Structural Requirements and Dynamics of Mitosin-Kinetochore Interaction in M Phase

Abstract

Mitosin is a 350-kDa human nuclear protein which transiently associates with centromeres and spindle poles in M phase. Ultrastructure studies reveal that it is located at the outer kinetochore plate. In this work, we explored the detailed structural basis and dynamics of the mitosin-kinetochore interaction. Two major regions important for targeting to centromeres were identified by analyzing different deletion mutants expressed in CHO cells: (i) the “core region” between amino acids 2792 and 2887, which was essential for the centromere localization of mitosin; and (ii) the internal repeats between residues 2094 and 2487, which cooperated with the core region to achieve strong mitosin-kinetochore interaction. The core region is characteristic of two leucine zipper motifs. Deletion of either motif abolished the centromere localization activity. In addition, Cys²⁸⁶⁴, adjacent to the second motif, was also essential for the activity of the core region. In contrast, the internal repeats alone were insufficient for centromere localization. We propose that this region may serve as a regulatory domain to facilitate interaction of the core region with the kinetochore. We showed that mitosin molecules entering nuclei after nuclear envelope breakdown (NEBD) were not assembled onto kinetochores efficiently, suggesting that the mitosin-kinetochore interaction is stabilized prior to NEBD. This result supports the idea of an ordered process for kinetochore assembly. Our data also suggest that mitosin might interact with chromatin in interphase. Evidence for coordinated regulation between the centromere-targeting and the putative chromatin-binding activities is also provided.

The kinetochore is a three-layer structure located at the centromere of chromosomes. It is required for separation of sister chromatids in eucaryotes during M phase. Recently, accumulated evidence has also shown that, in addition to the ability to ‘mechanically’ move the chromosomes, the kinetochore may also function as a sensor for the mitotic checkpoints to guarantee precise segregation of genetic materials to daughter cells (reviewed in references 1, 10, 21, and 25).

Dissecting its protein components is one of the approaches to understanding the molecular basis of the mammalian kinetochore functions. Since the identification of CENP-A (for centromere protein A), CENP-B, and CENP-C (9), novel centromere-associated proteins, including CENP-D, CENP-E, dynein, and mitosin (also named CENP-F), have also been characterized at the molecular level (reviewed in reference 22). Most of them have been shown by electron microscopy to reside in distinct regions of the kinetochore. For instance, CENP-B, a centromeric DNA-binding protein, associates with human centromeric DNA beneath the inner plate of the kinetochore (4, 18); CENP-A and CENP-C are components of the inner kinetochore plate (26, 33); mitosin/CENP-F is localized to the coronal surface of the outer kinetochore plate (24, 39); and the two motor proteins, dynein and CENP-E, are concentrated in the fibrous corona (5, 35). Recently, the family of centromere proteins has been further expanded by discoveries of the human homologues of yeast Mad2 (15) and Skp1 (3) and the murine homologue of yeast Bub1 (31). These evolutionarily conserved proteins transiently associate with centromeres and are involved in control of the mitotic checkpoints. Clearly, proteins involved in the structure and function of kinetochores must interact in a certain way in mature kinetochores. This issue, however, is presently not clear.

The structural characteristics of centromere proteins have been extensively studied. CENP-A, a histone H3 homologue of 17 kDa, targets the centromere via a histone H3-related domain (30). CENP-B (80 kDa) has been shown to utilize its N-terminal 125 amino acid residues to interact with the centromere (23, 38). The target of this centromere-binding domain is a 17-bp alphoid DNA (18). In addition to this, CENP-B also forms homodimers through a dimerization domain of 59 amino acid residues at its C terminus (13, 38). The centromere localization domain of CENP-C (140 kDa) lies in the homologous region with Mif2, a protein of budding yeast required for correct segregation of chromosomes (14, 36); this region overlaps with a potential DNA-binding domain (29, 36). CENP-E (312 kDa), a kinesin-like protein which is required for metaphase chromosome alignment (27, 34, 37), possesses two microtubule-binding domains (16); its centromere localization domain, however, has not been documented in detail.

Five functional regions have been identified in mitosin, a 350-kDa kinetochore protein. The region between amino acid residues 2961 and 3001 binds to the retinoblastoma protein (Rb) in vitro (39). Residues 2930 through 2958 contain a strong bipartite nuclear localization signal (NLS) (7, 40). Potentially, mitosin can also form homodimers through a C-terminal domain within the region from residue 2488 to 2925 (40). A polypeptide containing residues 2488 through 3113 is able to target the centromere when expressed in monkey kidney CV1 cells (40). Finally, the C-terminal portion from residue 2094 to 3113 has been shown to be capable of spindle pole localization during M phase (41).

In this study, we tried to address the detailed structural requirements for mitosin-kinetochore interaction and to analyze the biochemistry of such an interaction qualitatively. We defined the structural elements important for centromere localization of mitosin in detail. We analyzed the dynamics of mitosin-kinetochore interaction by an in vivo competition experiment. In addition, we showed that mitosin might contribute to nuclear organization, possibly through association with chromatin. The minimal polypeptide of mitosin sufficient for centromere localization can be used as a probe to isolate a gene(s) coding for its downstream target(s) on kinetochores. Connections between mitosin and other kinetochore components can therefore be further explored.

MATERIALS AND METHODS

Plasmid constructs.

Construction of most deletion mutants was based on pTN, which is able to express a FLAG-tagged mitosin mutant containing amino acids 2094 to 3113 under control of the tetracycline-responsive system (11, 41). All mutants expressed were thus tagged with a FLAG epitope at their N termini. Schematic diagrams of all constructs are shown in Fig. 1, 4A, and 5A. Sequencing was always performed to confirm faithful ligation at junctions during plasmid construction whenever analysis by restriction cleavage was not possible.

FIG. 1.

Determination of structural elements important for centromere localization of mitosin. (A) Diagrams of representative mutants used in this study. A diagram showing both structural characteristics of the full-length mitosin and restriction sites on its cDNA used for plasmid construction is displayed on top. Names of the constructs are printed on the left for each mutant. Numbers represent positions of amino acid residues. PKKKRKV indicates an insertion of the NLS of SV40 large T antigen. All the listed mutants were nuclear proteins in interphase as expected (data not shown). Their abilities for centromere localization (+, positive; −, negative) are summarized on the right. Dashed lines indicate common boundaries shared by multiple mutants. The essential core region is located between residues 2792 and 2887. (B) Sequence of the core region. Leucines in each putative leucine zipper are underlined. Arrowheads indicate the two cysteine residues on which we performed mutagenesis studies (Fig. 4).

FIG. 4.

Cysteine²⁸⁶⁴ is an essential amino acid for the centromere localization of mitosin. (A) Diagrams showing the point mutants used in this study. Both the plasmids were constructed based on pTCB-N (Fig. 1A). Restriction sites used for construction are shown. The positions of point mutations are also indicated. Numbers indicate positions of amino acid residues. (B) IIF images of the point mutants. CHO cells were transfected with pTCBC1-N or pTCBC2-N. Chromosome spreads and IIF staining were performed as described in the legend to Fig. 2. Similar to their parental mutant, mitosin-pTCB-N, both point mutants tended to aggregate in cells. Left panels, IIF images; right panels, chromosomes visualized by DAPI staining; middle panels, superimposed images. Panels 1 to 3 show a typical mitotic cell expressing mitosin-pTCBC1-N; strong centromere localization was observed despite the existence of aggregates. Panels 4 to 6 show a mitotic cell expressing mitosin-pTCBC2-N; no centromere localization was observed. Differences in the sizes of chromosomes are due to different extents of expansion in hypotonic buffer during sample preparation. Scale bar, 10 μm.

FIG. 5.

Mitosin colocalizes with nuclear DNA in interphase. (A) Diagrams of representative mitosin mutants used in this study. Restriction sites used for constructing the listed plasmids are shown at the top. The ability of each mutant to colocalize with DNA is summarized on the right. A putative chromatin-binding domain is speculated to exist between residues 2902 and 3037 and is indicated between dashed lines. (B) IF images of the mutants. CHO cells transfected with pTN, pTNE, pTND-N, and pTZ were fixed on coverslips and subjected to IIF microscopy as described in the legend to Fig. 3. All the listed mutants were nuclear proteins in interphase as expected. Panels 1, 3, 5, and 7, IF images of different mutants; panels 2, 4, 6, and 8, nuclear DNA stained by DAPI. Arrowheads indicate typical areas where mitosin mutants colocalize with condensed DNA. Typical cells at interphase expressing mitosin-pTN (panels 1 and 2), mitosin-pTNE (panels 3 and 4), mitosin-pTND-N (panels 5 and 6), and mitosin-pTZ (panels 7 and 8) are displayed. Scale bar, 20 μm.

To construct pTZ, the region coding for amino acids 2489 through 2901 was deleted as in EΔZ (40) from pTN. In pTNR, the EcoRI restriction fragment from nucleotides 6604 to 7150 was removed from pTN; amino acids 2178 to 2359 were therefore deleted, similar to the deletions in EΔR (40). pTND and pTNE were constructed from pTN to express mitosin mutants with truncated C termini as in EΔC2 and EΔC1 (40), respectively. To render nuclear localization of the mitosin mutant produced by pTND, a sequence coding for the NLS of simian virus 40 (SV40) large T antigen (7) (5′ CTAGGCCTAAGAAAAAGCGTAAAGTCA 3′/3′ CGGATTCTTTTTCGCATTTCAGTGATC 5′) was introduced in frame into the NheI site located between the FLAG-coding sequence and mitosin cDNA to form pTND-N. pTNH-N and pTNF-N contained both the NLS and 5′ coding sequence of mitosin as in pTND-N, but their 3′ coding sequences were truncated at EcoNI (at position 8775) and ScaI (at position 8660), respectively. pTG was constructed to express the same mutant as EΔG, i.e., the mutant containing amino acids 2488 to 3113 (40), but under control of the tetracycline-responsive system. Construction of pTC has been described previously (41). To construct pTCP, pTC was cleaved with BglII (at position 9561) and partially digested with PstI; a 3.6-kb fragment was then self-ligated; pTCP therefore contained the coding sequence from positions 8337 to 8974. To create pTS, pTCP was cleaved at the NheI site, treated with mung bean nuclease, and then digested with KpnI; the resultant 3.7-kb fragment, from which only the FLAG-coding sequence was deleted, was ligated to the 1.2-kb KpnI-StuI restriction fragment (positions 6349 to 7533) from pTN. To construct pTSE, pTND was cleaved with EcoNI, treated with mung bean nuclease, and then digested with BamHI; the resultant 4.2-kb fragment was ligated to the 0.3-kb ScaI-BamHI fragment (positions 8660 to 8974) from pTCP. pTCB was created by further deleting the 3′ coding sequence of pTS to the EcoNI site (at position 8775). In pTCB-N, the sequence encoding the NLS of large T antigen was introduced into pTCB in the same way as described above. pTNC was constructed by deleting the PvuII-EcoNI fragment (8482 to 8775) from pTN; their incompatible termini were removed with mung bean nuclease before ligation.

To further narrow down the core region, the coding region between 8448 and 8733 was amplified by PCR with primers 10p-18 (5′ GAAAATGAAGTTGTTGATC 3′) and 10p-rKB (5′ CGGATCCTCACAGATGGGCCACTTG 3′). The PCR fragment was cleaved with BamHI and then ligated to replace the StuI-BamHI fragment of pTN (7533 to 9664). The resulting plasmid was named pTKB. The PCR fragment in pTKB was sequenced. The sequence coding for the NLS of large T antigen was inserted as previously described to create pTKB-N.

Point mutations were introduced by PCR. To mutate cysteine²⁸⁰¹ to serine, the cDNA fragment between nucleotides 7080 and 8481, in which a deletion from 7533 to 8337 resided, was amplified from pTCB by PCR with primers 10p-7 (5′ AGTGGAGAACCTTGAAAG 3′) and 10p-rC1 (5′ CTGTTTAGAGGATTTGATCAAC 3′) (antisense primer; the mutated codon is underlined). The PCR product was cleaved with BglII (at 7255), and the fragment containing nucleotides 7255 to 8481 was cloned into pTCB-N to replace the corresponding BglII-PvuII fragment to form pTCBC1-N. To mutate cysteine²⁸⁶⁴ to serine, primers 10p-C2 (5′ ACTCTTCCTTGCTTATAAGC 3′) (sense primer, the mutated codon is underlined) and tet-rp (5′ ACTGCATTCTAGTTGTGGT 3′) (its target sequence is located in the vector) were used to amplify the 150-bp fragment from pTCB by PCR. After cleavage with BamHI, the PCR fragment containing the sequence from 8660 to 8775 was cloned into pTCB-N to replace the corresponding ScaI-BamHI fragment to create pTCBC2-N. The point mutations in pTCBC1-N and pTCBC2-N were both confirmed by sequencing. The PCR fragments in both plasmids were also sequenced. pTCBC1-N and pTCBC2-N were thus identical to pTCB-N, except for their point mutations.

Transfection.

Chinese hamster ovary (CHO) cells were maintained in Dulbecco’s modified Eagle medium (DMEM; Gibco) supplemented with 10% calf serum (Sijiqin Company, Hangzhou, China) in an atmosphere containing 5% CO₂. For stable expression, CHO cells were transfected and selected with G418 as described previously (41). G418-resistant colonies were then cultured as a whole in DMEM containing 0.2 mg of G418/ml. For transient expression, cells were assayed 48 h after transfection. To prevent unscheduled expression, all the transfected cells were maintained in DMEM containing tetracycline (1 μg/ml). For centromere localization, both transient and ectopic experiments generated similar results.

IIF studies.

For indirect immunofluorescence (IIF) staining, transfected cells were grown on coverslips overnight in the absence of tetracycline to induce expression before fixation in methanol for 15 min at −20°C. For preparation of metaphase chromosome spread, transfected cells were split onto glass coverslips to about 40% confluency, synchronized first in the presence of both thymidine (2 mM) and tetracycline (1 μg/ml) for 8 h, then cultured overnight in fresh DMEM containing nocodazole (0.4 μg/ml) but no tetracycline. The chromosome spread was then prepared as previously described (8). Samples on the coverslips were fixed in cold methanol. Immunostaining was performed with anti-FLAG M2 antibody (IBI) and fluorescein isothiocyanate-conjugated anti-mouse immunoglobulin G (Sino-American Biotechnology Company, Shanghai, China) as described previously (39). Nuclear DNA was stained by DAPI (4′,6-diamidino-2-phenylindole). Due to the difficulty of controlling the expression levels of different mutants in different individual cells, it was not possible to precisely measure the intensities of centromere-specific signals. Our results were therefore based on both extensive observation of 10 to 20 mitotic cells expressing each mutant and careful comparisons with records for other related mutants. Multiple representative images were recorded on Kodak Gold III (ASA400) or Lucky Pan SHD400 films with an Olympus BX50 fluorescence microscope. Prints were digitized using a Umax Vista-S6 scanner. Images for publication were organized by using Photoshop software and directly exported from a Tektronix Phaser 450 image printer.

RESULTS

Identification of the structural elements important for mitosin-kinetochore interaction.

To further understand the structural basis of control of the centromere association of mitosin, we performed detailed deletion analysis to define its centromere localization domain. In a previous study, my colleagues and I showed that a mutant containing amino acid residues 2488 to 3113 localizes to the centromere region (40). Attempts to further define the domain, however, were precluded due to strong cytoplasmic IF in mitotic cells expressing shorter mutants; the centromere-specific signals appeared to be overwhelmed by the background generated by free mitosin. Furthermore, mitotic cells expressing mutants were rare in many cases. We thus considered that if we could reduce the expression levels, subcellular targets of the mutants might be highlighted. Moreover, a lower expression level might also reduce the extent of cell cycle block caused by overexpression of mitosin (39). We therefore utilized the tetracycline-responsive system (11) to express FLAG-tagged mitosin mutants. The CHO cell line was used as the host cell line due to its relatively low chromosome number (2N = 20). Expression levels were indeed reduced when interphase CHO cells expressing identical mutants under control of either the cytomegalovirus promoter or the tetracycline-responsive promoter were compared by IIF microscopy (data not shown). Similar examination showed that the average levels of most mutants expressed in CHO cells were compatible; only mitosin-pTCP showed a relatively low expression level (data not shown). For IIF study, we prepared chromosome spreads by using nocodazole-arrested cells to achieve clearer results. To avoid a possible effect of different subcellular localization on centromere targeting, we introduced an NLS into mutants in which the intrinsic NLS had been deleted.

We first confirmed that results achieved in CHO cells were consistent with previous ones in CV1. Similar to results obtained in CV1 cells by using plasmids EΔN and EΔZ (40), a mitosin mutant expressed in CHO from pTN, named mitosin-pTN, exhibited strong centromere IF (Fig. 1A and Fig. 2, panels 1 to 3), while mitosin-pTZ lacked the ability for centromere localization (Fig. 1A and Fig. 2, panels 4 to 6). Further analysis showed that regions from amino acid residues 2488 to 2755 and from residues 2967 to 3113 were both dispensable because mitosin-pTS localized to centromeres effectively (Fig. 1A and Fig. 2, panels 7 to 9). Further deletion of residues 2756 to 2862 from pTS (which generated mitosin-pTSE), however, completely abolished centromere localization (Fig. 1A and Fig. 2, panels 10 to 12). On the other hand, while mitosin-pTNH-N (containing amino acids 2094 to 2901) was positive for centromere localization (Fig. 1 and data not shown), mitosin-pTNF-N (containing residues 2094 to 2862) was completely negative (Fig. 1A and Fig. 2, panels 13 to 15).

FIG. 2.

IIF images of representative mitosin mutants. CHO cells transfected with mitosin mutants were plated on coverslips, synchronized to prometaphase, and subjected to preparation of in situ chromosome spreads. Samples were then immunostained with anti-FLAG M2 monoclonal antibody and fluorescein isothiocyanate-conjugated goat anti-mouse immunoglobulin G. Chromosomes were visualized by DAPI staining. All the left panels are IIF images of different mutants. All the right panels are images of chromosomes. The middle panels consist of superimposed images. The names of plasmids used for transfection are listed on the left. Arrowheads indicate typical centromere IF of positive mutants. One typical mitotic cell is shown for each mutant. The insets in panels 16 and 17 are amplified images included to highlight the weak centromere-specific IF. Scale bar, 10 μm.

Further analysis showed that a critical region (which we refer to as the core region) was located within residues 2756 to 2901. Although mitosin-pTG (containing amino acids 2488 to 3113; Fig. 1A) was identical to the polypeptides expressed by EΔG (40), it did exhibit clearer centromere IF over the background due to lowered expression levels (data not shown). Centromere localization was also observed in mitotic cells expressing mitosin-pTC, a shorter mutant containing residues 2756 to 3113 (Fig. 1A and Fig. 2, panels 16 to 18). In contrast to positive mutants containing the internal repeats (such as mitosin-pTN and mitosin-pTS), both mitosin-pTG and mitosin-pTC bound weakly to kinetochores; the centromere-specific IF was only a little stronger than the cytoplasmic staining (Fig. 2, panels 16 to 18, and data not shown). Mitosin-pTC appeared even more unstable at centromeres since it was hardly detectable in cells that were well swelled in hypotonic buffer (data not shown).

The negative results from mitosin-pTCP (containing amino acids 2756 to 2966; Fig. 1A) in multiple experiments (data not shown), however, obscured the possibility that the core region might serve as a discrete centromere localization domain, implying a complex situation in terms of the centromere localization of mitosin. In contrast to mitosin-pTS, which localized to centromeres strongly (Fig. 2, panels 7 to 9), mitosin-pTCP lacks only the internal repeat region (Fig. 1A). We also noted that strong centromere IF but low cytoplasmic background actually correlated with all the tested positive mutants containing the internal repeats, for instance, mitosin-pTN (Fig. 1A and Fig. 2, panels 1 to 3) and mitosin-pTS (Fig. 1A and Fig. 2, panels 7 to 9), suggesting high binding affinities of these mutants for kinetochores. On the other hand, although mitosin-pTG and -pTC, which lacked the internal repeats, were both able to localize to centromeres, their affinities for kinetochores were low; the majority of the mutants appeared to be free forms or the cytoplasmic background. These observations suggested that the internal repeats were also important for mitosin-kinetochore interaction. Contribution of the internal repeats to centromere localization was further corroborated by mitosin-pTNR. This mutant contained only one chimeric repeat, in contrast to mitosin-pTN (Fig. 1A). However, it targeted centromeres more weakly than the latter did (Fig. 2, panels 1 to 3 and 19 to 21). Similar to the weakly positive mutant mitosin-pTC, the majority of mitosin-pTNR also appeared as free forms (background) (Fig. 2, panels 19 to 21). Both the repeat units were therefore required for maximal localization potential of mitosin to centromeres. In contrast to mitosin-pTC, however, the repeat region alone did not target the centromeres, as indicated by results from mitosin-pTZ, -pTSE, and -pTNF-N (Fig. 1A and 2) as well as a mutant containing the repeats only (data not shown). These results suggested that strong centromere localization of mitosin might require combination of both the internal repeat region and the core region. Apparently, as indicated by results from mitosin-pTG and -pTC, certain sequences adjacent to the core region were also involved in the centromere localization of mitosin.

The behaviors of mitosin-pTCB-N (Fig. 1A) confirmed our speculation. Indeed, this mutant targeted centromeres efficiently (Fig. 2, panels 22 to 24). Clear centromere IF with low background resembled the patterns produced by mitosin-pTN and -pTS. Between amino acid residues 2756 and 2901, there are two leucine zipper motifs, spanning residues 2797 to 2818 and 2866 to 2887 (Fig. 1B). Deleting either motif (as in the case of mitosin-pTSE and mitosin-pTNF-N) abolished centromere localization (Fig. 1A and 2). Indeed, as proved with mitosin-pTKB-N, a shorter region just spanning both the leucine zipper motifs was sufficient for the core region function (Fig. 1A and data not shown). The core region was therefore defined as residues 2792 to 2887 (Fig. 1B).

We noticed that many mutants tended to form huge amorphous aggregates in cells, probably due to destruction of molecular integrity by mutagenesis. Among these mutants were mitosin-pTS, -pTSE, -pTNF-N, -pTCB-N, and -pTKB-N (Fig. 2 and 3 and data not shown). All these mutants were correlated with C-terminal truncation. Such aggregates were frequently observed in cells at either interphase or M phase; for nuclear proteins, aggregates were also found in the cytoplasm (Fig. 2 and 3 and data not shown). In contrast, mutants such as mitosin-pTN, -pTNR, -pTNE, -pTNC, -pTG, and -pTC (41) behaved relatively normally in cells in interphase: their distribution in the nucleus was relatively uniform, and no aggregates were noticed in the cytoplasm (see Fig. 5) (data not shown). Aggregate formation added further complexity to our study. Fortunately, aggregation did not alter the centromere localization properties because centromere-specific IF was readily detected in positive mutants with an aggregation tendency, e.g., mitosin-pTS and -pTCB-N (Fig. 2, panels 7 to 9 and 22 to 24). Neither did the aggregation significantly alter the affinities of mutants for kinetochores, because these mutants exhibited strong centromere IF comparable with that of mitosin-pTN (Fig. 2, panels 1 to 3, 7 to 9, and 22 to 24). These observations allowed us to directly compare the centromere localization properties among different mutants.

FIG. 3.

Subcellular localization of mitosin-pTCB-N and mitosin-pTCB in the cell cycle. CHO cells transfected with pTCB-N and pTCB were grown, respectively, on coverslips overnight in the absence of tetracycline and fixed in cold methanol. Mitosin mutants were stained with anti-FLAG M2 monoclonal antibody and fluorescein isothiocyanate-conjugated goat anti-mouse immunoglobulin G (panels 1, 3, 5, and 7). Chromosomes were visualized by DAPI (panels 2, 4, 6, and 8). Both mutants tended to form aggregates in the majority of cells throughout the cell cycle. To highlight the centromere-specific localization of mitosin-pTCB-N, only mitotic cells with few aggregates are shown. (A) Diagrams of constructs used in this assay. Numbers indicate positions of amino acid residues. (B) Mitosin-pTCB-N forms strong centromere dots during M phase. Arrowheads indicate positions of centromere-specific IF. (C) Localization of mitosin-pTCB to centromeres is hardly detectable throughout M phase. A typical cell is shown at interphase (panels 1 and 2), at prophase (panels 3 and 4), at metaphase (panels 5 and 6), and at telophase (panels 7 and 8). Scale bar, 20 μm.

Binding of mitosin to kinetochores is hindered after NEBD.

To explore the dynamics of the mitosin-kinetochore interaction, we studied the ability of a competitor, mitosin-pTCB, to target to centromeres in unsynchronized cells. As a cytoplasmic protein, mitosin-pTCB would be sequestered outside the nucleus until nuclear envelope breakdown (NEBD). Then it would compete with endogenous mitosin for sites on kinetochores. Mitosin-pTCB-N and mitosin-pTCB differed by only the presence of an artificial NLS in the former (Fig. 3A). Mitosin-pTCB-N was therefore a nuclear protein in interphase (Fig. 3B, panels 1 to 2), while mitosin-pTCB was cytoplasmic (Fig. 3C, panels 1 to 2). Although both mutants formed huge aggregates in cells at different stages of the cell cycle, their centromere localization properties were not significantly affected, similar to other mutants discussed previously. Similar to the wild-type mitosin, mitosin-pTCB-N redistributed to the centromeres in prophase (Fig. 3B, panels 3 to 4). In contrast to this, mitosin-pTCB stayed in the cytoplasm in prophase; no detectable IF was observed in the nucleus in prophase (Fig. 3C, panels 3 to 4). In addition to that at prophase, strong centromere localization of mitosin-pTCB-N persisted in cells at metaphase (Fig. 3B, panels 5 to 6) and anaphase (panels 7 to 8). Mitosin-pTCB, however, was still hardly observed at centromeres, even after NEBD, e.g., in cells at metaphase (Fig. 3C, panels 5 to 6) and anaphase (panels 7 to 8). Nevertheless, mitosin-pTCB was capable of centromere localization: when it was exposed to kinetochores for a prolonged period, e.g., in mitotic cells arrested by nocodazole, positive signals at centromeres were observed (data not shown). These data indicated that mitosin molecules available after NEBD could no longer be incorporated into kinetochores efficiently.

Cysteine²⁸⁶⁴ is essential for the activity of the core region.

In addition to the two leucine zipper motifs in the core region, another notable feature is the two cysteine residues located at positions 2801 and 2864 (Fig. 1B). The first cysteine lies in the first leucine zipper motif, and the second one is close to the second motif. To examine if these cysteines were essential for the activity of the core region, we mutated the codon of each cysteine into serine based on construct pTCB-N. The resulting plasmids, pTCBC1-N and pTCBC2-N, were capable of expressing polypeptides identical to mitosin-pTCB-N except for the point mutations Cys²⁸⁰¹→Ser and Cys²⁸⁶⁴→Ser, respectively (Fig. 4A). Proper expression of these mutants was confirmed by immunoblotting (data not shown). We found that both mutants formed amorphous aggregates throughout the cell cycle (Fig. 4B and data not shown) like their parental mutant, mitosin-pTCB-N (Fig. 3B). Nevertheless, the point mutation at Cys²⁸⁰¹ did not affect the centromere localization of the mutant (Fig. 4B, panels 1 to 3). The mutation at Cys²⁸⁶⁴, however, completely abolished localization of the mutant to centromeres (Fig. 4B, panels 4 to 6). Cys²⁸⁶⁴ was therefore a critical residue for the activity of the core region. According to these results, Cys²⁸⁰¹ and Cys²⁸⁶⁴ are unlikely to form a disulfide bridge with each other.

Mitosin colocalizes with nuclear DNA in interphase.

The nuclear matrix has been shown to serve as sites of chromatin organization in the nucleus (2, 6). Mitosin/CENP-F is a nuclear matrix protein in interphase (17, 24). We frequently noticed that, in interphase, mitosin tended to closely colocalize with DAPI-stained nuclear DNA. Such a colocalization was not highlighted in wild-type mitosin and some deletion mutants (e.g., mitosin-pTN) due to the relatively homogeneous distributions of both mitosin and DNA (Fig. 5A and 5B, panels 1 to 2) (39–41). Nevertheless, in nuclei swollen by hypotonic buffer, colocalization of wild-type mitosin with nuclear DNA in CV1 cells was observed by IIF microscopy (data not shown). In addition, colocalization was obvious with certain mutants, such as mitosin-pTC (41), mitosin-pTNE (Fig. 5A and 5B, panels 3 to 4), and mitosin-pTZ (Fig. 5A and 5B, panels 7 to 8). In cells expressing these mutants, brightly stained DNA foci were frequently superimposed with strong IF of mitosin mutants. It appeared that these mutants somehow altered the chromatin organization in the nucleus because the distinct DNA foci were usually not observed in mock-transfected cells (Fig. 5B and data not shown). On the other hand, all the tested mutants with truncations to residue 2949 and further removed from the C terminus (i.e., mitosin-pTND-N and -pTCB-N) completely lost such a colocalization (Fig. 5A, 5B, panels 5 to 6, and 3B). These results implied that mitosin might interact with chromatin proteins or DNA. As summarized in Fig. 5A, a putative chromatin-binding domain of mitosin was speculated to lie between residues 2902 and 3037.

Deletion of the core region results in distribution of mitosin along chromosomes.

In prophase, mitosin/CENP-F dissociates from the nuclear matrix and redistributes to kinetochores (17, 24, 39). Therefore, activation of the centromere colocalization property should be accompanied by loss of its nuclear matrix-associating as well as the putative chromatin-binding activities. To test if there might be any direct connections among these events, we deleted the core region from mitosin-pTN to form mitosin-pTNC, which lacked residues 2804 to 2901 of mitosin-pTN (Fig. 6A). Mitosin-pTNC did not aggregate in interphase cells; it was distributed in the nucleus in a pattern similar to that of mitosin-pTZ (data not shown). To our surprise, mitosin-pTNC appeared to fail to completely dissociate from chromatin in M phase since bright foci of various sizes were observed along virtually all chromosomes (Fig. 6B). The number of foci on chromosomes varied from cell to cell, as shown in Fig. 6B, panels 1 to 3 and 4 to 6. We also noticed that some of the foci only partially associated with chromosomes (panel 2). In some cases, especially when the foci on chromosomes were dense, immunofluorescence could be observed at centromere regions of some, but not all, chromosomes (data not shown). Detailed study showed that such centromere colocalization was nonspecific, since, in many cases, IF at centromeres was not observed (panels 1 to 3). These results suggest that the core region might down-regulate the putative chromatin-binding activity of mitosin in prophase.

FIG. 6.

Mitosin-pTNC is distributed along chromosomes in M phase. Chromosome spreads prepared from CHO cells expressing mitosin-pTNC were subjected to IIF staining as described in the legend to Fig. 2. (A) Diagrams for mitosin-pTNC. Numbers indicate positions of amino acids. (B) IF images. Left panels, IIF images of transfected cells; right panels, chromosome images stained by DAPI; middle panels, superimposed images. Panels 1 to 3 show a mitotic cell bearing large foci along chromosomes; arrows indicate centromeres without IF label. Panels 4 to 6 show a mitotic cell bearing dense IF dots along chromosomes. Scale bar, 10 μm.

DISCUSSION

The structural basis for mitosin-kinetochore interaction.

By performing deletion analysis, we found that two major structural elements, the internal repeats from residues 2094 through 2487 and the core region located between residues 2792 and 2887, were critical for strong mitosin-kinetochore interaction. Their contributions to the centromere localization property, however, were clearly different. Although independent localization to centromeres was not detected, the core region was indeed essential for the centromere localization of mitosin: without this region (e.g., in mitosin-pTNC and -pTZ), or even when the region was partially deleted (e.g., in mitosin-pTSE and -pTNF-N) or point mutated (e.g., in mitosin-pTCBC2-N), the corresponding mutants never exhibited centromere localization. Without the repeat region, in contrast, some mutants containing the core region (e.g., mitosin-pTG and -pTC) were still able to manifest centromere localization in spite of their weak affinities. Clearly, some sequences in the vicinity of the core region could partially compensate for the function of the internal repeats. In spite of this, clearly strong centromere localization was observed only in mutants containing both the internal repeats and the core region (e.g., mitosin-pTN, -pTS, -pTCB-N, and -pTKB-N). In addition, we found that both the repeat units were required for maximal mitosin-kinetochore interaction. How the internal repeats coordinate with the core region to achieve strong centromere localization is still an intriguing mystery. Our data suggest that the internal repeats (and other minor elements adjacent to the core region), instead of binding to kinetochores directly, might facilitate and/or stabilize the core region-kinetochore interaction.

Neither the core region nor the internal repeats of mitosin show significant sequence homology with centromere localization domains of other centromere proteins. The core region contains two leucine zipper motifs. Deletion of either one abolished localization to centromeres. These two motifs could potentially form heterodimers with their downstream partners on the kinetochore. The core region also contains two cysteine residues. Cys²⁸⁶⁴, which lies adjacent to the second leucine zipper motif, was found to be an essential amino acid.

We have shown previously that there is a potential dimerization domain within residues 2488 through 2925 (40). Since this region includes the core region (residues 2792 to 2887), the latter could be merely a dimerization domain that was passively targeted to the centromeres via dimer formation with endogenous hamster mitosin. We excluded this possibility by studying the subcellular localization of mitosin-pTCB in unsynchronized CHO cells. If the core region (together with the internal repeats) did not dimerize with hamster mitosin, mitosin-pTCB would stay outside the nucleus until NEBD at prometaphase; otherwise, this mutant would be brought into the nucleus by endogenous mitosin prior to NEBD. We examined prophase cells expressing mitosin-pTCB carefully. Neither nuclear nor centromere localization was observed in these cells (Fig. 3C). Lack of centromere localization was not due to loss of this ability since, when cells expressing mitosin-pTCB were blocked at prometaphase by nocodazole, weak centromere localization was detected in chromosome spread (data not shown). Therefore, neither the core region nor the internal repeats were capable of homodimerization.

The minimal fragment capable of centromere localization with high affinity (e.g., mitosin-pTKB-N) will be used as a probe to clone the gene(s) coding for the downstream target(s) of mitosin. The molecular architecture of the kinetochore can thus be approached progressively.

The dynamics of mitosin-kinetochore interaction.

With the study of a cytoplasmic mutant, mitosin-pTCB, we explored the kinetics of mitosin-kinetochore interaction qualitatively. We found that bound mitosin did not exchange its binding sites easily with free forms after NEBD. Significant exchange could be observed only when mitosis was blocked to allow prolonged time of competition. This result implied that the assembly of mitosin into kinetochores was completed prior to NEBD. After this point, free mitosin could no longer be recruited into kinetochores effectively. One of the possible explanations is that, after NEBD, access to the bound mitosin or mitosin-binding sites is denied by another component(s) of kinetochores assembled after mitosin. It has long been suspected that functional kinetochores are assembled in multiple stages. In cells at interphase, precursors of kinetochores, or prekinetochores, are found as discrete foci which contain certain centromere proteins (e.g., CENP-A, CENP-B, and CENP-C) colocalized with alphoid satellite DNA (12, 19, 20). In contrast, mitosin is one of the components assembled onto kinetochores in prophase (39).

The core region may regulate the putative chromatin-binding activity of mitosin.

Several lines of evidence suggest that mitosin might bind to chromatin proteins or DNA in interphase. First, several mitosin mutants colocalized with brightly stained DNA foci in cells at interphase. Among cells expressing mitosin-pTZ, for instance, those bright DNA foci were observed only in transfected cells; mitosin-pTZ appeared to have reorganized chromatin distribution in these cells (Fig. 5B). This phenomenon was also observed in cells expressing mitosin-pTNE (Fig. 5B), mitosin-pTC (41), and several other mutants (data not shown). Second, our data suggested that there might be a chromatin-binding domain between residues 2902 and 3037. This region covers the previously identified in vitro Rb-binding domain and the intrinsic NLS (39, 40). Third, mitosin-pTNC exhibited both the colocalization with nuclear DNA in interphase and localization along chromosomes in mitotic cells. Last, during interphase, wild-type mitosin also colocalized with nuclear DNA in cells treated with hypotonic buffer (data not shown). The nuclear matrix is a complex structure implicated in multiple functions including chromatin organization (reviewed in references 2 and 6). The present results further imply that, as a nuclear matrix protein in interphase (17), mitosin/CENP-F might provide sites of attachment for proper chromatin organization. Detailed studies at the electron microscopic level are required to further test these speculations.

The chromosome localization of mitosin-pTNC, a mutant lacking only the core region of mitosin-pTN, provided possible insights into the regulation of different activities in mitosin. Activation of the core region (possibly by hyperphosphorylation of mitosin at or after the G₂/M transition [39]) might in turn deactivate the interaction of mitosin with chromatin for concerted behaviors of the molecule during the cell cycle (39). In this case, the putative chromatin-binding property of mitosin would be functional only in interphase. Nevertheless, colocalization of mitosin with nuclear DNA could have resulted from interaction with nuclear components other than chromatin. The aberrant association of mitosin-pTNC with chromosomes could also be due to artifacts generated by mutant proteins. Although several centromere proteins, including CENP-A (28, 32, 33), CENP-B (18, 23, 38), and CENP-C (29, 36), have been shown to possess DNA-binding activities, direct evidence is required to clarify if mitosin indeed binds directly to chromatin proteins or DNA in cells.