Human Chromatid Cohesin Component hRad21 Is Phosphorylated in M Phase and Associated with Metaphase Centromeres*

Sister chromatids duplicated in S phase are connected with each other during G2 and M phase until the onset of anaphase. This chromatid cohesion is essential for correct segregation of genetic material to daughter cells. Recently, understanding of the molecular mechanisms governing chromatid cohesion in yeast has been greatly advanced, whereas these processes in mammalian cells remain unclear. We report here biochemical and cytological analyses of human Rad21, a homologue of the yeast cohesin subunit, Scc1p/Mcd1p. hRad21 is a nuclear phosphorylated protein. Its abundance does not change during the cell cycle, and it becomes hyperyphosphorylated in M phase. Most hRad21 is not associated with chromatin when the nuclear envelope breakdown takes place in prophase. However, a detailed analysis of the spread chromosomes indicated that hRad21 remains associated with prometaphase-like chromosomes along their entire lengths. The mitotic chromatin-bound hRad21 becomes dissociated in a highly regulated manner because hRad21 remains specifically at the centromeres but disappears from the arm regions on metaphase-like chromosomes. Interestingly, hRad21 at the metaphase centromeres appears to be present at the inner pairing domain where the two sister chromatids are supposed to be in intimate contact. These results suggest that hRad21 has a critical role in chromatid cohesion in human mitotic cells.

Smc3p are found in budding yeast, Xenopus, and mammals (7)(8)(9). They are members of the SMC (structural maintenance of chromosome) protein family that is characterized by the presence of coiled-coil domains and ATPase domains (reviewed in Ref. 10). Very recently, two Scc3p homologues, SA1 and SA2, have been found in Xenopus and human cells (11). Interestingly, in Xenopus, two distinct classes of cohesin, termed xcohesin SA1 and x-cohesin SA2 , are present. These two complexes share Xenopus (X)SCC1, XSCC2, and XRAD21 and differ via containing either XSA1 or XSA2 (11). Immunodepletion of Xenopus cohesin from egg extracts led to a failure of chromatid cohesion (8). Therefore, the cohesin complex is likely to be conserved in all eukaryotes, including humans.
However, the precise roles of cohesin may be different between budding yeast and other eukaryotes. In budding yeast, Scc1p/Mcd1p abruptly dissociates from chromatin at the onset of anaphase (2,4). In contrast, ϳ95% of Xenopus cohesion molecules (XSMC1, XSMC3, and XRAD21) dissociate from chromatin at the entry of mitosis, much earlier than the metaphase-anaphase transition (8). Similarly, in indirect immunofluorescence (IF) 1 experiments, it has been found that human Smc1p and mouse Rad21 (called PW29) are mostly excluded from mitotic chromosomes (9,12). Therefore, most cohesins are apparently absent on metaphase chromatids in higher eukaryotes. It is not known what molecules or conditions are responsible for the chromatid cohesion immediately before anaphase. Two models have been proposed to explain the apparent inconsistency between the timings of the cohesin-chromatin dissociation and the mitotic chromatid separation (8). The first model proposes that cohesin molecules are responsible for interphase-specific chromatid cohesion and that some yet unidentified mitosis-specific cohesion machinery is responsible for the chromatid cohesion from prophase until the onset of anaphase. The second model hypothesizes that the same cohesin complex is required for both interphase-and mitosis-specific chromatid cohesions. However, in this model, the complex dissociates from chromatin in two steps, whereby most cohesin is released from chromatin at the entry into mitosis. The remaining cohesin connects chromatids in metaphase and dissociates from chromatin at the onset of anaphase. The recent discovery of Xenopus (X)SA proteins revealed that a small population of XSA1 is associated with the metaphase chromosomes formed in Xenopus cell-free extracts, thus supporting the second model (11).
In this paper, we describe the biochemical and cytological behaviors of hRad21. We aimed to better understand the roles of hRad21 particularly in metaphase. We observed a small but significant population of hRad21 associated with colcemid-in-duced mitotic chromosomes. These results suggest that the mitotic cohesion is mediated by cohesin, further underscoring the conserved mechanisms regulating chromatid cohesion and separation in eukaryotes.

EXPERIMENTAL PROCEDURES
Cloning of hRad21 cDNA and Preparation of Anti-hRad21 Antibodies-Full-length hRad21 cDNA was isolated from a HeLa cell cDNA library by reference to the published hRad21 cDNA sequence (6). The cloned cDNA was completely sequenced. The resulting sequence was identical to the published clone. The full-length hRad21 cDNA was then subcloned into pGEX5X-1 (Amersham Pharmacia Biotech). glutathione S-transferase-fused hRad21 was expressed in Escherichia coli. The recombinant protein contained in the inclusion body was denatured and purified using Prep Cell Model 491 (Bio-Rad). The purified protein was mixed with Freund's complete adjuvant and injected into rabbits to obtain the anti-hRad21 antisera. To generate the anti-C-hRad21 antibodies, an oligopeptide possessing a C-terminal amino acid sequence of hRad21 (QQAIELTQEEPYSD, amino acids 606 -619) was conjugated with keyhole limpet hemocyanine and was used to immunize the rabbits. The resulting antibodies were purified by affinity chromatography. Mouse anti-␣-tubulin monoclonal antibody was purchased from Sigma. Mouse anti-PCNA, anti-lamin B, and anti-cyclin B1 monoclonal antibodies were purchased from Santa Cruz Biotechnology. Anti-CENP-B antibody was a kind gift from Dr. H. Masumoto (Nagoya University).
Western Blotting-Western blotting was performed according to Ref. 13. Briefly, membranes were first pretreated for 1 h in Block-Ace solution (Dai Nippon Pharmaceuticals). Then all subsequent incubations and washes were carried out in 1ϫ TNT buffer (20ϫ TNT: 0.4 M Tris-HCl, 2.8 M NaCl, 1.0% Tween 20). Membranes were incubated for 1 h with primary antibodies at room temperature, followed by three washes. Then membranes were incubated for 30 min with horseradish peroxidase-conjugated anti-rabbit antibodies (Amersham Pharmacia Biotech), followed by three washes. Signals were detected using an ECL kit (Amersham Pharmacia Biotech).
In Vitro Transcription and Translation-In vitro transcription and translation experiments were performed using a TnT Quick Coupled Transcription/Translation System (Promega). 1 g of template plasmid DNA was used in each reaction. For autoradiography, proteins were labeled by addition of [ 35 S]methionine (1000 Ci/mmol at 10 mCi/ml; Amersham Pharmacia Biotech) to the reaction mixture. For the Western analyses, cold methionine (0.02 mM) was used during synthesis of the protein.
Indirect Immunofluorescence Experiments-Cells were grown on coverslips. Two fixation protocols were used, producing essentially similar results. In the first protocol, cells were fixed with 100% methanol for 20 min at Ϫ20°C. In the second protocol, cells were fixed with 4% paraformaldehyde at room temperature for 10 min, and then permeabilized with 0.1% Triton X-100 for 10 min. Fixed cells were pretreated with PBS containing 0.1% bovine serum albumin and 0.1% skim milk and then incubated with primary antibodies for 1.0 h at 37°C. Cells were washed by PBS three times and incubated with secondary antibodies for 1 h at 37°C. Coverslips were washed as above, mounted in mounting solution containing 0.25 g/ml propidium iodide (PI) or TOTO3, and examined by laser confocal microscopy (Zeiss).
Cell Cycle Analyses and Metabolic Labeling-HeLa cells were synchronized at early S phase by a thymidine and aphidicolin double-block protocol as described in (14). In the 32 P-metabolic labeling experiments, cells were incubated in a phosphate-free Dulbecco's modified Eagle's medium supplemented with [ 32 P]orthophosphate to a final concentration of 0.5 mCi/ml for 1 h prior to harvesting. Cells were lysed by the modified RIPA buffer. Cell lysates were pretreated with normal rabbit IgG. Then the precleared cell lysates were incubated with primary antibodies, and the antigen-antibody complex was immunoprecipitated by protein A-Sepharose (Amersham Pharmacia Biotech). Obtained proteins were fractionated on 6% SDS-PAGE, and transferred to a polyvinylidene difluoride membrane. The membrane was first subjected to autoradiography. Then the same membrane was analyzed for hRad21 by Western blotting using anti-yhRad21 antibodies. The signals were detected by a peroxidase immunostaining kit (Wako).
Cell Fractionation Analyses-1 ϫ 10 7 HeLa cells were washed with cold PBS, and collected. Cell pellets were suspended in five volumes of hypotonic buffer (10 mM HEPES, pH 7.5, 5 mM KCl, 1.5 mM MgCl 2 , and 1 mM dithiothreitol). Low speed cell pellets were resuspended in original cell volumes of hypotonic buffer and homogenized with a Dounce homogenizer using the tight pestle. Cells were examined with a light microscope to confirm that most cells had been disrupted. Cytoplasmic and nuclear fractions were separated by centrifugation at 1,000 ϫ g. Cytosolic fractions were obtained from the cytoplasmic fractions following high speed centrifugation at 15,000 ϫ g. Nuclear pellet fractions were washed twice with the hypotonic buffer and extracted by suspending the nuclei in hypotonic buffer containing either salt or detergents for 30 min on ice. Nuclear extract was separated from insoluble materials by high speed centrifugation. DNase I digestion of the crude nuclei was done according to Ref. 15.
Mitotic Spread Chromosome Analyses-HeLa cells were synchronized at early S phase following a thymidine and aphidicolin doubleblock protocol. To obtain metaphase-like chromosomes, cells were released from the block and cultured for 8 h, followed by an additional culture for 3 h in the presence of 0.5 g/ml colcemid. To obtain prometaphase-like chromosomes, cells were released from the block and cultured for 8.5 or 9 h, followed by an additional culture for 10 min in the presence of 0.5 g/ml colcemid. The cells were harvested, treated with a hypotonic buffer (10 mM Tris, 10 mM NaCl, 5 mM MgCl 2 ) for 15 min, and attached to a glass slides by Cytospin. The cells were fixed using cold methanol for 20 min, permeabilized with 0.1% Triton X-100 for 10 min, pretreated by blocking solution (0.1% bovine serum albumin and 0.1% skimmilk in PBS), and finally incubated with rabbit anti-hRad21 antibodies and mouse anti-CENP-B monoclonal antibodies. These antibodies were detected by Alexia Fluor 488-conjugated anti-rabbit Ig antibodies (Molecular Probes) and Cy3-conjugated anti-mouse Ig antibodies (Amersham Pharmacia Biotech), respectively. DNA was stained with TOTO3 (Molecular Probes). Images were captured by a lazer confocal microscope (Zeiss).

hRad21 Protein Does Not Change Its Abundance during Cell
Cycle-Budding yeast Scc1p/Mcd1p is a nuclear protein that associates with chromatin from late G1 through the metaphase-anaphase transition (2,3). Scc1p/Mcd1p abundance is strictly regulated in a cell cycle-dependent manner. The protein is absent in early G 1 , accumulates in S, G 2 , and metaphase, and declines in anaphase (2,3). In addition, SCC1/MCD1 mRNA is absent in early G 1 and most abundant in late G 1 /S (2, 3). hRad21 mRNA is most abundant in late S through G 2 (6). Therefore, we first investigated hRad21 protein levels throughout the cell cycle.
Full-length human Rad21 (hRad21) cDNA was isolated from a HeLa cell cDNA library. Nucleotide sequencing of the obtained cDNA yielded a sequence of complete identity to the published one (6), thus having an ORF that potentially encodes a protein of 631 amino acid residues with a calculated molecular mass of 72 kDa.
Two different rabbit antisera were raised, one (anti-hRad21 antibody) against the glutathione S-transferase-fused recombinant full-length hRad21 protein expressed in E. coli and the other (anti-C-hRad21 antibody) against a C-terminal synthetic oligopeptide (see "Experimental Procedures"). These antibodies were purified by affinity chromatography using the respective cognate antigens. Following hRad21 cDNA transfection both antibodies recognized the recombinant hRad21 protein that was expressed in the 293T cells (human kidney cells transfected with SV40 T antigen) (Fig. 1A). Both antibodies also recognized an endogenous protein in untransfected 293T cells that showed the same mobility in SDS-PAGE with that of the overexpressed hRad21 (Fig. 1B). These reactions were specific because preincubating the antibodies with the antigens prior to Western blotting prevented detection of hRad21 (Fig. 1, A and B).
The apparent molecular mass of the hRad21 protein (about 120 kDa), as estimated using SDS-PAGE electrophoresis, was greater than that predicted from its supposed amino acid sequence. Similar observations were made of budding yeast Scc1p (2), fission yeast rad21 protein (16), and Xenopus Rad21 homologue, XRAD21 (8). To confirm that the 120-kDa band indeed represents hRad21 protein, we transcribed and translated hRad21 cDNA in vitro using a rabbit reticulocyte lysate system. Recombinant proteins were synthesized in the presence of either [ 35 S]methionine or cold methionine and then analyzed by SDS-PAGE. Labeled proteins were detected by autoradiography, whereas unlabeled proteins were subjected to Western blotting with anti-hRad21 and anti-C-hRad21 antibodies. As shown in Fig. 1C, the [ 35 S]methionine-labeled hRad21 protein synthesized in vitro was detected as a 120-kDa protein band after SDS-PAGE electrophoretic migration. Because this 120-kDa protein was specifically recognized by both anti-hRad21 and anti-C-hRad21 antibodies (Fig. 1C), we concluded that it is indeed hRad21 protein.
HeLa cells were synchronized using a thymidine and aphidicolin double-block protocol (Fig. 2). The cells were harvested at intervals after release from the block and first extracted by 1% Triton X (Fractions T), followed by 0.5 M NaCl (Fractions N). The insoluble pellet fractions (Fractions P) were also examined. These fractions were analyzed using Western blotting. ␣-Tubulin served as the control cytoplasmic protein and was efficiently extracted to the T fractions, indicating that soluble proteins were successfully extracted with the Triton X-100 treatment (Fig. 2B). It is known that PCNA is insoluble during S phase (17). PCNA was detected in Fractions P most strongly at 0 -4 h and to a lesser level at 8 h (Fig. 2C), indicating that most of the cells were in S phase at 0 -4 h. In contrast, Cyclin B1 was most abundant at 8 and 10 h and suddenly disappeared at 12 and 14 h (Fig. 2B). These results indicate that the cells were in G 2 /M phase at 8 to 10 h and exited from M phase at 12-14 h. FACscan analysis strongly supported the interpretation of the cell cycle progression inferred from the marker proteins ( Fig.  2A). In these cells, the abundance of hRad21, as detected by anti-C-hRad21 antibody, remained essentially unchanged during the cell cycle. It is known that Scc1p/Mcd1p is proteolytically cleaved at the onset of anaphase (18). We reproducibly observed some more quickly migrating proteins in hRad21 Western blots. At this moment, we do not know whether these fragments were derived from hRad21 or whether they had any physiological relevance. However, we did not see an increase of these small fragments in M phase, suggesting that they were not related to the chromatid separation event.
We found that hRad21 was adequately extracted in 0.5 M NaCl but not in 1% Triton X-100 at all examined stages of the cell cycle (Fig. 2B). We also found that hRad21 was not present in Fractions P (data not shown). Furthermore, hRad21 was not extracted by DNase I (data not shown). These results indicate that hRad21 is not a soluble protein or a protein loosely binding to chromatin. hRad21 may be associated with nuclear structures at all cell cycle stages including M phase.
hRad 21 Is Hyperphosphorylated in M Phase-It has been shown that fission yeast Rad21p and budding yeast Scc1p are phosphorylated from S phase to anaphase (16). We next examined whether hRad21 is phosphorylated in a cell cycle-dependent manner. For this purpose, HeLa cells were synchronized following a thymidine and aphidicolin double-block protocol. The cells were harvested at intervals after being released from the block. Prior to each harvest, the cells were cultured for 1 h in the presence of [ 32 P]orthophosphate. A parallel culture was used for determining the cell cycle by FACscan analysis. hRad21 was immunoprecipitated using anti-hRad21 antibodies from cell extracts containing approximately the same number of cells. The immunoprecipitates were separated in an SDS-PAGE gel, and the proteins were blotted onto a membrane. First, the membrane was subjected to autoradiography (Fig. 3B). Then the same membrane was analyzed for total hRad21 levels using a Western analysis with anti-C-hRad21 antibodies (Fig. 3C). From these experiments, it was found that hRad21 is phosphorylated most intensely at 10, 12, and 14 h after release from the block. Because the FACscan analysis revealed that these samples were derived mostly from M phase cells, these results imply that hRad21 becomes hyperphosphorylated in M phase.
To test the hypothesis that hRad21 is hyperphosphorylated in M phase more vigorously, we repeated the experiment in a quantitative manner. HeLa cells were arrested at G 1 /S following a thymidine and aphidicolin double-block protocol. A portion of the cells were released from the block and cultured for 8 h and then arrested at the following metaphase by incubating FIG. 1. Specificities of anti-hRad21 antibodies used in this study. A, anti-hRad21 and anti-C-hRad21 antibodies were raised in rabbits, using the fulllength hRad21 protein and a C-terminal oligopeptide as antigens, respectively. After the antibodies were affinity-purified, the extract derived from 293T cells that had been transfected with the hRad21 expression vector, was examined by Western with (ϩ) or without (Ϫ) preabsorption of the antibodies with the cognate antigen. B, anti-hRad21 and anti-C-hRad21 antibodies recognize the endogenous hRad21 protein derived from untransfected 293T cells. 293T whole cell extracts were analyzed by Western blotting with (ϩ) or without (Ϫ) preabsorption of the antibodies to the cognate antigens. C, anti-hRad21 and anti-C-hRad21 antibodies specifically recognize the recombinant hRad21 protein synthesized in rabbit reticulocyte lysates. Protein synthesis in rabbit reticulocyte lysates was done in the the cells with colcemid for 4 h. These G 1 /S-arrested and metaphase-arrested cells, along with the control exponentially growing cells, were labeled with [ 32 P]orthophosphate for 2 h prior to each harvest. FACscan analysis of parallel cultures indicated that the two populations of cells were indeed arrested at G 1 /S and metaphase, respectively (Fig. 4A). hRad21 was immunoprecipitated from the exponentially growing, G 1 /S-arrested and metaphase-arrested cell lysates using anti-hRad21 antibodies. Three samples containing different protein amounts (1ϫ, 3ϫ, and 9ϫ; where 3ϫ and 9ϫ samples contained 3-and 9-fold more protein than 1ϫ samples, respectively) were analyzed for each type of immunoprecipitate using SDS-PAGE and blotted onto a membrane. The membrane was subsequently subjected to autoradiography (Fig. 4B), followed by Western blotting with anti-hRad21 antibodies (Fig. 4C). When the anti-hRad21-positive signals in the Western analysis were quantified, the 3ϫ and 9ϫ samples showed an almost 3-fold difference in their signal intensities, whereas the 1ϫ samples were undetectable. This apparently linear dose-response relationship observed in the Western blotting data for 3ϫ and 9ϫ samples indicates that the intensities reliably reflect protein abundance in this range of protein amounts. In contrast, autoradiography intensities of the 9ϫ samples displayed more than a 3-fold intensity difference relative to the 3ϫ samples and therefore seemed to lack a linear dose-response relationship. Nevertheless, when we calculated relative phosphorylation lev-els by dividing the autoradiography intensities by the Western blot intensities, we saw consistent cell cycle-specific changes in hRad21 phosphorylation levels both among the 3ϫ samples and among the 9ϫ samples (Fig. 4D). In both comparisons, the colcemid-arrested cells showed 3-fold increases in the specific phosphorylation levels of hRad21 compared with the G 1 /Sarrested cells. Therefore, we concluded that hRad21 is hyperphosphorylated in M phase.
In the autoradiographs of interphase immunoprecipitates, we observed three major and several minor 32 P-labeled bands additional to hRad21 (Figs. 3B and 4B). The apparent molecular masses of the major bands as determined from SDS-PAGE were about 150, 140, and 80 kDa (Fig. 3B). Because these three bands were not reactive with anti-hRad21 antibodies in the Western blotting analysis, we concluded that these three bands are not hRad21. We interpret them as being hRad21-associated phosphorylated proteins and have designated them p150, p140, and p80. Interestingly, the phosphorylation levels of hRad21 and its putative associated proteins (p150, p140, and p80) in the metaphase-arrested cells were significantly higher than those found with the interphase cells (Figs. 3B and 4B). Furthermore, additional phosphorylated proteins, with apparent molecular masses of 180, 95, and 85 kDa, were specifically found in the hRad21 immunoprecipitate derived from the metaphase-arrested cells. Again, these three proteins were interpreted as putative hRad21-associated proteins and designated p180, p95, and p85. Although we do not know the identity of these bands, the results suggest that cohesin components may be coordinately modified and/or regulated by phosphorylation during M phase.
hRad21 Is a Nuclear Protein Mostly Excluded from Chromosomes in Mitosis-We next examined the subcellular localization of hRad21 in the asynchronous HeLa cells by IF experiments using anti-hRad21 antibodies. HeLa cells were fixed with methanol, with or without being permeabilized in 0.1% Triton X-100. The samples were stained with TOTO3 to stain the DNA and examined by IF using anti-hRad21 and anti-␣tubulin antibodies. hRad21 was detected in interphase nuclei and was not present in nucleoli (Fig. 5). When the cells were pretreated with Triton X-100 prior to fixation, most of the ␣-tubulin was extracted to disappearance, whereas hRad21 remained. These results confirm the biochemical results described above and indicate that hRad21 is a nuclear protein associated with nuclear structures.
If hRad21 is involved in sister chromatid cohesion, as has been proposed, one would expect the subcellular localization of hRad21 to change in M phase. However, it has been reported that most Xenopus Rad21 is dissociated from prophase-like chromosomes (8). We therefore were interested in the intracellular distribution of hRad21 in metaphase cells. Two fixation methods, one employing cold methanol and the other involving paraformaldehyde (see "Experimental Procedures"), were used. We found these two protocols essentially gave rise to the same results, and the results obtained from methanol-fixed cells are shown in Fig. 6 and described below. Asynchronous HeLa cells stained with DNA dye TOTO3 were examined by IF using anti-␣-tubulin and anti-hRad21 antibodies (Fig. 6). Although some hRad21 seemed to be associated with chromatin during early prophase when the two centrosomes are not yet separated, a significant amount was already dissociated from chromatin (Fig. 6, row a). hRad21 was quite heterogeneously distributed in late prophase cells, with some fraction of it still apparently bound to chromatin (Fig. 6, row b). However, in metaphase and anaphase, it seemed that most hRad21 had completely dissociated from the chromatin and was associated with the spindles (Fig. 6, rows c-e). The microtubule-hRad21

FIG. 2. hRad21 protein levels do not change during cell cycle.
HeLa cells were synchronized at early S phase following a thymidine and aphidicolin double-block protocol. The cells were harvested at intervals after release from the block. A, FACscan analysis. The cells were stained with PI, and the DNA contents were analyzed by FACscan. B, Western blotting analyses for hRad21, cyclin B1, and ␣-tubulin. The cells harvested from the synchronized culture (indicated by the harvest time after block release), as well as an asynchronized exponentially growing culture (Expo), were first extracted with 1% Triton X-100 (T) and then with 0.5 M NaCl (N). The proteins were run in a SDS-PAGE gel and analyzed in Western blotting experiments using anti-C-hRad21, anti-cyclin B1, and anti-␣ tubulin antibodies. Cell cycle stages deduced from A and C, and the cyclin B1 abundance are shown above. C, the insoluble fractions (P) were analyzed for PCNA by Western blotting with anti-PCNA antibodies. association was particularly obvious in telophase and cytokinesis, because part of the hRad21 was colocalized at midbody. hRad21-chromatin association resumed during cytokinesis and completed in nascent two daughter cells.
To understand the relationship of the nuclear envelope breakdown and hRad21 dissociation from chromatin, we examined asynchronous cells by IF using anti-hRad21 and antilamin B antibodies (Fig. 7). At entry to mitosis, a major population of hRad21 was already excluded from chromatin of cells that had been already undergone nuclear envelope breakdown (Fig. 7, row a). It was not possible to determine whether this fraction of hRad21 had once associated with the chromatin and then dissociated from it in prophase or whether the fraction had never associated with the chromatin. However, some additional fraction of hRad21 apparently dissociated as the chromatin further condensed (Fig. 7, row b). At the exit of M phase, a fraction of hRad21 remained in the cytoplasm even when the nuclear envelope had apparently reformed (Fig. 7, row c). Therefore, it is suggested that hRad21 might be imported to nucleus actively. Eventually, nearly all hRad21 was localized inside the nuclei (Fig. 7, row d).
Significant Fractions of hRad21 Are Associated with Prometaphase Chromosomes and Metaphase Centromeres-Because of the background signals produced by the chromatin-unbound hRad21 in mitotic cells, it was difficult to determine whether hRad21 associates with condensed chromosomes in M phase or not. We therefore examined spread prometaphase-like and metaphase-like chromosomes to clarify whether hRad21 is present on condensed chromosomes. To this end, synchronized HeLa cells were treated with colcemid, and the spread chromosomes were analyzed by IF. We found that significant amounts of hRad21 associated along the axes of the less condensed prometaphase-like chromosomes (Fig. 8A, row a). These signals were not observed when the antibody was preabsorbed by the antigen and when chromosomes were stained by normal rabbit IgG, indicating that the signals were specific (data not shown). The hRad21 signals appeared heterogeneous along the length of chromosomes. To better understand the relative distribution of hRad21 on the chromosomes, we also localized centromeres using antibodies recognizing CENP-B, a protein that binds to centromeric alphoid DNAs. We found that hRad21 was localized at both the centromeres and the arm regions on prometaphase-like chromosomes (Fig. 8A, row a).
In contrast, when we analyzed further condensed metaphase-like chromosomes, we found that hRad21 became localized at centromeres (Fig. 8A, row b). Images under a high magnification revealed that hRad21 mostly disappeared from the arm regions and specifically localized at centromeres of metaphase-like chromosomes (Fig. 8B). Interestingly, when CENP-B signals were recognized as two spots, hRad21 was detected not only at CENP-B-positive regions but also at inner regions between the two sister CENP-B signals (Fig. 8B, rows  a and c). Taken together, these observations strongly suggest that hRad21 in M phase is dissociated from chromatin in two steps. The arm region-associated hRad21 is dissociated during the prophase to metaphase transition, whereas the centromere-bound hRad21 remains associated with metaphase-like chromosomes. DISCUSSION It is estimated that about 70% of total Scc1p/Mcd1p is associated with chromatin in budding yeast during metaphase (18), whereas vertebrate cohesin molecules are mostly excluded from chromatin in prophase (8,9). Therefore, it has not been determined whether cohesin plays a role in the mitotic chromatid cohesion in higher eukaryotes. In this study, we have found that a small but significant fraction of hRad21 is localized at centromeres of colcemid-induced metaphase-like chromosomes. Moreover, hRad21 and its potential associated proteins are hyperphosphorylated in M phase. These results highly suggest that hRad21 has a critical role in the chromatid cohesion and separation in M phase, as has been demonstrated in budding yeast.
In human and mouse cells, an electron microscopy study revealed that sister chromatid arm regions are morphologically distinguishable with each other prior to metaphase, whereas   FIG. 3. hRad21 and associated proteins are hyperphosphorylated in M phase. HeLa cells were synchronized at early S phase following a thymidine and aphidicolin double-block protocol, and the cells were harvested at intervals after release from the block. The cells were cultured for 1 h in the presence of [ 32 P]orthophosphate prior to each harvest. A, FACscan analysis. Synchronized cells (indicated by harvest times) and exponentially growing cells (Expo) were stained with PI, and the DNA contents were analyzed by FACscan. Because the samples were fixed at different times, absolute values for the PI signals are not consistent between the samples. B, autoradiography of labeled proteins. Harvests containing approximately the same number of synchronized and exponentially growing (Expo) cells were subjected to immunoprecipitation either with anti-hRad21 antibodies (I) or preimmune normal rabbit immunoglobulin (P). The immunoprecipitates were fractionated in an SDS-PAGE gel and blotted onto a membrane. The membrane was exposed to an x-ray film to visualize the labeled proteins. The position of hRad21 as identified by Western blotting in C is shown. The positions of four additional labeled proteins, p150, p140, p95, and p80, are also indicated. C, Western blotting of hRad21 with the same membrane analyzed as in B.
the centromeric regions are connected until the onset of chromatid segregation in anaphase (19). This implies that chromatids may be separated in a two-step process in mammalian cells. In the first step, the arm regions are disjoined during prophase to metaphase. At this stage, the remaining cohesion at centromeres serves as a hinge connecting two sister chromatids. In the second step that happens at the metaphase-anaphase transition, the centromere cohesion is dissolved, and the two liberated chromatids separate and move toward centrosomes by pulling forces exerted by spindles. Such a two-step cohesion-separation event has been suggested cytologically in a variety of animals such as Drosophila, chicken, muntjac, mouse, and human (reviewed in Ref. 20).
The changes of hRad21 distribution that we have found with mitotic chromosomes are remarkably well correlated with these cytological observations. We have found that hRad21 is present along the entire length of chromosomes during prophase to prometaphase. However, the arm-associated hRad21 disappeared and only the centromere-associated hRad21 remains on the colcemid-induced metaphase-like chromosomes. Therefore, the dynamics of hRad21 distribution on mitotic chromosomes perfectly corresponds to the morphological changes of cohesion between mitotic chromatids. These results suggest that cohesin containing hRad21 is responsible for the chromatid cohesion in M phase. After this paper was submitted for publication, it was reported that a small population of SA1 protein is associated with metaphase chromosomes both in Xenopus egg cell-free extracts and Chinese hamster ovary cells (21). Even more recently, mouse Scc1/Rad21 protein expressed ectopically in HeLa cells was shown to be specifically associated with centromeric regions in metaphase chromosomes (22). Our study indicates that the endogenous hRad21 protein behaves as the ectopically expressed protein and further substantiates the likelihood of a highly regulated distribution of cohesin complex on vertebrate mitotic chromosomes as revealed in the above-mentioned studies.
Scc1p/Mcd1p has been shown to associate with specific regions along the chromosome arms, especially at centromereproximal regions (23)(24)(25). Especially, yeast cohesion becomes further preferentially localized at centromere-proximal regions on metaphase-like chromosomes induced by a microtubule inhibitor methyl 2-benzimidazolecarbamate (23). Therefore, in both yeast and humans, cohesin is enriched at centromereproximal regions just prior to the onset of anaphase. It is tempting to speculate that these particular cohesin fractions associated with centromeres are the target of the APC (anaphase promoting complex)-dependent pathway that triggers anaphase. The recent study by Waizenegger et al. (22) supports this hypothesis by analyzing the biochemical and cytological behaviors of HeLa cells cohesin and mitotic processes in Xenopus cell-free extract.
Interestingly, we found that the centromere-associated hRad21 on the metaphase-like chromosomes appeared to be present between two sister CENP-B signals in some cases. Molecular cloning of the alphoid repeats on human chromosome 21 identified two different classes of alphoid repeats, ␣21-I and ␣21-II (26). ␣21-I contains the CENP-B box to which CENP-B binds, but ␣21-II does not. Correspondingly, it has FIG. 4. hRad21 and its putative associated proteins are hyperphosphorylated in M phase. A, exponentially growing (Expo), G 1 /S-arrested (G1/S), and colcemid-arrested (Meta) HeLa cells were analyzed for their DNA contents by FACscan. B, extracts containing the same 32 P counts of cells were subjected to immunoprecipitation with either anti-hRad21 antibodies (1ϫ, 3ϫ, and 9ϫ), or with preimmune normal rabbit immunoglobulin (P). Three different amounts of protein (1ϫ, 3ϫ, and 9ϫ) were analyzed for each type of immunoprecipitates. Immunoprecipitates were separated by SDS-PAGE and blotted onto a membrane. The membrane was then exposed to an x-ray film to visualize the labeled proteins. been reported that ␣21-I is precisely colocalized with CENP-B signals on spread metaphase-like chromosomes, but ␣21-II distributes over a broader area. Our observation that hRad21 distributed in broader regions that included the sister CENP-B signals and the inner paring domain located between the two CENP-B signals, suggests that a subset of alphoid DNAs sim-ilar to ␣21-II may be involved in sister chromatid cohesion at centromeres.
Several aspects of the behavior of hRad21 are different from that of yeast Scc1p/Mcd1p. We found that hRad21 levels do not change during the cell cycle. This finding is different from that of budding yeast, wherein Scc1p/Mcd1p levels have been found to fluctuate throughout the cell cycle. We found that only a small fraction of hRad21 associated with prophase chromatin. The remaining and major population of hRad21 did not locate to chromatin in the earliest prometaphase cells, wherein the nuclear envelope breakdown had already occurred. We could not determine whether the major fraction of hRad21 we observed had once bound with chromatin and dissociated in early prophase or whether it had been unbound to chromatin throughout the cell cycle. Nevertheless, we conclude that only a minor fraction of hRad21 contributes to chromatid cohesion in M phase, a situation different from that of yeast Scc1p/ Mcd1p (18). This observation suggests that hRad21 may play a role additional to chromatid cohesion. In support of this notion, the mitotic hRad21 protein, the majority of which is dissociated from chromatin (Figs. 6 and 7), was resistant to detergent extraction (Fig. 2B). Therefore, it is suggested that this "free" form of hRad21 is associated with other nuclear structures, such as nuclear matrix and spindles. We noted that a fraction of hRad21 appeared to closely associate with spindle midzones and midbody in anaphase and telophase, respectively. This behavior is reminiscent of that of the chromosome passenger proteins INCENP (inner cen/tromere proteins) (27). It has been recently proposed that a primary function of INCENP involves chromosome segregation processes (28). It may be interesting to investigate the possible roles of hRad21 in later stages of mitosis.
It is not known how the dissociation of hRad21 from mitotic chromatin is regulated. In budding yeast, Scc1p/Mcd1p is cleaved proteolytically in a manner dependent on Esp1p at the onset of anaphase. Similarly, it has been reported very recently that a small fraction of mammalian Scc1/Rad21 is proteolytically cleaved at the metaphase-to-anaphase transition (22). In this study, we could not demonstrate a similar proteolysis that FIG. 5. hRad21 is a nuclear protein associated with nuclear strucutres. Asynchronous HeLa cells were stained with TOTO3 (red) and analyzed by IF using anti-hRad21 (green) and anti-␣-tubulin antibodies (blue). Before cold methanol fixation, cells were pretreated (ϩ) or not (Ϫ) with 0.1% Triton X-100.  Asynchronous HeLa cells were stained with DNA dye TOTO3 (red) and examined by IF using anti-hRad21 (green) and anti-lamin B antibodies (purple). a, a prometaphase cell with the nuclear envelope disassembled; b, a prometaphase cell with further condensed chromosomes; c, two nascent daughter cells in which the nuclear envelopes apparently reformed; d, two nascent daughter cells in which most hRad21 is localized in nuclei. The scale bar shown in d represents 10 m. specifically happened to anaphase hRad21. However, we have found that hRad21 is phosphorylated in a cell cycle-dependent manner. Significantly, hRad21 and its potentially associated proteins, p180, p150, p140, p95, p85, and p80, were correspondingly phosphorylated most heavily in M phase. Recently, it has been reported that Xenopus Scc3 homologues (XSA1 and 2) are phosphorylated by Cdc2-cyclin B in vitro. Phosphorylated SA proteins show low affinities for chromatin, and thus it has been proposed that this protein modification might prevent unbound SA proteins in M phase from reassociating with chromatin (11). Interestingly, p140 and p150, which were coimmunoprecipitated with anti-hRad21 in this study, showed apparent molecular masses in SDS-PAGE similar to those of human SA pro-teins (140 kDa) (11), suggesting that p140 and p150 may be human SA proteins. Xenopus cohesin contains a fifth unidentified component having an apparent molecular mass of 95 kDa in SDS-PAGE (8). p95 that was also coimmunoprecipitated with anti-hRad21 in this study may be the human counterpart of this Xenopus cohesin component. Identities of the other phosphorylated proteins (p180, p85, and p80) revealed in this study are not known. p180, p95, and p85 are particularly interesting because these phosphorylated proteins were not detected in the interphase hRad21 immunoprecipitates (Fig.  4B). These proteins may be phosphorylated strictly in mitosis or associated with the cohesin complex specifically in mitosis. Taken together, these results suggest that numerous compo- nents of human cohesin are coordinately modified by phosphorylation in M phase, which may regulate the function of cohesin. Future investigations of hRad21 should undoubtedly be aimed at understanding how arm-associated and centromereassociated fractions of hRad21 are differentially regulated.