Recently, we have witnessed tremendous progress in understanding the molecular anatomy of the APC in yeast and vertebrate experimental systems. The need to elucidate APC function and identify its natural substrates in human somatic cells has recently become apparent from studies demonstrating a potential link between APC-dependent proteolysis and cancer. Thus, kinetochore-associated APC regulators Mad2, Mad3, and Bub1 were found down-regulated or mutated in subsets of tumors and directly implicated in contributing to genomic instability (3, 16, 30). Molecular cloning of human securin (73) unexpectedly revealed the identity of Pds1 with the PTTG oncogene overexpressed in several types of human cancer (9, 55). Our own results showed that APC-Cdh1 assembly is controlled by the pRb-E2F tumor suppressor pathway which is frequently deregulated during multistep tumorigenesis (32). Here we have generated novel experimental tools allowing positive or negative modulation of the APC ubiquitin ligase activity by conditional manipulation of APC-Cdh1 assembly or ablation of Cdh1 function by neutralizing antibodies, respectively. We present data supporting an essential role of periodic oscillation of the APC-associated ubiquitin ligase activity for proliferation and genome integrity of human cells.

Plasmids and gene transfer. Human Cdh1 cDNA was tagged on the amino terminus with myc epitope and subcloned into the pBI tetracycline-responsive plasmid (Clontech). Expression plasmids coding for puromycin resistance, i.e., pBabePuro, for a constitutively active mutant of the retinoblastoma protein, i.e., pRbΔcdk, and for cyclin B1-luciferase were reported previously (32, 36). The cycE-Luc reporter plasmid (pCE −3565/+263) (1) was a gift from P. Jansen-Dürr. Plasmids 6xE2F-Luc containing six E2F-responsive elements in front of the TATA box and Myc-Luc containing four Myc-responsive E boxes cloned into the pG12-promoter vector and pCMV-LacZ reporter plasmids were previously described (56).

Cell culture. U-2-OS-TA, a U-2-OS human osteosarcoma cell line with stably integrated tetracycline-regulated transcriptional activator and a neomycin resistance gene (37), was transfected with the pBI plasmid containing myc-Cdh1 along with the pBabePuro plasmid in a 10:1 ratio. Puromycin-resistant clones were isolated and cultured in Dulbecco's modified Eagle's medium with 10% fetal calf serum (FCS), G418 (400 μg/ml), puromycin (1 μg/ml), and tetracycline (2 μg/ml). Induction of myc-Cdh1 expression by removal of tetracycline was performed according to procedures previously published (37). Synchronization of cells in S phase was achieved by addition of aphidicolin (5 μg/ml) into culture medium for 18 h. Metaphase-arrested cells were obtained by incubating the cells in the presence of nocodazole (40 ng/ml) for 12 h.

Immunochemical techniques. Antibodies used in this study included rabbit polyclonal antibodies to human Cdc20 (SC 8358; Santa Cruz), p27 (PC52; Calbiochem), mitosin (14C10; GeneTex), and CENP-F (NB500-101; Novus Biologicals). Rabbit sera to human Cdh1 and Cdc27 were previously described (13, 24). Rabbit sera against cyclin A and human Pds1 were obtained from M. Pagano and H. Zou, respectively. Mouse monoclonal antibodies included CC03 to cyclin B1 (Calbiochem); PL-5 to human Plk1 (15); 9E10 to myc epitope (gift from G. Evan); CTR-453 to pericentriolar proteins (donated by M. Bornens); KH-95 to E2F-1 (Santa Cruz), DCS-141 to Mcm7 (C. S. Sørensen, C. Lukas, E. R. Kramer, C. Gieffers, J.-M. Peters, J. Bartek, and J. Lukas, unpublished data); HE-12 and HE-172 to cylin E (34), and A78720 to Aim1 (Transduction Laboratories). Monoclonal antibody to Skp2 was provided by W. Krek. Immunoprecipitation, immunoblotting, and in situ immunocytochemical techniques including detection of bromodeoxyuridine (BrdU) incorporated into newly synthesized DNA were described earlier (33, 35). Cyclin E-associated kinase activity using histone H1 as a substrate was assessed essentially as previously described (56).

Microinjection. Affinity-purified rabbit antibody to Cdh1 (Sat105 [13]) or purified nonimmune rabbit immunoglobulin G (IgG) (Sigma) was microinjected (2 mg/ml) into R-12 cells (34) grown on glass CELLocate coverslips (Eppendorf) and synchronized in G₀ by incubation for 48 h in serum-free medium. Alternatively, expression plasmids for myc-Cdh1, cyclin E, and pRbΔcdk (25-μg/ml needle concentration) were coinjected in combinations specified in the figure legends. The cells were subsequently stimulated by the addition of medium containing 10% FCS supplemented with BrdU (100 μg/ml). All microinjections were performed with a Zeiss-AIS system exactly as previously described (33, 34).

Reporter assays. U-2-OS Cdh1 cells were electroporated with 1 μg of the reporter plasmids containing Myc, cyclin E, and E2F promoter fragments regulating the expression of the luciferase gene, together with 0.5 μg of the internal control pCMV-LacZ. The cells were harvested after 48 h of culture with or without tetracycline as indicated in the figure legends. The resulting luciferase and β-galactosidase activities were measured with a Lumat LB 9501 luminometer (Berthold) and an UltroSpec 2000 spectrophotometer (Pharmacia Biotech), respectively.

RT-PCR. U-2-OS Cdh1 cells were induced to express myc-Cdh1 by removal of tetracycline from the culture medium for the time specified in figure legends. Total RNA was isolated using a Triazol reagent kit (Gibco-BRL) according to manufacturer's instructions. Conditions for reverse transcription (RT) and PCR amplification including the primer sequences for cyclin E and the porphobilinogen deaminase (PBGD) housekeeping gene have been described in detail (56).

Flow cytometry. Cells were trypsinized and fixed in 70% ice-cold methanol for 20 min, washed in phosphate-buffered saline, and incubated for 30 min at 37°C in propidium iodide (PI) buffer (10 mM Tris-HCl [pH 7.5], 5 mM MgCl₂, 50 μg of PI per ml, and 10 μg of RNase A per ml). The stained cells were acquired by the FACSCalibur flow cytometer (Becton Dickinson), and the DNA content was analyzed using CellQuest software.

In vitro ubiquitination assay. Cyclin B fragments (amino acids 13 through 110) from sea urchins were iodinated as previously described (24). [³⁵S]methionine- and [³⁵S]cysteine-labeled Xenopus Pds1, Plx, and geminin and human Aim1, cyclin A, and Cdc20 proteins were prepared by coupled transcription-translation reactions in rabbit reticulocyte lysate (Promega). To obtain highly pure Cdh1-activated APC, the inactive APC core was purified from interphase Xenopus egg extracts. Under such conditions, APC is not phosphorylated and thus not bound to Cdc20. Since Xenopus eggs do not contain any endogenous Cdh1, such APC was highly and specifically activated by baculovirus-expressed human Cdh1. Conversely, specific APC-Cdc20 ubiquitin ligase was generated by using APC purified from mitotic Xenopus oocytes, which contain high cyclin B-Cdc2 activity essential to modify the APC core to a status activatable by baculovirus-expressed human Cdc20. The in vitro ubiquitination reaction was performed essentially as previously described (25). Samples were analyzed by sodium dodecyl sulfate-5 to 15% polyacrylamide gel electrophoresis and phosphorimaging.

Generation of human cell lines allowing conditional manipulation of APC-Cdh1 assembly during the cell cycle. Human U-2-OS cells engineered to express myc-tagged Cdh1 in a tetracycline-dependent manner rapidly accumulated the myc-Cdh1 protein upon removal of tetracycline from culture medium (Fig. 1A). Densitometric measurements revealed that the degree of myc-Cdh1 overexpression varied among different clones, ranging from a five- to eightfold increase compared to the endogenous protein (fivefold in clone A6, shown as an example in Fig. 1A). All phenotypic and biochemical consequences of the myc-Cdh1 induction were identical and essentially reproduced in both higher- and lower-expressing clones, and unless stated otherwise, the data shown below were obtained from experiments performed with clone A6.

FIG. 1 Conditionally elevated myc-Cdh1 associates with endogenous APC irrespective of cell cycle position. (A) U-2-OS Cdh1-inducible cells (clone A6) were grown in the absence of tetracycline and at the indicated time points, the abundance of endogenous and (more ...)

Immunoprecipitation analyses verified that the myc-Cdh1 protein induced in exponentially growing U-2-OS cells was functional, as it readily associated with the cellular APC, immunopurified through its Cdc27 subunit (Fig. 1B). Moreover, induction of myc-Cdh1 resulted in formation of stable APC complexes also in cells synchronized in S phase by aphidicolin (Fig. 1C), which means under conditions when APC and Cdh1 normally do not associate due to CDK-dependent phosphorylation of the Cdh1 subunit (10, 24, 32). We concluded that the degree of elevation in Cdh1 protein conditionally achieved in our cell lines saturates the cellular capacity to modify Cdh1 during S phase and results in stable, nonoscillating APC-Cdh1 association throughout the cell cycle. These results suggest that Cdh1 protein levels relative to other APC subunits and to its upstream regulators such as cyclin A (32) must be tightly controlled and must not exceed a certain threshold to avoid untimely APC activation.

The next question was whether the ectopic Cdh1 also affects recruitment of Cdc20, another activating subunit, into APC complexes. Interestingly, Western blotting analysis revealed that following myc-Cdh1 induction, Cdc20 protein rapidly disappeared and remained virtually undetectable as long as the cells grew without tetracycline (Fig. 1D). Reduction of the Cdc20 protein was caused by acceleration of its turnover since inhibition of the proteasome restored the Cdc20 protein levels to the values observed in exponentially growing cells (Fig. 1D). The rapid disappearance of the Cdc20 protein following myc-Cdh1 induction indicated that mammalian Cdc20 is degraded in an APC-Cdh1-dependent manner in vivo. Collectively, these data demonstrated that the APC activity in our inducible cell lines was largely (if not completely) generated by APC-Cdh1 complexes.

Nonperiodic APC-Cdh1 association prevents G₂ and M events and causes endoreplication. Next we asked how the U-2-OS cells would respond to transient or prolonged loss of periodicity of APC-dependent proteolysis. Following removal of tetracycline from the culture medium, the cells rapidly ceased to multiply (Sørensen et al., unpublished data). Detailed flow cytometry analysis revealed that induction of myc-Cdh1 dramatically deregulated cell cycle progression at multiple transition points. First we noticed that such cells transiently but reproducibly accumulated in G₁ (Fig. 2A). Detailed measurements of G₁ length in cells released from metaphase arrest confirmed that induction of myc-Cdh1 approximately doubled the time required for progression through G₁ and delayed entry into S phase for 6 to 8 h (Fig. 2B). However, cells continuously expressing myc-Cdh1 proved unable to sustain a stable G₁ arrest and, between 36 and 48 h after induction, forced progressive accumulation of cells with a 4 N DNA content (Fig. 2A). Multiparameter flow cytometry combining measurements of DNA distribution with BrdU incorporation also revealed that the rate of nucleotide incorporation in myc-Cdh1-expressing cells was 2.1-fold slower compared to that in the S-phase cells from the same clone not expressing the myc-Cdh1 transgene. Consistently, when released from aphidicolin-mediated early S-phase arrest, myc-Cdh1-expressing cells traversed the S phase more slowly compared to the control cells (Sørensen et al., unpublished data). Nevertheless, this slow but uninterrupted DNA synthesis led to a progressive increase in cellular ploidy culminating between days 4 and 5, when the majority of the cells expressing myc-Cdh1 acquired DNA content higher than 4 N (Fig. 2A). The cells at this stage contained enlarged nuclei (Fig. 2C) consistent with the increased DNA content detected by flow cytometry. Immunostaining with antibodies specific for pericentriolar proteins revealed that such cells failed to separate centrosomes and create bipolar microtubule-generating centers (Fig. 2C), an event which otherwise marks the transition between G₂ phase and initial stages of mitosis (39). Together with complete lack of any morphological signs of mitosis (Sørensen et al., unpublished data), these data suggest that deregulated activation of APC-Cdh1 ubiquitin ligase in U-2-OS cells precluded productive cell division and promoted slow but continuous DNA replication, uncoupled from mitosis.

FIG. 2 Induction of myc-Cdh1 prevents cell division and promotes endoreplication. (A) U-2-OS Cdh1 cells were stimulated to express myc-Cdh1 by removal of tetracycline (−Tet). At the indicated time points after myc-Cdh1 induction, the cells were fixed, (more ...)

Ultimately, prolonged induction of ectopic Cdh1 appeared to be lethal, as evidenced by progressive cell detachment detectable after 5 days of myc-Cdh1 induction (Sørensen et al., unpublished data). However, shorter exposure of cells to active APC-Cdh1 was partially reversible, albeit with serious consequences for genomic integrity. Readdition of tetracycline and silencing of the myc-Cdh1 expression after 12 to 48 h of its induction restored cell division in a significant proportion of cells (Fig. 2D) (Sørensen et al., unpublished data). Superimposing of fluorescence-activated cell sorting histograms revealed that, although such cells resumed cell cycle progression, they appeared to either decrease or increase the ploidy compared to cells never exposed to an excess of ectopic Cdh1 (Fig. 2D). Collectively, these results indicate the critical importance of periodic oscillation of the APC-Cdh1 ubiquitin ligase activity for coordinating S phase with cell division and consequently for protecting the stability of the genome in mammalian cells.

APC-Cdh1 induces destruction of mitotic regulators. Western blotting analysis revealed that a number of known or putative APC substrates regulating entry into, progression through, and exit from mitosis, such as cyclin B1, cyclin A, securin Pds1, the polo-like kinase Plk1, the kinetochore-associated proteins mitosin-CENP-F, the Aurora-like midbody-associated protein Aim1 (Fig. 3A), and Cdc20 (Fig. 1D), were quantitatively degraded following myc-Cdh1 induction. The degradation was proteasome dependent (Fig. 3A, right panels; Fig. 1D) and relatively fast; for instance, cyclin B1 protein began to decline at around 3 h and reached a nadir at 6 h after myc-Cdh1 induction (Fig. 3B) (Sørensen et al., unpublished data). Thus, it appears that in mammalian cells, APC-Cdh1 can induce rapid degradation of many known or putative APC substrates directly involved in regulating cell division. In several independent experiments, we have noticed that myc-Cdh1-induced destruction of cyclin A appeared to be less quantitative and delayed compared to that of cyclin B1 (Fig. 3A) (Sørensen et al., unpublished data). This phenomenon was especially pronounced when myc-Cdh1 was expressed in cells arrested in S phase by aphidicolin (Fig. 3B), suggesting that S-phase cells may contain a subpopulation of cyclin A specifically protected from APC-mediated proteolysis.

FIG. 3 APC-Cdh1 ubiquitinates and targets for destruction multiple regulators of cell division. (A) U-2-OS Cdh1 cells were induced to express myc-Cdh1 by the removal of tetracycline (−Tet), and the abundance of the indicated proteins (all APC substrates) (more ...)

To directly address the issue of the apparently broad substrate recognition by the human APC-Cdh1 ubiquitin ligase, we reconstituted the APC-dependent ubiquitination reaction in vitro. Consistent with the data obtained with the inducible clones, cyclin B, Pds1, Aim1, and Plx (Xenopus homolog of human Plk1 kinase) were efficiently ubiquitinated by APC activated by Cdh1 (Fig. 3C). Also other APC substrates such as geminin, cyclin A, and Cdc20 were efficiently modified by Cdh1-activated APC in an in vitro ubiquitination assay (Sørensen et al., unpublished data). Direct comparison of APC ubiquitin ligase activities catalyzed by Cdh1- and Cdc20-activating subunits, respectively, revealed that, while APC-Cdh1 efficiently ubiquitinated all substrates tested, APC-Cdc20 was less efficient in modification of Plx and Aim1 (Fig. 3C). We concluded that once assembled, APC-Cdh1 is capable of ubiquitinating and priming for rapid destruction a broad range of key regulators of multiple steps of mammalian cell division.

Failure to inactivate APC-Cdh1 at the G₁/S transition promotes E2F-dependent up-regulation of cyclin E-Cdk2. Contrary to the destruction of mitotic regulators, induction of myc-Cdh1 promoted progressive up-regulation of cyclin E-Cdk2 (Fig. 4A), which in mammalian cells represents the major CDK activity responsible for initiation of DNA replication (18, 28, 45, 46). This was achieved by at least two distinct mechanisms. First, starting from 24 h after myc-Cdh1 induction, we observed a rapid decline of the p27 protein, a potent inhibitor of Cdk2 kinase activity (51, 65), accompanied by accumulation of Skp2, the F-box protein recruiting p27 to the SCF ubiquitin ligase (Fig. 4A) (4, 62). We did notice the transient drop of the Skp2 protein shortly after replating the cells into tetracycline-free medium; nevertheless, the progressive increase of the Skp2 levels after 24 h was reproducibly observed in several independent experiments and distinct U-2-OS-Cdh1 clones (Sørensen et al., unpublished data). This was not due to stabilization of the Skp2 protein, whose turnover was actually somewhat accelerated (Sørensen et al., unpublished data), potentially explaining its initial transient decline (Fig. 4A). The gradual accumulation of Skp2 was likely a consequence of stimulation of its transcription, which was detected in myc-Cdh1-expressing cells (Fig. 4B). Second, myc-Cdh1 induction lead to a marked increase of the cyclin E protein itself (Fig. 4A). Pulse-chase assays did not show any significant changes in cyclin E protein stability (Sørensen et al., unpublished data). However, RT-PCR analysis (Fig. 4B) and Northern blot analysis (Sørensen et al., unpublished data) revealed that the transcription of cyclin E mRNA increased approximately fivefold in cells expressing myc-Cdh1. Consistently, reporter assays performed with a cyclin E promoter fragment showed marked superactivation in cells induced to express myc-Cdh1 (Fig. 4C). This could not be explained solely by the myc-Cdh1-induced changes in the cell cycle position since similar results were obtained when the expression of ectopic Cdh1 was induced in cells arrested by aphidicolin in early S phase (Fig. 4D). Cyclin E expression is controlled by a concerted action of the two major G₁ transcriptional activities, namely, those associated with E2F and Myc (reference 56 and references therein). Several pieces of evidence suggested that the accelerated transcription of cyclin E reflected the increased activity of endogenous E2F. First, the Cdh1-induced activation of the cyclin E promoter could be fully repressed by coexpression with pRbΔcdk (Sørensen et al., unpublished data), a phosphorylation-deficient retinoblastoma protein which was previously demonstrated to be a potent repressor of all known E2F isoforms (34). Second, parallel luciferase measurements revealed that induction of ectopic Cdh1 led to a severalfold increase of the activity associated with endogenous E2F but not with Myc transcription factors (Fig. 4C). Finally, induction of myc-Cdh1 led to an increased accumulation of at least one member of the E2F transcription factor family, namely, E2F-1 (Fig. 4A). Also in this case, active APC-Cdh1 did not significantly affect E2F-1 protein stability (Sørensen et al., unpublished data) but rather promoted increased transcription of the E2F-1 gene (Fig. 4B). Taken together, these data strongly suggested that nonperiodic activation of APC-Cdh1 created conditions permissive for gradual accumulation and elevated activity of the E2F transcription factors and consequently for increased synthesis of at least some cell cycle regulators capable of initiating and maintaining DNA synthesis, which, in the concomitant absence of cell division, likely generated the major force promoting progressive endoreplication.

FIG. 4 Constitutive APC-Cdh1 assembly stimulates E2F-dependent superactivation of cyclin E-Cdk2. (A) U-2-OS Cdh1 cells were induced exactly as described in the legend for Fig. 3A. The kinase activity associated with cyclin E was measured using histone H1 as (more ...)

Functional status of APC-Cdh1 ubiquitin ligase modulates the ability of cyclin E-Cdk2 to initiate DNA replication. The data obtained with our inducible cell lines showed that, on one hand, persistent APC-Cdh1 interaction delayed both the passage through G₁/S transition and the progression through S phase. On the other hand, induction of myc-Cdh1 gradually increased the activity associated with the S-phase-promoting cyclin E-Cdk2 holoenzyme and caused progressive overreplication of the genome. Although counterintuitive at first glance, this complex consequence of nonperiodic APC-Cdh1 activity on DNA synthesis could be reconciled if Cdk2, by itself capable of supporting limited DNA replication, requires an APC-sensitive auxiliary factor to set a correct timing of initiation and subsequent rate of DNA synthesis. To test such predictions, we employed an inverse approach to our previous experiments, and rather than overexpressing ectopic Cdh1, we ablated the function of the endogenous protein by means of a specific anti-Cdh1 neutralizing antibody. We performed these experiments in immortalized rat fibroblasts (R-12 cell line) which could be synchronized by mitogen depletion and thus provide an accurate system to monitor events at the G₁/S transition (33, 37). First we tested the capability of Sat105, a highly specific anti-Cdh1 antibody (13; Sørensen et al., unpublished data) to interfere with Cdh1 function. Indeed, when added in the in vitro ubiquitination reaction, Sat105 progressively inhibited Cdh1-dependent polyubiquitination of cyclin B in a dose-dependent manner (Fig. 5A). This effect was specific for APC-Cdh1 because APC-Cdc20 complexes were not inhibited even by high doses of these antibodies (Fig. 5A). Moreover, Sat105 inhibited endogenous APC-Cdh1 activity also in vivo. When microinjected into G₁ fibroblasts, which naturally contain highly active APC-Cdh1, Sat105 significantly stabilized a cyclin B1-luciferase reporter protein (Fig. 5B), which serves as a sensitive marker for APC-dependent proteolysis (32).

FIG. 5 Inactivation of APC-Cdh1 cooperates with cyclin E-Cdk2 to initiate DNA replication. (A) APC-dependent ubiquitination reaction of radiolabeled cyclin B (marked by asterisk) specifically catalyzed by Cdh1 and Cdc20, respectively, was reconstituted in vitro (more ...)

Having validated the ability of Sat105 to functionally neutralize Cdh1, we next tested whether inactivation of APC-Cdh1 could affect the length of G₁. Indeed, microinjection of Sat105 into synchronized rat fibroblasts consistently resulted in a moderate (15 to 20%) increase in cells incorporating BrdU when compared to control-injected cells (Sørensen et al., unpublished data). To specifically address whether the functional status of APC-Cdh1 influences the S-phase-promoting potential of cyclin E-Cdk2, we generated a model system based on two recent findings obtained in our laboratories. First, rat fibroblasts deprived of de novo expression of E2F target genes by means of a constitutively active allele of the retinoblastoma protein (pRbΔcdk) undergo G₁ arrest, which could be rescued by coexpression of cyclin E from a heterologous promoter (6, 23, 34). Second, pRbΔcdk greatly stimulates assembly of APC-Cdh1 holoenzyme even in the presence of active cyclin E-Cdk2 (32). Hence, expression of pRbΔcdk together with cyclin E generates a defined system in which S phase is generated largely by the cyclin E-Cdk2 kinase activity in the concomitant presence of functional APC. Importantly, under such conditions, coinjection of Sat105 significantly accelerated S-phase entry compared to control cells injected with nonspecific antibody (Fig. 5C). More detailed analysis at multiple time points revealed that under these experimental conditions, the Sat105-microinjected cells accumulated in S phase approximately 6 h earlier compared to control cells (Sørensen et al., unpublished data). Conversely, both the mitogen-induced S phase and accelerating effect of ectopic cyclin E on G₁/S transition were significantly reduced upon comicroinjection with expression plasmid for myc-Cdh1 (Fig. 5D). Taken together, these data indicate that cancellation of the APC ubiquitin ligase activity at the G₁/S transition and throughout S phase allows accumulation of at least one unstable protein, which augments cyclin E-Cdk2-mediated initiation of DNA replication.

Periodic assembly of APC with Cdh1 represents an essential step in coordinating multiple cell cycle transitions in human cells. Our data demonstrate that periodic regulation of APC-Cdh1-mediated proteolysis appears crucial for productive cell division of mammalian cells. Failure to inactivate APC-Cdh1 ubiquitin ligase in a timely manner led to unscheduled destruction of key G₂/M regulators such as mitotic cyclins, Plk1, and Pds1 and consequently precluded cell cycle progression beyond the G₂ phase. Our finding that Cdc20 represents a bona fide in vivo substrate for APC-Cdh1 ubiquitin ligase is fully consistent with the recent report, published while this manuscript was being prepared, that ubiquitination of Cdc20 in an in vitro-reconstituted reaction is specifically catalyzed by APC-Cdh1 (49). The observation that human APC-Cdh1 catalyzes ubiquitination and promotes destruction of the above G₂/M regulators with similar rapid kinetics was unexpected, given the evidence from lower eukaryotes arguing for more restricted substrate specificity conferred by Cdc20 or Cdh1, respectively (57, 66). For instance, yeast Cdh1, even when overexpressed, induced complete destruction of Clb2 yet only partial elimination of Pds1. Our results suggest that the situation is different and that the APC-Cdh1 ubiquitin ligase may regulate abundance of a broad range of proteins containing appropriate recognition signals such as the destruction box or the newly identified KEN box (49). This notion is further supported by recent data from budding yeast cells in which Pds1 proteolysis in G₁ has been shown to depend on Cdh1, not Cdc20 (54). In contrast, APC-Cdc20 appears to have a more restricted substrate specificity, at least in vitro where it can ubiquitinate Aim1 and Plk1 less well than APC-Cdh1. Cdc20 and Cdh1 may therefore differ mostly in their mode of regulation and the resulting temporal difference in their ability to activate the APC, whereas their substrate specificities may be more similar than previously proposed. In addition, we identified mitosin-CENP-F and Aim1 as two novel KEN box-containing proteins specifically degraded in an APC-Cdh1-dependent manner. It is intriguing that mitosin-CENP-F localizes to kinetochores (53, 72), suggesting that kinetochore-associated proteins may not only help to sequester and inhibit APC upon activation of the spindle checkpoint (69) but could actually be themselves targeted for APC-mediated proteolysis. This observation further underscores the need for a tight control of APC-Cdh1 activity, to ensure timely and spatially correct segregation of sister chromatids in anaphase. The Aurora and Ipl1-like midbody-associated protein Aim1 has been recently found to play an essential role in cytokinesis in mammalian cells (64). Our finding that Aim1 protein is also degraded in an APC-Cdh1-dependent manner helps explain the periodic accumulation of this protein during the cell cycle (64) and suggests that Cdh1, through regulation of Aim1 protein turnover, supports timely exit from mitosis.

Besides the negative effect on cell division, conceptually consistent with findings made in yeast, Drosophila, and Xenopus laevis (69), our results revealed a novel aspect of APC-mediated proteolysis, namely, that dissociation of Cdh1 from the APC core at the G₁/S transition appears to play a central role in protecting cells from genomic instability by helping to restrict chromosomal duplication to only one round per cell division cycle. We could demonstrate that untimely activation of APC-Cdh1 during S phase interfered with the natural oscillation of some factors that were rate limiting for G₁/S transition, such as cyclin E or E2F-1, and as such likely contributed to uninterrupted DNA synthesis and severe rereplication of the genome. Moreover, de novo refiring of at least some origins without intervening mitosis could be further augmented by APC-Cdh1-mediated destruction of the newly emerging class of APC substrates such as geminin, which at least in Xenopus embryos functions as a potent inhibitor of initiation of DNA replication (38). This, together with high cyclin E-Cdk2 activity could overpower a potential lack of other APC targets, which may normally accumulate at the G₁/S transition and cooperate with cyclin E-Cdk2 to efficiently initiate DNA replication in a timely manner. Although the identity of such protein(s) remains elusive, our in vivo APC-Cdh1 neutralization data strongly argue for its existence (Fig. 5C and D). One plausible candidate is Dbf4, which together with its catalytic subunit Cdc7 was demonstrated as an essential factor for S-phase entry in various organisms (5, 11, 47, 67), and yeast Dbf4 was recently shown to be a bona fide substrate for APC-mediated proteolysis (11, 47). Finally, several other observations opened up the possibility that the persistent replication-promoting activity in cells expressing ectopic Cdh1 could be facilitated also by a potential cross talk between APC and SCF ubiquitin ligases. High and nonperiodic APC-Cdh1 activity apparently supported progressive accumulation of Skp2 F-box protein, a crucial SCF component, and thus created conditions permissive for a rapid turnover of p27 CDK inhibitor that would otherwise restrain S-phase-promoting CDK activities. How APC-Cdh1 induces accumulation of Skp2 remains to be established. The reported requirement of Skp2 for S phase (71), the sharp increase of Skp2 accumulation at the G₁/S transition (31), and increased transcription of the Skp2 gene in myc-Cdh1-expressing cells (this study) indicate that Skp2 expression could be regulated, at least in part, by E2F.

One plausible explanation for the gradually increasing E2F-mediated transcription could be the degradation of cyclin A observed in myc-Cdh1-expressing cells. In such a scenario, APC-Cdh1 would not directly induce E2F-dependent transcription but rather would prevent its physiological silencing, which normally occurs at the end of S phase following the completion of genome duplication. It has been previously demonstrated that cyclin A-Cdk2 phosphorylates the E2F-heterodimerizing partner DP-1 and thereby cancels the ability of E2F–DP-1 complexes to bind DNA (26). It has also been shown that failure to displace E2F from DNA results in the inability to exit S phase (27), the phenotype very similar to what we describe here as a consequence of the deregulated APC-Cdh1 activity. Hence, one plausible way to gradually amplify cellular E2F activity in our system would be based on autostimulation of E2F-1 synthesis, promoted by reduced levels of cyclin A upon myc-Cdh1 induction.

The involvement of cyclin A as a mediator of some crucial phenotypic consequences of nonperiodic APC-Cdh1 assembly raised an important issue of complexity in terms of the functional interplay between APC-Cdh1 and E2F-cyclin A. We have previously demonstrated that, during S phase, E2F-regulated genes including cyclin A prevent Cdh1 from interacting with the APC core and thus ensure that the APC ubiquitin ligase would not be switched on prematurely (32). The data presented in this study show that approximately fivefold elevation of Cdh1 protein was sufficient to revert this dependency and induce destruction of a major pool of cellular cyclin A. However, our observation that S-phase cells retained detectable amounts of cyclin A at the time when induction of myc-Cdh1 nearly completely eliminated other APC substrates suggests the existence of a fraction of cyclin A specifically protected against APC-dependent proteolysis. Indeed, factors capable of interfering with vertebrate cyclin A but not cyclin B destruction have been recently reported in the literature (12). Availability of such an APC-insensitive pool of cyclin A would be consistent with the reported role of cyclin A-Cdk2 in phosphorylating Cdh1 and thus securing that cells beyond the G₁/S transition would not prematurely reactivate destruction of other APC targets (32). In fact, our observation that despite steadily increasing cyclin E-Cdk2 activity in myc-Cdh1-expressing cells, Cdh1 remained bound to functional APC further supports cyclin A but not cyclin E as a direct Cdh1 kinase in mammalian cells. The interspecies differences between mammals and the genus Drosophila, in which cyclin E-Cdk2 has been proposed as a physiological fizzy-related kinase, could be explained by at least several distinctions in cyclin-CDK biochemistry including the dominant catalytic partners of cyclin A and its subcellular localization (61). Finally, very recent advances in the proteolysis field indicate that degradation of mammalian cyclin A might be unexpectedly complex; during this study, Nakayama and colleagues reported that cyclin A but not cyclin B can be efficiently ubiquitinated by SCF^Skp2 ubiquitin ligase (42). Based on our finding that elevation of myc-Cdh1 stimulates transcription of the Skp2 F-box protein and given that Skp2 and cyclin A proteins physically interact (71), it is plausible that in our system, as well as during unperturbed cell cycle progression (42), the steady-state levels of cyclin A result from a delicate balance between its E2F-regulated transcription and its protein turnover controlled by both APC and SCF ubiquitin ligases.

Taken together, the cellular mechanisms involved in regulating the periodicity of APC-Cdh1 assembly in mammalian cells appear to play a crucial role in coupling DNA replication with productive cell division. In fact, Cdh1 appears to join the small number of cell cycle genes such as the myc, p53, cyclin E, and p21 genes, the deregulation of which appears capable of inducing genomic instability. It is tempting to speculate that at least some mechanisms involved in control of APC-Cdh1 assembly could be deregulated during multistep tumorigenesis, for instance in those human tumors that aberrantly accumulate mitotic B-type cyclins during G₁ (20). Our finding that mitosin-CENP-F is targeted by APC-Cdh1 for destruction opens up a possibility to search for potential defects affecting stability of kinetochore-associated proteins and raise the question whether these could explain at least some aspects of aneuploidy detected in the majority of human tumors. Finally, our observation that APC-Cdh1 directly determines the level of the human Pds1 (PTTG) and Aim1 proto-oncogenes raises the possibility that aberrations in posttranscriptional regulation including proteolysis might account for their elevated expression in human malignancies (9, 55, 63).

We are grateful to C. Giefers, H. Zou, M. Pagano, M. Bornens, W. Krek, and G. Evan for donating diverse antibodies and to H. Zou, T. McGarry, K. Helin, P. Jansen-Dürr, J. Cleveland, and Y. Terada for cDNAs and reporter constructs.

This work was supported by grants from the Danish Cancer Society, the Human Frontier Science Program (RG-299/97), the Danish Medical Research Council (9600821), and the Austrian Industrial Research Promoting Fund (FFF 3/12801).