NuMA-FL, and GFP-NuMA-T2055E (gift from P. Gönzcy; Kotak et al., 2013). Cells were transfected with 0.5–2 µg of plasmid DNA using Lipofectamine 2000 (Invitrogen) and analyzed 24 h after transfection. For experiments involving siRNA and plasmid DNA transfection, the siRNA and DNA were added to the cells 48 and 24 h before analysis, respectively.

Cells were fixed with 4% formaldehyde/0.2% glutaraldehyde (in PBS) for 15 min at room temperature. Slides were washed for 5 min with PBS/NH₄Cl (50 mM), washed twice with PBS, and incubated overnight with primary antibodies diluted in permeabilization buffer (20 mM Pipes, 0.5% saponin, 5% BSA, 137 mM NaCl, and 2.7 mM KCl, pH 6.8) at 4°C in a wet chamber. The slides were washed, incubated at room temperature with secondary antibodies diluted in permeabilization buffer for 45 min, and mounted on Vectashield with DAPI (Vector Laboratories). For immunofluorescence of HeLa GFP-PP1α (GA4.3), cells were fixed with 4% formaldehyde and 0.1% Triton X-100, and then slides were washed with PBS and mounted on Vectashield with DAPI. Alternatively, cells were fixed with methanol at −20°C for 4 min, washed with PBS, blocked for 1 h with 1% PBS/BSA, and incubated overnight with primary antibodies (in PBS/1% BSA) at 4°C in a wet chamber. After PBS washes, slides were incubated with secondary antibodies (in PBS/1% BSA) for 45 min at room temperature, washed with PBS, and mounted on Vectashield with DAPI. The following primary antibodies were used: anti-NuMA (rabbit; 1:500; Abcam), anti-Gαi (mouse; 1:200; Santa Cruz Biotechnology, Inc.), anti-p150^glued (mouse; 1:400; BD), anti–Aurora B (rabbit; 1:1,000; McClelland et al., 2007), antiphospho–Aurora B T232 (rabbit; 1:1,000; Rockland Inc.), antiphospho–Aurora A T288 (rabbit; 1:1,500; Cell Signaling Technology), anti–α-tubulin (mouse; 1:1,000; Sigma-Aldrich), anti–γ-tubulin (rabbit; 1:1,000; Sigma-Aldrich), anti-MPM2 (mitotic proteins; mouse; 1:1,000; Abcam), and anti-LGN (rabbit; 1:400; Sigma-Aldrich). For secondary antibodies, cross-adsorbed Alexa Fluor–conjugated antibodies (1:400; Invitrogen) were used.

Images for immunofluorescence were acquired using either an A1r spectral confocal microscope (Nikon) or an LSM800 Airyscan confocal microscope (ZEISS). The A1r spectral (inverted Ti Eclipse) confocal microscope is equipped with four PMTs, including two highly sensitive detectors (GaAsP) for green and red channels. Experiments were performed with a 60× 1.4 NA CFI Plan Apochromat Lambda oil objective, and NIS Elements AR software (v.4.20.01; Nikon) was used to set acquisition parameters. The LSM800 Airyscan confocal microscope was equipped with two high-sensitive GaAsP photomultiplier tube detectors. Experiments were performed with a Plan Apochromat 63× 1.40 NA oil objective, and ZEN software (v.2.3; ZEISS) was used to set acquisition parameters. Live-cell imaging was performed for 8–12 h at 37°C on the Ti widefield microscope (Nikon) equipped with an environmental chamber using a 60× 1.3 NA oil objective and a CoolSNAP HQ camera (Roper Scientific) at a sampling rate of 3 min, recording at each time point of 12 z stacks separated by 1 µm using NIS Elements AR software. The recorded images were quantified using FIJI software (ImageJ; National Institutes of Health) and Imaris (7.7.1; Bitplane). TIFF or PNG images were mounted into figures using CS Photoshop and CS Illustrator (Adobe).

The spindle rotation phenotype was scored in time-lapse videos by analyzing mean changes in angle/position of the metaphase spindle over time through live-cell imaging. To determine the spindle angle in fixed metaphase cells, cells were stained for γ- or α-tubulin. The x, y, and z coordinates of the spindle poles were obtained, and the spindle angle was calculated by applying the spindle pole coordinates to the trigonometric function shown in Fig. S3 A. The levels of cortical protein were measured by overlaying a 3-pixel-wide line on the cortex of a metaphase cell to obtain the mean intensity (Fig. S3 B). To quantify chromosomal Aurora B and phosphor–T232 Aurora B in metaphase, the DAPI image was used to build a mask for measuring the mean intensity value of the fluorescence signal on chromatin, and the same mask was used to correct for background (Fig. S3 C). To obtain the line profile of cortical proteins, a 10-pixel-wide × 6-µm-long line scan overlapping the cortex was used (Fig. S3 D). The line profiles of individual cells were first averaged within an experiment before calculating the mean between experiments. To quantify cortical NuMA levels around the entire cortex, we generated an intensity profile in metaphase using ImageJ (Fig. S3 E). To obtain the volume of NuMA at spindle poles, we applied a threshold function to the immunofluorescence images and performed the measurement using Imaris software (Fig. S3 F). To quantify GFP-PP1α at kinetochores, we determined the ratio between GFP signal intensities at kinetochores and normalized it with the cytoplasmic GFP signal (Fig. S3 G).

For cell extracts, cells were harvested in lysis buffer (150 mM KCl, 50 mM Tris, pH 7.4, 5 mM MgCl₂, 5% glycerol, 1% Triton X-100, 2 mM β-mercaptoethanol supplemented with EDTA-free protease inhibitor, and PhosSTOP phosphatase inhibitor [Roche]) and incubated on ice for 1 h. After centrifugation for 20 min at 13,000 rpm, the protein concentration in the supernatant was determined in a Bradford assay (Bio-Rad Laboratories) using a UV/Vis Spectrophotomer (MicroDigital). Alternatively, to quantify total NuMA levels, cells were directly harvested in Laemmli sample buffer and boiled for 10 min before mechanical shearing through a 26G syringe tip. The following primary antibodies were used: anti-NuMA (rabbit; 1:1,000; Abcam), anti-PP2Ac (rabbit; 1:1,000; EMD Millipore), anti-PP1αc (rabbit; 1:1,000; Cell Signaling Technology), anti-p97 (rabbit; 1:1.000; a gift from H. Meyer; Cao et al., 2003), anti-p37 (rabbit; 1.15 µg/µl; this study), and anti–Repo-Man (rabbit; 1:2,000; Cell Signaling Technology). The following HRP-conjugated secondary antibodies were used: ECL anti–rabbit IgG (GE Healthcare) and anti–mouse IgG (Bio-Rad Laboratories). Proteins were separated by SDS-PAGE and transferred onto nitrocellulose membranes by wet blotting. Protein bands were detected using the ECL Prime Western Blotting Detection reagent (GE Healthcare) and a PXi/PXi Touch luminescence detector (Syngene). Protein bands were quantified with ImageJ as described in the method outlined at http://lukemiller.org/index.php/2010/11/-analyzing-gels-and-western-blots-with-image-j/.

The full-length p37 cDNA was subcloned by Gateway technology (Invitrogen) into the pDEST–maltose-binding protein (MBP) vector. Recombinant MBP-tagged p37 (MBP-p37) protein expression was induced in BL21 Escherichia coli and purified using standard column purification on an amylose resin (New England Biolabs, Inc.) column. Purified MBP-p37 was used to immunize two rabbits with a standard immunization protocol (Moravian-Biotechnology). The anti-p37 serum was affinity purified on membrane strips carrying bacterially expressed GST-p37 antigen.

Outliers in fluorescent intensity quantifications were determined and removed from data using the modified Thompson Tau method. Paired two-tailed t tests, one-way ANOVA tests, Mann-Whitney tests, and Kruskal-Wallis tests were run on PRISM (7.02; GraphPad Software). Graphs were plotted in PRISM and mounted in Illustrator. Box blots show median, lower, and upper quartiles (line and box), 10th and 90th percentiles (whiskers), and means (+).

Fig. S1 is related to Figs. 1 and 2 and shows quantification of spindle angle in p37-depleted cells, characterization of spindle rotation, characterization of the p37 antibody, cortical LGN and NuMA levels in LGN and LGN/p37 depletions, MPM2 data, and levels of Gαi and LGN in control and calyculin A–treated cells. Fig. S2 is related to Figs. 3 and 4 and shows chromosomal Aurora B and phospho–T232 Aurora B levels in p37-depleted cells, Western blotting of the lysates of controls and the codepletion of Sds-22 and Repo-Man with p37, corresponding quantifications and Gαi and LGN staining in the Repo-Man and Repo-Man/p37 depletions, and NuMA line profile after depletion of Repo-Man with a second independent siRNA. Fig. S3 is related to the Image acquisition, live-cell imaging, and image quantification section of the Materials and methods and illustrates how the data presented in the manuscript were quantified.

Authors are grateful to P. Gönczy (Swiss Federal Institute of Technology in Lausanne, Lausanne, Switzerland), Q. Du (Augusta University, Augusta, Georgia), L. Trinkle-Mulcahy (University of Ottawa, Ottawa, Canada), I. Cheeseman, H. Meyer (University of Essen, Essen, Germany), and the MitoCheck consortium for reagents and cell lines, the Bioimaging platform of the Medical Faculty of the University of Geneva for microscopy support, and M. Bollen (University of Leuven, Leuven, Belgium), H. Meyer, and the members of the Gotta and Meraldi laboratories for helpful discussions, and Daniela Harry for technical support.

Work in the Gotta laboratory is supported by a Swiss National Science Foundation project grant (31003A_156013) and the University of Geneva, and in the Meraldi laboratory by a Swiss National Science Foundation project grant (31003A_160006) and the University of Geneva.

The authors declare no competing financial interests.

Author contributions: the project was initiated and directed by P. Meraldi and M. Gotta. All experiments were carried out by B.H. Lee with help from F. Schwager. B.H. Lee, P. Meraldi, and M. Gotta interpreted the data and wrote the manuscript.