CENP-B Interacts with CENP-C Domains Containing Mif2 Regions Responsible for Centromere Localization*

Recently, human artificial chromosomes featuring functional centromeres have been generated efficiently from naked synthetic alphoid DNA containing CENP-B boxes as a de novo mechanism in a human cultured cell line, but not from the synthetic alphoid DNA only containing mutations within CENP-B boxes, indicating that CENP-B has some functions in assembling centromere/kinetochore components on alphoid DNA. To investigate whether any interactions exist between CENP-B and the other centromere proteins, we screened a cDNA library by yeast two-hybrid analysis. An interaction between CENP-B and CENP-C was detected, and the CENP-C domains required were determined to overlap with three Mif2 homologous regions, which were also revealed to be involved in the CENP-C assembly of centromeres by expression of truncated polypeptides in cultured cells. Overproduction of truncated CENP-B containing no CENP-C interaction domains caused abnormal duplication of CENP-C domains at G2 and cell cycle delay at metaphase. These results suggest that the interaction between CENP-B and CENP-C may be involved in the correct assembly of CENP-C on alphoid DNA. In other words, a possible molecular linkage may exist between one of the kinetochore components and human centromere DNA through CENP-B/CENP-B box interaction.

Recently, mammalian artificial chromosomes have been generated efficiently from naked genomic alphoid DNA and a synthetic alphoid DNA containing CENP-B boxes as a de novo mechanism in a human cultured cell line, but not from alphoid DNA lacking CENP-B boxes or from synthetic alphoid DNA only containing point mutations within the CENP-B boxes (45)(46)(47)(48). On these artificial chromosomes, CENP-A, CENP-B, CENP-C, and CENP-E assemble like functional centromeres on normal human chromosomes. Mammalian artificial chromosomes are maintained stably through cell divisions (46 -48). These data indicate that the CENP-B/CENP-B box interaction has some crucial roles for assembling other essential centromere/kinetochore components on the naked alphoid DNA.
The purpose of the present study was to investigate whether any interactions exist between CENP-B and the other centromere proteins. Using yeast two-hybrid analysis, interactions between CENP-B and CENP-C were detected and the required CENP-C domains were determined to overlap with three Mif2 homologous regions. Overproduction of CENP-B without the CENP-C interaction domains caused abnormal assembly of CENP-C at G 2 and cell cycle delay at metaphase.
The CENP-C gene was cloned with screening of a human HeLa S3 5Ј-STRETCH PLUS cDNA library probed with the PCR product CENP-C2185-2847. CENP-C1-850 was cloned by PCR with the cDNA library and TaKaRa Ex Taq (TaKaRa). All cloned DNAs were sequenced with an AutoRead sequencing kit (Amersham Biosciences) and a dRhodamine dye terminator kit (Applied Biosystems) following the protocol from the manufacturer.
Two-hybrid Analysis-Yeast transformation and ␤-galactosidase assays were performed according to the procedure of Kitagawa et al. (32). For screening, 25 g of a mouse testis cDNA library in pGADGH was transformed into yeast strain L40 containing LexACB132-599. Yeast cells corresponding to 5 ϫ 10 6 transformants were plated onto Yc/ THULL plates (ϪTrp/ϪHis/ϪUra/ϪLys/ϪLeu) containing 50 mM 3-aminotriazole to screen for transcriptional activation of the HIS3 gene. After incubation for 3 days at 30°C, yeast colonies were isolated and streaked on Yc/UTL plates (ϪTrp/ϪUra/ϪLeu) to obtain single colonies. Three colonies were analyzed using a ␤-galactosidase filter assay.
Isolation of HeLa Nuclei and Preparation of Soluble Chromatin-90% confluent HeLa cells were washed twice with PBS 1 and once with isolation buffer (3.75 mM Tris-HCl, pH 8.0, 0.05 mM spermine, 0.125 mM spermidine, 0.5 mM EDTA, 20 mM KCl, 0.1 mM phenylmethylsulfonyl fluoride, 0.5 mM DTT, 40 M z-Leu-Leu-Leu-H aldehyde (Peptide Institute, Japan)). Isolation buffer containing 0.1% digitonin (Wako) was then added with incubation on ice for 10 min. Cells were collected with a cell scraper, centrifuged at 4°C for 5 min at 1,000 rpm with an RL-131 (Tomy), and suspended in isolation buffer containing 0.1% digitonin before homogenization with a Dounce homogenizer. Nuclei were collected by centrifugation at 4°C for 5 min at 1,000 rpm, re-suspended in wash buffer (20 mM HEPES, pH 8.0, 0.5 mM EDTA, 20 mM KCl, 0.1 mM phenylmethylsulfonyl fluoride, 0.5 mM DTT, 40 M z-Leu-Leu-Leu-H aldehyde) at 1 ϫ 10 8 nuclei/ml, and transferred to 1.5-ml tubes. Nuclei were collected by centrifugation for 5 min at 3,000 rpm with an MRX-150 (Tomy) and re-suspended in buffer A (5 mM HEPES, pH 8.0, 1.5 M aprotinin, 10 M leupeptin, 1 mM DTT, 40 M z-Leu-Leu-Leu-H aldehyde). Nuclei were collected by centrifugation and re-suspended again with buffer A containing 1 mM CaCl 2 . Microccocal nuclease (Amersham Biosciences) was added to 200 units/ml, followed by incubation for 5 min at 37°C. The enzyme reaction was stopped by adding EGTA to 2 mM, and the supernatant suspension was recovered by centrifugation at 4°C for 5 min at 8,000 rpm. After precipitation in buffer A containing 2 mM EGTA and sonication with a Sonifier 450 (Branson), the supernatant was recovered by centrifugation at 4°C for 5 min at 8,000 rpm.
Immunoprecipitation with Anti-CENP Antibodies-Aliquots of 150 l of soluble chromatin were transferred to siliconized tubes and mixed with 150 l of 2ϫ IP buffer (55 mM HEPES, pH 8.0, 600 mM NaCl, 4 mM MgCl 2 , 4 mM ATP, 1.5 M aprotinin, 10 M leupeptin, 1 mM DTT, 40 M z-Leu-Leu-Leu-H aldehyde, 0.2% Nonidet P-40). Antibodies against CENP-A, -B, and -C were added, and samples were rotated at 4°C for 16 h. Protein G-Sepharose (Amersham Biosciences) or Dynabeads M-280 sheep anti-mouse IgG (Dynal) were washed twice with IP buffer (30 mM HEPES, pH 8.0, 300 mM NaCl, 2 mM MgCl 2 , 2 mM ATP, 1.5 M aprotinin, 10 M leupeptin, 1 mM DTT, 40 M z-Leu-Leu-Leu-H aldehyde, 0.1% Nonidet P-40), blocked with IP buffer containing 5% BSA at 4°C for 30 min, added to samples, and rotated for 1 h. Sepharose was collected by centrifugation at 2,000 rpm for 1 min and washed five times with IP buffer. The magnetic Dynabeads were collected with Dynal MPC-M (Dynal) for 1min and washed five times with IP buffer. Beads were suspended in SDS sample buffer and boiled for 5 min before electrophoresis.
Isolation of Cell Lines Expressing CENP-B Polypeptides-To establish cell lines expressing the CENP-B N-terminal polypeptide (CB1-160) or CENP-B (CB1-599), HeLa cells were transfected with pMTCB1-160c and pSV2neo or pMTCENP-B and pSV2neo by co-precipitation with calcium phosphate, and resistant colonies were selected with 400 g/ml Geneticin.
Cell Cycle Analysis of Cells Expressing CENP-B Polypeptides-For synchronization, CB1-160-expressing cells were treated with 2.5 mM thymidine for 20 h, washed twice, and incubated for 10 h in the absence of the drug. Cells were then incubated in the presence of 5 g/ml aphidicolin for 20 h. After its removal, 0.6 M cadmium chloride was added to the medium. Cells were harvested every 1 and 4 h, fixed onto coverslips, and stained with DAPI, and then the mitotic cells were counted. To analyze the cell cycle of CB1-561-expressing cells, 293S cells were transfected with pMTCENP-B or pMTCB1-561c by co-precipitation with calcium phosphate. At 1.5 and 2.5 days following the induction of the polypeptides, cells were fixed and indirect immunofluorescent staining was performed. To determine the cell cycle of these polypeptide-expressing cells, nuclei were counterstained with propidium iodide and the DNA content was measured with an IPLab (Signal Analytics Corp.).
Transfection of Expression Plasmids and Indirect Immunofluorescent Staining-To express RFP-CENP-C deletion mutants, tetON HeLa (Clontech) cells were grown on poly-D-lysine-coated coverslips and transfected with Polyfect transfection reagent (Qiagen) using the protocol from the manufacturer for HeLa cells. Cells were fixed with 75% acetone at Ϫ20°C for 30 min. The cells were then fixed again with periodate-lysine-paraformaldehyde at 4°C for 30 min, and finally washed three times in PBS for 5 min. Monoclonal anti-CENP-B (2D8D8), monoclonal anti-T7tag (Novagen) or guinea pig polyclonal anti-CENP-C antibodies were applied to the coverslips for 1 h at 37°C, followed by washing three times in PBS for 5min each. Fluorescein isothiocyanate-or rhodamine-conjugated anti-mouse IgG or anti-guinea pig IgG antibodies in PBS containing 0.2% BSA and 0.2% skim milk were then applied to the coverslips for 1 h at 37°C. After washing three times in PBS for 5 min each, the coverslips were stained with 1 g/ml DAPI in PBS for 5 min and rinsed again in PBS. Samples were finally mounted and analyzed using a Zeiss microscope equipped with a Photometrics cooled CCD camera (PXL 1400-C1) coupled to an image analysis system (IPLab, Signal Analytics Corp.). In Figs. 8 -10, images were taken at 0.2-m z-steps using Zeiss microscope equipped with a cooled CCD camera (MicroMax, Princeton Instruments) coupled to Meta-Morph (Universal Imaging) and used to produce deconvolved recon-structions of the image stacks with AutoDeblur (AutoQuant Imaging). Images are presented as maximum intensity projections. In Fig. 7B and 9B, z-stack images were processed with HazeBuster (VayTek).
Antibodies-The anti-CENP-A antibody was generated as described previously (47,51). To prepare anti-CENP-C antibodies, CENP-C 630 -943 polypeptides were purified by the inclusion body method, and then fractionated in a 13% polyacrylamide gel including 0.1% SDS with the Prep Cell System (Bio-Rad). They were then used to immunize rabbits, guinea pigs, and mice. Immunoglobulin G fractions were obtained using protein A-Sepharose (Amersham Biosciences) for the rabbit serum (CRa1) and the guinea pig serum (CGP2), and protein G-Sepharose (Amersham Biosciences) for the mouse monoclonal antibody. To prepare monoclonal anti-CENP-B antibodies, mice were immunized with purified CENP-B polypeptides and then the immunoglobulin G fractions were obtained with protein G-Sepharose. To prepare anti-CENP-B antibody against CENP-B COOH-terminal region (BC1), rabbits were immunized with purified CENP-B541-599 polypeptides. Then, the immunoglobulin G fractions were purified by affinity for CENP-B541-599 polypeptide. The CENP-B polyclonal antibody, BN1, was generated as described previously (32)

Detection of CENP-B and CENP-C Interactions Using a
Yeast Two-hybrid System-Using a mammalian artificial chromosome assay, de novo centromere assembly occurred with the alphoid DNA depending on the existence of functional CENP-B boxes. Therefore, we analyzed possible interactions between CENP-B and centromere proteins using a yeast two-hybrid system. We introduced bait plasmids encoding a LexA DNA binding domain fused with the CENP-B central and COOHterminal domains (LexACB132-599) into yeast L40 cells. By two-hybrid screening from a prey plasmid encoding a GAL4 activation domain fused with a mouse cDNA library, a truncated mouse CENP-C containing amino acids 671-907 was obtained (Fig. 1A). We confirmed that the yeast cells expressing the GAL4 activation domain fused with a human CENP-C (GADCENP-C) could grow as well as positive controls express- ing GADCB368 -599 in which the CENP-B homodimerization domain was included ( Fig. 1A) (32). To exclude artifactual transcriptional activation in the yeast, the bait and prey plasmids were exchanged but the interaction between CENP-C and CENP-B368 -599 was still detectable (Fig. 1B). CENP-C homotypic interaction was also detected by two-hybrid analysis (Fig.  1B). However, we could not detect any interactions between CENP-A and CENP-B or between CENP-A and CENP-C (Fig.  1, A and B).
CENP-B Interaction with CENP-C Involves Its Acidic Region-To determine which domain of CENP-B interacted with CENP-C, we generated CENP-B deletion mutants and performed assays with the two-hybrid system. When plasmids encoding LexACB132-556 or LexACB132-465 and GAD-CENP-C were co-transformed, colony formation on selective plates and a high level of ␤-galactosidase activity were detected (Fig. 2). In contrast, no colony formation and a low level of ␤-galactosidase activity were observed with GADCENP-C and LexACB132-405 or the acidic region deletion mutant LexACB132-599⌬405-465 (Fig. 2).
To confirm the interaction between the CENP-B acidic region and CENP-C in HeLa cells, we generated cell lines expressing tagged CENP-B containing only the N-terminal DNA binding domain, (CB1-160), or full-length CENP-B (CB1-599). Ando et al. reported that CENP-A, -B, and -C co-immunoprecipitated with the chromatin complex corresponding to the position of the tetra-to pentanucleosome units solubilized with micrococcal nuclease treatment (51). Solubilized chromatin fractions were prepared from the tagged CENP-B-expressing cell lines under more restrictive conditions using micrococcal nuclease treatment and sonication (Fig. 3A), and immunoprecipitation analyses was carried out with the anti-CENP-C antibody. Endogenous CENP-A, -B, and -C co-immunoprecipitation was confirmed (Fig. 3A). From the efficiency of our immunoprecipitation assay, ϳ20 -25% of the total endogenous CENP-B was estimated to be included in the same chromatin complex with CENP-C (data not shown).
CENP-B binds to CENP-B boxes in alphoid DNA via an N-terminal 125 amino acid domain, and therefore the CB1-160 peptide also specifically binds to chromosomal alphoid DNA containing CENP-B boxes (32, 34, 36; see Fig. 8). If the complex formation including CENP-B and CENP-C were mediated only by a presumptive independent interaction of CENP-C to alphoid DNA, then CB1-160, which also bound to the CENP-B box, must have been co-immunoprecipitated with CENP-C as well as endogenous CENP-B or tagged full-length CENP-B (CB1-599). CB1-599 also co-immunoprecipitated with CENP-C as well as endogenous CENP-B, whereas CB1-160 did not (Fig. 3B). Moreover, the amount of co-immunoprecipitated endogenous CENP-B was decreased when CB1-160 was overproduced. Thus, CB1-160 may compete with endogenous CENP-B for binding to the CENP-B box. These results indicate that the interaction between CENP-B and CENP-C does indeed exist and was lost without the CENP-C interaction domain in CENP-B in cells.
Direct interaction between CENP-B and CENP-C was then confirmed with polypeptides generated by in vitro transcription and translation (Fig. 3C). CENP-C and CB1-160 (lane 1) or CB1-599 (lane 2) were, respectively, mixed and immunoprecipitated with anti-CENP-C antibody (lanes 5 and 6) or control IgG (lanes 3 and 4). CB1-599 co-immunoprecipitated with CENP-C, whereas CB1-160 did not (Fig. 3C). All these results as well as the results from the two-hybrid analyses indicate the possible molecular interaction between one of the kinetochore components, CENP-C, and the human centromere DNA-binding protein CENP-B.
CENP-C Mif2 Homologous Regions Are Involved in CENP-B Interactions-To determine the interaction domain of CENP-C, we generated deletion mutants for the two-hybrid analysis. Interaction between CB132-599 and CENP-C (CC728 -943) was identified (Figs. 1A and 4). The latter domain contains two regions that are highly conserved with S. cerevisiae Mif2 (20). We therefore generated truncated mutants deleting either region of Mif2. None of the resultant polypeptides, CC728 -886, CC728 -786, CC886 -943, or CC787-943, interacted with CENP-B on two-hybrid analysis (Fig. 4). Unexpectedly, however, CC1-727 lacking both of the two COOH-terminal Mif2 blocks did show some interaction activity. Analysis of deletion constructs from the COOH-terminal region showed that CC1-429 and CC283-429, but not CC1-283, could interact (Fig. 4). Therefore, CC283-429 containing another Mif2 region (block1) also had the capacity to interact with CENP-B. Weak ␤-galactosidase activity was detected with CC430 -727, but no colonies were formed on selective plates (Fig. 4). These results indicate that CENP-C interacts with CENP-B through two regions, 283-429 and 727-943, both of which contain Mif2 homologous regions.
Mif2 Regions of CENP-C Are Involved in Its Localization to Centromeres-In previous reports, centromere targeting of CENP-C was indicated to be the result of the central region (amino acids 478 -537) by Yang et al. (26) and the carboxylterminal region (amino acids 584 -943) by Lanini and McKeon (27). Full-length and truncated CENP-C polypeptides fused with RFP were transiently expressed in HeLa cells to analyze their localization (Fig. 5 and Table I). RFP-CENP-C (fulllength) was detected at the centromere (97.4% of cells), and abnormal localization was only observed at a very low frequency (2.6% of cells) (Fig. 5 and Table I). Truncated CENP-C, RFPCC1-727, RFPCC728 -943, and RFPCC1-429 containing the CENP-B interaction domains were detected at the centromere like full-length CENP-C, although abnormal localizations of RFPCC1-429 to nucleoli was more frequent at 5.8% or to whole nuclei at 2.3%. RFPCC1-283 containing no CENP-B interaction domain did not show any signal (Table I)  cleoli together (Fig. 5 and Table I). RFPCC430 -727 was localized at centromeres with a very low frequency (0.5%) and in the whole nucleus in 95% cells, consistent with an earlier report for CENP-C 421-591 (28). The localization signals of truncated CENP-C polypeptides at the centromere drastically decreased with replacement of the RFP tag with a FLAG tag (data not shown). Thus, the stability of truncated CENP-C polypeptides may be influenced by the tagged sequences. RFP without the fusion protein showed no specific affinity for the centromere, and thus the RFP tag is very effective for CENP-C detection. These results indicate that the localization of CENP-C at the centromere mainly depends on the domains containing the three Mif2 homologous regions, although a weak centromere localization activity exists in 430 -727.
Overproduction of a Truncated CENP-B Polypeptide Lacking the CENP-C Interaction Domain Affects Cell Cycle Progression at Mitosis-To examine whether the existence of CENP-B lacking the CENP-C interaction domain on the centromeric alphoid locus affected the assembly of the centromere structure or cell cycle progression, we induced overproduction of the CENP-B N-terminal polypeptide (CB1-160) in a stable HeLa cell line (HBN160-E cells). Within 24 h after induction, a 5-10-fold excess of the CB1-160 polypeptide was detected with the anti-CENP-B antibody, as compared with the amount of endogenous CENP-B (Fig. 6A). When the cells were synchronized in the G 1 /S phase using drugs, and then the number of mitotic cells was counted with every 4 h (Fig. 6B) or 1 h (data not shown) after simultaneous release from the G 1 /S block and induction of CB1-160 expression, the majority of tested cells, both normal HeLa and HBN160-E cells, entered mitosis within 12 h-13 h. After 16 h, the mitotic index decreased in control cells, but HBN160-E cells still showed a high rate (Fig. 6B). Similar results were obtained when the induction of CB1-160 expression was started before the G 1 /S block release, indicating that CB1-160-overproducing cells progressed through the cell cycle normally from the G 1 /S phase to the beginning of mitosis, but then took longer to exit.
To further analyze details of the mitotic state of the CB1-160-overproducing cells, we observed random cultured HBN160-E cells at 2-h intervals. Five of the HBN160-E cells in the uninduced state (arrowheads) in this microscopic field changed their shape to round mitotic cells and 2 h later, all finished cell division and became flat (Fig. 6C). However, many of the HBN160-E cells in the induced state for CB1-160 overproduction took over 2 h to complete cell division (Fig. 6C, small arrows) and two cells stayed in mitosis for 8 -10 h (Fig.  6C, arrows). The chromosome DNA alignments of these two cells indicated that one cell was arrested at prometaphase (Fig.  6C, inset, lower cell) and the other was arrested just before metaphase because of the existence of a few chromosomes apart from the metaphase plate (Fig. 6C, inset, large arrowheads). Indeed, the mean mitotic period of individual dividing cells was prolonged to 71.7 min (n ϭ 22) in the CB1-160overproducing case, compared with 55.2 min (n ϭ 24) for cells without induction, as estimated by time lapse analysis, and 9.1% of the cells were arrested for longer than 6 h (Fig. 7A). These results indicate that overproduction of CB1-160 causes cell cycle arrest or at least a delay between prometaphase and metaphase.
Microtubule and Kinetochore Interactions become More Unstable in the CB1-160-overproducing Cells-In metaphase-arrested HBN160-E cells, lagging or mal-oriented chromosomes were observed. To examine the possibility that the interaction between the centromere/kinetochore and spindle microtubules was affected by the expression of CB1-160, we analyzed the sensitivity of the cells to the microtubule-depolymerizing drug, colcemid (Fig. 7A). Even without colcemid treatment, HBN160-E cells took a longer time (71.7 min) to complete mitosis than normal HeLa cells (47.0 min). On treatment with colcemid at a low concentration, 5 ng/ml, the average mitotic length of dividing HBN160-E cells was increased by almost 30 min (101.1 min) compared with cells without colcemid treatment, whereas that of normal HeLa cells increased by only 6.6 min (53.6 min). Moreover, ϳ10% of cells did not complete mitosis within 6 h of observation and this number increased in HBN160-E cells under the above conditions. Mono-oriented (mal-oriented) or lagging chromosomes were detected at a 4-fold higher rate (27% of cells) in CB1-160-overproducing cells than that in parental HeLa cells (7%) (Fig. 7B). Microtubule and kinetochore interactions thus presumably become unstable in the CB1-160-expressing state.
Duplication and Assembly of CENP-C and CENP-E at the Kinetochore Were Affected by Overproduction of CB1-160 -To analyze whether the proper kinetochore structure was formed in CB1-160-overproducing cells, we observed the localization of the kinetochore components, CENP-C and CENP-E. In normal cells, CENP-B and CENP-C were found to co-localize at centromeres in nuclei through the G 1 to S phases, the patterns for CENP-B and CENP-C overlapped each other and the sizes of each staining spot varied among the centromeres in a nucleus (Fig. 8A). The smaller (small arrows) or the larger (large arrows) sizes of CENP-C signals correlated with those of CENP-B signals (Fig. 8A). However, during the G 2 phase, each centromere spot is duplicated (52). Duplication of the CENP-C spots was also observed at the G 2 phase, and the compact and paired round dots of CENP-C became uniform among each centromere and located within the CENP-B-stained areas (Fig. 8A, arrowheads and insets). In the CB1-160-overproducing cells, al-though endogenous CENP-B detected with anti-CENP-B COOH-terminal region still existed at the centromeres overlapped with the signals of CB1-160, the signals of endogenous CENP-B were decreased and expanded to the broader areas as compared with those of CENP-B in non-CB1-160-producing cells (Fig. 8B). Thus, both the immunoprecipitation (Fig. 3B) and the cytological analyses indicated that CB1-160 competes well with endogenous CENP-B for binding to the CENP-B box. CENP-C existed at these centromeres as compact dots through the G 1 to S phases (data not shown). However, in the G 2 phase, CENP-C-positive structures were not compact or uniform but abnormally stretched and varied in size, and sometimes multiple dots appeared on a single centromere (Fig. 8C, insets). Such abnormal duplication of CENP-C was observed during  the centromere duplication period in 70% of G 2 phase cells overproducing CB1-160.
In the CB1-160-overproducing cells at prometaphase, CENP-C signals were also abnormal on mono-oriented or maloriented chromosomes (no CENP-C signal on lagging chromosome in Fig. 9A), in some cases being stretched and only one side of the paired centromere regions being detected by the anti-CENP-C antibody (data not shown). In normal HeLa cells at prometaphase, strong and compact dotlike CENP-E signals almost co-localized with, but were slightly peripheral to, the CENP-B signals (Fig. 9B). In mitosis-arrested HBN160-E cells, the majority of CENP-E signals on lagging (or mal-oriented) chromosomes were also abnormal. In some cases, a pair of CENP-E signals was both localized to one side of paired CENP-B signals and in other cases, CENP-E signals were enlarged or diffused (Fig. 9B, inset). In normal HeLa cells at late prometaphase, the intensity of CENP-E signals was much stronger than that of metaphase cells. In HBN160-E prometaphase cells containing lagging chromosomes, CENP-E signals on the lagging chromosomes were strong and abnormally large, whereas those on the chromosomes at the metaphase plates were almost undetectable (Fig. 9B). In normal HeLa cells, lagging chromosomes appeared at a lower rate (less than 5%), but we did not observe such abnormal localization of CENP-C and CENP-E on these lagging chromosomes. These results demonstrate that the duplication and assembly of CENP-C and CENP-E at the kinetochore were affected by overproduction of CB1-160.
A CENP-B Polypeptide Lacking Only the Homo-dimerization Domain Causes Cell Cycle Arrest in the G 2 Period-CB1-160 includes neither the CENP-C interaction domain nor the homodimerization domain. Analyses of the effects of a CENP-B C-del polypeptide (CB1-561) lacking the homodimerization domain but containing that for CENP-C interaction were therefore performed. Despite repeated experiments, we could not obtain any stable transformants with inducible expression of CB1-561. Therefore, we analyzed cells transiently expressing CB1-561 in 293S cells. Compared with CB1-160, CB1-561 was localized in a larger region in nuclei forming aggregates. Strong or weak CENP-C multiple signals overlapped with and surrounded these abnormal assemblies of CB1-561 sites (Fig.  10A). However, non-overlapped CENP-C signals in variable sizes were also increased in far beyond the expected number of centromeres in whole nucleus in high level of CB1-561-expressing cells (Fig. 10A, lower). The excess of CB1-561 in the nuclei or the absence of the homodimerization domain at the alphoid sites caused the abnormal accumulation or the multiple nucleation of CENP-C sites. Next, we analyzed the cell cycle state of CB1-561-expressing cells by observation of samples stained with indirect immunofluorescence. At 1.5 days after the transfection, the proportion of late S-G 2 cells was increased compared with normal or full-length CENP-B-expressing cells (data not shown). After 2.5 days, the numbers of late S-G 2 cells were increased further and no metaphase cells were observed (Fig. 10B), indicating that expression of CB1-561 caused cell cycle arrest in the G 2 period. Overproduction of both CB1-160 and CB1-561 affects the assembly or duplication of CENP-C in G 2 and/or cell cycle progression in G 2 /M.

Assemblies of Kinetochore Proteins on Specific Alphoid
DNA-CENP-A, -B, and -C exist on alphoid DNA at normal human centromeres throughout the cell cycle and were reported to be co-immunoprecipitable with the chromatin complex corresponding to the position of the tetra-to pentanucleosome units (51,54). However, in CENP-B knockout mouse cells, functional structures of pre-existing kinetochores are maintained without CENP-B (41)(42)(43). No CENP-B was detected at the centromere on the Y chromosome or neocentromeres on fragmented chromosome arms lacking centromeric alphoid DNA (15,37). It was also reported that mistargeting CENP-A to non-centromeric regions by overexpression caused misassembly of CENP-C to the same ectopic loci, but without a function as centromeres (55). Thus, CENP-C is able to assem- FIG. 7. Expression of CB1-160 increases sensitivity to colcemid. A, sensitivity of the CB1-160-expressing cells to colcemid. The CB1-160 polypeptide was expressed for 12 h, and then colcemid (0, 5, and 10 ng/ml) was added and the mitotic length of each cell was measured every 20 min. The columns show the average mitotic length of normal HeLa cells (black) and HBN160-E cells (shaded). The lines show the ratio of cells that could complete mitosis within the observation period (6 h). B, immunostaining of a typical normal HeLa cell at metaphase, an HBN160-E cell arrested for more than 2 h at metaphase, and an HBN160-E cell at anaphase. Cells were stained with DAPI (blue), and BN1 (red) and anti-tubulin (green) antibodies. Images of BN1 and tubulin staining were captured separately and merged to give pseudo-color pictures. Co-localization is indicated by yellow. The two insets show zooms of the CENP-B signals at the spindle ends, as defined by the white rectangles. The arrowhead indicates a mono-oriented chromosome, and the arrow indicates a laggard chromosome. ble at the centromere by a pathway other than through interaction with CENP-B. Despite these discrepancies, CENP-B and CENP-B boxes still exist in centromeres of all normal human and mouse autosomes and X chromosomes. The mechanisms that specify the location of centromeres are not simple. Because, on the de novo artificial chromosomes derived from the transfection of naked alphoid DNA into cells, not only CENP-B but also CENP-A, CENP-C and CENP-E assemble (46,47). Moreover, synthetic alphoid DNA repeats with only two nucleotides exchanged within the CENP-B boxes lose their ability to assemble mammalian artificial chromosomes and active centromere/kinetochore structures as well as the binding of CENP-B, whereas synthetic alphoid DNA repeats containing canonical CENP-B boxes still have these activities (48). Thus, CENP-B and CENP-B boxes are necessary for at least de novo assembly of centromere components on naked alphoid DNA.
In the present study, among centromere proteins, we have demonstrated that CENP-B interacts with CENP-C using a yeast two-hybrid system. The interaction between CENP-B and CENP-C involves two domains on CENP-C at amino acid residues 283-429 and 727-943, both of which contain Mif2 homologous regions. This molecular interaction between CENP-B and CENP-C was confirmed by immunoprecipitation with solubilized chromatin prepared from HeLa cells expressing truncated CENP-B polypeptides and with in vitro translated polypeptides, and by cytological observation of the truncated polypeptides.
Therefore, although CENP-C is able to assemble at the preexisting centromere without CENP-B, the molecular interaction between CENP-C and CENP-B does indeed exist in human cells. Thus, we speculate that the interaction may specify and increase the probability of the assembly of CENP-C on alphoid DNA containing CENP-B boxes, or alternatively provide a molecular linkage between one of the kinetochore components and human centromere DNA through the CENP-B/CENP-B box interaction.
Overexpression of CENP-B Deletion Mutants Affects the Assembly of CENP-C and Cell Cycle Progression through G 2 /M-Overproduction of the truncated CENP-B polypeptide containing the domain sufficient for specific localization to the centromere (32,35,36) caused abnormal assembly and duplication of CENP-C at G 2 . An excess amount of the truncated CENP-B polypeptide in human cells might affect the centromere structure by competing with normal CENP-B for binding to the CENP-B box on alphoid DNA. Our results indicate that defects in the proper assembly of CENP-C during the prekinetochore duplication period, as a result of overproduction of the CENP-B truncated polypeptide, explain the prolonged mitotic periods observed and the increase in lagging or maloriented chromosomes. The hypersensitivity of the cells to the microtubule-depolymerizing drug and abnormal assembly of CENP-E also indicated that the same defects in kinetochore assembly might have occurred. These kinetochore abnormalities monitored by the mitotic checkpoint mechanism may result in the prolonged mitotic periods. In the CB1-561 polypeptideexpressing cells, both abnormal accumulation of overlapping CENP-C and non-overlapped CENP-C in variable sizes were also observed with no mitosis. Although some differences in phenotype were observed between the cells which overproduced the CB1-160 and CB1-561 polypeptides, defects occurred in the same G 2 period when the CENP-C sites duplicate.
Microinjection of an anti-CENP-C antibody into cultured human cells induced cell cycle arrest at the G 2 /M phase and a reduction in kinetochore size (18). Interestingly, when Mif2 was overexpressed in budding yeast, chromosomes mis-segregated during mitosis and cells accumulated in the G 2 and M phases of the cell cycle (19). In S. cerevisiae and chicken DT40 cells, temperature-sensitive Mif2 and CENP-C mutations in the domain homologous to human CENP-C (Mif2 block 2) showed similar phenotypes (19,56). Thus, the abnormal dupli- cation of CENP-C and cell cycle arrest at G 2 caused by overproduction of truncated CENP-B polypeptides indicate a functional significance for the interaction between CENP-B and CENP-C at the kinetochore duplication period.

Molecular Interactions between CENP-B and CENP-C Domains Containing Mif2
Homologous Regions-The present study demonstrated an interaction between CENP-B and CENP-C involving two domains on CENP-C at amino acid residues 283-429 and 727-943, both of which contain Mif2 homologous regions. Overproduction of truncated CENP-B polypeptides showed similar phenotypes to temperature-sensi-tive Mif2 and CENP-C mutations in the domain homologous to human CENP-C (Mif2 block 2) in S. cerevisiae and chicken DT40 cells (19,56).
Under the conditions in which we demonstrated the interaction between CENP-B and CENP-C, a truncated mouse CENP-C containing the COOH-terminal Mif2 regions highly conserved between humans and mice was also isolated from a mouse cDNA library by two-hybrid screening with CENP-B as the bait. However, negative results for an interaction have been described previously (57). We experienced that the expression of CENP-B caused a weak inhibitory effect on yeast cell growth, so that some bias of 10 Ϫ3 -10 Ϫ4 arose for the transformation efficiency of the CENP-B cDNA clone when HeLa cDNA library and a certain amount of CENP-B cDNA were mixed and used for the transformation. The apparent anomaly might be caused by this bias or differences in the bait and reporter constructs.
CENP-C Domains and Multiple Assembly Pathways for Centromere Components-Our results indicate that CENP-C has two main domains for targeting centromeres, 283-429 and 728 -943. CENP-C728 -943 is included in the 584 -943 region reported by Lanini and McKeon (27). CENP-C dimerization activity also exists in this region (820 -943) (58) and allowed homotypic interaction, as detected by our two-hybrid analysis. The results suggest that not only the interaction between CENP-B or other assembly pathway components, but also the self-dimerization activity, must be involved in the centromere localization of CENP-C.
Several groups have reported that CENP-C has DNA binding activity but that this lacks sequence specificity (25,26). Recently, Politi et al. (59) showed that CENP-C and CENP-B associate with the same types of alphoid arrays using a chromatin immunoprecipitation assay, but in distinct non-overlapping centromere domains by ultrastructural analysis. In their study, the alphoid DNA binding activity of CENP-C was only detected with polypeptides containing both the DNA binding domain and the CENP-B interaction domains described in our present study. CENP-C also exists at neocentromeres where no alphoid sequence was detected (15). In contrast, CENP-B has specific binding activity to CENP-B boxes in centromeric type I alphoid DNA, and therefore overlapped with type I alphoid DNA sites throughout the cell cycle (33,34,60). Multiple molecular interactions may influence the affinity of CENP-C for the specific sites.
Goshima et al. reported that both CENP-A and hMis12 are required for the localization of hMis6 to the kinetochore, although CENP-A and hMis12 assemble at the kinetochore independently. They suggested that there were different pathways for the assembly of kinetochore components by hMis12 and CENP-A (3). Liu et al. reported that the kinetochore localizations of hBUB1, hBUBR1, hROD, hZW10, CENP-E, and CENP-A were not affected but CENP-F, MAD1, and MAD2 were affected by RNAi of CENP-I (hMis6), and the authors suggested that CENP-I specifies a discrete branch of the kinetochore assembly pathway (4). Our previous studies including chromatin immunoprecipitation analyses showed that the information carried by the primary sequence of the ␣21-I alphoid DNA and CENP-B box is important for de novo formation of functional centromere chromatin accompanied by binding of CENP-A, CENP-B, CENP-C and CENP-E (48,61). Thus, there might be several pathways for the assembly of kinetochore components.
We demonstrated an interaction between CENP-B and CENP-C in the present study. These results suggest that CENP-B may be involved in the assembly of the centromere/ kinetochore structure by mediating the linkage of the essential component CENP-C to specific alphoid DNA containing CENP-B boxes. De novo centromere assembly mechanisms on the naked centromere DNAs were observed not only with a human cultured cell line but also with the most characterized centromeres of two yeasts, S. cerevisiae, and S. pombe (62,63). Complex components of the centromere/kinetochore structure may assemble to the specific centromere DNA in each species through multiple interactions and assembly pathways.