Nucleolin functions in nucleolus formation and chromosome congression

A layer around chromosomes during mitosis was first observed by light microscopy about 80 years ago (Sharp, 1929), and subsequent electron microscopic studies revealed a layer of closely packed fibrils and dense granules around each chromosome (Gautier et al., 1992a; Gautier et al., 1992b). The layer, designated the `chromosome periphery' (Hernandez-Verdun and Gautier, 1994), appears at prometaphase and disappears at telophase during mitosis. It contains numerous components, including nuclear and nucleolar proteins as well as several RNAs (Van Hooser et al., 2005).

Some nucleolar proteins were found to bind to the surface of chromosomes throughout mitosis, and then become incorporated into the newly formed nucleoli at the end of telophase (Azum-Gélade et al., 1994; Dundr et al., 1997). Therefore, the chromosome, and especially its periphery, has been proposed to act as a means of transport for various proteins, including nucleolar proteins. Besides its transportation function, the chromosome periphery may also play active roles in chromosome structure. As an example, the chromosome peripheral protein Ki-67 is involved in the formation of higher-order chromatin structures (Takagi et al., 1999; Kametaka et al., 2002; Scholzen et al., 2002; Saiwaki et al., 2005). In addition, it has been speculated that the chromosome periphery is involved in the progression of mitosis, possibly by participating in the protection and stabilization of the chromosomes. An example for this scenario is the chromosome peripheral protein complex Ku, which appears to protect chromosome telomeres (Bertuch and Lundblad, 2003).

In our recent proteome analysis, 107 proteins were identified in highly purified human metaphase chromosomes (Uchiyama et al., 2005). Furthermore, the observation that more than a quarter of the identified proteins were classified as chromosome peripheral proteins strongly suggests that these proteins are related to either chromosome organization or mitotic progression. Human nucleolin, a 710 amino acid protein, is one of the major and most extensively studied nucleolar proteins. An analysis of its amino acid sequence revealed the presence of three different structural domains, the N-terminal, central and C-terminal domains (Ginisty et al., 1999). The N-terminal domain is composed of highly acidic regions and contains multiple phosphorylation sites. The central domain contains four RNA-binding domains. The C-terminal domain is rich in glycine, arginine and phenylalanine residues. Nucleolin is highly conserved in many eukaryotic species. Proteins with homology to human nucleolin have been identified in Saccharomyces pombe (Gulli et al., 1995), Xenopus laevis (Caizergues-Ferrer et al., 1989), Arabidopsis thaliana (Didier and Klee, 1992) and peas (Tong et al., 1997).

The nucleolus is the site responsible for ribosome biogenesis (Hernandez-Verdun, 2005). As visualized by electron microscopy, the nucleolus consists of three main components, designated fibrillar centers (FCs), dense fibrillar components (DFCs) and granular components (GCs). The relatively opaque FCs in mammalian nucleoli are surrounded by the highly contrasted fine strands of DFCs. In turn, the DFCs are surrounded by GCs, which fill out the peripheral parts of the nucleolus. Subsequent studies revealed that the FCs harbor copies of rRNA genes, and it is well established that the initial pre-rRNA resides in the DFCs, whereas late processing events take place in the GCs. Finally, pre-ribosomal particles are translocated to the cytoplasm (Shaw and Doonan, 2005).

Nucleolin is an abundant nucleolar protein localized in the DFCs and GCs of nucleoli (Biqqioqera et al., 1990). Nucleolin plays essential roles in rDNA transcription, rRNA maturation, ribosome assembly and nucleocytoplasmic transport (Ginisty et al., 1999; Srivastava and Pollard, 1999). As a multifunctional protein, nucleolin has also been proposed to be a nuclear matrix-binding protein (Gotzmann et al., 1997), to interact with telomerase (Khurts et al., 2004) and to be involved in the regulation of apoptosis (Mi et al., 2003), intestinal cell differentiation (Turck et al., 2006) and remodeling of nucleosomes (Angelov et al., 2006). Although many functions of nucleolin have been clarified, previous studies have fallen short of demonstrating its functions during mitosis.

Nucleolin is highly phosphorylated (Olson et al., 1975; Rao et al., 1982) and its phosphorylation is highly regulated during the cell cycle (Peter et al., 1990; Morimoto et al., 2005). Extensive phosphorylation by casein kinase 2 (CK2) occurs at interphase and by CDC2 during mitosis (Peter et al., 1990), and this regulated phosphorylation of nucleolin probably regulates nucleolin functions during the cell cycle (Ginisty et al., 1999). It has been demonstrated that phosphorylation of nucleolin at interphase is correlated with active rRNA transcription (Suzuki et al., 1987). Despite a previous study describing the localization of nucleolin phosphorylated by CDC2 (Dranovsky et al., 2001), no information about its functions during mitosis has been obtained to date.

Here, we demonstrate that nucleolin localizes to the nucleoli, which are located in the peripheral region including the vicinity of the outer kinetochore of chromosomes, in interphase nuclei and during mitosis. Staining with an antibody specific for nucleolin phosphorylated by CDC2 reveals that nucleolin is also associated with the spindle poles from prometaphase to anaphase. Using the RNA interference (RNAi) technique, cells depleted of nucleolin show disorganization of the nucleoli accompanied by loss of the nucleolar proteins fibrillarin, B23 and Ki-67. The main finding of this study is disruption of chromosome congression in cells depleted of nucleolin. We show that this defect is due to improper kinetochore attachments, resulting in reduced tension, and syntelic attachments. Depletion of nucleolin also appears to affect spindle assembly.

Although the localization of nucleolin has previously been described (Azum-Gélade et al., 1994; Medina et al., 1995; Dundr et al., 1997; Gotzmann et al., 1997; Piñol-Roma, 1999), its dynamic localization during the cell cycle has not been reported. To investigate this, immunostaining for nucleolin was examined throughout the cell cycle (Fig. 1A). At interphase, nucleolin was prominently localized at the outer layer of the nucleoli, and present at a lower concentration throughout the nucleoplasm. At prophase, most of the nucleolin remained in the nucleolar remnants. When the nuclear membrane and nucleoli were disrupted at prometaphase, nucleolin began to disperse throughout the cytoplasm and localize at the chromosome periphery. At metaphase, nucleolin was localized at the chromosome periphery and slightly distributed in the cytoplasm. At anaphase, some nucleolin remained associated with the chromosome periphery, while some became packaged into cytoplasmic particles known as nucleolus-derived foci (NDFs) (Dundr et al., 1997). At telophase I (early telophase), nucleolin was still associated with chromosomes, and a fraction of it accumulated into NDFs and pre-nucleolar bodies (PNBs) (Dundr et al., 1997). The NDFs disappeared at telophase II (late telophase) and were replaced by PNBs.

Nucleolin was evenly detected in the chromosome periphery in metaphase chromosome spreads (Fig. 1B). It was also detected between the two chromatids of each chromosome and in the kinetochores. The detailed localization of nucleolin in the kinetochore was clarified by staining with an antibody against aurora kinase B (Aurora-B) as a centromeric heterochromatin marker, the anti-centromere autoimmune serum (CREST) as an inner kinetochore marker and an antibody against Mad2 as an outer kinetochore marker. Interestingly, the signals for nucleolin were separate from those for Aurora-B or CREST, but adjacent to those for Mad2, suggesting that nucleolin is localized in the vicinity of the outer kinetochore. When fibrillarin, another chromosome peripheral protein, was double-stained with nucleolin in metaphase chromosome spreads, the two proteins overlapped in the chromosome periphery and were colocalized in the vicinity of the outer kinetochore. The localization of nucleolin in the vicinity of the outer kinetochore was confirmed by line scans through kinetochore pairs (Fig. 1C), and by measuring the distances between the centers of the immunofluorescence signals for nucleolin and those for Aurora-B, CREST, Mad2 and fibrillarin (Fig. 1D).

We depleted human nucleolin in HeLa cells by using RNAi. A nonspecific small interference RNA (siRNA) was used for control transfections. Two different siRNA oligonucleotides (siRNA1 and siRNA2) were used to exclude any potential off-target effects of the RNAi. The cellular expression levels of nucleolin were reduced to 15.5% and 39.1% after transfection with siRNA1 and siRNA2, respectively (Fig. 2A). No significant reductions in the expression levels of fibrillarin, B23, β-actin and α-tubulin were detected. These data are consistent with a recent report that depletion of nucleolin by siRNA did not considerably affect the expression levels of other proteins, including B23, aprataxin and β-tubulin (Becherel et al., 2006). To further increase our confidence in the data obtained from the siRNA experiments, we checked the expression levels of several proteins, including cyclin B1 and hnRNP-U in total cell extracts, and CDC20, CDC27 (a subunit of the anaphase-promoting complex), Aurora-B, BubR1 and Hec1 (one of the subunits of human NDC80 complex) in mitotic cell extracts. The results revealed that the expression levels of these proteins were not considerably affected by nucleolin RNAi (data not shown). RNAi-mediated depletion of nucleolin resulted in dramatic reductions in its fluorescence intensities in both the nucleoli and chromosome periphery (Fig. 2B). For siRNA1 depletion, the average intensities of nucleolin in the nucleoli and chromosome periphery were decreased by 66.2±7.8% and 73.3±7.0%, respectively, compared with the values in control cells. For siRNA2 depletion, the average intensities of nucleolin in the nucleoli and chromosome periphery were decreased by 59.1±7.4% and 63.1±8.6%, respectively, compared with the values in control cells.

To compare the localization of nucleolin with those of other nucleolar proteins, two different DFC markers, namely fibrillarin (Fomproix et al., 1998) and Ki-67 (Kill, 1996), and one marker for both DFCs and GCs, namely B23 (Biqqioqera et al., 1990), were stained in the nucleoli. Nucleolin was mainly localized at the outer layer with fainter staining at the center of the nucleoli (Fig. 2C). By contrast, fibrillarin appeared as discrete dot-like foci within the nucleoli. B23 was almost uniformly dispersed over the whole nucleoli (Fig. 2D). In a previous study, Ki-67 was detected in large numbers of discrete foci throughout the nucleoplasm during early G1 (Kill, 1996). In the present study, Ki-67 was localized in the nucleoli during S and G2 phase (Fig. 2E), consistent with that previous study (Kill, 1996). In cells depleted by siRNAs targeting nucleolin, the spherical shape of the nucleoli became irregular, as revealed by the weak nucleolar staining of nucleolin, fibrillarin and B23 (Fig. 2C,D). In the nucleolin-depleted cells, Ki-67 was dispersed throughout the nuclei without any obvious staining in the nucleoli (Fig. 2E). After nucleolin depletion, the relative fluorescence intensity of nucleolin in the nucleoli was considerably reduced (Fig. 2B). Unexpectedly, the average intensities of several other nucleolar proteins, including fibrillarin, B23 and Ki-67, in the nucleolin-depleted nucleoli were decreased by 69.4%, 30.9% and 68.1% compared with the corresponding values in control cells. These fluorescence reductions were probably due to dispersion of fibrillarin and B23 by nucleolin RNAi, although the expression levels of these two proteins were not considerably changed (Fig. 2A). These data suggest that nucleolin is responsible for maintaining the shape and functional compartmentalization of nucleoli, since the spherical structures of the nucleoli and DFCs, which were marked by fibrillarin or Ki-67 staining, were highly disorganized.

During mitosis, nucleolin depletion induced the accumulation of mitotic cells. The mitotic index of nucleolin-depleted cells was approximately two-fold higher in cells after RNAi with siRNA1 (5.7±0.4%) than in control cells (2.7±0.2%) at 72 hours after transfection (Fig. 3A). Similar effects were seen in the samples transfected with siRNA2. Upon depletion of Aurora-B by RNAi, checkpoint proteins, including BubR1 and Mad2, are rarely recruited to the kinetochores and the spindle checkpoint is silenced, triggering mitotic exit (Vagnarelli and Earnshaw, 2004). To determine whether the spindle checkpoint can be activated in nucleolin-depleted cells, we used siRNAs to deplete nucleolin alone or nucleolin and Aurora-B simultaneously. Using immunostaining and immunoblotting, we confirmed that both proteins were depleted to the same levels as those observed for the single siRNA treatments (data not shown). In cells only depleted of nucleolin, the mitotic index increased considerably compared with that in control cells (Fig. 3B). However, in cells simultaneously depleted of both nucleolin and Aurora-B, the mitotic index was similar to that in control cells, suggesting that the mitotic delay observed in nucleolin-depleted cells resulted from spindle-checkpoint activation. Consistent with our hypothesis, quantification of the distinct mitotic stages revealed that the nucleolin-depleted cells showed a two-fold increase in the population of prometaphase cells (Fig. 3C). To further confirm this finding, the spindle-checkpoint proteins Bub1, BubR1 or Mad2 were double-stained with α-tubulin in control and nucleolin-depleted cells. At prometaphase, the three spindle checkpoint proteins were recruited to the kinetochores in control cells (Fig. 3D-F). When the cells proceeded to metaphase, Bub1, BubR1 and Mad2 showed either a lack of association with the kinetochores of aligned chromosomes or were present at low levels (Fig. 3D-F). In nucleolin-depleted cells, Bub1, BubR1 and Mad2 were detected at the kinetochores of misaligned chromosomes, but not at those of chromosomes aligned at the spindle equator (Fig. 3D-F). Taken together, these data suggest that nucleolin depletion activates the spindle checkpoint.

After incubation with the siRNAs targeting nucleolin, signals for nucleolin were rarely detected in mitotic cells (Fig. 2B, Fig. 4A) or chromosome metaphase spreads (Fig. 4B). Interestingly, signals for both fibrillarin (Fig. 4B) and B23 (data not shown) in the chromosome periphery also disappeared after nucleolin depletion. Nucleolin depletion mainly induced two types of pronounced defects in chromosome congression: misalignment and non-alignment. We categorized the defects in chromosome alignment into two groups, designated misaligned chromosomes (one to ten chromosomes were not aligned at the metaphase plate, whereas the majority of the chromosomes were aligned) and non-aligned chromosomes (>10 chromosomes were not aligned at the metaphase plate or chromosomes were scattered throughout the cytoplasm). In the misalignment phenotype, several chromosomes remained near the spindle poles, although the other chromosomes aligned at the spindle equator (Fig. 4Ci). In the non-alignment phenotype, most of the chromosomes remained dispersed (Fig. 4Cii). Mitotic cells depleted by siRNA1 exhibited 16.2±5.5% and 16.9±1.8% of the misalignment and non-alignment phenotypes, respectively (Fig. 4D). Similar to siRNA1, mitotic cells depleted by siRNA2 exhibited 21.1±2.4% and 8.5±1.2% of the misalignment and non-alignment phenotypes, respectively. By contrast, the control cells showed very few populations of each phenotype (2.5±1.5% and 3.4±1.3%, respectively). As a positive control for chromosome congression, we examined the effects of depletion of Aurora-B, a mitotic kinase known to be required for chromosome congression (Hauf et al., 2003). Most of the nucleolin-depleted cells contained only one or a few misaligned chromosomes, and only 33.6% of the cells had five or more misaligned chromosomes (Fig. 4E). These findings demonstrate that human nucleolin is involved in normal chromosome congression.

Failure of chromosome congression may be due to an inability of the kinetochore-microtubule interactions to maintain sufficient tension. To verify this hypothesis, the interkinetochore distances, which reflect the accurate tension force (Waters et al., 1996), were measured in kinetochore pairs by staining Aurora-B in the middle of the kinetochores and CENP-A on either side of Aurora-B (Fig. 5A). We found that the average interkinetochore distance decreased from 1.4±0.2 μm in control metaphase cells to 1.1±0.2 μm for non-aligned chromosomes in nucleolin-depleted cells (P<0.001) (Fig. 5B). There was also a 21.4% reduction in centromere stretching relative to the resting length of 1.1±0.1 μm in control prometaphase cells. As a positive control, the interkinetochore distance of cells treated with colcemid was reduced by 34.2% (0.9±0.2 μm) compared with that in control metaphase cells. We also measured the interkinetochore distance of aligned chromosomes in the misalignment phenotype. Unexpectedly, we found an interkinetochore distance of 1.2±0.2 μm, representing reduced sister chromatid stretching compared with that in control metaphase cells (P<0.001).

Nucleolin depletion resulted in a reduction of stretched kinetochores on misaligned or non-aligned sister chromatids, which could indicate that these kinetochores are incorrectly bound to microtubules. When cells were stained with CREST and α-tubulin antibodies, bioriented kinetochore pairs with kinetochore fibers terminating at the kinetochores were frequently observed in control cells (Fig. 5C, enlargement). In the misalignment phenotype, the chromosomes at the spindle equator were attached to microtubules oriented from the opposite spindle poles (Fig. 5D, enlargement a). Although the kinetochores were found to attach to microtubules in nucleolin-depleted cells, the chromosomes often became syntelically attached, i.e. both kinetochores were attached to the same spindle pole (Fig. 5D, enlargement b). Similarly, the dispersed chromosomes also presented syntelic attachments in the non-alignment phenotype (Fig. 5E, enlargements a, b). These defects in biorientation prompted us to analyze the stability of the kinetochore-microtubule attachments. Cells were incubated in ice-cold medium for 10 minutes before fixation to depolymerize all microtubules except stable kinetochore fibers (Rieder, 1981). Control metaphase cells exhibited high levels of tubulin fluorescence. The kinetochore fibers in the nucleolin-depleted cells were considerably diminished (Fig. 5F). In addition, chromosomes located near the spindle poles lacked any associated microtubule staining, indicating that few stable kinetochore-microtubule attachments were present on these chromosomes (Fig. 5F). In conclusion, nucleolin is involved in correct kinetochore-microtubule attachments.

To define the underlying mechanism responsible for the prometaphase mitotic delay in nucleolin-depleted cells, HeLa cells stably expressing GFP-tagged histone H1.2 (GFP-H1.2) (Gambe et al., 2007) were monitored from prophase. In total, 69.2% of control cells completed mitosis within 1 hour (Fig. 6A,C). Congression was rapid in these control cells and, after nuclear envelope breakdown, all the chromosomes moved quickly towards the spindle equator to achieve full alignment. The situation was quite different in nucleolin-depleted cells. After more than 180 minutes of monitoring, 54.4% of the cells contained various numbers of chromosomes that were unable to migrate to the spindle equator and remained in the vicinity of the spindle poles during congression (Fig. 6B,C). Although nucleolin-depleted cells spent a similar time in prophase to control cells (mean time: 9 minutes), most of the cells showed a very extended prometaphase (mean time: 238 minutes). Nevertheless, metaphase onset was not seen in 42.4% of nucleolin-depleted cells. As time progressed, the chromosomes in these cells gradually became decondensed and then aggregated to form mini-nuclei, a hallmark of cell death. These aberrations were not owing to the experimental conditions, because normal mitotic progression was detected in all the control cells. Only 12% of nucleolin-depleted cells completed cell division with a significant mitotic delay at prometaphase.

To further characterize the defects in chromosome congression after nucleolin depletion, we performed RNAi in HeLa cells stably expressing GFP-tagged CENP-A (GFP-CENP-A) (Kunitoku et al., 2003). In control cells, paired sister kinetochores congressed to the spindle equator and oscillated until anaphase onset (Fig. 6D). In nucleolin-depleted cells, some kinetochore pairs remained at the spindle poles for the duration of the analysis. Frequent oscillation of sister kinetochores was observed, and they moved back and forth between the spindle poles and the spindle equator. These findings indicate that nucleolin plays important roles in both the establishment and maintenance of bioriented kinetochore microtubule attachments.

The specificity of the TG-3 antibody, which recognizes nucleolin phosphorylated by CDC2 during mitosis (phospho-nucleolin), was confirmed in a previous study (Dranovsky et al., 2001). We confirmed its specificity by immunoblotting and immunostaining (Fig. 7A,B). Following transfection of nucleolin siRNAs, quantitative immunoblotting of cell extracts revealed that the phospho-nucleolin level was reduced to 18.7% of the control level. The TG-3 staining pattern in metaphase chromosome spreads was the same as that using an anti-nucleolin antibody (data not shown). We observed phospho-nucleolin fluorescence at the chromosome periphery and spindle poles in mitotic cells fixed with 4% paraformaldehyde (PFA) (data not shown). The polar localization of phospho-nucleolin was facilitated by detergent treatment before the PFA fixation (data not shown). The association of phospho-nucleolin with the spindle poles was best observed when the cells were fixed with methanol at -20°C (Fig. 7C). With this protocol, labeling of the chromosomes was also observed. Phospho-nucleolin appeared in the vicinity of the spindle poles at prometaphase when the nuclear envelope began to disassemble. It became most prominent at metaphase when the chromosomes became aligned at the spindle equator. At anaphase, when the chromosomes segregated, the staining of the spindle poles became more dispersed. At telophase, phospho-nucleolin appeared within the reforming nuclei and disappeared from the spindle poles. B23, a major phosphoprotein in interphase nucleoli, is associated with the poles of the mitotic spindle and chromosome periphery during mitosis (Zatsepina et al., 1999). Consistent with this previous report, we found that B23 was localized at the spindle poles at prometaphase and metaphase following Triton-X 100 extraction before fixation (Fig. 7D). The polar localization of B23 disappeared after nucleolin depletion (Fig. 7D), although the polar localization of another spindle pole protein, CLASP1, was not affected by the depletion of nucleolin (data not shown).

A multipolar spindle phenotype was defined as the presence of more than two spindle poles within a cell (Fig. 7Ev). The frequencies of disorganized and multipolar spindles were 12.9±7.2% and 11.3±6.4% in nucleolin-depleted mitotic cells, compared with only 1.2±1.3% and 3.8±1.1% in control cells, respectively (Fig. 7F), indicating that the normal function of nucleolin is involved in maintaining the mitotic spindle morphology. When nucleolin-depleted cells were stained for α-tubulin and γ-tubulin, multiple centrosomes were found in the multipolar spindle cells (Fig. 7G), suggesting that nucleolin is involved in regulating centrosome duplication to ensure the formation of a bipolar spindle.

Since the localizations of both fibrillarin and B23 in the nucleoli or chromosome periphery appeared to be dependent on the presence of nucleolin, it is possible that nucleolin, fibrillarin and B23 are present in a complex in vivo, and that the absence of one of these components could influence the stability of the complex. To investigate this possibility, we immunoprecipitated total cell extracts and synchronized mitotic cell extracts with an anti-nucleolin antibody. We used a double-thymidine block to achieve the mitotic cell synchronization. DNA staining with DAPI demonstrated that, after removal of the thymidine block, about 60% of the cells passed through mitosis over 10 hours (data not shown). Consistent with previous reports (Yanagida et al., 2001; Yanagida et al., 2004), physical associations of nucleolin with fibrillarin (Fig. 8A) and B23 (data not shown) were detected in the total cell extracts. Fibrillarin and B23 remained associated with nucleolin in complexes in the synchronized mitotic cells (Fig. 8B).

When nucleolin was depleted, dispersed nucleoli were observed. Simultaneously, nucleolar proteins including fibrillarin, B23 and Ki-67 showed no systematic localizations in the nucleoli. Similar to nucleolin, fibrillarin and B23 have been identified in ribonucleoprotein complexes (Yanagida et al., 2001), although the relationship between Ki-67 and nucleolin remains unclear. Since Ki-67 is also dispersed in nucleoli after nucleolin depletion, we hypothesize that the apparent nucleolar dispersions of these three proteins in nucleolin-depleted cells probably arise as a consequence of disorganization of the nucleolar structure. It has been reported that mutations in genes encoding nucleolin homologues in fission yeast (GAR2) and Arabidopsis (AtNUC-L1) have severe effects on the organization of nucleoli (Léger-Silvestre et al., 1997; Pontvianne et al., 2007). However, the function of nucleolin in the organization of nucleoli has not been reported in animal cells. The present study is the first description of an essential role for human nucleolin in maintaining the nucleolar structure.

Nucleoli were recently shown to be dynamic subnuclear bodies, because the residence times of most nucleolar proteins are in the order of tens of seconds (Phair and Misteli, 2000; Chen and Huang, 2001). In reality, the structure of a nucleolus is the total sum of the molecules entering and leaving a given subnuclear space, with the residence times of individual molecules governing its overall organization (Olson and Dundr, 2005). These residence times are determined by interactions between nucleolar molecules. Thus, when the amount of nucleolar nucleolin is reduced by RNAi, fibrillarin, B23 and Ki-67 could lose their affinities for other nucleolar components, which finally results in their shorter residence times in the nucleolus. Fibrillarin is involved in all the major post-transcriptional activities in ribosome synthesis, the first steps of rRNA processing, pre-rRNA modification and ribosome assembly (Tollervey et al., 1993). Similar to nucleolin depletion, microinjection of anti-fibrillarin antibodies can also induce defects in nucleolus formation (Fomproix et al., 1998).

In conclusion, all these data suggest that nucleolar proteins, including nucleolin and fibrillarin, which are involved in ribosome biogenesis, play important roles in nucleolus formation. Consistent with an earlier hypothesis that a subset of ribonucleoprotein complexes remains assembled and becomes incorporated into the new nucleolus (Medina et al., 1995; Piñol-Roma et al., 1999), associations of nucleolin with fibrillarin and B23 were found during the cell cycle. It seems likely that these interactions between nucleolin and nucleolar proteins, including fibrillarin and B23, are of physical significance, because we have demonstrated that nucleolin is required in order to retain fibrillarin and B23 in the nucleoli and chromosome periphery.

In nucleolin-depleted HeLa cells, chromosome congression to the spindle equator was delayed. This could be due to activation of the spindle checkpoint in response to improper kinetochore-microtubule attachments, such as those demonstrated for chromosomes stuck at the spindle poles during prometaphase. Recent data have shown that Aurora-B plays an important role in correction of improper microtubule-kinetochore interactions, such as syntelic or merotelic attachments (Hauf et al., 2003; Cimini et al., 2006). Activation of Aurora-B during mitosis can correct chromosome attachment errors by selective disassembly of kinetochore-microtubule fibers, thereby activating the spindle checkpoint until all the chromosomes make proper bipolar attachments to the mitotic spindle. It is therefore probable that nucleolin depletion does not activate the checkpoint directly but rather causes an increase in syntelic attachments, which are changed into unattached kinetochores by Aurora-B.

Electron microscopy studies have revealed that the plus ends of spindle microtubules terminate in the outer kinetochore of vertebrate cells (McEwen et al., 1998; VandenBeldt et al., 2006), suggesting that nucleolin covering the outermost layer of kinetochores could affect the kinetochore-microtubule attachments. We found that cells depleted of nucleolin were delayed at prometaphase with misaligned or non-aligned chromosomes and defects in chromosome biorientation. Aligned chromosomes exhibited reduced centromere stretching and diminished kinetochore-microtubule fibers. Our data suggest that nucleolin is involved in efficient kinetochore-microtubule interactions and/or the stabilization of such interactions. These interactions must be correct to allow normal chromosome congression and segregation. We did observe syntelic attachments in the misaligned or non-aligned chromosomes of nucleolin-depleted cells (Fig. 5D,E), and these may hinder the formation of proper attachments. Our evidence reveals that nucleolin is not required for kinetochore assembly. Specifically, the evidence for this conclusion includes the following: (1) nucleolin depletion does not affect the localizations of kinetochore proteins, including Aurora-B (data not shown), Bub1 (Fig. 3D), BubR1 (Fig. 3E), CENP-A (Fig. 5A), CENP-E (data not shown), Hec1 (data not shown) and Mad2 (Fig. 3F); and (2) kinetochores can capture microtubules in the absence of nucleolin (Fig. 5D,E).

One possibility is that disruption of the nucleoli per se perturbes chromosome congression, mitotic progression and/or spindle formation. However, depletion of certain nucleolar proteins, such as UBF (Rubbi and Milner, 2003) and TIF-IA (Yuan et al., 2005) localized in the nucleolus organizer region during mitosis (Roussel et al., 1993; Prieto and McStay, 2005), by antibody microinjection or genetic inactivation did not lead to mitotic delay or defects in chromosome congression, although disorganized nucleoli were observed. Therefore, our reported mitotic abnormalities are not caused by nucleolar disorganization but rather depletion of nucleolin. To the best of our knowledge, this is the first report that a chromosome peripheral protein can contribute to chromosome congression during mitosis.

Movement of a chromosome on the mitotic spindle is mediated by the formation of a bundle of microtubules that tethers the kinetochore on the chromosome to a spindle pole. Mechanical linkages between chromatids and microtubules are crucially important for spindle assembly (Cassimeris and Skibbens, 2003). Since nucleolin is important for the stability of kinetochore-microtubule attachments, we suspect that the mitotic spindle disorganization observed in nucleolin-depleted cells is the consequence of a defect in the kinetochore-microtubule attachments. Consistent with this hypothesis, abrogation of kinetochore-microtubule attachments by depleting certain kinetochore proteins, including CENP-F (Holt et al., 2005), hMis12 (Goshima et al., 2003), NDC80 complex (Bharadwaj et al., 2004), CENP-A (Regnier et al., 2005) and Shugoshin (Salic et al., 2004), can lead to overt spindle defects.

Protein phosphorylation has been proposed as an important control mechanism for the events leading to the initiation and completion of mitosis (Vandre et al., 1984). Numerous phosphoproteins relocate to the spindle poles during mitosis (Vandre et al., 1984; Vandré and Burry, 1992). Meanwhile, nucleolin is highly phosphorylated by CDC2 during mitosis (Belenguer et al., 1990; Peter et al., 1990). Here, we report for the first time that nucleolin is also present in the vicinity of the spindle poles during mitosis. Using standard PFA cell fixation, we were able to observe both nucleolin localizations, and found that the stronger signal emerged from the chromosome periphery, whereas the weaker one arose from the spindle poles. In the vast majority of mitotic cells fixed with aldehydes, it was obvious that the polar nucleolin labeling was masked by the cytoplasmic mass. This may explain why the presence of nucleolin at the spindle poles has not yet been described. Therefore, application of a fixation procedure that is known to preserve the cytoskeleton makes it possible to detect the nucleolin signals in the spindle poles.

In the present study, we found that the frequencies of cells with multipolar spindles were much higher among nucleolin-depleted cells than among control cells, and this phenomenon is probably caused by the existence of supernumerary centrosomes. It has been reported that B23 is involved in regulating centrosome duplication in the cell cycle (Okuda, 2002; Wang et al., 2005). Since nucleolin depletion resulted in a dramatic reduction in the spindle pole localization of B23, it is possible that nucleolin also plays a role in centrosome duplication through its interaction with B23, which is also known to be phosphorylated by CDC2 kinase at the onset of mitosis (Peter et al., 1990). Furthermore, hyperphosphorylated B23 isoforms are specifically associated with the spindle poles (Zatsepina et al., 1999). Here, we report similar behavior for nucleolin, since its phosphorylated isoform was associated with the spindle poles. Furthermore, the nucleolin in the vicinity of the spindle poles resisted extraction with detergent before fixation, whereas that in the chromosome periphery largely disappeared under the same conditions (data not shown), suggesting that the properties of these two types of nucleolin are different. Therefore, mapping and mutation of the nucleolin in the vicinity of the spindle poles will be necessary to investigate the functions of this particular, yet previously undescribed, nucleolin isoform.

HeLa cells were grown in Dulbecco's modified Eagle's medium (DMEM; GIBCO BRL) supplemented with 10% fetal bovine serum (Equitech-Bio Inc.) and penicillin.

Mitotic arrest was induced by adding 100 ng/ml colcemid to the culture medium and incubating the cells for 16 hours. For mitosis synchronization, cells were treated twice with 2.5 mM thymidine (Sigma) for 16 hours each, separated by a 10-hour exposure to culture medium, and then released from the block by incubation with culture medium. After approximately 10 hours, approximately 60% of the cells passed through mitosis.

A nonspecific siRNA duplex (Qiagen) was used for control transfections. Two siRNA duplexes (siRNA1: 5′-AGAGUUUGCUUCAUUCGAATT-3′; and siRNA2: 5′-GACAGUGAUGAAGAGGAGGTT-3′) targeting human nucleolin were used to deplete nucleolin. Previously reported siRNA duplexes for Aurora-B (Ditchfield et al., 2003) were used to deplete Aurora-B. HeLa cells were transfected with siRNA duplexes using Lipofectamine 2000 (Invitrogen), and then subjected to immunoblotting and immunostaining analyses at 72 hours after the transfections.

The antibodies against the following proteins were used as follows: nucleolin (mouse monoclonal; 1:100 dilution; Santa Cruz Biotechnology); Ki-67 (mouse monoclonal; 1:200 dilution; DakoCytomation); α-tubulin (mouse monoclonal; 1:200 dilution; Calbiochem); α-tubulin (rabbit polyclonal; 1:200 dilution; Abcam); γ-tubulin (rabbit monoclonal; 1:1000 dilution; Sigma); BubR1 (mouse monoclonal; 1:500 dilution; BD Transduction Laboratories); Bub1 (mouse monoclonal; 1:50 dilution; Chemicon); CENP-E (mouse monoclonal; 1:100 dilution; Abcam); CENP-A (mouse monoclonal; 1:50 dilution; Abcam); Aurora-B (rabbit polyclonal, 1:1000 dilution; Abcam); fibrillarin (rabbit polyclonal; 1:100 dilution; Abcam); anti-centromere autoimmune serum (CREST; (human monoclonal; 1:100 dilution; Cortex Biochem); Hec1 (mouse monoclonal; 1:100 dilution; ABR-Affinity BioReagents); Mad2 (rabbit polyclonal; 1:100 dilution; Covance), B23 (goat polyclonal; 1:50 dilution; Santa Cruz Biotechnology), hnRNP-U (mouse monoclonal; 1:250 dilution; Abcam); CDC27 (mouse monoclonal; 1:250 dilution; Abcam); CDC20 (mouse monoclonal; 1:250 dilution; MBL); cyclin B1 (rabbit monoclonal; 1:250 dilution; Sigma); and β-actin (mouse monoclonal; 1:500 dilution; Sigma). The TG-3 antibody against phosphorylated nucleolin was obtained from Peter Davies (Albert Einstein College of Medicine in Yeshiva University, New York). The anti-CLASP1 antibody was obtained from William Earnshaw (Wellcome Trust Centre for Cell Biology, University of Edinburgh).

All cell preparations were examined using a deconvolution microscopy system (DeltaVision; Applied Precision) and oil immersion objectives. The fluorescence intensities of the signals were analyzed with the Image J program (http://rsb.info.nih.gov/ij/download.html). The interkinetochore distances were measured as previously described (Waters et al., 1996). The distances were determined in at least five cells (40 kinetochore pairs).

For living cell imaging, the medium was replaced with a CO₂-independent medium supplemented with 10% fetal bovine serum, penicillin-streptomycin and L-glutamine (Invitrogen), and covered with mineral oil immediately before analysis. Cells were maintained at 37°C using a heated stage. Images of cells expressing GFP-H1.2 were collected under an inverted microscope (Olympus IX81) using a 40×NA 1.4 planApo objective. The acquisition parameters, including exposure, focus and illumination, were controlled by the MetaMorph software (Universal Imaging). Single focal plane images were collected by a camera at 3-minute intervals. To capture images of cells expressing GFP-CENP-A, three z-sections were acquired at 1-μm steps at 10-second time intervals. The z-stacks were projected using the MetaMorph software. All subsequent analyses and processing of images were performed using the MetaMorph software.

HeLa cells on poly-L-lysine-coated coverslips were fixed in 4% PFA in PBS for 10 minutes at 37°C and then permeabilized in 0.2% Triton X-100 in PBS for 10 minutes. Next, the cells were blocked with 1% BSA in PBS for 30 minutes, and sequentially incubated with primary and secondary antibodies diluted in 1% BSA in PBS for 1 hour. All the incubation steps were carried out at 37°C. For some antibodies (CENP-E and γ-tubulin), immunostaining analyses were carried out as previously described (Forgues et al., 2003). For staining of Bub1 and phosphorylated nucleolin, cells were fixed in methanol for 10 minutes at -20°C. Immunostaining for B23 was performed as previously described (Zatsepina et al., 1999). Immunostaining of metaphase chromosome spreads was performed as previously described (Ono et al., 2003), except that the cells were treated with 75 mM KCl at room temperature for 15 minutes rather than at 37°C for 30 minutes.

Following extraction of transfected cells with SDS sample buffer, the protein extracts were separated by SDS-PAGE and transferred to PVDF membranes for immunoblotting. The immunoblots were blocked with 0.1% BSA in TBST (0.1% Tween 20, 25 mM Tris-HCl pH 7.4, 137 mM NaCl, 25 mM KCl) and then sequentially incubated with primary and secondary antibodies diluted in TBST. Finally, the immunoreactive protein bands were detected by NBT/BCIP solution (Roche) supplemented with AP buffer (100 mM Tris-HCl pH 9.5, 100 mM NaCl, 1 mM MgCl₂).

Immunoprecipitation was carried out using a mouse monoclonal antibody against nucleolin, or normal mouse IgG (Santa Cruz Biotechnology) as a control. The antibodies were bound to protein G-Sepharose (GE Healthcare) containing 0.1% BSA, 150 mM NaCl, 20 mM Tris-HCl pH 8.0 and 0.05% Tween 20 by rocking at 4°C overnight. HeLa cells were harvested by centrifugation at 310 g for 5 minutes, rinsed with PBS and centrifuged again at 310 g. Cell pellets were stored at -80°C until analysis. The cells were extracted in 5 volumes of buffer A [10% cell lysis buffer (Cell Signaling), 0.1% SDS, protease inhibitor cocktail (Roche)] for 30 minutes, and centrifuged at 15,100 g for 30 minutes. To reduce the amount of proteins nonspecifically binding to the beads, the supernatant was first rocked with beads for 30 minutes at 4°C. After centrifugation, the supernatant was incubated with antibody-bead complexes for 3 hours at 4°C. Next, the beads were collected by centrifugation and washed five times in buffer A. Finally, protein-bead complexes were solubilized by boiling, centrifuged, subjected to SDS-PAGE and analyzed by immunoblotting.

This work was supported by Grants-in-Aid for Scientific Research from the Ministry of Education, Science, Culture, Sports, Science and Technology of Japan to K.F., S.M. and S.U. and by an Industrial Technology Research Grant Program in 2003 from the New Energy and Industrial Technology Development Organization (NEDO) of Japan to S.M. We thank Peter Davies for the anti-phospho-nucleolin antibody, William C. Earnshaw for the anti-CLASP1 antibody and Toru Hirota for the HeLa cells stably expressing GFP-CENP-A.