Dynamic changes of NuMA during the cell cycle and possible appearance of a truncated form of NuMA during apoptosis

Nuclear-mitotic apparatus (NuMA) protein was first discovered as a nuclear matrix-associated protein that was able to translocate to the spindle pole regions at mitosis (Lydersen and Pettijohn, 1980). Several proteins were reported with a similar cell cycle-dependent redistribution pattern and with a very close molecular mass, such as SPN antigen (Kallajoki et al., 1991), centrophilin (Tousson et al., 1991), 1F1/1H1 antigen (Compton et al., 1991), SP-H antigen (Maekawa et al., 1991), p240 antigen (Yang et al., 1992), and W1 antigen (Tang et al., 1993). cDNA and peptide sequencing led us to conclude that all represent the same protein or a family of closely related proteins (Compton et al., 1992; Yang et al., 1992; Tang et al., 1993; Kallajoki et al., 1993; Maekawa and Kuriyama, 1993). The first proposed and the most descriptive name, NuMA, was suggested to refer to this protein (Compton et al., 1992; Cleveland, 1995).

The predicted structure of NuMA (∼230 kDa) consists of the proline-rich globular headand tail-domains, and a central discontinuous α-helical rod region with the heptad repeats of coiled-coil arrangements, similar to the intermediate-filament family of proteins (reviewed by Stewart, 1990). Experimental evidence has suggested that the C-terminal globular tail contains a nuclear localization signal motif and a spindle association domain (Compton and Cleveland, 1993; Tang et al., 1994). All the reports essentially agree that NuMA dissociates from condensing chromatin during early prophase, remains association with the spindle pole regions at metaphase and anaphase, and redistributes into the reforming nucleus in telophase. Rapid relocation of NuMA to the mitotic centrosomes has led to the hypothesis that its poleward movement is based on a microtubule minus end-directed motor protein (Compton et al., 1992). Cells microinjected with antibody against NuMA showed aberrant multipolar spindles and micronucleation, indicating that NuMA is required for normal spindle assembly and for completion of mitosis (Kallajoki et al., 1991, 1993; Yang and Snyder, 1992). Expression of tailless NuMA in DNA-transfected cells resulted in post-mitotic micronucleation, further suggesting that the role of NuMA is in normal reassembly of a single nucleus following completion of mitosis (Compton and Cleveland, 1993). Furthermore, evidence showing that NuMA was colocalized and coprecipitated with splicing factors raised the possibility that NuMA in interphase nuclei may be in close contact with the pre-mRNA splicing apparatus (Zeng et al., 1994).

While much is now known about the behaviour of NuMA as a mitosis-dependent relocated protein, until recently it has not been well documented whether there are biochemical changes in NuMA correlated with its morphological dynamics in the cell cycle, although evidence has shown that NuMA’s mitosis-specific interaction with the mitotic spindle relies on the Cdc2 phosphorylation sites in NuMA (Compton and Luo, 1995). We report here that interphase NuMA appears to be a 220 kDa protein, and that it alters during mitosis to yield sequentially two additional forms with apparent molecular masses of 240 kDa and 180 kDa. The orderly alterations involve phosphorylation, dephosphorylation and proteolytic cleavage. Moreover, we have developed a double immunofluorescence staining technique that allows us to observe subtle spatial changes of NuMA in relation to chromatin condensation, chromosome movement and mitotic spindle dynamics. These morphological changes correlate temporally with the post-translational modifications of NuMA in the cell cycle. Interestingly, if the 180 kDa truncated NuMA was induced in interphase, nuclear architecture was collapsed and nuclear apoptotic phenomena were observed. These findings suggest that NuMA functions in interphase cells to retain nuclear structural integrity. After the onset of apoptosis, NuMA cleavage seems to be a common consequence that may contribute to nuclear disruption during the cell-death pathway.

Human cervical carcinoma HeLa, T leukemia CEM, and myeloblastic leukemia HL60 cell lines were cultured in RPMI 1640 medium supplemented with 10% heat-inactivated fetal bovine serum (Gibco-BRL). HeLa cells were passaged with trypsin-EDTA to maintain subconfluency. CEM and HL60 cells were subcultured twice a week.

The procedures for preparation of CEM nuclei as immunogen, and for production of hybridomas were identical to those reported previously (Chen et al., 1991). Hybridoma supernatants were pre-screened by a dot-immunobinding assay similar to that described (Pai et al., 1995). These selected hybridomas were rescreened by indirect immunofluorescence microscopy. Monoclonal antibody (mAb) HS3 stained the spindle pole regions of metaphase cells and its target antigen was found to be identical to NuMA (Compton et al., 1992; Yang et al., 1992).

Seven overlapping restriction fragments derived from a partial NuMA cDNA (see below) (corresponding to NuMA amino acid residues 350-630, 350-896, 632-934, 879-934, 898-1067, 936-1156, 1151-1347; using the numbering system of Yang et al., 1992) were subcloned into a T7 expression vector, pRSET (Invitrogen). The purified overexpressed fusion proteins were mixed and used as immunogens to raise mAbs against different regions of NuMA protein. A batch of mAbs was obtained. HS87 mAb recognized fusion proteins derived from NuMA cDNA fragments encoding amino acid residues of 350-896 and 632-934. HS87 alone was used throughout this study, but using combined mAbs of HS3, HS17, HS74, HS87, HS99, and MY121 for immunoprecipitation. All the antibodies are able to detect the 220 kDa, 240 kDa, and 180 kDa forms of NuMA on immunoblots.

HS3 was used to immunoscreen a human hepatoma HA22T λgt11 cDNA library, as followed by a procedure previously reported (Fang and Yeh, 1993). The cDNA inserts were subcloned into the EcoRI site of a pBluescript II vector (Stratagene). Sequence analysis of a 4.2 kb insert revealed that it is identical to the nucleotides 1048-5214 of NuMA cDNA (amino acids 350-1738) as reported (Yang et al., 1992). An EcoRI-HindIII fragment (1.6 kb) of NuMA cDNA encoding amino acid residues 350-896 that contains the HS87 epitope, and an XhoI-digested pBluescript KS+ vector were filled in and ligated. Clones with the correct orientation were selected. Nested deletions from the 3′-end were generated using the Exo/Mung system (Strategene).

An XbaI linker (Clontech) containing translational stop codons in all reading frames was added to the ligation mixtures. The ligation products were used to transform E. coli XL1-Blue. The transformants carrying the plasmids were induced by 1 mM isopropyl thio-β-D-galactopyranoside (Strategene). Cells were pelleted and lysed in reducing Laemmli sample buffer. After boiling for 5 minutes, samples were subjected to SDS-PAGE and immunoblotting. Clones positive or negative to HS87 were scored and their exact deletion sites were determined by sequencing. The 5′-end nested deletion was performed similarly using a plasmid carrying NuMA cDNA encoding amino acid residues 350-789 as the starting material but without adding XbaI-linker during ligation.

Exponentially growing HeLa monolayers were trypsinized and washed once with phosphate-buffered saline (PBS). The cells were seeded in 60 mm culture dishes (5 ml/dish) at 1×106 cells/ml with 0.4 μg/ml nocodazole (Sigma) for M-phase synchronization. For M-phase release experiments, the rounded-up mitotic HeLa cells, collected after treatment with nocodazole (0.4 μg/ml) for 24 hours, were washed twice in PBS and recultured in fresh medium. After incubation of the cells at 37°C for various time periods (0-36 hours) in M-phase synchronization and release experiments, cells in suspension and monolayer were harvested together, washed twice in PBS, and resuspended in PBS at 4×107 cells/ml. Aliquots of cells from the same sample were subjected to immunofluorescence staining and immunoblot analysis. At some time points as indicated, cells remaining in suspension or adhesion were analysed separately.

Stock solutions of cycloheximide (7 mM; Boehringer Mannheim) in PBS and of okadaic acid (OA) (100 μM; Sigma), 6-dimethylaminopurine (6-DMAP) (0.25 M; Sigma), 5,6-dichloro-1-β-D-ribofuranosylbenzimidazole (DRB) (0.5 M; Sigma), camptothecin (150 μM; Sigma), staurosporine (1 mM; Sigma), and A23187 (10 mM; Calbiochem) in dimethyl sulfoxide were stored in aliquots at −70°C. All of the stock were added directly to the HeLa or HL60 cell cultures at a final concentration as indicated in each experiment. All culturing conditions and assay procedures were similar to those described above for nocodazole treatment.

Purified anti-human cyclin B1 mAb (1 μg; PharMingen, cat. no. 14541A) in 500 μl RPMI 1640 containing 10% fetal bovine serum was tumbled at 4°C overnight with 40 μl of a 50% slurry of Protein A-Sepharose (Pharmacia LKB) suspended in NET buffer (0.5% NP-40, 1 mM EDTA, 1 mM EGTA, 150 mM NaCl, 5 mM MgCl2, 50 mM Tris-HCl, pH 8.0). The Sepharose beads were spun down and washed three times with NET buffer. About 1×108 M-phase arrested HeLa cells (treated with 0.4 μg/ml nocodazole for 24 hours) were lysed in 1 ml of extraction buffer (0.5% Triton X-100, 150 mM NaCl, 0.2 mM EGTA, 10 mM EDTA, 10 mM KH2PO4, 5 mM MgCl2, 20 mM Tris-HCl, pH 7.4, 10% glycerol, 1 mM phenylmethylsulfonyl fluoride, 5 mM NaF, 0.4 mM Na3VO4). After incubation at 4°C for 10 minutes, the cell lysates were clarified by centrifugation at 10,000 g for 10 minutes and stored at −70°C. A 20 μl sample of the cell lysate and 200 μl of NET buffer were added to the above anti-cyclin B1 antibody-bound beads. After rotation for 4 hours at 4°C, beads were washed three times with NET buffer and then suspended in kinase reaction buffer (50 mM Tris-HCl, pH 7.4, 10 mM MgCl2, 1 mM dithiothreitol). These beads coupled with cyclin B1/Cdc2 complex were used for the in vitro kinase assay.

Protein A-Sepharose beads were incubated with the mixed hybridoma supernatants of HS3, HS17, HS74, HS87, HS99, and MY121 (250 μl of each) as described above for preparation of the anticyclin B1 antibody-bound beads. Interphase or DRB (0.25 mM) treated HeLa cells were lysed in reducing Laemmli sample buffer (2×107 cells/ml) and boiled for 5 minutes. These cell lysates were diluted by adding nine volumes of NET buffer. The diluted cell lysate (100 μl) was incubated with the anti-NuMA antibody-bound beads for 4 hours at 4°C. Essentially all NuMA from 2×105 HeLa cells was immunoprecipitated under these conditions. After washing, NuMA-bound beads were mixed with the cyclin B1/Cdc2-bound beads in 40 μl kinase reaction buffer with or without 5 mM ATP. In some experiments, a Cdc2 kinase inhibitor 6-DMAP (Jessus et al., 1991) at a final concentration of 5 mM was added to the kinase reaction buffer. For all experiments, the kinase reaction was performed at 30°C for 30 minutes. Bound proteins were eluted with reducing Laemmli sample buffer (boiled for 5 minutes) and subjected to SDS-PAGE and immunoblotting. For 32P-labelling experiments, 2.5 mM ATP and 5 μCi of [γ-32P]ATP were supplemented in the kinase buffer and the reaction time was for 1 hour as described previously (Pai et al., 1995). The same filter was analyzed with immunoblotting and with the Molecular Dynamics ImageQuant phosphoimaging system following an 8 hour exposure.

HeLa cells after drug treatment were suspended in PBS and deposited on microscopic slides by cytospin centrifugation. These cells were fixed with 3% formaldehyde in PBS for 20 minutes at room temperature, followed by permeabilization with −20°C acetone for 3 minutes and washing once with PBS. Indirect immunofluorescence staining was performed with HS87 hybridoma supernatant or anti-α-tubulin mAb (Sigma, Cat. No. T-9026) and then with fluorescein isothiocyanate (FITC)-conjugated goat anti-mouse IgG (Cappel). Preparation of the mounting fluid followed a protocol modified from that described (Johnson and Nogueira Araujo, 1981). A 1 ml sample of ice-cold PBS (pH 7.4) containing 10 mg of p-phenylenediamine (Sigma) was added to 9 ml of ice-cold glycerol. The solution was satisfactory for use after storage at −20°C in dark until its color became yellow. About 25 μl of the mounting fluid was added to the sample before examination by fluorescence microscopy (Olympus BX50 equipped with BX-FLA).

HeLa cells were treated with OA or DRB and HL60 cells were treated with camptothecin, staurosporine, cycloheximide, or A23187 as indicated. About 2×106 cells were lysed in 200 μl DNA extration buffer containing 10 mM Tris-HCl, pH 8.0, 100 mM NaCl, 25 mM EDTA, 0.5% SDS, 0.5 mg/ml proteinase K (Caelles et al., 1994). After overnight incubation at 37°C, 500 μl ethanol and 20 μl of 7.5 M CH3COONH4 were added to the lysate. Precipitated DNA was dissolved in 50 μl TE buffer (10 mM Tris-HCl, 1 mM EDTA, pH 7.5) containing 50 μg/ml RNase A. After incubation at 37°C for 2 hours, DNA was analyzed on a 2% agarose gel and visualized by ethidium bromide staining (5 μg/ml).

Cells were dissolved in Laemmli sample buffer and boiled with 5% β-mercaptoethanol for 5 minutes. Proteins were separated by 6% SDS-PAGE and transferred to nitrocellulose filters. The filters were reacted with HS87 hybridoma supernatant and then with peroxidaseconjugated goat anti-mouse IgG (Bio-Rad) as described (Chen et al., 1991). The specific proteins were visualized by use of the enhanced chemiluminescence system (Amersham).

In attempts to search for nuclear proteins that exhibit dynamic changes during mitosis, mAbs were generated against purified nuclei and screened for those with variable M-phase staining patterns. One mAb, HS3, was particularly noted since it bound to the spindle pole region of cells at metaphase and anaphase but showed a speckled staining pattern in telophase cells. This mAb was used to screen a human cDNA library. The nucleotide sequence of one isolated cDNA, HS3 clone 1 (4.2 kb), was determined. Comparing the deduced amino acid sequence with the protein data bank revealed that the HS3 antigen is NuMA (Fig. 1A). We then raised a batch of mAbs against the expressed fusion proteins which corresponded to the long helix of NuMA in order to study the topology of NuMA in vivo during different stages of mitosis. Comparing all of the mAbs, HS87 showed the finest staining features throughout mitosis (see Fig. 2), indicating that epitope recognized by HS87 was not interfered by other cellular components at different phases of the cell cycle. This epitope was defined in a region between amino acid residues 645 and 683 (Fig. 1). The HS87 antibody was then used throughout this study.

We found that an antifade reagent, para-phenylenediamine (Johnson and Nogueira Araujo, 1981), acts as a DNA dye as well as a fluorochrome with absorption characteristics at the wavelength of blue light, the same range as FITC. Both paraphenylenediamine and FITC can be excited using one exciting filter that transmits blue light; emitted lights were yellowish brown and green, respectively. This method allowed us to observe both DNA (yellowish brown) and the FITC-antibody stained antigen (green) simultaneously. It was used for comparing the relative positions of NuMA and DNA in HeLa cells at interphase (Fig. 2A,C) or during mitosis with heterogeneous cell populations (Fig. 2B,D-N). In interphase, NuMA and DNA were mingled together (Fig. 2C); note that NuMA was sparse in the nucleolar region where there is the least density of DNA (Thiry, 1992). During mitosis, NuMA was dissociated from the condensing chromatin at early prophase, in which nucleoli were still visible (Fig. 2D); it formed speckle-like aggregates at late prophase when nucleoli were disassembled but the nuclear envelope was still intact since no cytoplasmic NuMA was detected at this time (Fig. 2E). As cells progressed into early prometaphase with the nuclear envelope disintegrated and chromosomes not yet fully condensed, a filamentous structure was observed (Fig. 2F). This was presumably the stage when NuMA was captured by the mitotic spindle, identified by staining cells of the same stage with an anti-tubulin antibody (Fig. 2O). NuMA was then accumulated at the newly duplicated centrosomes (Fig. 2G) and later at the separated centrosomes (Fig. 2H). Filamentlike NuMA was also found when the two centrosomes were already moving apart (Fig. 2I). Two prominent microtubule asters were detected at the same mitotic stage (Fig. 2P). It was noticed that a significant amount of diffuse NuMA existed in the cytoplasm throughout prometaphase (Fig. 2F-I); however, at metaphase most NuMA appeared to be concentrated at the polar regions to form two bright crescents which were completely away from the chromosomes (Fig. 2J). As the chromosomes segregated at anaphase, the staining intensity of NuMA at the spindle pole regions decreased (compare Fig. 2J with Fig. 2K,L). Filamentous NuMA reappeared again at late anaphase (Fig. 2M) with a staining pattern similar to the staining of microtubule arrays at late anaphase (Fig. 2Q). Eventually, NuMA was distributed in a dotted manner around the entire periphery of the telophase chromosomes (Fig. 2N).

We then explored whether the distinct relocation of NuMA in mitotic cells reflects mitosis-specific modifications of NuMA. Adherent interphase HeLa cells showed only the 220 kDa protein (Fig. 3A, lane 1), whereas rounded-up mitotic cells had three types of NuMA with apparent molecular masses of 240 kDa, 220 kDa, and 180 kDa (Fig. 3A, lane 2). Nocodazole, which arrests cells at prometaphase, was then used for a timecourse study. The 240 kDa form of NuMA appeared within 8 hours of treatment and gradually accumulated afterwards, in contrast to the 220 kDa protein which was completely lost after 32 hours of treatment (Fig. 3A, lanes 3-9). The 180 kDa form of NuMA appeared at a relatively later time compared to that of the 240 kDa NuMA. After treatment for 12 hours, mixed mitotic (40%) and interphase (60%) cells were observed (Fig. 3C). At this time point, amounts of the 240 kDa and 220 kDa NuMA were about equal (Fig. 3A, lane 5). Since the 220 kDa NuMA was contributed mainly by the interphase cells of the mixed populations, the 240 kDa NuMA was likely to be the one that aggregated into the small foci observed in the mitotic cells. At 24 and 36 hours, mitotic cells with bright dots became prominent and so did the 240 kDa NuMA (Fig. 3A, lanes 7 and 9; D,E) suggesting that the bright dots consist of the 240 kDa NuMA. After treatment with nocodazole for 24 hours, about 5% of the cells remained adherent. Microscopic examination revealed that these cells had a distinct nucleolar structure and an intact nuclear envelope, but with DNA and NuMA distributed less homogeneously than those of the untreated cells, presumably at the G₂/M transition (Fig. 3F). However, only 220 kDa NuMA was found in these adherent cells even when they had been exposed to nocodazole for 24 hours (Fig. 3A, lane 10). The prometaphase-arrested HeLa cells (24 hours after nocodazole treatment), predominantly with the 240 kDa form (Fig. 3B, lane 2; D), were released from the nocodazole block in order to examine how the three forms of NuMA evolved during the exit from mitosis. While 240 kDa NuMA decreased gradually, 180 kDa NuMA appeared soon after release and increased gradually, but declined to trace amounts when most cells had adhered to the culture plates at 36 hours after withdrawing the nocodazole (Fig. 3B, lanes 2-9). In contrast, 220 kDa NuMA accumulated after the release but remained throughout interphase. Since microtubules are regenerated after removal of nocodazole, typical metaphase or anaphase cells with NuMA at the bipolar centrosomes were enriched at 4 hours after the release (compare Fig. 3G with Fig. 2B). At this time point, more than 50% of the cells had already progressed into interphase with a distinct nucleolar structure (Fig. 3G), and such mixed populations showed mainly the 180 kDa and 220 kDa forms of NuMA (Fig. 3B, lane 5). Since 220 kDa NuMA was contributed mainly by the interphase cells, 180 kDa NuMA was therefore likely to be present in the metaphase and anaphase cells. This is in contrast to the 240 kDa form, which is predominant in prometaphase. After release for 8 hours (Fig. 3B, lane 10; H), the adherent cells, morphologically resembling the interphase cells, had only the 220 kDa form of NuMA (Fig. 3B, lane 12; I), whereas the cells remaining in suspension contained all three types of NuMA (Fig. 3B, lane 11). When normally rounded-up mitotic HeLa cells (Fig. 3J, lane 2) were cultured without drug treatment, the 240 kDa form of NuMA disappeared completely within 2 hours and the 180 kDa form remained for longer but became almost insignificant as most cells had attached to the plates at a time point of 6 hours (Fig. 3J, lanes 3-12). Taken together, these findings suggest that the 240 kDa and 180 kDa forms are mitosis-specific and not the products of drug side effects. Additionally, 240 kDa NuMA appears earlier than 180 kDa NuMA during mitosis but both disappear before exit from mitosis. For simplicity, we named the 220 kDa, 240 kDa, and 180 kDa proteins as the interphase (I), mitotic (M), and release (R) forms of NuMA, respectively.

It is known that okadaic acid (OA), a potent inhibitor of protein phosphatases-1 and -2A, activates Cdc2 kinase activity resulting in the transient induction of mitosis-specific events (Yamashita et al., 1990). We found that both M- and R-forms of NuMA were induced in HeLa cells treated with OA (22 nM) for 24 hours (Fig. 4A, lane 5) and the M-form appeared earlier than the R-form in a time course study for OA (data not shown). Since human NuMA has four Cdc2 kinase recognition motifs near the carboxyl terminus (Yang et al., 1992), it is likely that NuMA is a substrate for Cdc2 kinase. To test this, we examined whether the immunoprecipitated I-form of NuMA could be phosphorylated in vitro by purified Cdc2 kinase. The Cdc2 kinase and cyclin B complex, obtained from M-phase HeLa cell lysate, in the presence of ATP was able to convert the I-form of NuMA into the slowly migrating bands, which were distributed broadly compared to the M-form induced in vivo by nocodazole or OA (Fig. 4A, lanes 1, 2, 4 and 5). Presumably these broad bands represent NuMA being phosphorylated at the four Cdc2 kinase recognition sites with different levels (see also Fig. 5F, lane 4). This mobility shift was partially prevented if 6-dimethyl-aminopurine (6-DMAP), which inactivates M-phase promoting factor (MPF) activity by inducing tyrosine phosphorylation of Cdc2 kinase (Jessus et al., 1991), was added into the reaction mixture (Fig. 4A, lane 3). Hence, Cdc2 kinase phosphorylates the I-form into the multiple forms of NuMA in vitro, however, under the in vivo M-phase conditions (i.e. the presence of phosphatase) the extent of NuMA phosphorylation seems to be less. We also used [γ-32P]ATP to confirm that the slowly migrating band appearing in the Cdc2 kinase reaction mixture was due to incorporation of 32P into NuMA protein (Fig. 4B,C). Interestingly, two unknown Cdc2 substrates were coimmunoprecipitated with either NuMA or Cdc2/cyclin B (see asterisks in Fig. 4C, lane 4).

The protein sequence of NuMA has a number of potential casein kinase II (CKII) phosphorylation sites (S/TXXD/EXX; where additional acidic residues in the X position increase the phosphorylation rate; reviewed by Pinna, 1990). We then tested the effects on NuMA of a CKII inhibitor, 5,6-dichloro-1-β-D-ribofuranosylbenzimidazole (DRB) (Zandomeni and Weinmann, 1984). Strikingly, the R-form of NuMA appeared in the DRB (0.25 mM) treated interphase HeLa cells (Fig. 5A, lanes 3 and 4). After a 12 hour treatment, cells did not have detectable amounts of the M-form but showed a fast moving band similar to the R-form of NuMA (Fig. 5A, lanes 2 and 3), and these cells were rounded-up with abnormal features (Fig. 5B). Most cells showed that NuMA was dissociated from chromatin and resembled morphologically the late prophase cells (note that nucleoli had been disassembled). Some cells were micronucleated and only one cell in this field showed normal M-phase characteristics (arrowhead). Perhaps cells which have already progressed into M-phase are insensitive to DRB. To test this, DRB and nocodazole were added to the cell culture simultaneously. As predicted, the number of intact M-phase cells increased and only the non-mitotic cells exhibited the abnormal features (Fig. 5C). Cell viability at this time point (12 hours) remained higher than 80% but decreased to about 50% when the combined treatment with nocodazole and DRB was extended to 24 hours, at which time the M-form of NuMA disappeared (Fig. 5A, lane 5). If nocodazole was added to the culture before adding DRB, protection was more obvious (Fig. 5A, lanes 6-11; D,E), especially when the time lag was increased to 24 hours, in which case the effects of DRB on NuMA were completely abolished (Fig. 5A, lanes 10 and 11; E). Collectively, the I-form converted into the R-form in DRB treated interphase cells, but the M-form in prometaphase cells was insensitive to DRB treatment. Since the in vitro phosphatase treatment failed to convert the I-form into the R-form, although it worked well to convert the M-form into the I-form (data not shown), the large difference in the molecular masses of R-form and I-form of NuMA apparently did not rely on the CKII-dependent phosphorylation. Perhaps it is more likely that DRB activates a protease that cleaves the I-form into the R-form and this cleavage is normally pursued only in mitosis.

Taking advantage of the four Cdc2 phosphorylation sites clustered near NuMA’s C terminus, we designed experiments to clarify whether the R-form is a cleaved product. The mixed I- and R-forms of NuMA from DRB treated HeLa cells were immunoprecipitated for the in vitro Cdc2 kinase action. The I-form but not the R-form exhibited slowly migrating behavior after the reaction (Fig. 5F). Similarly, using in vitro [γ-32P]ATP labelling, we demonstrated that only the R-form of NuMA failed to incorporate 32P in the Cdc2 kinase assay (Fig. 5G,H, lane 4). These results support the idea that the R-form is a truncated form of NuMA because the Cdc2 sites near the C terminus are missing. However, whether the N terminus remains intact is not known.

Since a proportion of interphase cells showed micronucleation when their R-form of NuMA was induced by DRB (Fig. 5) or by OA (data not shown), we wondered whether the appearance of truncated NuMA in interphase cells might correlate with DNA fragmentation. To test this, HeLa cells were treated with DRB (0.25 mM) for different time intervals or with OA at various concentrations for 24 hours. We found that DNA fragmentation paralleled the accumulation of the R-form of NuMA in the treated interphase cells in a timeand dose-dependent manner (Fig. 6A). DNA fragmentation was not observed in the mitotic cells in which chromatin had already been condensed and was therefore not in contact with the truncated NuMA (Fig. 6, lane CM).

We then examined whether the R-form of NuMA could be induced in HL60 cells after stimulation by many known apoptosis-inducers such as camptothecin, staurosporine, cycloheximide, and A23187 (Gong et al., 1993; Bertrand et al., 1994; Matsubara et al., 1994). As expected, the extent of NuMA degradation was comparable with the degree of DNA fragmentation (Fig. 6B). It appears that once cells are in apoptosis, though being activated via diverse signaling pathways, NuMA cleavage is a common consequence.

In the normally rounded-up HeLa cells, which contained a mixed population of mitotic cells but with prophase cells as the major type (Fig. 2B), the I-form of NuMA was the dominant species compared to the M-form and the R-form (Fig. 3A, lane 2). Note that all of these cells exhibited features of DNA condensation and NuMA aggregation, though to different degrees (Fig. 2B). These observations suggest that dissociation of NuMA from the progressively condensing chromatin occurs before the phosphorylation of NuMA into the M-form. Interesting questions arise: which molecule(s) is responsible for the initiation of DNA condensation and how is the signal transduced to NuMA for its subsequent modification? Perphaps after initiation of chromatin compaction by RCC1 together with its associated small Ras-like GTP-binding protein Ran (reviewed by Dasso, 1993), topoisomerase II (Swedlow et al., 1993), and the SMC protein family (reviewed by Peterson, 1994), the I-form of NuMA is released from the prophase chromatin and subsequently subject to phosphorylation by MPF. Our observation that the M-form of NuMA peaks at prometaphase and declines soon after release of the block (Fig. 3B), is consistent with the fact that MPF activity abruptly arises at prophase and is inactivated at the metaphase-anaphase transition (reviewed by King et al., 1994). Moreover, after release of the prometaphase-arrest, a phosphatase activity appears to convert the M-form into the I-form.

The transition from metaphase to anaphase is induced by activation of a ubiquitin-dependent proteolytic pathway which causes cyclin B degradation and sister chromatid separation (Glotzer et al., 1991; Holloway et al., 1993). Several pieces of evidence suggest that the R-form of NuMA is also derived from protease cleavage. First, the R-form of NuMA failed to be phosphorylated by Cdc2 kinase (Fig. 5F-H), indicating that it lacks the carboxyl end containing four of the consensus phosphorylation sites for Cdc2 kinase. Second, micronucleation appeared in interphase HeLa cells if the R-form of NuMA had been induced by DRB treatment (Fig. 5A,B), a phenomenon similar to the observation that expression of tail-less NuMA results in the post-mitotic micronucleation in daughter cells (Compton and Cleveland, 1993). Third, if the R-form of NuMA is indeed a truncated protein, its fate must be irreversible and destined to complete destruction before exit from mitosis. This is fully supported by the experimental evidence that NuMA was present at much reduced levels at interphase compared to M-phase (Fig. 3A, lanes 1 and 2), and that when cells had been released from metaphase-arrest, the R-form of NuMA was gradually accumulated in the beginning and degraded completely afterwards (Fig. 3B). Consistent with this interpretation, we found that degraded forms with molecular masses lower than 180 kDa appeared along with accumulated R-form (Fig. 3B, lanes 5-7). Only the I-form of NuMA remained throughout once it had appeared at the late stage of mitosis.

Careful examination of normally progressing mitotic cells at the single cell level revealed that the fluorescence signals of NuMA at the spindle pole regions diminished once cells had passed the metaphase-anaphase transition (Fig. 2J-L). This phenomenon has also been observed by other laboratories (Kallajoki et al., 1991; Tousson et al., 1991; Compton et al., 1992; Yang et al., 1992; Tang et al., 1993). Additionally, the spindle pole (or spindle microtubule) targetting domain of NuMA has been defined at its C-terminal region (Maekawa and Kuriyama, 1993; Compton and Cleveland, 1993; Tang et al., 1994), and expression of the truncated NuMA lacking its globular tail results in the failure of binding the mutant NuMA to the mitotic spindle (Compton and Cleveland, 1993). Thus, the M-form of NuMA, with or without processing through the I-form, may change into the R-form mainly at the spindle pole region at a time beyond the metaphase-anaphase transition, resulting in loss of its ability to remain at the spindle pole and diffusion of the truncated NuMA into the surrounding cytoplasm (see Fig. 2L). This is also supported by the time course study, that the R-form of NuMA was insignificant in nocodazole induced M-phase HeLa cells (Fig. 3B, lane 2) in which NuMA was not able to be collected at the spindle pole region (Fig. 3D), but the R-form was increased quickly after release from the block (Fig. 3B, lanes 3-5), especially when bipolar distribution of NuMA became apparent (Fig. 3B, lane 5; G).

Study of the SP-H and SPN antigens has shown that the M-phase form of NuMA co-sediments with purified taxolstabilized microtubules (Maekawa et al., 1991; Kallajoki et al., 1992). Further work has demonstrated that the C-terminal onethird of the SP-H antigen remains bound to microtubules in vitro (Maekawa and Kuriyama, 1993). On the other hand, NuMA, 1F1/1H1 and SPN antigens have been observed to be associated in vivo with not only the pericentrosomal region but also its nearby emanated astral microtubules in prometaphase (Lydersen and Pettijohn, 1980; Compton et al., 1992; Kallajoki et al., 1992). Additionally, centrophilin appears to be present in the midbody along the interzonal microtubules at late telophase (Tousson et al., 1991). With the highly selected monoclonal antibody, HS87, and the newly established double-staining technique, we were able to observe the subcellular redistribution of NuMA during mitosis and have provided strong evidence that filamentous patterns of NuMA distribution exist at prometaphase and anaphase (see Fig. 2B,F,I,M). We suggest that NuMA is transported on microtubules at these stages, implying that a temporally regulated, reversible translocation of NuMA might proceed efficiently along microtubules, using as yet unidentified minus endand plus end-directed motor proteins. The M-form of NuMA might move toward the minus end and subsequently concentrate at the spindle pole. Once the M-form of NuMA reverts to the I-form (after passing the metaphase-anaphase transition), it could move away from the spindle pole, presumably by moving along the microtubules toward the plus end, and return to the chromosomes at late anaphase. Eventually, a dotted distribution pattern of NuMA can be observed at the periphery of the telophase chromosome mass (Fig. 2B,N).

Treatment of cells with nocodazole, a microtubule disrupting agent, has led to the detection of centrophilin being intimately associated with the kinetochores of chromosomes (Tousson et al., 1991). Existence of NuMA at the centromere/kinetochore is also supported by immunofluorescence staining of purified chromosomes with the 1F1/1H1 monoclonal antibodies (Compton et al., 1991). It seems likely that without the microtubule arrays, a significant amount of NuMA binds firmly with the progressively condensing chromosomes at the centromeric area, and others are aggregated into the NuMA foci which are dispersed throughout cytoplasm (see Fig. 3C-E). None of them can relocate to the pericentrosomal region simply by diffusion.

The most notable nuclear events of apoptotic cells are chromatin condensation, internucleosomal DNA cleavage, and nuclear fragmentation. These three features have been found in cells in which the R-form of NuMA has been induced inappropriately, deviating from the normal mitotic pathway (Figs 5, 6). In the normal cell cycle, the R-form of NuMA is absolutely restricted to the mitotic stage at which chromatin is in the compact form, but is fated to be degraded completely before chromatin returns to its decondensing form in interphase. It seems likely that the appearance of apoptotic nuclei correlates with accumulation of the R-form of NuMA during interphase. NuMA is structurally similar to the family of intermediate filament proteins which undergo coiled-coil based dimerization to form the elementary building blocks for assembly into the framework via the head and the tail domains of the proteins (reviewed by Heins and Aebi, 1994). Thus, protein-protein interactions through the tripartite regions of intermediate filament proteins are essential for maintaining the filamentous network. We have provided evidence which suggests that the R-form of NuMA is a truncated product, at least lacking the tail domain (Fig. 5F-H). The phenomenon that interphase nuclei collapse could be due to the presence of functionally incapable NuMA, i.e. lacking its tail domain and with the loss of its abilities to interact with other macromolecules, resulting in an increase in its solubility (data not shown) and deteriorating nuclear structure (Figs 5, 6). Recently, in vitro evidence has shown that NuMA binds specifically to the specialized DNA sequences, termed MARs or SARs (for matrix or scaffold attachment regions) (Ludérus et al., 1994), suggesting that NuMA might play a physiological role in the organization of chromatin into loop domains. Nevertheless, it is possible that once cells are in the apoptotic pathway, the induced NuMA cleavage might have a role in the nuclear structural changes, such as detachment of the chromatin loops from the nuclear matrix, that would lead subsequently to micronucleation. Consistent with this, micronucleation has been observed in transfectants that overexpress the tail-less form of NuMA (Compton and Cleveland, 1993). Whether NuMA degradation facilitates the production of oligonucleosomal laddering, a hallmark of apoptosis, would be an interesting question.

Accumulated evidence has suggested that protease activities serve as essential mechanisms in the commitment to apoptosis, such as those of interleukin-1β-converting enzyme (ICE/ced-3) (Yuan et al., 1993; Miura et al., 1993), ICE/ced-3 related proteases (Lazebnik et al.,1994; Wang et al., 1994; Fernandes-Alnemri et al., 1994; Tewari et al., 1995; Munday et al., 1995; Nicholson et al., 1995), calpain (Squier et al., 1994), granzyme B (Heusel et al., 1994), and serine proteases (Weaver et al., 1993; Zhivotovsky et al., 1994). Specific cleavage of a nuclear enzyme poly(ADP-ribose) polymerase (PARP) has been proven as an early marker in apoptosis (Kaufmann et al., 1993; Lazebnik et al., 1994). The findings that CPP32/Yama/apopain, a member of the ICE/ced-3 family, cleaves the death substrate PARP(Tewari et al., 1995; Nicholson et al., 1995) and that the pro teolytic activity of Yama can be inhibited by CrmA, a wellknown cell-death inhibitor (Ray et al., 1992; Tewari et al., 1995; Tewari and Dixit, 1995), indicate that CPP32/Yama/apopain may be a common protease involved in the cell death pathway. However, how the PARP cleavage attributes to the observed events of nuclear structural destruction during apoptosis is not clear, especially when evidence has shown that germline inactivation of PARP has no effect on either mouse development or chromatin stability (Wang et al., 1995). Our findings that the appearance of the R-form of NuMA in interphase cells correlated well with apoptotic events suggests that NuMA might be one of the nuclear structural targets for a death protease. It is important to check whether NuMA could be a substrate of CPP32/Yama/apopain. Our studies reveal some interesting features of the putative protease. The activity of this protease peaks at a time point after cells pass the metaphase-anaphase transition and disappears before mitotic exit (Fig. 3B). However, the protease in interphase cells can be activated by apoptotic inducers resulting in NuMA degradation along with DNA fragmentation (Fig. 6). Moreover, MPF activity appears to counteract directly or indirectly the activation of the protease as supported by evidence that: (i) the R-form of NuMA always appears later than the M-form (note that the M-form of NuMA relies on MPF and, therefore, reflects the in vivo MPF activity), or at time periods when the M-form is declining (see Fig. 3A,B); and (ii) nocodazole induced MPF activity abolishes the DRB effect (Fig. 5A-E). Thus, DRB is an interphase-specific inducer of apoptosis and does not work on mitotic cells. Importantly, the degraded fragments of NuMA are similar in both mitosis and apoptosis even when different cell lines were used (compare Fig. 3B, lanes 5-7 with Fig. 6B, lanes 3 and 5). If these cleavage sites on NuMA turn out to be identical, a common protease may be shared in the machinery for both mitosis and apoptosis.

We are grateful to Szecheng J. Lo for his valuable advice and critical comments on the manuscript, to Hung-Kai Chen for helpful discussions, and to Tsung-Sheng Su for providing us with the hepatoma cell line HA22T cDNA library. This work was supported by Research grants NSC 82-0420-B010-105, NSC 83-0412-B010-067, and NSC 84-2331-B010-015 from the National Science Council of the Republic of China.