Chromosome Association of Minichromosome Maintenance Proteins in Drosophila Mitotic Cycles

For antibody staining, embryos were fixed and stained with purified primary antibodies, and diluted 1:50–1:500 using standard procedures. Specificity of rabbit polyclonal antibodies against each MCM has been described previously (Su et al., 1996). Anti–β-tubulin antibody (Amersham, Arlington Heights, IL) was used at 1:100 dilution. Primary antibodies were detected either with fluorescent secondary antibodies (Jackson ImmunoResearch Laboratories; West Grove, PA) or by histochemical staining using the peroxidase reaction. DNA was stained with 10 mg/ml bisbenzamid (Hoechst 33258). WGA (Molecular Probes, Inc., Eugene, OR) was used at 500 ng/ml.

Stable versions of mitotic cyclins were constructed essentially as described previously (Sigrist et al., 1995). Briefly, stable cyclin A lacks amino acids (aa) 2–170. Stable cyclin B lacks aa 3–47 and has a glycine as the second aa residue, before aa 48 from the wild-type sequence (Sprenger et al., 1997). Both cyclins were produced from hsp70 promoter and carried as homozygous lines. w⁶⁷, the isogenic stock used for carrying the transgenes, was used as a control in all heat shock experiments. Drosophila embryos were collected for 2 h and aged for either 2 h at room temperature (rt; to reach cycle 14) or for 5 h at rt (to reach cycle 16) before heat shock. Embryos were heat shocked by floating grape juice–agar plates on water at 37° for 30 min. After recovery at rt for 2 h, embryos were fixed for 20 min in a two-phase mixture of heptane and fix (PBS + 10% formaldehyde) according to standard procedures, and were processed for antibody staining as described above. The level of induced stable cyclins is about two- to three fold over that of endogenous cyclins. Cell cycle and developmental stages of embryos were confirmed by the morphology of embryos and the density of cells.

Embryos were collected from heterozygous dap^3x41 parents for 1 h and aged at 25° for 5.5 h to reach cycle 16 before fixing. This is a mutant resulting from imprecise excision of the transposon from the dap³ line (de Nooij et al., 1997). A Cyo wg-LacZ balancer was used to identify the genotype to embryos by immunostaining for β-galactosidase (Amersham antibodies). Homozygotes were identified by the lack of β-galactosidase staining.

Identification and characterization of three Drosophila MCMs, MCM2, MCM4, and MCM5, have been described (Treisman et al., 1995; Feger et al., 1995; Su et al., 1996, 1997). Here, we first describe their localization during embryonic divisions as detected by immunostaining of fixed Drosophila embryos. Fig. 1, A and B, shows nuclear staining for DmMCM2 during G₂ of cycle 14, dispersal of staining during mitosis 14, and reaccummulation in telophase 14 (DmMCM4 and DmMCM5 staining was indistinguishable from DmMCM2; not shown). Although all interphase cells exhibited nuclear MCM staining (see below), these early cell cycles lack a G₁ phase. Fig. 4 documents that DmMCM5 is nuclear during G₁ of cell cycle 17 (similar observations were made for DmMCM2 and DMCM4; not shown). We conclude that these proteins are nuclear in G₁, S, and G₂, and that they are dispersed from chromatin upon entry into mitosis. This cell cycle pattern of localization is similar to the behavior of vertebrate MCMs (reviewed in Chong et al., 1996).

During the embryonic cycles that lack a G₁ phase, relocalization of MCMs to the nuclear region occurred as nuclei exit mitosis (Fig. 1). Association was first detected on the mitotic chromosomes at late anaphase/telophase (Fig. 1, C–H, white arrows). Relocalization of MCMs precedes and overlaps that of nuclear pore complexes (NPCs), as visualized by staining with WGA (Fig. 1, D and G; Finlay et al., 1987; see also Stafstrom and Staehelin, 1984). Therefore, this relocalization, at least initially, most likely results from binding of MCMs to chromatin and not to nuclear import. Similarly, localization of MCMs to chromatin precedes complete nuclear reformation in Xenopus (Coue et al., 1996). In both syncytial and postblastoderm cycles, nuclear MCM immunofluoresence increased further as cells progressed into interphase (e.g., Fig. 1, F–H, black arrows). This increase is possibly caused by import through NPCs: While only some MCMs contain sequences with obvious similarity to nuclear localization signals, they form complexes with each other, and analyses in Xenopus have documented import into intact nuclei (Madine et al., 1995a).

Next, we asked what mitotic event brings about chromosome association of MCMs. Chromosome association of MCMs correlates temporally with progression from metaphase to telophase. Progression beyond metaphase in many organisms is driven by proteolysis, the substrates of which include mitotic cyclins. Possibly, it is the degradation of one or more of these proteins that allows MCMs to associate with chromosomes. To ask if mitotic cyclins are involved, we analyzed chromatin association of MCMs when cyclin degradation is prevented.

Deletion of the “destruction box” at the amino termini of mitotic cyclins results in their stabilization in many systems, including Drosophila (Ghiara et al., 1991; Holloway et al., 1993; Surana et al., 1993; Sigrist et al., 1995). Expression of stable cyclins from heat-inducible promoters and the consequences during postblastoderm embryonic divisions of Drosophila have been described (Sigrist et al., 1995). Similar experiments using our stable cyclin constructs reproduced these observations (Sprenger et al., 1997; see Materials and Methods). To restate briefly, in the presence of stable cyclins, cells enter mitosis but they arrest with condensed chromosomes. In a mitotic arrest by stable cyclin A (cyclin A^s), the chromosomes and spindle have metaphase appearance and the sister chromatids remain together. In contrast, in an arrest caused by stable cyclin B (cyclin B^s) or coproduction of cyclins A^s and B^s, sister chromatids are separated as they would be at the meta– phase–anaphase transition. Importantly, endogenous cyclins are degraded normally in an arrest caused by stable cyclins; i.e., cyclin B is degraded in an arrest caused by cyclin A^s, and cyclin A is degraded in an arrest caused by cyclin B^s (Sigrist et al., 1995).

To test if MCM–chromosome association requires cyclin degradation, we analyzed three Drosophila MCMs after induction of stable cyclins in cell cycles 14 and 15. All three MCMs dispersed from the nucleus upon entry to mitosis in the presence of stable cyclins, as they do in normal mitoses (not shown). During arrests in M14 and M15, reassociation with chromosomes occurred in the presence of cyclin B^s but not cyclin A^s (Fig. 2; data not shown). In the presence of cyclin A^s, DmMCM2 staining was uniform throughout the cell, except for exclusion from the region occupied by chromosomes (Fig. 2, panel 2, arrowheads). This is similar to what is seen in metaphases of wild-type embryos and heat-shocked w⁶⁷ controls lacking the transgenes (not shown). In contrast, in the presence of cyclin B^s, no such clearing of the DmMCM2 stain from the chromosomal region was seen. In most cells, colocalization of DmMCM2 stain and DNA stain is observed, suggesting concentration of DmMCM2 on chromosomes; this is similar to chromosomes in late anaphase of wild-type mitoses (e.g., Fig. 1, F–H, white arrows). The intensity of DmMCM2 stain on the chromosomes varied from cell to cell; this may be because cells of a Drosophila embryo undergo M14 and M15 asynchronously (Foe, 1989) and had been arrested in mitosis for various lengths of time at fixation.

While we cannot rule out that the lack of DmMCM2 stain on chromosomes that we report here and in Fig. 1 is caused by epitope masking at certain stages of mitosis, this notion is not consistent with the strong cytoplasmic staining of DmMCM2 seen at the same stages (e.g., in A^s arrest in Fig. 2, panel 2).

These results indicate that cyclin A^s is able to prevent DmMCM2 association with chromosomes, while cyclin B^s lacks the ability to do so under our experimental conditions. Consistent with this idea, when cyclins A^s and B^s were coexpressed, DmMCM2 was not associated with chromosomes (Fig. 2, panel 3, arrowheads). Staining for DmMCM4 and DmMCM5 showed that binding of these MCMs paralleled that of DmMCM2 (data not shown).

Our data show that all requirements for MCM–chromosome association are met during a mitotic arrest with cyclin B^s during embryonic mitoses M14 and M15. These mitoses are followed immediately by S phase and no gap phase. Thus, many activities required for MCM–chromosome association and DNA replication are likely to be present at these times. Later, upon the completion of M16, many epidermal cells withdraw from the cell cycle for the remainder of embryogenesis and enter a prolonged gap phase (G₁17; Foe et al., 1993). When cells were arrested in M16 with cyclin B^s, DmMCM2 (Fig. 3,D) stain was excluded from the region occupied by chromosomes (Fig. 3 C). These data indicate that in a cyclin B^s arrest in M16, all requirements for MCM–chromosome association are not met, whereas they were met in cyclin B^s arrest in M14 or M15.

The difference between M16 and earlier mitoses is also evident in careful analysis of unperturbed embryos. We found that chromosomes in late anaphase/telophase of M16 did not accumulate MCMs, whereas chromosomes at the same stage in earlier mitoses had acquired the MCM stain (compare bracketed nuclei in Fig. 4,B to those in Fig. 4,A). We conclude that chromosome association of MCMs does not occur or is greatly delayed in M16. One possible basis for the difference we have observed between M16 and earlier mitoses is the presence of an inhibitor of MCM–chromosome association at this terminal mitosis. The data described below are consistent with this idea. Note, however, that once cells finish M16 and enter the next interphase, nuclear MCM staining is evident, presumably as a result of nuclear import (Fig. 4,C; see also Fig. 1, F–H, and text).

We also examined embryos arrested in M16 because of production of cyclin A^s or coproduction of cyclins A^s and B^s, or because of a mutation in fizzy, which is required for mitotic degradation of cyclins (Sigrist et al., 1995). We did not see chromosome association of DmMCM2 in any of these cases (data not shown).

Analysis of dacapo (dap) mutants suggests that this gene has a role in MCM–chromosome association in M16. dap encodes a p21/p27 type cdk inhibitor (Dap) whose developmentally regulated expression just before M16 contributes to the withdrawal of epidermal cells from the cell cycle (de Nooij et al., 1997; Lane et al., 1997). In homozygous dap mutants, many epidermal cells enter an additional S phase (S17) immediately after M16. Analysis of DmMCM2 in dap homozygotes showed that many late anaphase/telophase chromosomes acquire DmMCM2 in M16, unlike those in M16 of wild-type or heterozygous sibling embryos (Fig. 5). This is illustrated most clearly in embryos in which the DNA has been stained with the fluorescent dye Hoechst 33258 and DmMCM2 has been stained immunohistochemically. When the MCMs colocalize with DNA, the histochemical stain efficiently quenches the fluorescent staining (e.g., see Fig. 4). Focusing on the bright cells, one can see that the later stages of mitosis are not quenched in cell cycle 16 of wild-type or heterozygous dap embryos, but they are quenched in homozygous dap mutant embryos (Fig. 5). For example, late anaphase/telophase nuclei in wild type/heterozygotes are devoid of histochemical DmMCM2 stain and are clearly visible when visualized for DNA (Fig. 5,A, boxes, and panels 1–3). In contrast, late anaphase/telophase nuclei in homozygous dap mutants have acquired the DmMCM2 stain and are barely visible when visualized for DNA (Fig. 5,B and panels 4–6). Not all anaphase/telophase figures acquire MCM stain in dap mutants (e.g., Fig. 5 B, arrowhead). Likewise, not all cells of the epidermis undergo S17 in a homozygous dap mutant (de Nooij et al., 1997; Lane et al., 1997). Direct analysis of DmMCM2 immunofluorescent staining also revealed association of MCMs with late anaphase/telophase nuclei of homozygous dap mutants (not shown). These data suggest that expression of Dap in cycle 16 normally inhibits chromosome association of DmMCM2.

A summary of data on MCM–chromosome interactions is shown schematically in Fig. 6 and described below.

The normal cycle of association and dissociation of MCM proteins from chromatin is thought to play a role in the cell cycle oscillations in the competence of DNA for replication. To investigate whether the mitotic association of Drosophila MCMs to chromatin is controlled by the destruction of cyclins, we have analyzed the ability of stabilized cyclins to block this association. Our results show that stable cyclin A, but not stable cyclin B, can block this mitotic association. Additionally, a difference in the behavior of MCMs during mitoses that are followed immediately by S phase (M14 and M15) and a mitosis that is followed by a prolonged G₁ quiescence (M16) indicates that MCM association may require other factors in addition to the destruction of mitotic cyclins.

The association of MCM proteins to condensed chromosomes in M14 and M15 arrested with stable cyclin B demonstrates that this association requires neither decondensation of the chromosomes nor assembly of a nuclear membrane. This finding is consistent with the observed kinetics of association of MCMs with chromosomes upon exit from mitosis in unperturbed cycles of vertebrates (Todorov et al., 1994; Coue et al., 1996; Schulte et al., 1996) and Drosophila (this report). The ability of stable cyclin A to block MCM association to mitotic chromosomes suggests either that the normal cyclin A can inhibit this association or that the deletion of the destruction box and flanking sequences produced a novel activity. While we can not presently eliminate the latter alternative, we suggest that inhibition of MCM association to chromatin by cyclin A:cdc2 kinase might contribute to the failure of MCMs to associate with chromosomes during metaphase. The activation of the proteolytic machinery and destruction of cyclin A, and possibly other inhibitors of MCM– chromosome association, would then time the beginning of the assembly of replication proteins onto DNA.

How might cyclins or cyclin-dependent kinases control MCM behavior? Some members of the MCM family contain putative cdk phosphorylation sites. In Xenopus extracts, the phosphorylation state of a MCM homologue, XMCM4, has been shown to change in concert with mitotic kinase activity (Coue et al., 1996), and cdks can direct phosphorylation of XMCM4 and promote its dispersal from the nucleus in vitro (Hendrickson et al., 1996). Direct inhibition of MCM–chromosome association by mitotic cyclin/cdk phosphorylation is suggested by these analyses. However, our efforts to detect differential modification of MCMs on the basis of mobility on denaturing and native gels have been negative (Su, T.T., and P.H. O'Farrell, unpublished data). In addition, we cannot exclude more indirect paths of regulation. For example, association of Xenopus MCMs to chromatin requires CDC6 and ORC2 (Coleman et al., 1996). Thus, cdks could influence the activity of these proteins and thereby affect MCM–chromosome association indirectly.

A differential ability of Drosophila cyclins A and B to block MCM–chromosome association might be relevant to previous findings that cyclin A, but not cyclin B, prevents rereplication of DNA in G₂ in Drosophila. That is, rereplication in cyclin A mutants occurs in the presence of cyclin B (Lehner and O'Farrell, 1990; Sauer et al., 1995). Perhaps inhibition of MCM–chromosome association by cyclin A contributes to the block to rereplication.

In many ways, our results parallel recent findings in Xenopus extracts (Mahbubani et al., 1997) in which MCMs have been shown to contribute to replication licensing factor (RLF) activity (Blow and Laskey, 1988; Chong et al., 1995; Madine et al., 1995a). (a) RLF becomes active at metaphase–anaphase transition; Drosophila MCMs binding to chromosomes follows metaphase–anaphase transition. (b) A cdk-dependent activity renders RLF inactive in metaphase; cyclin A appears to prevent chromosome association of Drosophila MCMs, presumably until its degradation. (c) 6-Dimethylaminopurine, which prevents RLF activation, has been shown to stabilize a mitotic cyclin; we show here that stabilization of a mitotic cyclin prevents MCM–chromosome association. Thus, our results complement the biochemical data in Xenopus and, additionally, provide evidence for the role of mitotic kinases in regulation of MCM behavior in vivo.

Before cycle 16 of the embryo, each mitosis is followed immediately by S phase. We suggest that any necessary preparation for DNA synthesis must occur during mitosis in these cycles to allow immediate progression into S phase subsequently. Accordingly, we found that binding of MCMs to chromosomes occurs during mitosis in precycle 16 mitoses we have examined. In contrast, M16 is followed by a long gap phase in epidermal cells. Thus, at this stage in embryogenesis, there is an extended postmitotic period during which cells might prepare for DNA synthesis and binding of MCMs to chromosomes might occur at any time before S phase. Indeed, at this stage, we find that MCMs do not bind chromosomes in M16.

The change in MCM–chromosome association seen at M16 is unlikely to be a direct outcome of the transition from maternal to zygotic control of MCM transcription. This is because in Drosophila embryos, maternal–zygotic transition occurs earlier, during cycle 14 (Foe et al., 1993). The switch from maternal to zygotic transcription of DmMCM2, DmMCM4, and DmMCM5 also occurs at this time (Treisman et al., 1995; Su, T.T., and P.H. O'Farrell, unpublished observations). Furthermore, mutants in MCM genes show no defect until later stages, indicating that the maternal supply of MCM gene products is sufficient for these early stages we have examined (Feger et al., 1995; Treisman et al., 1995).

While the result in M16 appears to contradict reports of MCM–chromosome association during mitosis in mammalian cell cycles that contain a G₁, we wish to stress that G₁17 in Drosophila epidermal cells lasts for the rest of embryogenesis and terminates only upon hatching of the larvae. Thus, G₁17 may be different from G₁ of cycling cells and resemble a quiescent state. Indeed, although MCMs appear to be chromatin-bound in G₁ of cycling mammalian cells, they are downregulated in G₀ cells (Schulte et al., 1996).

We conclude that, unlike in earlier mitoses, all requirements for chromosome association of MCMs are not met in M16. Similarly, in an arrest with cyclin B^s during M16, DmMCM2 was not present on chromosomes, unlike in cyclin B^s arrests at earlier mitoses. The inability of M16 chromosomes to bind MCMs appears to depend on dap, which is expressed for the first time in G₂16 in epidermal cells. In dap mutants, DmMCM2 binds chromosomes in M16 (this report) and cells enter S phase instead of G₁17 (de Nooij et al., 1997; Lane et al., 1997). Our data, therefore, implicate Dap as an inhibitor of chromosome association of MCMs.

Dap is a specific inhibitor of cyclin E:cdk2 complexes, and its role in promoting the arrest of cycle 17 cells in G₁ is thought to result from its contribution to the downregulation of cyclin E (Lane et al., 1997). Indeed, downregulation of cyclin E is known to be essential for G₁ quiescence in cycle 17, and expression of cyclin E is sufficient to drive ectopic S phase (Knoblich et al., 1994). This raises the possibility that Dap inhibits MCM–chromosome association by inhibiting cyclin E:cdk2. Perhaps cyclin E:cdk2 promotes chromosome association of MCMs. The behavior of MCMs in mitoses up to and including M16 is consistent with this idea (Fig. 6) First, cyclin E is present continuously in embryonic cycles in which S follows M (Knoblich et al., 1994); MCMs associate with mitotic chromosomes after metaphase in these cycles (Fig. 6,A). Second, cyclin E activity is absent in M16, which is followed not by S but by G₁17 (Knoblich et al., 1994); likewise, MCMs do not associate with M16 chromosomes (Fig. 6,B). In dap mutants, M16 is followed by S17, and cyclin E activity is thought to persist (Fig. 6 C). On the basis of these data, we propose that mitotic cyclin degradation creates a permissive state for MCMs to associate with chromosomes, but that cyclin E:cdk2 also contributes positively to this association. Observations in mammalian cultured cells suggests that mitotic association of MCMs and chromosomes can occur in cell cycles that include a G₁ and lack cyclin E during mitosis (Schulte et al., 1996; Sherr, 1994). We do not know the basis for these different behaviors, but we suggest that the activity provided by cyclin E in Drosophila may be supplied otherwise, perhaps by a redundant function, in mammalian cells.

In contrast to our proposal that cyclin E:cdk2 stimulates MCM–chromosome association, a recent report argued that addition of high levels of cyclin E to Xenopus extracts can inhibit MCM binding to sperm chromatin (Hua et al., 1997). However, this report emphasized the role of cyclin E in the inhibition of DNA rereplication. We concur with this emphasis and argue that cyclins make both positive and negative inputs to DNA replication and the assembly of proteins (such as MCMs) at origins. In Drosophila, both cyclins A and E are capable of driving G₁ cells into S phase, yet both cyclins A and E are able to inhibit postreplicative cells from rereplicating their DNA (Sauer et al., 1995; Sprenger et al., 1997; Follette, P.J., R.J. Duronio and P.H. O'Farrell, manuscript in preparation; reviewed in Su et al., 1995, and Wuarin and Nurse, 1996). In yeast, cyclins have both stimulatory and inhibitory inputs into DNA replication (reviewed in Diffley, 1996; Nasmyth, 1996). In Xenopus, depletion of the endogenous cyclin E demonstrates that replication in Xenopus extracts requires cyclin E (Fang and Newport, 1991; Jackson et al., 1995), yet high levels of exogenous cyclin E block the binding of MCM protein to newly added chromatin (Hua et al., 1997). These observations are consistent with the idea that cdk activity has at least two inputs into cycles of replication, one positive (perhaps leading to binding of MCMs to chromatin) and one negative (possibly inhibiting rebinding of MCMs to chromatin).

In conclusion, our data suggest a role for mitotic cyclin degradation in chromosome binding of replication factors, MCMs. Degradation of mitotic cyclins is essential for the completion of mitosis. The completion of mitosis thus appears to be intimately linked to preparation for DNA replication by a common event, degradation of mitotic cyclins. As a result, the two major phases of the cell cycle, mitosis and DNA replication, may be viewed, not as discrete phases that must be ordered, but rather, as events that are coupled such that completion of one may occur concurrently with preparation for the other.

We thank Anita Sil, Peter Follette, and Smruti Vidwans for critical reading of the manuscript; Wallace Marshall for WGA; Drs. de Nooij and Hariharan for dacapo stocks; and Frank Sprenger for sharing stable cyclin stocks before publication.

This work was supported by a National Institutes of Health (NIH) postdoctoral fellowship (GM15032) to T.T. Su and NIH grant (2-RO1-GM37193) to P.H. O'Farrell.

aa

amino acid

M16

mitosis 16

MCM

minichromosome maintenance (protein)

NPC

nuclear pore complex

RLF

replication licensing factor

rt

room temperature