Centrosome separation driven by actin-microfilaments during mitosis is mediated by centrosome-associated tyrosine-phosphorylated cortactin

Cortactin is an actin-filament-binding protein (Olazabal and Machesky, 2001; Weed and Parsons, 2001). It has initially been described as a tyrosine-phosphorylated protein in cells infected with v-Src and is a direct substrate of Src kinase (Kanner et al., 1990; Wu et al., 1991). Its association with F-actin stabilizes the microfilament network (Huang et al., 1997; Olazabal and Machesky, 2001; Pant et al., 2006). Of the all the different domains in cortactin, the N-terminal acidic region, the F-actin-binding repeats, the helix domain and the SH3 domain at the C-terminus are relatively conserved from invertebrates to vertebrates (Weed and Parsons, 2001). The proline-rich domain between the helix domain and C-terminal SH3 domain, however, is only conserved in vertebrates, but not between vertebrates and invertebrates (Weed and Parsons, 2001). The proline-rich domain of Drosophila melanogaster has only 15% homology with its counterpart in mouse. The proline-rich domain of murine cortactin has six identified phosphorylation sites (four tyrosine sites and two serine sites) (Huang et al., 1997; Head et al., 2003; Lua and Low, 2005; Martin et al., 2006). All four identified tyrosine phosphorylation sites are missing in Drosophila cortactin (Genebank: AAF55840). The current knowledge regarding the function of cortactin and its tyrosine phosphorylation mostly concerns actin-network organization, regulation of cell motility and formation of lamellipodia and membrane ruffles (Lua and Low, 2005).

During mitosis the complex sequence of events is dominated by the formation and activity of the microtubule spindle. The initiation of the microtubule spindle is governed by the centrosome, also known as the microtubule-spindle-organization center (Bornens, 2002; Rieder et al., 2001; Kellogg et al., 1994; Joshi, 1993; Mishima et al., 2004). Consequently, the separation and positioning of centrosomes determines the formation and orientation of spindles. The centrosome cycle progresses in a precise order: (1) the centrosome is duplicated at G1-S transition, (2) cohesion between two duplicated centrosomes is broken in G2 phase, (3) two centrosomes migrate apart in G2-M transition and the separated centrosomes initiate the spindle microtubules to form bipolar spindle in early mitotic phase (Meraldi and Nigg, 2002; Tsou and Stearns, 2006). Therefore, the separation and positioning of centrosomes are crucial for the correct formation of a bipolar spindle, which governs the chromosome segregation and the fate of daughter cells.

In order to drive the centrosomes to opposite ends of nucleus, motion force needs to be generated. It has been reported that actin microfilaments and myosin II are required for centrosome separation and positioning during mitosis, thus making them ideal candidates for providing the migration force (Rosenblatt et al., 2004; Burakov et al., 2003; Waters et al., 1993). When F-actin is depolymerized by addition of latrunculin, centrosome separation is blocked and cells enter mitosis with unseparated centrosomes (Uzbekov et al., 2002). In HeLa cells, one major pathway for centrosome separation that occurs prior to nuclear-envelope breakdown is dependent on an intact actin cytoskeleton (Whitehead et al., 1996). These studies indicate or imply that actin-filaments provide the essential driving force for centrosome separation, which means that a regulatory attachment of actin-filaments to centrosome is required at G2-M transition. Here we show that cortactin phosphorylation at tyrosine residues is induced in mitotic cells and phospho-cortactin (P-cortactin) is associated with centrosomes or spindle poles at G2-M transition until mitotic anaphase. We found that the cortactin C-terminal region, which can bind to centrosomes but does not mediate the F-actin association, inhibited the centrosome separation prior to nuclear envelope breakdown. We also found P-cortactin to be anchored at the tips of actin fibers in interphase cell. These observations indicate that centrosome or spindle-pole-associated P-cortactin provide the anchor on the centrosome or spindle pole for F-actin attachment.

In our study of hormone-induced 3T3-L1 adipocyte differentiation and mitotic clonal expansion (Jin et al., 2000), an accumulated tyrosine-phosphorylated protein was identified as P-Tyr-cortactin (data not shown). In hydroxylurea synchronized 3T3-L1 cells, we observed cell-cycle-dependent phosphorylation of cortactin at Tyr421 (Fig. 1A,B). Levels of P-Tyr-cortactin were low in interphase, G1-S transition or S phase cells. However, it was dramatically increased after cells reached G2 phase. The P-cortactin was then dephosphorylated when the cells finished mitosis and returned to G1 phase (Fig. 1B). Cortactin protein levels were constant throughout the cell cycle (Fig. 1B). Similar phosphorylation and dephosphorylation on Tyr466 and Tyr482 was also observed (data not shown).

In the mitotic phase, the spindle structure is predominantly formed by microtubules, whereas no distinct structure is formed by actin-filaments. As an F-actin-associated protein, cortactin was evenly distributed in cytoplasm of mitotic cells (Fig. 2A and supplementary material Fig. S1). However, P-Tyr-cortactin was almost exclusively located at the mitotic spindle poles, identified by staining for γ-tubulin (Fig. 2B) (Joshi, 1993). Individual antibodies against cortactin phosphorylated at Tyr421 (P-Tyr421-cortactin), Tyr466 (P-Tyr466-cortactin) or Tyr482 (P-Tyr466-cortactin) (anti-P-Tyr421-cortactin, anti-P-Tyr466-cortactin or anti-P-Tyr482-cortactin antibodies, respectively) were used to stain spindle poles (Fig. 2B). When mitotic spindles were revealed using anti-α-tubulin antibody, P-Tyr-cortactin was found to be located exactly at the spindle poles (Fig. 2C). P-cortactin, which is associated with the spindle poles, was phosphorylated at least on one of the three Src kinase sites, at Tyr421, Tyr466 or Tyr482 (Fig. 2B) (Kanner et al., 1990; Wu et al., 1991; Huang et al., 1997). However, this association did not enrich the spindle poles with just phosphorylated tyrosine residues (Fig. 2D). The spindle poles were distinct for P-Tyr-cortactin (Fig. 2B,C). Even the overexposed immunofluorescence images show that P-Tyr-cortactin was only detected at spindle poles (supplementary material Fig. S1).

Besides the staining for P-Tyr-cortactin at centrosomes and spindle poles, we detected an interaction between P-Tyr-cortactin and the centrosomal protein γ-tubulin (Fig. 1C). Anti-P-Tyr421-cortactin antibody immunoprecipitated γ-tubulin in mitotic cells and anti-γ-tubulin antibody immunoprecipitated P-Tyr-cortactin (Fig. 1C). In addition, centrosomal fraction isolated by sucrose-density gradient from nocodazole-arrested mitotic cells contained P-Tyr-cortactin (Fig. 1D) (Moudjou and Bornens, 1994). Furthermore, in cells expressing eGFP-tagged cortactin, GFP-cortactin fusion protein was visualized at the spindle pole (supplementary material Fig. S1).

The detection of both cortactin and P-Tyr-cortactin in isolated centrosomes (Fig. 1C,D) and at centrosomes or spindle poles (Fig. 2 and supplementary material Fig. S1) supported the conclusion that it is the tyrosine-phosphorylated form of cortactin, revealed by anti-P-cortactin antibodies in immunofluorescence staining. To further verify the specificity of anti-P-cortactin antibodies for P-Tyr-cortactin, four identified tyrosine phosphorylation sites were mutated to phenylalanine (Tyr421Phe, Tyr466Phe, Tyr475Phe and Tyr482Phe) in murine cortactin (Huang et al., 1997; Head et al., 2003; Martin et al., 2006) (supplementary material Fig. S2). Anti-P-Tyr421-cortactin and anti-P-Tyr466-cortactin antibodies (both from Sigma) were able to detect tyrosine-phosphorylated exogenous wild-type cortactin fused to GFP (GFP-cortactin) as well as endogenous cortactin (supplementary material Fig. S2). Exogenous mutated GFP-cortactin fusion protein containing the four Phe substitutions was not detected by either antibody. Most importantly, tyrosine phosphorylation on exogenous GFP-cortactin containing only Tyr421 was detected by anti-P-Tyr421-cortactin antibody but not anti-P-Tyr466-cortactin antibody, whereas tyrosine phosphorylation on exogenous GFP-cortactin containing only Tyr466 was detected by anti-P-Tyr466-cortactin antibody but not anti-P-Tyr421-cortactin antibody (supplementary material Fig. S2). These results unequivocally established the specificity of anti-P-Tyr421-cortactin and anti-P-Tyr466-cortactin antibodies for their respective P-Tyr-cortactin. Moreover, two additional antibodies against P-Tyr421-cortactin and P-Tyr466-cortactin from BioSource, which have been characterized (Head et al., 2003), were used for immunofluorescence staining and compared with the antibodies in our current study. As shown in supplementary material Fig. S2, the use of antibodies from either supplier (Sigma nad BioSource) resulted in identical staining of centrosomes or spindle poles. Our observation that GFP-cortactin localizes at spindle poles together with the confirmed antibody specificity (as characterized by the mutated cortactins; supplementary material Figs S1, S2), it seemed extremely unlikely that the described P-Tyr-cortactin immunostaining found at centrosomes or spindle poles was due to non-specific staining of other proteins. In fact, no centrosome-associated P-cortactin was observed by immunofluorescence in 3T3-L1 cells whose cortactin expression had been suppressed (supplementary material Fig. S3). This observation clearly points against non-specific centrosomal staining when using anti-P-cortactin antibodies in immunofluorescence analysis.

The centrosome has its own duplication cycle: at G1-S transition one centrosome is duplicated into two centrosomes; from S to early G2 phase the duplicated centrosomes are cohered; at G2-M transition the cohesion between two centrosomes is broken and two centrosomes migrate apart to form the spindle poles at opposite ends of nucleus (Meraldi and Nigg, 2002; Tsou and Stearns, 2006; Mishima et al., 2004). Accordingly, there was little P-Tyr-cortactin in G1-phase, G1-S-transition and S-phase cells (Fig. 1B). Neither was centrosome(s) in G1 and S phase associated by P-cortactin (Fig. 3A,B). At G2-M transition, when chromatin was visibly condensed, the nuclear envelope was still intact and centrosomes had not separated, P-cortactin was already associated with centrosomes (Fig. 3B). From G2-M transition to metaphase the P-cortactin association was increased as centrosomes were separated and spindles were formed (Fig. 3A,B). When spindles gradually disintegrated during the transition from anaphase to telophase, the association of P-cortactin with spindle poles was also diminished (Fig. 3A,B). Dissociation of P-cortactin from telophase centrosomes correlated with the dephosphorylation of P-cortactin after mitosis (Fig. 1B). The same association and dissociation for P-Tyr466-cortactin and P-Tyr482-cortactin was also observed during mitosis in 3T3-L1 cell (data not shown).

Of four tyrosine phosphorylation sites (Tyr421, Tyr466, Tyr475, Tyr482) in murine cortactin (Kanner et al., 1990; Wu et al., 1991; Huang et al., 1997), we detected phosphorylation at all three Src kinase sites (Tyr421, Tyr466, Tyr482) in centrosome-associated P-cortactin (Fig. 2B). The tyrosine phosphorylation pattern that determines centrosomal association for P-cortactin is currently under investigation. Src kinase is activated by autophosphorylation on Tyr418 (Roskoski, Jr, 2004). It has been described previously that, during mitosis, Src kinase is activated and associated with centrosomes (David-Pfeuty and Nouvian-Dooghe, 1990; David-Pfeuty et al., 1993; Kuga et al., 2007). We found also that activated Src kinase is dynamically associated with mitotic spindle poles during mitosis in 3T3-L1 cells (supplementary material Fig. S3). The association of P-cortactin with centrosomes or spindle poles correlated closely with the association of activated Src kinase (supplementary material Fig. S3). This suggested that regulation of tyrosine phosphorylation of cortactin occurs on the centrosomes or spindle poles. In the spindle pole, the associated P-cortactin was located at the centrosome but not in the peripheral α-tubulin mass (Fig. 3C). As revealed by eGFP-tagged centrin 2, P-cortactin within the centrosome was in the pericentriolar material but not associated directly to the pair of centrioles (Fig. 3D).

Centrosome separation prior to nuclear envelope breakdown is driven by actin-filaments but not microtubules (Rosenblatt et al., 2004; Burakov et al., 2003; Uzbekov et al., 2002; Whitehead et al., 1996; Waters et al., 1993). Since cortactin is an F-actin-associated protein and its tyrosine phosphorylated form was associated with centrosomes prior to their separation (Fig. 3B), the function of centrosome-associated P-cortactin in centrosome separation driven by actin-filaments was examined. HeLa cells were arrested at G1-S transition by thymidine (Fig. 4A). By 8 hours after thymidine release, the synchronized cells progressed through S phase and reached G2 phase (Fig. 4A). During this time the centrosomes were not separated and the average distance between two cohesive centrosomes was less than 2 μm (Fig. 4C). Ten hours after thymidine release cells progressed into mitosis and centrosome separation was observed (Fig. 4A). The average distance between two separated centrosomes in the synchronized cell population was ∼10 μm (Fig. 4C). The centrosome separation process in a synchronized cell population can be better illustrated by plotting the centrosome-separation ratio (distance between two centrosomes vs nuclear diameter) for individual cells (Fig. 4B,D). Before reaching G2-M transition (4-8 hours after thymidine release), the centrosome-separation ratio was less than 0.1 in most cells (Fig. 4D). At G2-M transition (10 hours after thymidine release), centrosome separation was clearly visible and the centrosome-separation ratio was shifted towards 1.0 (Fig. 4A,D).

However, in the presence of cytochalasin B, which depolymerizes actin-filaments, centrosome separation and mitotic entry were delayed or blocked (Fig. 4A, Fig. 5A). Cells progressed normally through S phase to G2 phase in the presence of cytochalasin B (4-8 hours after thymidine release), but could not reach G2-M transition at 10 hours after thymidine release. The centrosome separation in the presence of cytochalasin B was blocked as most cells in the synchronized cell population had low a centrosome-separation ratio (Fig. 4D). F-actin depolymerization appeared to inhibit centrosome separation prior to nuclear envelope breakdown. The same results were also obtained with the treatment of latrunculin A, another F-actin depolymerization reagent (data not shown).

Interestingly, the cohesion between two duplicated centrosomes was broken when cells were treated with nocodazole to depolymerize microtubules (Fig. 4A). In the presence of nocodazole, two centrosomes were randomly separated even through thymidine-synchronized cells were still in S phase (Fig. 4A,C,D). Without the restriction from intact microtubule system the centrosome separation appeared to be a default process, because depolymerization of microtubules by using nocodazole or of both microtubules and microfilaments by using nocodazole and cytochalasin B induced centrosome separation at any cell-cycle phases (Fig. 4A,C,D). Microtubules appeared to provide the cohesion between two duplicated centrosomes and restrict the premature centrosome separation.

Actin-filament depolymerization by cytochalasin B not only blocked centrosome separation, but also inhibited mitotic entry in HeLa cells (Fig. 5A). Chromatin condensation signals the mitotic entry and can be indicated by increased phosphorylation of histone 3 (Ajiro et al., 1996). Histone 3 phosphorylation was clearly blocked by treatment with cytochalasin B (Fig. 5A). Twelve hours after thymidine release, control HeLa cells completed the mitosis as cyclin B1, which is highly expressed in G2-M cells and degraded after the completion of mitosis (Smits and Medema, 2001), started to decrease (Fig. 5A). However, cyclin B1 levels in cytochalasin-B-treated cells remained high 12 hours after thymidine release (Fig. 5A). P38, which is activated at spindle-assembly check point (Takenaka et al., 1998), was less activated in cytochalasin-B-treated cells (Fig. 5A). Thus, actin-filaments depolymerization inhibited the mitotic entry at stage of centrosome separation and chromatin condensation prior to spindle assembly.

In the temporal order, cortactin phosphorylation was induced in G2 phase and concomitantly associated with centrosome (Fig. 1B, Fig. 3B and Fig. 5B). It might be phosphorylated by activated Src kinase at the centrosomes (supplementary material Fig. S3). The centrosomes isolated from HeLa cells that were blocked at G1-S transition by thymidine contained no P-cortactin, whereas centrosomes isolated from mitotic HeLa cells showed association with P-cortactin (Fig. 5C). Since the biochemical isolation of centrosomes could not completely eliminate other cellular organelles, the association of P-cortactin with cohesive centrosomes and separated centrosomes was further analyzed in immunofluorescence studies using anti-P-cortactin antibodies. The average distance between two cohesive centrosomes was 1.8 μm (Fig. 4C). If P-cortactin mediates the attachment of actin-filaments that pull centrosomes apart, most centrosomes that are separated by more than 1.8 μm should be associated with P-cortactin. As shown in Fig. 5D, there is a clear correlation between P-cortactin association and centrosome separation. Almost all separated centrosomes were associated with P-cortactin, whereas two-thirds of the cohered centrosomes were not associated with P-cortactin (Fig. 5D).

The Tyr phosphorylation sites in cortactin are located at the C-terminal proline-rich domain. We therefore created a truncated version of the cortactin C-terminus lacking its F-actin-binding repeats (Fig. 6A). Thus, the phosphorylated cortactin C-terminus can still bind to centrosomes albeit without providing binding for F-actin. By competing-off wild-type P-cortactin from centrosomes, the truncated cortactin C-terminus should have a dominant-negative effect on P-cortactin-mediated centrosome separation (Fig. 6C). Centrosomes labeled with the eGFP-tagged cortactin C-terminus were observed in HeLa cells expressing the fusion protein (Fig. 6B), and centrosome separation was clearly also inhibited (Fig. 6D,E).

Since the association of cortactin with centrosome depends on tyrosine phosphorylation, the substitution of tyrosine residues for phenylalanine prevents interaction of the cortactin C-terminus with the centrosome and should abolish its dominant-negative effect on centrosome separation (Fig. 6C). In HeLa cell expressing the eGFP-tagged mutated cortactin C-terminus, no association with centrosomes was observed (Fig. 6B), and centrosome separation progressed normally when they were thymidine-synchronized (Fig. 6D,E). Therefore, only the wild-type cortactin C-terminus but not that containing mutated Tyr residues has the ability to inhibit centrosome separation.

The dominant-negative effect of the cortactin C-terminus on centrosome separation indicated that the F-actin-binding capacity of cortactin is essential for P-cortactin in centrosome separation (Fig. 6D,E), because P-cortactin associated with a centrosome might serve as the binding site for actin-filaments in order to pull centrosomes. In COS-7 or NIH3T3 cells, actin stress fibers can be readily stained by phalloidin and visualized (Fig. 7 and supplementary material Fig. S5). Cortactin per se did not exhibit any distinct pattern of actin fiber association (Fig. 7A,C). However, P-cortactin was exclusively present at the ends of actin fibers (Fig. 7A,C). P-cortactin anchored the actin fibers to the focal adhesion points, which were revealed by paxillin staining (Fig. 7B). All three Src kinase sites in cortactin (Tyr421, Tyr466, Tyr482) were phosphorylated in the P-cortactin that had associated with focal adhesion points (Fig. 7B). Results from cortactin knockdown cells, Src-kinase-transfected cells and cells expressing GFP-cortactin fusion protein provided evidence for the specificity of P-cortactin at focal adhesion points (supplementary material Fig. S4).

During mitosis, P-cortactin was no longer associated with paxillin because focal adhesion points were most probably disintegrated (Fig. 7D). Instead, most P-cortactin was localized at the spindle poles (Fig. 7D). In fact, the spindle-pole-associated P-cortactin was so concentrated that the suitable microscopic exposure time for P-cortactin at spindle poles was not enough to reveal its presence at actin fiber tips (Fig. 7D and supplementary material Fig. S5). Although no direct attachment of visible actin fibers to mitotic centrosomes or spindle poles was visualized – because of the disintegration of actin stress fibers in mitotic phases – we detected interaction between P-Tyr-cortactin and actin by immunoprecipitation, and the centrosome or spindle pole was the most prominent site of P-cortactin in mitotic cells (Fig. 7D,E). Inhibition of centrosome separation by the wild-type cortactin C-terminus – which can associate with centrosomes but not with F-actin – clearly supports the role of P-cortactin as an anchor for actin-filaments in order to pull centrosomes apart.

Owing to the high visibility of the microtubule spindle during mitosis, the function of centrosome in spindle formation is well elucidated (Bornens, 2002; Rieder et al., 2001; Kellogg et al., 1994; Joshi, 1993; Mishima et al., 2004). However, much less studied is how the duplicated centrosomes migrate and position themselves on opposite sides of the nucleus before the microtubule spindle is assembled. In contrast to the hypothesis that microtubules are extended between two centrosomes to push them apart, our results from the nocodazole treatment suggests that the microtubules stabilize the cohesion between two duplicated centrosomes and prevent premature centrosome separation (Fig. 4). Other research has also found that the movement of centrosome separation is not owing to the interaction of microtubules from opposing asters (Waters et al., 1993).

That actin-filaments drive centrosome separation before spindle assembly has been suggested by other researchers and is supported by our current study (Rosenblatt et al., 2004; Burakov et al., 2003; Uzbekov et al., 2002; Whitehead et al., 1996; Waters et al., 1993) (Fig. 4). Since the association of P-Tyr-cortactin with centrosomes precedes centrosome separation (Fig. 5) and actin fibers anchor on P-Tyr-cortactin (Fig. 7), actin-filaments might attach to P-cortactin on centrosomes and then provide the force to drive centrosomes apart. Inhibition of centrosome separation by the wild-type cortactin C-terminus – which can associate with centrosomes but cannot bind to actin-filaments – strongly supports the idea that P-cortactin anchors actin-filaments to centrosomes (Fig. 6). The accumulation of P-Tyr-cortactin on centrosomes attracts actin-filaments during mitosis (Figs 2,3). Based on these observations, we conclude that the actin-filaments attached to P-cortactin on unseparated centrosomes pull the centrosomes apart before the nuclear envelope breaks down (Fig. 8).

Centrosome separation happens at G2-M transition and, presumably, the exact temporal regulation of F-actin attachment to the centrosome is crucial for mitotic progression. Association of P-Tyr-cortactin but not cortactin per se with the centrosome meets this prerequisite for the regulated F-actin attachment. Both tyrosine phosphorylation of cortactin and the association of P-cortactin with the centrosome starts at G2-M transition, just before the centrosomes separate (Figs 1, 3, 5).

All three identified Src kinase phosphorylation sites in cortactin (Tyr421, Tyr466, Tyr482) are phosphorylated in centrosome-associated P-cortactin (Fig. 2B). Activated Src kinase was also dynamically associated with the centrosome and/or spindle poles during mitosis (supplementary material Fig. S3) (David-Pfeuty and Nouvian-Dooghe, 1990; David-Pfeuty et al., 1993; Kuga et al., 2007). Kinetically, the association and dissociation of P-Tyr-cortactin with the centrosome or spindle pole was closely correlated to that of activated Src kinase during mitosis (supplementary material Fig. S3). The role of Src kinase as a regulator of cell growth is well established. As an oncogene product, high Src kinase activity promotes cell transformation. In cancer cells, many oncogene kinases are highly activated and their regulation is disrupted. As such, the regulation of cortactin phosphorylation can also be expected to be disrupted. With abnormal cortactin phosphorylation, the incidences of abnormal centrosome separation and abnormal spindle formation would be increased, which would lead to abnormal chromosome segregation. It is well-known that polyploidy or aneuploidy is a common event in cancer cells.

The amino acid sequence of murine cortactin contains a stretch of ten Tyr residues in the C-terminal region (amino acids 421-499; positions 421, 433, 442, 460, 466, 475, 482, 485, 497, 499). These Tyr residues are relatively conversed between mouse, human and chicken, but not Drosophila or other invertebrates. There are only five Tyr residues (Tyr449, Tyr486, Tyr495, Tyr507, Tyr509) in Drosophila. However, none of those positions correspond to the Tyr phosphorylation sites identified in murine cortactin (Tyr421, Tyr466, Tyr475, Tyr482). Given the important regulatory functions for Tyr421, Tyr466 and Tyr482 phosphorylation, the phosphorylation of other tyrosine residues requires further investigation. It may well be that the combination of phosphorylation on certain tyrosine residues determines the cellular localization of P-cortactin and regulates its function in vertebrate cell division.

In conclusion, our results indicate a new mechanism and regulation of F-actin-mediated centrosome separation, as well as a new function for cortactin and the regulation of its phosphorylation. Cortactin tyrosine phosphorylation is often associated with dynamics of actin networks (Weed and Parsons, 2001; Daly, 2004; Lua and Low, 2005). Regulation of cortactin phosphorylation might be involved in the reorganization of the cytoskeleton during mitosis. As actin fibers were also found to be anchored at focal adhesion points through P-cortactin (Fig. 7), P-cortactin could be a general anchoring site for actin fibers. In addition, P-cortactin association was highest in metaphase spindle poles (Fig. 3B and supplementary material Fig. S3). Its function at mitotic spindles requires further investigation.

Anti-cortactin, anti-phosphotyrosine, anti-cyclin B1, anti-p38, anti-actin and anti-eGFP antibodies were purchased from Santa Cruz Biotechnology. Antibody against phosphorylated histone 3 was from Upstate. Anti-paxillin antibody was from Transduction Laboratories. Anti-P-Tyr482-cortactin antibody was from Chemicon. Anti-P-Tyr421-cortactin, anti-P-Tyr466-cortactin, anti-α-tubulin, anti-γ-tubulin antibodies, anti-Myc tag, horseradish-peroxidase-conjugated, FITC-conjugated and rhodamine-conjugated secondary antibodies, DAPI, phalloidin, cytochalasin B and nocodazole were from all from Sigma. Two additional anti-P-Tyr421-cortactin, anti-P-Tyr466-cortactin antibodies were from BioSource (Head et al., 2003). The Leica laser scanning confocal microsystem, including the Leica TCS SP2 confocal microscope, Leica confocal scanner and Leica confocal acquisition software were used with the HCX PL APO 1bd. BL 63.0×/1.4 oil objective at 1.4 numerical aperture at a working temperature of 22°C. The fluorescence medium used was Sigma's DABCO.

3T3-L1, HeLa or COS-7 cells were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% calf serum. To synchronize 3T3-L1 cells, they were serum-starved for 48 hours, fed with serum-containing medium supplemented with 0.12 mM hydroxylurea for another 24 hours and then released by washing the cells with fresh medium. Cells were harvested for flow cytometry and western blot analysis.

To synchronize HeLa cells, they were treated with 2 mM thymidine for 24 hours, released for 12 hours by replacing the medium and then treated again with 2 mM thymidine for 24 hours and released by replacing the medium. The released cells were treated with 5 μM cytochalasin B, 100 ng/ml nocodazole or both. At each time point (indicated in figures) the cells were harvested for flow cytometry analysis and immunofluorescence staining.

For flow cytometry analysis, cells were trypsinized from culture plates, fixed in 75% ethanol or 4% paraformaldehyde, treated with RNaseA, and stained with propidium iodide. Immunofluorescence staining, immunoprecipitation and western blot analysis were carried out following a previously described protocol (Huo et al., 2003).

3T3-L1 cells were arrested at mitotic phase by treating them with 1 μM nocodazole for 24 hours and 1 μg/ml cytochalasin B for the final 1 hour. Centrosomes were isolated by one round of 60% sucrose cushion centrifugation and the second round of 30%, 50% and 70% sucrose step-gradient centrifugation (Moudjou and Bornens, 1994). The second gradient was 5 ml and the fractions were collected from the top by 1 ml/fraction for fractions 1-3, and 0.2 ml/fraction for fractions 4-12.

For isolation of centrosomes at G1-S transition or of mitotic centrosomes from HeLa cell, cells were blocked at G1-S transition by addition of 2 mM thymidine for 24 hours or by addition of 100 ng/ml nocodazole for 24 hours. At the final hour, 1 μg/ml cytochalasin B and 100 ng/ml nocodazole were added to depolymerize the cytoskeleton.

Murine cortactin cDNA was isolated from 3T3-L1 cells. The cortactin N-terminal comprises amino acids 1-329 that contain the N-terminal acidic region and six-and-a-half actin binding repeats. The cortactin C-terminal comprises amino acids 330-546, containing the helix, proline-rich region and SH3 domain. CT(Y/F) is the cortactin C-terminus with Phe substitutions for Tyr421, Tyr466, Tyr475, Tyr482. The C-terminus was fused to GFP to yield the GFP-CT fusion protein or was Myc-tagged to yield CT-myc-tagged cortactin C-terminal protein. Thymidine-synchronized HeLa cells were transfected with cortactin C-terminal fusion protein Invitrogen Lipofectamin 2000 following manufacturer's protocol. The transfected cells were synchronized again and harvested for immunoflurescence staining 10 hours after the thymidine-release. GFP or Myc-tag immunofluorescence were visualized to identify the expression of transfected protein and γ-tubulin was stained to reveal centrosomes or spindle poles.

This work was supported by grants 90208007 and 30521005 from the China National Nature Sciences Foundation, and 2002CB513000 and 2006CB910700 from the Ministry of Sciences and Technology of China.