The Nuclear Localization of γ-Tubulin Is Regulated by SadB-mediated Phosphorylation

Background: γ-Tubulin moderates gene expression by accumulating in the nucleus during early cell division. Results: SadB-mediated phosphorylation of Ser385-γ-tubulin exposes the nuclear localization signal of γ-tubulin. Conclusion: SadB kinases regulate the cellular location of γ-tubulin and in this way control cell growth. Significance: All knowledge on tubulins can aid the design of more efficient chemotherapeutic agents.

Cell Culture, Fractionation, Transfection, and Cell Cycle Analysis-NIH3T3 cells and U2OS cells were cultured, transfected, and fractionated as described (2,8,21). In brief, fractionated cells were lysed in buffer containing 0.1% Triton X-100 (BADT), and the supernatant was the cytosolic fraction. Thereafter, the supernatant of lysed nuclei was the nuclear membrane fraction, and the remaining pellet was the chromatin fraction (21). The amount of microtubule components attached to the nuclear membrane varies a lot between experiments (8). The purified fractions were analyzed by Western blotting using ␣-tubulin and histone as molecular markers for the cytosolic and nuclear fractions, respectively (8). Alternatively, the different fractions were resuspended in 1ϫ gtub buffer (50 mM Tris, (pH 7.5), 150 mM NaCl, 1 mM dithiothreitol, 1 mM EGTA, 1 mM MgCl 2 , 0.1 mM GTP, 0.5% Triton X-100, 0.1 mM Na 3 VO 4 , 2 g/ml aprotinin, 10 M leupeptin, 1 g/ml pepstatin, and 1 mM phenylmethylsulfonyl fluoride) and immunoprecipitated, as described (2,8). For cell synchronization, serum-arrested NIH3T3 cells were released for different time periods, as described previously (2,8). To arrest cells at early S phase, U2OS cells were presynchronized by treatment with 2 mM thymidine for 16 h and then washed and incubated in normal growth medium for 8 h, followed by a second 2 mM thymidine treatment for an additional 16 h. Then cells were harvested (0 h) or washed and incubated in normal growth medium for various times (22). Otherwise, to arrest cells in G1, U2OS cells were kept at confluence during 48 h. Cell cycle progression was analyzed by flow cytometry (2).
To obtain more equal protein levels of the various ␥-tubGFP mutants, the following DNA amounts were used in transfection experiments: 2 g of Ser 385 -C␥-tubGFP, 1 g of Ala 385 -C␥-tubGFP, 3 g of Asp 385 -C␥-tubGFP, or 1 g of GFP. In addition, U2OS cells were simultaneously transfected and thymidine-presynchronized as mentioned above.
Microscopy-NIH3T3 or U2OS cells were cultured and fixed as reported (2). Cells were incubated (1 h) with primary antibody, washed, and incubated (1 h) with Alexa488-or Cy3-labeled secondary antibody (Jackson), as described (2). Images were captured using an Olympus Bx51 or an Olympus IX73 microscope. Nearly simultaneous confocal GFP/differential interference contrast imaging sequences were collected using a Zeiss Axio Observer microscope (ϫ63, 1.4 numerical aperture plan-apochromat lens) with a stepper motor control for z-positioning and an Upgrade kit Axio Observer camera. Time lapse images were captured every 2 min. Time intervals of the mitotic processes were determined by counting film frames. Images were processed using ImageJ software. A minimum of 100 cells were counted in each sample, and the percentage of cells containing higher phospho-Ser 385 (Ser(P) 385 )-␥-tubulin levels in the centrosomes or nucleus was calculated.
RACE-PCR Analysis-Total RNA was isolated from U2OS cells as described elsewhere (2). For RACE-PCR, first-strand cDNA was synthesized using a Smart RACE kit (Clontech) and total RNA. The following targets were used to identify the N and C termini of SadB S : 5Ј-TCGGGCCGGGACCAAGGGCA-CCATGT-3Ј and 5Ј-TCGGGCCTCCTTGGGCGTCAGTCT-CCC-3Ј (N terminus); 5Ј-CCTGGTTCTGGAGCACGTCTC-GGG-3Ј and 5Ј-CGGGTGCAGGGGTCTTGGGGTCTTA-CTC-3Ј (C terminus). The entire SadB S sequence was then cloned into the HindIII/EcoRI sites of pcDNA3 (Invitrogen) with the FLAG epitope introduced 5Ј of hSadB S or into BamHI/ EcoRI sites of pGEX2T. We encountered difficulties in cloning SADB S , because Escherichia coli DH5␣ frequently deleted the kinase domain, thereby causing a frameshift mutation creating a nonsense codon. One of the recovered mutants was SadB S
Gene Expression Analysis and Luciferase Assays-Total RNA isolation was performed as described previously (8). mRNA expression array analysis was performed using the human Illumina platform. Data were normalized using quantile normalization, and the analysis of differential expression was performed using a linear model fitting (LIMMA packages) as described previously (23). Heat maps representing the expression intensity were drawn using the R function heatmap.2 in the gplots package (G. R. Warnes, B. Bolker, and T. Lumley, gplots: Various R programming tools for plotting data, R package version 2.6.0). Unsupervised clustering was performed using the R function hclust (method ϭ "ward"). Luciferase assays were performed on transfected U2OS cells, as described elsewhere (8).
To increase the affinity of the rabbit polyclonal anti-SADB antibody (2) in Western blot analysis, we mixed it 5:1 with rabbit polyclonal anti-N-terminal SadB (Abcam) antibody.
Statistical Analysis-All data are expressed as mean Ϯ S.D. (n Ͻ 4) or S.E. (n Ն 4), and Student's paired t test was used to analyze differences. Cell cycle profiles were assessed using FlowJo (Tree Star, Inc.). Western blotting bands were quantitated with ImageJ software.

RESULTS
Two Biochemically Different Pools of ␥-Tubulin in Mammalian Cell Lines-To visualize the cellular localization of ␥-tubulin during cell cycle (8,10,11), we performed immunofluorescence analysis with previously characterized antibodies (8,9) of a synchronized cell population (Fig. 1, A and B). Upon S phase entry, ␥-tubulin accumulated in the nucleus (8,10,11) and remained nuclear throughout S phase (Fig. 1A). During G 2 phase, a successive decrease of chromatin-bound ␥-tubulin was observed, reaching the lowest level in mitotic chromosomes (Fig. 1A). Finally, at early mitosis, ␥-tubulin was found in the cytosol and centrosomes (Fig. 1, A and B). In most eukaryotic cells, ␥-tubulin links a growing microtubule to a ␥TuRC by interacting with GCP2 and GCP3 (4). To study a possible association of other ␥TuRC components to chromatin, we analyzed the localization of GCP2 and GCP3 by immunostaining (Fig.  1B) and of GCP2 and ␣-tubulin by Western blotting (Fig. 1C). Although the localization of ␥-tubulin varied (Fig. 1, A and B), GCP2 and GCP3 occurred only in the cytosol, centrosomes, and mitotic spindles (Fig. 1B). Analysis of biochemical fractionations (21) confirmed that GCP2 and ␣-tubulin only associated with ␥-tubulin immunoprecipitates from cytosolic and nuclear membrane fractions (Fig. 1C). These results indicate that the composition of chromatin-associated ␥-tubulin complexes differs from ␥TuRC.
Increased Expression of SadB Augments the Size of the Nuclear ␥-Tubulin Pool-To identify the domain in ␥-tubulin important for its effect on S-phase entry (8), we tested the effect of various ␥-tubulin mutants tagged with green fluorescence protein (GFP; ␥-tubGFP) on the cell cycle. This revealed that U2OS cells transfected with ␥-TUBULIN shRNA accumulated in G 1 phase (n ϭ 3, p Ͻ 0.05), an effect that was rescued by ectopic expression of either an sh-resistant ␥-TUBULIN gene (n ϭ 3, p Ͻ 0.05) or an sh-resistant C ␥-tubulin terminus (C␥-tubGFP; n ϭ 3, p Ͻ 0.05) but not by expression of the shresistant ␥-tubulin N terminus (N␥-tubGFP) (8), suggesting that the ␥-tubulin domain that determines optimal cell cycle progression is the C-terminal region of ␥-tubulin ( Fig. 2A).
We have previously shown that the C terminus of ␥-tubulin (␥-tubulin(334-452)) encompasses the DNA-binding domain and the NLS, the latter of which includes residues Arg 399 , Lys 400 , Arg 409 , and Lys 410 (Fig. 2B) (1,8). Phosphorylation of residues near an NLS is often done to induce a conformational change to unmask the NLS and thereby trigger translocation of a cytosolic protein to the nucleus (24). Applying that strategy, we found the SadB phosphorylation motif Ser-X-Ser (Ser 383 -Ile 384 -Ser 385 ) 14 amino acids upstream of Arg 399 (Fig. 2B) (8).
To investigate whether hSadB affects the nuclear ␥-tubulin pool, we cloned hSADB Short (SADB S ) (GenBank TM accession number HQ830199) from U2OS cells. Expression of the SADB gene product, SadB S , was decreased in U2OS cells expressing SADB shRNA (Fig. 2C) (2). Human SadB S and mouse SADB S (2) protein sequences showed 99% homology, with only two amino acids differing: Ser 5 and Val 62 in the former but Ala 5 and Ile 62 in the latter (Fig. 2D). We have previously demonstrated that SADB kinases localized to centrosomes, but the exact cellular localization of the various SADB isoforms has not yet been elucidated. Biochemical fractionations of U2OS cells showed disparate localization of SadB L and SadB S . Endogenous SadB L and recombinant FLAG-SADB L localized to cytosolic and chromatin fractions (15), whereas endogenous SadB S and recombinant FLAG-SadB S were found mainly in chromatin fractions (Fig. 2E). Notably, elevated expression of mouse SADB L or hSadB S increased the pool of chromatin-bound ␥-tubulin (Fig. 2E). A densitometric analysis of Western blots containing the different cellular fractions showed that increase protein levels of SadB S caused a 67.2% (29.3% Ϯ 4.4 control and 49.0% Ϯ 1.7 SadB S ; n ϭ 3) rise of the amount of endogenous ␥-tubulin in the chromatin (Fig. 2E). Immunofluorescence analysis of endogenous SadB and FLAG-SadB S and HA-SADB S confirmed that SadB S localized not only to centrosomes (1) but also to the nucleus (Fig. 2F). Furthermore, compared with control cells, increased expression of SadB S raised the nuclear ␥-tubulin content by 84% in U2OS (n ϭ 8) and with 72% (n ϭ 8) in NIH3T3 cells, creating distinct ␥-tubulin domains in the nuclei of these cells (Fig. 2F). Thus, augmented nuclear ␥-tubulin lev-els coincided with an accumulation of cells in S phase ( Fig. 2G; n ϭ 3, p Ͻ 0.01) (4). These findings suggest a role for SadB S in regulation of the nuclear localization of ␥-tubulin.
To examine the endogenous Ser(P) 385 -␥-tubulin levels and their potential dependence on SadB in cells, we prepared an anti-Ser(P) 385 antibody that recognized purified bacterially produced full-length human ␥-tubulin that had been phosphorylated in in vitro by purified bacterially produced SadB S , and notably, this signal was reduced by phosphatase treatment or incubation with the kinase-dead mutant SadB S ⌬61-198 (Fig. 3D). To study a possible link between phosphorylation of Ser 131 -␥tubulin and of Ser 385 , we analyzed the effect of phosphorylation on Ser 385 in a non-phosphorylatable Ser 131 3 Ala ␥-tubulin FIGURE 1. Nuclear ␥-tubulin is not associated with GCP2 or GCP3. A, cells were synchronized in G 0 /G 1 by keeping cell confluence during 48 h (G 0 /G 1 ) or in early S phase by double thymidine block and released for 5 h (S) or 9 h (G 2 /M). Cell cycle progression was monitored by determining the DNA content of cells by flow cytometry (graphs) and by analyzing the protein levels of the G 1 marker cyclin D, the S-phase marker cyclin A, and the G 2 /M marker cyclin B in cell lysates (Tot. lysate) by Western blotting (right panels). A and B, localization of endogenous ␥-tubulin was examined by immunofluorescence staining with anti-␥tubulin (green; rabbit) and anti-GCP3 (red; mouse) or anti-␥-tubulin (green; mouse) and anti-GCP2 (red; rabbit) antibodies, and nuclei were detected using DAPI (blue) in U2OS cells (n ϭ 3-5). Scale bars, 10 m. C, cells (20 ϫ 10 6 ) were biochemically divided into the following cell fractions: cytosolic (C), nuclear membrane (N), and chromatin (CH). Each fraction was subjected to immunoprecipitations (IP) with an anti-␥-tubulin antibody and developed by Western blotting (WB) with an anti-GCP2 antibody (top), and then reprobed with ␣-tubulin (␣Tub) and ␥-tubulin (␥Tub). Aliquots of the lysates used in the immunoprecipitations were run as loading controls (Total lys. fract.) and analyzed by Western blotting (n ϭ 3). Models depict the cellular distribution of centrosomes/microtubules (Cent/MT), membranes (Membr.), and chromosomal (Chrom.) elements in the analyzed biochemical fractions. (Fig. 3E) and a phosphomimetic Ser 131 3 Asp ␥-tubulin (Fig.  3F). We found that phosphorylation of Ser 385 was reduced in the Ala 131 -␥-tubulin mutant but restored in the Asp 131 -␥-tubulin mutant, suggesting that phosphorylation of Ser 131 is a prerequisite for the phosphorylation of Ser 385 . However, the band detected by the anti-Ser(P) 385 antibody had a higher molecular mass than expected (60,000 Da) and was not detected by total anti-␥-tubulin antibody (Fig. 3, D-F) but was recognized by anti-Ser(P) 131 antibody (Fig. 3E) (2). Altogether, the results indicate that in vitro SadB kinases phosphorylate ␥-tubulin on Ser 385 and on Ser 131 .
To determine whether the expression levels of SadB or ␥-tubulin affected the phosphorylation levels of Ser 385 -␥-tubulin, we varied their protein levels in U2OS cells. Analysis of U2OS cellular extracts using the anti-Ser(P) 385 antibody showed an ϳ60,000 Da band, which levels were affected upon increased recombinant SadB S levels ( Fig. 3G) or reduced endogenous SadB (Fig. 3H) or ␥-tubulin expression (Fig. 3I). These data show that the 60 kDa band is ␥-tubulin and that SadB S regulates the cellular levels of Ser(P) 385 -␥-tubulin.
We subsequently examined endogenous Ser(P) 385 -␥-tubulin levels in synchronized NIH3T3 cells. Western blotting of chromatin fractions of NIH3T3 cells with Ser(P) 385 antibody revealed accumulation of endogenous Ser(P) 385 -␥-tubulin during early S phase (Fig. 4A). The 60 kDa band recognized by anti-Ser(P) 385 antibody was phosphatase-sensitive (Fig. 4B), FIGURE 2. An increased level of SadB S augments the nuclear pool of ␥-tubulin. A, the DNA content was determined by flow cytometry of non-synchronous U2OS cells expressing shRNA (shcontrol; shC) or sh-␥-TUBULIN (sh␥TUB) and co-transfected with one or the following constructs: GFP, wild-type ␥-tubGFP (WT-␥-Tub), N␥-tubGFP (N␥-Tub 1-333 ), or C␥-tubGFP (C␥-Tub 334 -452 ), as indicated. The data on each cell population are presented as the proportion of cells in G 1 (n ϭ 3). Bottom, cell extracts were analyzed by Western blotting. Arrowheads and arrows show GFP and endogenous ␥-tubulin, respectively (n ϭ 3). B, structure of wild-type human ␥-tubulin (h␥Tub) constructs depicting the NLS and phosphorylated Ser 385 (P). The region surrounding Ser 385 is conserved in the indicated species. Boldface letters, identity; blue letters, polar or hydrophobic conservation. C, total lysate of U2OS cells transfected with a control vector, SADB shRNA, or FLAG-SadB S and analyzed by Western blotting (WB) using an anti-SadB antibody, followed by anti-␣-tubulin (n ϭ 3). D, structure of human SadB S (top) and mouse SADB S (bottom) constructs showing the amino acids that differ between the two isoforms in boldface type. E, U2OS cells (1 ϫ 10 6 ) expressing control, FLAG-tagged human SadB S , or mouse SADB L vectors were biochemically divided into cytosolic (C), nuclear membrane (N), and chromatin (CH) fractions, as in Fig. 1C, and analyzed by Western blotting (WB; n ϭ 5). F, localization of endogenous human and mouse SadB (eSadB and eSADB, respectively), FLAG-SadB S , HA-SADB S , and endogenous ␥-tubulin was examined by immunofluorescence staining with SADB S , FLAG, or HA (green) and ␥-tubulin (red). Nuclei were detected using DAPI (blue) in transfected human U2OS and mouse NIH3T3 cells (n ϭ 5). Fluorescence intensity of endogenous nuclear ␥-tubulin staining in cells expressing FLAG-SadB S or HA-SADB S was quantified relative to control cells. Scale bars, 10 m. G, flow cytometry was performed to determine DNA content in control NIH3T3 or U2OS cells transfected with FLAG-SadB S (SadB S ) or HA-SADB S (SADB S ), as indicated. The percentage of S phase cells is shown in each panel (n ϭ 3). and increased Ser(P) 385 -␥-tubulin levels coincided with a rise in nuclear ␥-tubulin and E2F1 proteins (Fig. 4A). In contrast, decreased SADB levels reduced the transient increase in nuclear ␥-tubulin and Ser(P) 385 -␥-tubulin and also delayed S-phase entry (Fig. 4C) (2). Finally, immunofluorescence analysis of S-phase-synchronized U2OS cells (Fig. 1A) showed that Ser(P) 385 -␥-tubulin localized to centrosomes ( Fig. 4D; 10 Ϯ 2%; n ϭ 3) and chromatin ( Fig. 4E; 23 Ϯ 3%; n ϭ 3). Together, our data indicate that the transient increase in nuclear ␥-tubulin during early S phase is caused by SadB-mediated phosphorylation of Ser 385 .

The Nuclear Localization of ␥-Tubulin
AUGUST 1, 2014 • VOLUME 289 • NUMBER 31 tions were detectable at 14 h (Fig. 4, A and F), implying a possible interrelationship. In support of this view, in vitro phosphorylation in Ser 385 was altered in an Ala 131 -␥-tubulin and Asp 131 -␥-tubulin mutants (Fig. 3, E and F), and endogenous Ser(P) 385 -␥-tubulin had a centrosomal localization (Fig. 4D). Based on our results, we postulate that phosphorylation of Ser 131 (1, 6) liberates ␥TURCs from ␣␤-tubulin dimers (2) and in this way allows the transient phosphorylation of Ser 385 -␥tubulin (Fig. 4, A and C).   (n ϭ 3). Numbers on the Western blot indicate variations on ␥-tubulin expression relative to control. To adjust for differences in protein loading, the protein concentration of ␥-tubulin was determined by its ratio with histone for each treatment. Bottom, graph illustrates the mean value of the Ser(P) 385 -␥-tubulin signal in chromatin fractions from control shRNA-or SadB shRNA-transfected NIH3T3 cells (mean Ϯ S.E.; n ϭ 5). D and E, localization of endogenous Ser(P) 385 -␥-tubulin was examined by immunofluorescence staining with anti-Ser(P) 385 -␥-tubulin (green; rabbit) and anti-␥tubulin (red; mouse), and nuclei were detected using DAPI (blue) in human S phase-synchronized U2OS cells that were released for 5 h. D, a U2OS cell containing higher Ser(P) 385 -␥-tubulin levels in the centrosomes. E, a U2OS cell containing higher nuclear levels of Ser(P) 385 -␥-tubulin and nuclear ␥-tubulin. Arrows show the location of centrosomes (n ϭ 3). Inset, higher magnification. Scale bars, 10 m. F, top, Ser(P) 131 -␥-tubulin cell content of a cell population treated as in A was examined by immunoprecipitation (IP) of Ser(P) 131 -␥-tubulin from synchronous NIH3T3 cell extracts, followed by Western blot (WB) with ␥-tubulin Ab. Total ␥-tubulin levels were examined in cell extracts by immunoprecipitation with anti-␥-tubulin (n ϭ 3). Bottom, flow cytometry was performed to determine DNA content in NIH3T3 cells. The percentage of S phase cells is shown in each panel (n ϭ 3).
To identify the underlying cause of the low expression levels of Asp 385 -␥-tubGFP, we analyzed the effect of a proteosomal inhibitor, MG132, on the expression of the mutant protein. In comparison with the ␥-tubGFP expression, the expression levels of Asp 385 -␥-tubGFP were not affected by MG132 (Fig. 6A), implying that premature proteosomal degradation is not the cause of the low expression. To further elucidate the reason of the low protein expression of Asp 385 -␥-tubGFP, we evaluated the effect of the various ␥-tubGFP mutants on the activity of the nuclear ␥-tubulin downstream target E2F1 by performing an assay using luciferase reporter plasmids containing E2F-binding sites (19). The luciferase activity measured in U2OS cells transfected with E2F1 (18) was reduced upon increased levels of ␥-tubGFP or of the various Ser 385 -␥-tubGFP mutants, but both Asp 385 -␥-tubGFP and Ala 385 -␥-tubGFP exhibited a stronger moderating effect on E2F1 activity than wild-type ␥-tubGFP (Fig. 6B). Considering these findings, we hypothesized that Ser 385 -␥-tubulin may have a regulatory impact on transcription depending on its phosphorylation status.
To identify functional difference between the various Ser 385 -␥-tubulin mutants, an mRNA expression array was performed on control GFP-transfected and the various ␥-tubGFP-mutanttransfected cells, and the impact of ␥-tubulin on RNA expression was examined. By comparison with control cells (GFPtransfected), expression of Asp 385 -␥-tubGFP correlated with the up-regulated expression of 162 genes, but the expression of 251 genes was down-regulated ( Fig. 6C; p Ͻ 1 ϫ 10 Ϫ3 ) in an expression pattern that differed from the expression profile found in cells expressing ␥-tubGFP or Ala 385 -␥-tubGFP (Fig.  6C). However, we identified the two ␥-tubulin isoforms, TUBG1 and TUBG2 (25), among the Asp 385 -␥-tubGFP repressed genes, suggesting that the observed low expression levels of Asp 385 -␥-tubGFP may depend on a transcriptional feedback mechanism.
To further visualize the effect of the various mutants on known ␥-tubulin downstream targets (8), the expression of the 20 most differentially regulated E2F-controlled genes upon decreased expression levels of ␥-tubulin (8) were represented in a heat map (Fig. 6D). Indeed, the expressions of genes such as RB1, DUSP2, and RBM38 are affected by the phosphorylation levels of Ser 385 -␥-tubulin, demonstrating that the various recombinant ␥-tubulin proteins alter gene expression differently (Fig. 6, C and D).
Ala 385 -␥-Tubulin and Asp 385 -␥-Tubulin Affect Cell Cycle Progression-The level of chromatin-bound ␥-tubulin is lowest during G 1 and mitosis (Fig. 1, A and B) (8). Consequently, the decrease in chromatin-bound ␥-tubulin upon expression of the Ala 385 -␥-tubGFP mutant may cause accumulation of cells in the G 1 phase. Western blot analysis of U2OS cells expressing Ala 385 -␥-tubulin revealed increased cytosolic levels of the G 1 -S markers cyclin E and cyclin A, whereas there were low or undetectable amounts of the G 2 -M markers cyclin B and phosphorylated histone H1 (Fig. 5A). Immunofluorescence analysis showed that localization of WT-␥-tubGFP and Ala 385 -␥-tubGFP differed, being clearly nuclear for the former but mostly cytosolic for the latter. Moreover, chromatin-bound WT-␥-tubGFP accumulated in defined nuclear sites (Fig. 5B). Despite the low expression of Asp 385 -␥-tubGFP, the phosphomimetic mutant increased the chromatin-bound levels of endogenous ␥-tubulin and of the G 2 -M markers phosphorylated histone H1 and cyclin B (Fig. 5A). Together, these findings support the involvement of Ser 385 -␥-tubulin in regulation of cell cycle progression (8).
To further characterize the involvement of Ser 385 -␥-tubulin in cell division, we examined the cell cycle profile of a population of U2OS cells expressing the various ␥-tubGFP mutants (Fig. 7A). Compared with cells expressing GFP, those express-  Fig. 2E. Anti-GFP antibody immunoprecipitates of cell lysates prepared as described in Fig. 2E but expressing Asp 385 -␥-tubGFP were analyzed by WB. Arrows and arrowheads indicate the GFP and endogenous ␥-tubulin, respectively. B, the cellular location of ␥-tubGFP, Ala 385 -␥-tubGFP (S385AGFP), and Asp 385 -␥-tubGFP (S385DGFP) was determined by immunofluorescence analysis, and nuclei were detected using DAPI (blue); scale bars, 10 m.
To determine the effect of Ser 385 in the location of the nuclear ␥-tubulin C terminus, we transiently expressed Ser 385 -C␥-tubGFP, Ala 385 -C␥-tubGFP, or Asp 385 -C␥-tubGFP in NIH3T3 cells (8). Immunofluorescence analysis showed a constitutive nuclear localization of these mutants (Fig. 7B), which supports the idea that the N terminus of ␥-tubulin masks the NLS. However, Ala 385 -C␥-tubGFP and Ser 385 -C␥-tubGFP localized to the nucleus, the centrosomes, and the microtubules, but Asp 385 -C␥-tubGFP was mostly in the nucleus and in the centrosomes (Fig. 7B). Although the presence of Asp 385 -C␥-tubGFP was detected in single cells by immunofluorescence, neither anti-GFP nor anti-␥-tubulin antibodies detected the C-terminally tagged GFP Asp 385 -C␥-tubulin mutant by immunostaining or Western blotting analysis (Fig. 7, C and D).
Our findings indicate that the conformation of nuclear ␥-tubulin differs from the cytosolic pool, in a similar way as described previously for nuclear actin (26).
Moreover, we found that the total cytosolic amount of C␥-tubGFP increased by 16% when the Ala 385 -C␥-tubGFP mutant was expressed (Fig. 7C), altogether suggesting that the cellular phosphorylation levels of Ser 385 -␥-tubulin play a role in determining the cellular localization of ␥-tubulin.

Asp 385 -C␥-tubGFP Is Not Associated with Microtubule
Components-In an attempt to achieve equal expression of the various C␥-tubGFP mutants, we tested various transfections protocols. Upon simultaneous transfection and presynchronization of U2OS cells with thymidine (22), the various C␥-tubGFP mutants were more evenly expressed (Figs. 7C and 8A). Western blot analysis of GFP immunoprecipitates with anti-␥tubulin or -GFP antibodies detected two distinct bands. The expected relative molecular mass of C␥-tubGFP is 43,200 Da, but the observed protein size varied from 43 to 60 kDa. Both Ser 385 -C␥-tubGFP and Ala 385 -C␥-tubGFP mutants were detected as a 43 and a 60 kDa band, whereas the phosphomimetic mutant, Asp 385 -C␥-tubGFP, was only detected as a single 60 kDa band (Fig. 8A). In addition, analysis of the various C␥-tubGFP immunoprecipitates disclosed a Ser 385 -dependent association with ␣-tubulin and GCP2 (Fig. 8A), which provides  3). B, after the double thymidine block treatment, U2OS cells expressing the indicated constructs were released for 9 h. Localization of the C␥-tubGFP mutants was examined by immunofluorescence staining with anti-␣-tubulin (␣Tub; red) and nuclei DAPI staining (blue) in human U2OS cells incubated for 9 h (n ϭ 3-5). Scale bars, 10 m. C, differential interference contrast (DIC)/fluorescence images of time lapse from a U2OS cell with chromatin-bound Asp 385 -C␥-tubGFP that arrests in metaphase. Images were collected every 2 min. The image series shows chosen frames of the location of Asp 385 -C␥-tubGFP (n ϭ 14). D, after double thymidine block treatment, U2OS cells expressing Asp 385 -C␥-tubGFP (D385; green) were released for 24 h before being fixed. Nuclei were stained with DAPI (blue). Images show a representative dead U2OS cell that expresses Asp 385 -C␥-tubGFP (n ϭ 4). a potential explanation for the observed location of Ser 385 -C␥-tubGFP and Ala 385 -C␥-tubGFP in microtubules (Fig. 7B). In contrast, the phosphomimetic mutant, Asp 385 -C␥-tubGFP, neither formed tubular structures nor associated with microtubule components (Fig. 7B), indicating that Ser 385 regulates the binding of ␥-tubulin to microtubules. However, both Ser 385 -C␥-tubGFP and Ala 385 -C␥-tubGFP are found in the nuclear compartment and are detected by Western blotting as a 60 kDa band, indicating that the conformation state of Asp 385 -C␥-tub-GFP can transiently be induced by environmental factors (27).
To ascertain whether the accumulation of cells in G 2 /M (Fig.  7E) is a consequence of mitotic failure due to an aberrant association of the C␥-tubGFP mutants with mitotic chromosomes, we analyzed the effect of constitutive chromatin-bound ␥-tubulin in mitotic cells by immunofluorescence studies (Fig. 8B). We found that upon mitosis entry, the Ser 385 -C␥-tubGFP and the Ala 385 -C␥-tubGFP mutant proteins were detached from the chromatin, whereas the Asp 385 -C␥-tubGFP remained chromatin-bound during metaphase (Fig. 8, B and C); only cells with non-chromatin-bound Asp 385 -C␥-tubGFP progressed through mitosis (supplemental Videos S1 and S2). However, cells with chromatin-bound Asp 385 -C␥-tubGFP were unable to divide. The cells remained in metaphase during several h, and with time, the amount of chromatin-bound Asp 385 -C␥-tubGFP decreased, and cytosolic aggregates were formed. Finally, cells formed blebs and died (Fig. 8, C and D; 28 Ϯ 4%, n ϭ 4). Mitotic cells expressing the Ser 385 -C␥-tubGFP divided normally (supplemental Videos S3 and S4). Together, these results demonstrate that the cellular localization of ␥-tubulin controls mitotic progression.

DISCUSSION
Knowledge concerning the nuclear function of ␥-tubulin is still limited, and one of the remaining questions is how cytosolic ␥-tubulin becomes nuclear. To address this issue, we studied the presence of ␥TuRC components GCP2 and GCP3 and microtubules in different cellular compartments. In mammalian cell lines, we find neither GCP2, GCP3, nor ␣-tubulin in the nuclear fractions of the studied cells, suggesting a localizationdependent association of ␥-tubulin with microtubules and ␥TuRC components. Moreover, we aimed to identify the mechanism responsible for import of ␥-tubulin from the cytosol to the nucleus. Here, we found that the known microtubule (20) and centrosome (2) regulator, SadB, phosphorylates a serine residue near the ␥-tubulin NLS (8). Indeed, despite the differences in the C-terminal regions of SadB S and SadB L (2) or in their cellular location, SADB L and SadB S phosphorylate ␥-tubulin Ser 385 and thereby govern the size of the nuclear pool of ␥-tubulin. In support of that finding, overexpression of phosphomimetic mutants or SadB kinases increased the nuclear ␥-tubulin pool. A decrease in SADB/SadB levels or expression of the non-phosphorylatable mutants reduced the nuclear pool of ␥-tubulin. Also, the increase in nuclear ␥-tubulin caused by overexpression of SadB S , ␥-tubulin, or the phosphomimetic mutants increased the number of cells in S phase. Accordingly, S phase entry was delayed by decreases in levels of SadB (2) or ␥-tubulin protein or expression of non-phosphorylatable mutants, and these effects were reverted upon introduction of the corresponding RNAi-resistant gene (2). Finally, in the absence of the N terminus, the Ser 385 -C␥-tubGFP and Ala 385 -C␥-tubGFP mutants localized to the centrosome, to the nucleus and to regions where microtubules are concentrated in comparison with the Asp 385 -C␥-tubGFP that localized to centrosome and nucleus. These observations and the fact that SadB kinases and the Ser 385 motif are conserved among species strongly suggest that this phosphorylation plays a central role in the nuclear localization of ␥-tubulin during cell division.
Microtubule formation requires longitudinal stabilization by nucleation onto ␥TuRC and a GTP-dependent conformational change of ␣␤-tubulin dimers (1,4,6). Although the domains and activities of ␤and ␥-tubulin GTPases are similar, the conformations of ␥-tubulin-GDP and ␥-tubulin-GTP are almost identical and resemble the curved depolymerized state of ␣␤-tubulins (1,4,6). Thus, phosphorylation of Ser 385 may facilitate a conformational change that unmasks ␥-tubulin's NLS and releases ␥-tubulin from GCP2, GCP3, and microtubules. Several lines of evidence support this model. First, anti-total ␥-tubulin and -GFP antibodies do not recognize the chromatinbound phosphomimetic Ser 385 3 Asp ␥-tubulin mutants in immunofluorescence studies, but the antibodies detect immunoprecipitates of Asp 385 -␥-tubulin and Asp 385 -C␥-tubGFP by Western blotting. Anti-total-␥-tubulin recognizes Ser(P) 385 -␥tubulin in membranes containing cell lysates with higher concentrations of the phosphorylated protein. Furthermore, we can with an anti-Ser(P) 385 -␥-tubulin antibody detect, in a phosphatase-dependent manner, endogenous levels of Ser(P) 385 -␥tubulin or in vitro phosphorylated Ser 385 -␥-tubulin. Second, Ser(P) 385 -␥-tubulin and the ␥-tubulin C terminus undergo a size shift in SDS gels. Finally, only Ser 385 -C␥-tubGFP and Ala 385 -C␥-tubGFP form tubular structures and associate with ␣-tubulin and GCP2. Altogether, the results reported here support the existence of different ␥-tubulin conformational states that may aid ␥-tubulin to bind structurally distinct proteins and in this way provide ␥-tubulin with the observed functional properties (8, 10 -14, 27).
Regarding the mechanism by which ␥-tubulin regulates cell cycle progression, we think that the presence of nuclear ␥-tubulin at early S phase turns off the mediated gene transcription of E2Fs (8,9). Cell cycle progression is driven by the timely expression of cell cycle genes, such as cyclins (30). In most eukaryotes, there are three main waves of transcription, which coincide with the transition points G 1 -to-S, G 2 -to-M, and M-to-G 1 (30). Interference with these tran-scription waves will inevitable disturb cell cycle progression as the expression of necessary genes is impeded. The transient phosphorylation of Ser 385 at early S phase ends the first E2F-mediated transcriptional wave facilitating S phase execution (8), but the constitutive presence of nuclear Ser(P) 385 -␥-tubulin affects the mediated transcriptional waves of the following E2Fs (31) and thus cell cycle progression. This implies that the balance between chromatin-bound and microtubule-associated ␥-tubulin may form a cellular sensor for transducing cytoskeletal alterations between compartments that can directly modulate gene expression during the cell cycle.
We propose that the fluctuating activities of SadB during G 1 and S regulate the phosphorylation levels of ␥-tubulin at Ser 131 . In this way, SadB enhances ␥-tubulin polymerization at the nascent centriole and inhibits acentriolar formation of centrosomes elsewhere in the cell. However, phosphorylated Ser 131 -␥-tubulin also reduced astral microtubule nucleation at the centrosomes (2, 4), which probably facilitates the accessibility of SadB to Ser 385 at the G 1 to S phase transition. Phosphorylation of Ser 385 -␥-tubulin in the centrosomes then triggers a conformational change in ␥-tubulin that releases this protein from GCP2, GCP3, and microtubules. In the nucleus, ␥-tubulin puts an end to the activities of E2Fs, ensuring that the centrosomes and the chromosomes are replicated only once per cell cycle (2,8,16,30).
Together, our data indicate that the transient increase in nuclear ␥-tubulin during S phase is caused by SadB-mediated phosphorylation of Ser 385 , which induces a conformational change in ␥-tubulin that leads to its nuclear accumulation during S phase. This identifies SadB as a multifunctional cell cycle regulator that triggers centrosome replication and S phase progression in mammals by controlling phosphorylation of Ser 131 (2) and Ser 385 in ␥-tubulin.