Autoinhibitory mechanism controls binding of centrosomin motif 1 to γ-tubulin ring complex

Most microtubule-organizing centers, including centrosomes and the Golgi complex, require γ-tubulin ring complexes (γTuRCs) to initiate microtubule growth and to organize microtubules into ordered arrays (Lüders and Stearns, 2007; Petry and Vale, 2015; Wu and Akhmanova, 2017). γTuRCs are assembled as a macromolecular structure whose core components are γ-tubulin and five other γ-complex proteins (GCPs), GCP 2–6 (Kollman et al., 2011). In each γTuRC, several γ-tubulin small complexes, which are a tetrameric assembly of two γ-tubulins, GCP2 and GCP3, are arranged with additional γ-tubulins and GCP 4–6 to form an asymmetric, microtubule-incompatible structure (Kollman et al., 2008; Liu et al., 2020; Wieczorek et al., 2020a; Wieczorek et al., 2020b; Consolati et al., 2020). Recently, it was proposed that the binding of α/β-tubulin heterodimers to the γ-tubulins may facilitate the transformation of the asymmetric structure into a symmetric, microtubule-compatible geometry (Thawani et al., 2020).

The microtubule-nucleating activity of γTuRCs is tightly controlled in a spatiotemporal manner. For instance, most γTuRCs are found in the cytosol where they exhibit little or no microtubule-nucleating activity (Moudjou et al., 1996; Bauer et al., 2016). However, it remains unclear how this nucleating activity is regulated. In human cells, CDK5RAP2 (also known as Cep215), a protein that localizes to both centrosomes and the Golgi, contains a short sequence that specifically interacts with γTuRCs and stimulates their microtubule-nucleating activity (Fong et al., 2008; Wang et al., 2010; Choi et al., 2010). This sequence, called the γTuRC-mediated nucleation activator (γTuNA), consists of ∼30 residues and is located within the first half of a sequence motif known as the centrosomin motif 1 (CM1; Sawin et al., 2004; Zhang and Megraw, 2007; Choi et al., 2010). To date, all γ-tubulin complex (γTuC)-recruiting proteins in organisms ranging from yeast to humans have been found to contain, as the γTuC-binding element, either CM1 alone or CM1 in combination with a second motif called the Spc110/Pcp1 motif (SPM; Lin et al., 2014). Furthermore, in addition to the canonical γTuNA, a bipartite γTuNA has recently been identified in TPX2, a critical factor that, in association with the protein complex augmin, induces γTuRC-dependent nucleation of branched microtubules (Alfaro-Aco et al., 2017).

Although the γTuNA motif is a significant element that binds to and stimulates γTuRCs in both yeast and human cells, the mechanism that regulates its binding to γTuRCs remains unclear. In this report, we describe a previously unrecognized autoinhibitory mechanism within CM1 that controls γTuNA binding to γTuRCs. We found that this autoinhibition is relieved by phosphorylation of the autoinhibitory domain at Ser112. The physiological importance of this autoinhibition is significant because the loss of autoinhibition in CDK5RAP2 results in a significant increase in the growth of centrosome- and Golgi-derived microtubules, which affects centrosome positioning, the connection between centrosomes and the Golgi, and causes defects in Golgi assembly and cell polarization. We have also identified the centrosomal kinase NIMA-related kinase 2 (Nek2) as a candidate kinase that phosphorylates CDK5RAP2 at Ser112 to relieve autoinhibition.

Alignment of CM1 sequences from CDK5RAP2 and its orthologs in lower eukaryotes revealed that CM1 consists of two conserved segments (aa 59–88 and 96–114 of CDK5RAP2) separated by a short linker sequence (Fig. 1 A). While the first conserved segment (59–88) is recognized as the γTuNA, the function of the second conserved segment (96–114) has remained unclear. To investigate the function of the second conserved segment, we constructed two CDK5RAP2 fragments, 1–100 and 1–126, containing the γTuNA alone and the entire CM1, respectively, and compared their γTuRC-binding activity by performing transient-expression and immunoprecipitation assays. We found that the γTuNA-only construct (1–100) robustly coprecipitated γ-tubulin, but this coprecipitation activity was diminished by >90% in the entire CM1 (1–126; Fig. 1 B). Similar results were obtained when we compared the γTuRC-binding activity of the fragments 51–100 and 51–150, both of which lack the amino-terminal region upstream of the γTuNA (Fig. 1 C). These results indicate that extending the γTuNA to include the second conserved segment (96–114) strongly inhibits γTuRC-binding activity.

To further investigate the effect of the second conserved segment, we performed a mutational analysis. First, we individually mutated three conserved residues within this region and tested the mutants for their γTuRC-binding activity in coimmunoprecipitation experiments. The mutation K101A did not cause any change in γTuRC-binding activity (Fig. 1 D), whereas I104A drastically increased the binding activity and E105A moderately increased the activity (Fig. 1 D). We then created and tested the double mutant I104A/E105A, which exhibited further enhanced activity compared with the single mutant I104A (Fig. 1 E). Moreover, we engineered the double mutation I104A/E105A into full-length CDK5RAP2 and found that the mutation increased the γTuRC-binding activity of CDK5RAP2 by ∼12-fold (Fig. 1 F). In agreement with the binding results, the entire CM1 construct showed a much lower activity of γTuRC stimulation than the γTuNA protein (Fig. S1). Collectively, these results support the conclusion that the second conserved region harbors inhibitory activity toward the γTuNA. Therefore, we hereafter refer to the second conserved region as the γTuNA-inhibitor (γTuNA-In).

To understand the mechanism of autoinhibition, the interaction between the γTuNA and γTuNA-In domains was further studied. Co-expression of the γTuNA construct (FLAG-1–100) with the γTuNA-In construct (GFP-90–126) or the γTuNA-In mutant I104A/E105A was performed to validate the interaction. The immunoprecipitation of the γTuNA robustly coprecipitated the γTuNA-In but failed to coprecipitate the I104A/E105A mutant or GFP alone (Fig. 2 A). In direct-binding assays, the γTuNA-In recombinant protein readily bound to the γTuNA (Fig. 2 B). In a previous study, it was shown that within the γTuNA, Phe75 is crucial for γTuRC binding, and the F75A mutation abolished the binding activity (Choi et al., 2010). Therefore, the F75A mutant of the γTuNA was tested, and its binding activity toward γTuNA-In was significantly reduced (Fig. 2 C). Together, these results reveal the critical involvement of Ile104–Glu105 and Phe75 in the interaction between the γTuNA and γTuNA-In domains.

We established stable hTERT RPE-1 (RPE-1) sublines with inducible expression of CDK5RAP2. In these cells, epitope-tagged CDK5RAP2 and its I104A/E105A mutant were expressed at minimal levels in the absence of doxycycline, and upon induction, the proteins were expressed at levels similar and close to those of the endogenous protein (Fig. 3 A). Furthermore, we used siRNA-mediated knockdown to effectively suppress the expression of endogenous CDK5RAP2 in these cells by over 90% (Fig. 3 A). The I104A/E105A-expressing cells did not exhibit any overt defects in cell proliferation.

Quantitative analysis of microtubule regrowth after cold-induced microtubule depolymerization revealed that, relative to wild-type CDK5RAP2-expressing cells, I104A/E105A-expressing cells showed significant increases in centrosome-based and Golgi-based regrowth (∼2.2- and ∼5.4-fold, respectively; Fig. 3 B). Microtubule nucleation was also slightly increased in the cytosol in I104A/E105A-expressing cells. We also evaluated microtubule regrowth after nocodazole-induced depolymerization. After nocodazole washout, microtubules regrew from Golgi ministacks throughout the cytoplasm and from centrosomes (Fig. S2). After a short regrowth, we could unequivocally identify microtubules nucleated on the Golgi ministacks and those nucleated on the centrosomes (Fig. S2), and the results corroborate those of the assay after cold-induced depolymerization. Furthermore, in the I104A/E105A-expressing cells, the intensity of γ-tubulin on centrosomes was moderately (∼20%) higher than that in wild-type CDK5RAP2-expressing cells, although the Golgi intensity of γ-tubulin did not differ significantly (Fig. 3 C). Together, our results indicate that CDK5RAP2-CM1 autoinhibition exerts a strong control on γTuRC stimulation on centrosomes and the Golgi complex.

In our assays of cells expressing wild-type and I104A/E105A mutant CDK5RAP2 (Fig. 3, B and C), we observed that expression of the I104A/E105A mutant increased the proportion of cells in which centrosomes appeared either above or below the nuclei; in most of these cells, the centrosomes were situated below the nuclei (Fig. S3). Therefore, we analyzed the effects of transgene expression on the subcellular position of centrosomes. We measured the distances between the centrosomes and the cell centroid and adopted a published method of expressing the distances relative to the effective cell radius to minimize the influence of cell size and shape variations (Hale et al., 2011). Compared to cells expressing wild-type CDK5RAP2, cells expressing the I104A/E105A mutant showed a significant reduction in the distance between centrosomes and the cell centroid (28.9 versus 20.7% of the effective radius in cells expressing wild-type and I104A/E105A CDK5RAP2, respectively; Fig. 4, A and B). Conversely, centrosomes were within 20% of the effective cell radius from the cell centroid in ∼37.0% of the cells expressing the wild-type protein versus ∼56.4% of the cells expressing the I104A/E105A mutant (Fig. 4, A and B).

To assess whether Golgi-derived microtubules participate in centrosome centralization, we treated cells with brefeldin A (BFA) to disassemble Golgi complexes. We found that Golgi disassembly did not affect centrosome centralization induced by the I104A/E105A mutant (Fig. 4, A and B). Collectively, our results indicate that in cells expressing the I104A/E105A mutant, the overgrowth of centrosome-derived microtubules causes centrosome relocation toward the cell center and that Golgi-derived microtubules are dispensable for this centralization process.

Re-expression of the I104A/E105A mutant in RPE-1 cells resulted in the dissociation of the Golgi complex and centrosomes, as evidenced by an increase in the average distance between them from 1.53 μm in cells expressing wild-type CDK5RAP2 to 2.83 μm in cells expressing the I104A/E105A mutant, and the proportion of cells expressing the mutant protein with an increased centrosome–Golgi distance (distance larger than the average distance in cells expressing wild-type protein) was also significantly higher (Fig. 4, A and C). These findings suggest that the activity of centrosome-based microtubule growth is a crucial determinant of the relative positions of the Golgi and centrosomes, given that the Golgi organizes an asymmetric microtubule array that radiates from the cis- to trans-networks (Efimov et al., 2007).

During Golgi assembly, Golgi-derived and centrosome-derived microtubules facilitate, respectively, the fusion of Golgi ministacks dispersed in the cytoplasm (referred to as the G stage) and the translocation of the fused stacks toward the centrosome (referred to as the C stage; Miller et al., 2009). We disassembled Golgi complexes in RPE-1 cells re-expressing wild-type and I104A/E105A CDK5RAP2 by treating the cells with nocodazole, and after washing out the nocodazole, we monitored the reassembly process. In cells expressing the wild-type protein, both G-stage and C-stage assemblies occurred after nocodazole washout, and an intact Golgi was formed 1 h after washout (Fig. 5 A). By contrast, in I104A/E105A-expressing cells, although Golgi ministacks were clustered in the cytoplasm, as in cells expressing the wild-type protein, a considerable slowing was measured in both the translocation of Golgi patches toward centrosomes and the formation of the Golgi near centrosomes (Fig. 5 A). Therefore, the loss of CM1 autoinhibition impaired Golgi assembly at the C-stage but not at the G-stage.

During cell polarization, both the Golgi and the centrosomes are reoriented toward the leading edge; this process involves Golgi-derived microtubules and Golgi-centrosome association (Miller et al., 2009; Rivero et al., 2009; Yadav et al., 2009; Hurtado et al., 2011). We induced cell polarization through scratch wounding on monolayer cultures of the aforementioned RPE-1 sublines rescued with wild-type and I104A/E105A CDK5RAP2, and we examined the reorientation of the Golgi and centrosomes. In cells rescued with the I104A/E105A mutant, reorientation of both centrosomes and Golgi complexes was markedly slower than that in cells rescued with the wild-type protein (Fig. 5 B). This result was quantified by measuring, at several time points post-wounding, the angles required for reorienting the centrosomes and Golgi complexes to face straight toward the wound edge (Fig. 5 B). The time-course analysis revealed that cells expressing the I104A/E105A mutant showed delayed reorientation of both centrosomes and Golgi complexes; at 7 h post-wounding, the angles for centrosomes and Golgi complexes were ∼67.3° and ∼53.1°, respectively, whereas, in cells expressing the wild-type protein, the angles were ∼31.8° and ∼30.4° for centrosomes and Golgi complexes, respectively (Fig. 5 B). Thus, the loss of CM1 autoinhibition interferes with the reorientation of both centrosomes and the Golgi complex.

Nek2A has been reported to cause phosphorylation of CDK5RAP2 (Cervenka et al., 2016). Among the sites of Nek2A-induced phosphorylation, Thr102 and Ser112 are located within the γTuNA-In region (Fig. 1 A). While Ser112 is conserved or conservatively substituted in the metazoan orthologs of CDK5RAP2, Thr102 is not conserved. Moreover, phosphorylation at Thr102 was reported not to affect the γTuRC-binding activity of CM1 (Hanafusa et al., 2015). Thus, we investigated the function of Ser112 phosphorylation by substituting the residue with Asp to create a phosphomimetic mutant (S112D) and with Ala to create a non-phosphorylatable mutant (S112A). In pulldown assays, wild-type CDK5RAP2 and the S112A mutant exhibited similar γTuRC-binding activity, whereas the S112D mutant exhibited significantly higher binding activity (Fig. 6 A). Furthermore, the phosphomimetic mutation increased the activity to a similar level as the I104A/E105A mutation (Fig. 1 F and Fig. 6 A). Next, we tested the effect of S112D on γTuNA-In binding to the γTuNA. In contrast to wild-type γTuNA-In, the S112D mutant showed drastically diminished binding activity toward the γTuNA (Fig. 6 B), which agrees with the augmented binding of γTuRCs by CM1(S112D). These results strongly suggest that Ser112 phosphorylation disrupts the interaction between the γTuNA-In and the γTuNA to counteract CM1 autoinhibition.

To characterize the physiological effects of Ser112 phosphorylation, we created stable RPE-1 sublines inducibly expressing CDK5RAP2(S112D) and used these sublines, along with wild-type CDK5RAP2 sublines, in several assays. Prior to the assays, we depleted endogenous CDK5RAP2 by siRNA transfection and induced expression of wild-type CDK5RAP2 and the S112D mutant to levels similar to endogenous CDK5RAP2 in parent cells (Fig. 6 C). First, we measured microtubule regrowth after cold-induced depolymerization and found that S112D-expressing cells showed significantly higher regrowth activities on both centrosomes and Golgi complexes than wild-type-expressing cells (∼1.9- and ∼3.0-fold, respectively; Fig. 6 D). These results are consistent with those obtained from the rescue by the I104A/E105A mutant (Fig. 3 B), indicating that S112D and I104A/E105A have similar effects in disrupting autoinhibition. We then examined the effects of S112D on centrosome positioning, Golgi reassembly, and centrosome and Golgi reorientation. Similar to I104A/E105A, the S112D mutation reduced the distance between centrosomes and the cell centroid (Fig. S4), impaired Golgi reassembly after nocodazole washout (Fig. S5 A), and interfered with the scratch-wounding-induced reorientation of centrosomes and Golgi complexes (Fig. S5 B).

To confirm the phosphorylation of Ser112 by Nek2A, we first examined the interaction between Nek2A and CDK5RAP2. We detected robust binding of Nek2A and CDK5RAP2 in a coimmunoprecipitation assay (Fig. 7 A). After the phosphorylation reaction, proteins were resolved by SDS-PAGE, and the bands corresponding to the CM1 protein CDK5RAP2 1–140 were tryptically digested and analyzed by mass spectrometry. Phosphorylation of Ser112 was readily found in the Nek2A-treated sample (Fig. 7 B), but it was not detected in the sample treated in the absence of Nek2A. Therefore, Nek2A catalyzed Ser112 phosphorylation. To further assess the involvement of Nek2 in the regulation of cellular microtubule nucleation, we performed a microtubule regrowth assay on cells transfected with control or nek2-targeting siRNA; the transfection of the nek2-targeting siRNA effectively suppressed Nek2 expression (Fig. 7 C). The knockdown of Nek2 significantly reduced centrosome-based regrowth but did not show any obvious effect on Golgi-based regrowth (Fig. 7 D). These results revealed that Nek2 is involved in microtubule nucleation on centrosomes but is dispensable for Golgi-based nucleation.

We have revealed here that within CM1 a conserved segment located adjacent to the γTuNA serves as an autoinhibitory element and we have named this segment the γTuNA-In. This CM1 autoinhibition is highly potent: it blocked >90% of the γTuRC-binding activity of the γTuNA. Therefore, the γTuNA-In exerts a stringent on-site control of the γTuNA function. Mechanistically, we found that the γTuNA-In binds to the γTuNA and, furthermore, the binding requires Phe75, a γTuNA residue that is critically involved in the interaction between the γTuNA and γTuRCs (Choi et al., 2010). Notably, the I104A/E105A mutation within the γTuNA-In abolished the γTuNA-In binding to and inhibition of the γTuNA, thus providing control for establishing the autoinhibitory effect. Our findings suggest that CM1 is folded into a closed form in which the γTuNA-In blocks the γTuRC-binding site on the γTuNA. Recently, it was shown that the γTuNA exists as a dimer and, further, dimerization is required for binding to and activation of γTuRCs (Rale et al., 2022; Wieczorek et al., 2020a). It is also possible that the γTuNA-In inhibits the γTuNA activity toward γTuRCs by disrupting γTuNA dimerization.

We found that the loss of CM1 autoinhibition in CDK5RAP2 significantly enhanced microtubule nucleation on centrosomes and the Golgi, but only caused a moderate increase or no change in γ-tubulin levels on centrosomes and the Golgi, respectively (Fig. 3). These data suggest that in mammalian cells, CDK5RAP2 plays a significant role in stimulating rather than attaching γTuRCs to centrosomes and the Golgi. Several γTuRC-binding proteins are known to be involved in recruiting γTuRCs to centrosomes and the Golgi, including Nedd1 (also known as GCP-WD), pericentrin, AKAP450, and myomegalin variant 8, in addition to CDK5RAP2 (Lüders et al., 2006; Haren et al., 2006; Fong et al., 2008; Wang et al., 2014; Lin et al., 2014). Nedd1 was reported to tether γTuRCs to centrosomes and spindle microtubules, while AKAP450, in association with myomegalin variant 8, plays a key role in γTuRC attachment to the Golgi (Lüders et al., 2006; Haren et al., 2006; Rivero et al., 2009; Wang et al., 2014). Our data suggest that upon the removal of CM1 autoinhibition, CDK5RAP2 primarily binds to and stimulates γTuRCs present on centrosomes or the Golgi. Therefore, CDK5RAP2 binds to γTuRCs in a tightly controlled manner on centrosomes and the Golgi.

In interphase cells, centrosomes are typically located near but not at the cell centroid. Loss of CDK5RAP2-CM1 autoinhibition repositioned centrosomes toward the cell center and dissociated them from the Golgi (Fig. 4). Centrosome positioning results from the balancing of forces associated with centrosome-derived microtubules and weaker, myosin-generated forces (Burakov et al., 2003). Specifically, growing microtubules pushing against the cell cortex generate forces that drive the centrosome toward the cell center; conversely, antagonistic pulling forces generated on microtubules by cortical dynein direct the centrosome toward the cell periphery. The loss of CDK5RAP2-CM1 autoinhibition enhanced microtubule nucleation on centrosomes (Fig. 3) and thereby increased the number of centrosome-derived microtubules. Furthermore, computational simulations have indicated that an increase in microtubule numbers from centrosomes increases the net pushing forces, which would direct centrosomes toward the cell center (Zhu et al., 2010; Letort et al., 2016). Notably, because of centrosome repositioning toward the cell center, the centrosomes were detached from the Golgi in cells expressing the CM1 autoinhibition-defective mutant of CDK5RAP2 (Fig. 4). This detachment supports the existence of centrosome-independent mechanisms underlying Golgi positioning, and such mechanisms have been implicated in several studies (Magdalena et al., 2003; Taverna et al., 2016).

Golgi-derived microtubules act together with centrosome-derived microtubules in several Golgi functions, such as Golgi assembly, structural organization, and reorientation (Rivero et al., 2009; Miller et al., 2009). By using the CM1 autoinhibition-deficient mutant of CDK5RAP2 (i.e., the I104A/E105A mutant), we found that the aberrantly increased growth of Golgi-derived microtubules impaired not only Golgi reassembly after nocodazole-induced disassembly but also Golgi reorientation. These phenotypes resemble those observed following the knockdown of the DNA polymerase δ catalytic subunit, a γTuRC inhibitor that resides on the Golgi complex (Shen et al., 2017). It was reported that centrosome dissociation from the Golgi also impairs Golgi reassembly after disassembly and interferes with the centrosome and Golgi reorientation during cell polarization (Hurtado et al., 2011). Therefore, the defects in Golgi assembly and cell polarization observed in I104A/E105A- or S112D-expressing cells could result from both the aberrant growth of Golgi-derived microtubules and the dissociation of centrosomes from the Golgi.

Protein phosphorylation is a commonly seen mechanism that dissociates autoinhibitory domains from their targets, thus counteracting autoinhibition (Pufall and Graves, 2002). Nek2 induces CDK5RAP2 phosphorylation at Ser112, a residue located within the γTuNA-In, in cells (Cervenka et al., 2016) and catalyzes Ser112 phosphorylation in vitro (Fig. 7 B). The phosphomimetic mutation of Ser112 showed a similar effect as the mutation of Ile104-Glu105, a dipeptide within the γTuNA, to disrupt the interaction between the γTuNA and the γTuNA-In, which leads to an “open,” active state of CM1. Therefore, Ser112 phosphorylation is sufficient to relieve autoinhibition. Nek2 is a centrosomal kinase that acts at the onset of mitosis to induce centrosome disjunction (Fry et al., 1998). We found that silencing Nek2 expression compromised centrosome-based microtubule nucleation but did not affect Golgi-based nucleation (Fig. 7 D). Our data suggest that Nek2 likely phosphorylates Ser112 on centrosomes. However, further studies are required to determine whether the phosphorylation of Ser112 is cell-cycle-regulated and whether Nek2 is involved in centrosome maturation by phosphorylating this residue.

Given that the γTuNA-In is conserved from human to yeast and Ser112 is conserved or conservatively substituted in metazoan CM1 sequences (Fig. 1 A), the autoinhibition characterized here for CM1 is a conserved mechanism and the phosphorylation-mediated removal of the autoinhibition is likely to be conserved in metazoans. It has been recently shown that in Drosophila, the centrosomal isoform of the centrosomin contains a region amino-terminal to CM1 that inhibits the γTuRC-binding activity of CM1 (Tovey et al., 2021). However, this sequence is not conserved in mammalian CDK5RAP2. Moreover, the corresponding region in human CDK5RAP2 does not inhibit the CM1 activity (Tovey et al., 2021; Fig. 1 B). Therefore, although CM1 possesses a conserved autoinhibitory element, there are unique features of CM1 regulation in different species.

CDK5RAP2 fragment constructs were generated from a CDK5RAP2 cDNA (Fong et al., 2008) by using standard molecular cloning techniques, and CDK5RAP2 mutants were created through Single-Primer Reactions IN Parallel site-directed mutagenesis (Edelheit et al., 2009). To prepare inducible RPE-1 sublines, GFP-CDK5RAP2 and its mutants were cloned together with a Kozak sequence into pRetroX-tight-pur (Clontech) by using the Gibson assembly method (Gibson Assembly Master Mix, New England Biolabs). The GFP nanobody plasmid was acquired from Dr. Kazuhisa Nakayama (Kyoto University, Kyoto, Japan; plasmid #61838; Addgene; Katoh et al., 2015).

His₆-tagged recombinant proteins were expressed in Escherichia coli BL21 (DE3), purified using Ni²⁺-nitrilotriacetic acid resin (Qiagen), dialyzed against phosphate-buffered saline (137 mM NaCl, 2.7 mM KCl, 4.3 mM Na₂HPO₄, 1.47 mM KH₂PO₄, pH 7.4) supplemented with 10% glycerol, and stored in −80°C. Antibodies against CDK5RAP2 and GFP (both used at 1:500 dilution) have been described previously (Fong et al., 2008; Au et al., 2017). The following antibodies were purchased: anti-γ-tubulin (1:1,000, GTU88), anti-FLAG (1:1,000, rabbit polyclonal), anti-β-actin (1:5,000, AC-15), and anti-α-tubulin (1:600, DM1A), Sigma-Aldrich; anti-GM130 (1:600, monoclonal), BD Biosciences; anti-TGN46 (1:600, sheep polyclonal), Serotec; anti-pericentrin (1:300, C-16), Santa Cruz Biotechnology; and DyLight or Alexa Fluor secondary antibodies (1:600), Thermo Fisher Scientific.

All cell lines used here were purchased from American Type Culture Collection. HEK293T (CRL-11268) and Phoenix-AMPHO (CRL-3213) cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM, Gibco) and RPE-1 (CRL-4000) cells were cultured in DMEM/Ham’s F12 (1:1) medium; both media were supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin. Cells were grown in a humidified environment containing 5% CO₂ at 37°C and were free of mycoplasma contamination. Stable RPE-1 sublines with Tet-on inducible CDK5RAP2 expression were generated using pRetroX-tight-pur and pRetroX-Tet-On Advanced (Clontech). Plasmids were transfected using polyethylenimine (Polysciences) into HEK293T and Phoenix-AMPHO cells and siRNAs were transfected using Lipofectamine RNAiMAX (Thermo Fisher Scientific). siRNAs targeting cdk5rap2 (5′-UGG​AAG​AUC​UCC​UAA​CUA​A-3′) were used as previously reported (Fong et al., 2008). To disrupt the Golgi, cells were treated with 5 μg/ml BFA in a culture medium for 2 h (Vinogradova et al., 2012).

RPE-1 cells grown on glass coverslips were fixed for 15 min with methanol at −20°C or 4% paraformaldehyde in PHEM buffer (60 mM PIPES-KOH, 25 mM HEPES, pH 6.9, 10 mM EGTA, 2 mM MgCl₂) containing 0.5% Triton X-100 at room temperature. After fixation, the cells were sequentially stained with primary antibodies and Alexa Fluor-conjugated secondary antibodies, and nuclear DNA was stained with Hoechst 33258 (Sigma-Aldrich). To stain γ-tubulin on the Golgi, cells were extracted with a saponin-containing buffer (0.1 M PIPES-KOH, pH 6.9, 2 M glycerol, 5 mM MgCl₂, 2 mM EGTA, 0.1% saponin) before methanol fixation, and the anti-γ-tubulin signal was visualized by sequential staining with two secondary antibodies (goat anti-mouse followed by donkey anti-goat) labeled with the same dye (Wang et al., 2014). Epifluorescence images were captured using a microscope (Axio Observer Z1, Carl Zeiss) equipped with an ORCA-Flash 4.0 camera (Hamamatsu Photonics) and operated using ZEN microscope software (Version 2011, Carl Zeiss). Confocal images were acquired on a Leica TCS SP8 confocal microscope equipped with supersensitive HyD detectors and using a 63× oil objective and were processed by using the LAS X (Leica Application Suite X) Software Platform and the Fiji package of ImageJ.

The subcellular position of centrosomes was determined according to a reported method (Hale et al., 2011). Briefly, the cell boundary was outlined using the Freehand tool of ImageJ software, after which the cell centroid and the size of the outlined cell area were determined by the software. The cell area was converted into a circular shape to derive the effective cell radius, r_eff, by the formula r_eff = √(A/π), where A is the cell area. Distances between centrosomes and the cell centroid were expressed relative to r_eff. To determine distances between centrosomes and the Golgi, we outlined the Golgi area and measured the shortest distances from the centrosomes to the Golgi outline. If the centrosome was embedded in the Golgi area, the distance was considered to be 0.

Microtubule regrowth, Golgi reassembly, and Golgi reorientation were measured as reported (Xing et al., 2016; Shen et al., 2017). To examine microtubule regrowth, cells were first chilled on ice water for 1 h to depolymerize microtubules and then rewarmed at 37°C to allow the regrowth; the regrowth was terminated by cell fixing. To quantify regrowth, Golgi-associated microtubules were counted and the intensities of centrosome-based microtubule asters were measured; data are presented after background subtraction.

In the Golgi reassembly assays, the Golgi complexes were first disassembled by treating cells with 2.5 μg/ml nocodazole for 2 h at 37°C. After washing out the nocodazole with prechilled medium, Golgi assembly was initiated at 37°C and allowed to proceed at various times. Fluorescence images were processed using ImageJ software. Background subtraction was performed and Golgi or Golgi fragments were selected according to GM130 signals using the Freehand tool. The areas of the selected Golgi or Golgi fragments were measured using the Analyze Particles tool in the ImageJ software.

In cell polarization assays, a confluent monolayer of cells was scratched with a pipette tip to create a gap (Shen et al., 2017), and at various times after scratching, the cells were fixed for immunofluorescence microscopy. To examine centrosome and Golgi reorientation, two lines were drawn from the nucleus centroid—one perpendicular to the wound edge and the other to the centrosome or the centroid of the Golgi complex (Xing et al., 2016)—and the angle between the two lines was measured; this angle represents the degree of reorientation required to position the centrosome or the Golgi facing straight toward the wound edge.

Cell extracts were prepared at 4°C in lysis buffer (50 mM HEPES, pH 7.4, 150 mM NaCl, 1 mM EGTA, 1 mM MgCl₂, 1 mM dithiothreitol, Roche Complete Protease Inhibitor Cocktail) containing 0.5% IGEPAL CA-630 (Sigma-Aldrich) and clarified by centrifugation in a microcentrifuge. The extracts were incubated with the following beads as indicated for 2 h at 4°C with rotation: anti-FLAG M2-coupled beads (Sigma-Aldrich), M-280 streptavidin beads (Invitrogen), or Protein A-agarose beads (Thermo Fisher Scientific) coupled with primary antibody. The beads were subsequently collected by centrifugation and extensively washed in lysis buffer containing 0.1% IGEPAL CA-630, and the bound proteins were eluted and analyzed through SDS-PAGE and immunoblotting. To test protein binding in vitro, recombinant proteins were incubated in lysis buffer containing 0.1% IGEPAL CA-630 and 2 mg/ml bovine serum albumin at 4°C for 2 h, and then the proteins were immunoprecipitated and immunoblotted.

The recombinant protein His₆-FLAG-SBP-CDK5RAP2(1–140; 1 μg) was incubated with human Nek2 protein (Abcam) in a kinase buffer (50 mM Tris-HCl, pH 7.7, 10 mM MgCl₂, 10 mM NaF, 1 mM dithiothreitol, 0.1 mM ATP) at 30°C for 1 h. Reactions were terminated by adding SDS-PAGE sample buffer and heating at 70°C for 10 min. The proteins were then separated by SDS-PAGE and the gels were stained by GelCode Blue (Thermo Fisher Scientific). The bands corresponding to the 1–140 protein were excised for in-gel tryptic digestion. The resulting peptides were extracted, desalted using StageTips (Thermo Fisher Scientific), and analyzed by mass spectrometry using a UHPLC system coupled to a Q Exactive Plus Hybrid Quadrypole-Orbitrap mass spectrometer (Thermo Fisher Scientific). The acquired mass spectra were analyzed using Proteome Discoverer 2.5 software (Thermo Fisher Scientific) to identify the phosphorylation site of the CDK5RAP2 protein.

Statistical analyses were performed using GraphPad Prism 8 software, and all datasets were collected from at least three independent experiments. P values in experiments containing two groups were calculated using two-tailed unpaired Student’s t test, P values in experiments containing multiple groups were calculated using one-way ANOVA, and P values in experiments containing two or more groups at different time points were calculated using two-way ANOVA. Data distribution was assumed to be normal, but this was not formally tested. P ≥ 0.05 was considered statistically not significant.

Fig. S1 shows microtubule nucleation of γTuRCs in the presence of γ-TuNA (1–100) or CM1 (1–126). Fig. S2 examines microtubule regrowth after nocodazole washout in cells expressing CDK5RAP2 wild-type or I104A/E105A mutant. Fig. S3 shows centrosome positions relative to the nucleus in cells expressing CDK5RAP2(I104A/E105A). Fig. S4 shows centrosome positioning in cells expressing CDK5RAP2(S112D). Fig. S5 shows the results of Golgi reassembly and cell polarization assays using S112D-expressing cells.

The data are available from the corresponding author ([email protected]) upon any reasonable request.

This work was supported by grants from the Research Grants Council (General Research Fund 16100820 and Theme-based Research Scheme T13-605/18-W) and the Innovation and Technology Commission (ITCPD/17-9) of Hong Kong and the Shenzhen Science and Technology Committee Research Grant (SGDX20210823103200005). F.K.C. Au was supported by the Postdoctoral Fellowship Scheme (PDFS2021-6S03) of the Research Grants Council of Hong Kong.

Author contributions: R.Z. Qi and S. Yang designed the research project; S. Yang, F.K.C. Au., G. Li, and J. Lin performed the experiments; and R.Z. Qi, S. Yang, and X.D. Li wrote the manuscript.