|Home | About | Journals | Submit | Contact Us | Français|
Differentiation of trophoblast giant cells in the rodent placenta is accompanied by exit from the mitotic cell cycle and onset of endoreduplication. Commitment to giant cell differentiation is under developmental control, involving down-regulation of Id1 and Id2, concomitant with up-regulation of the basic helix-loop-helix factor Hxt and acquisition of increased adhesiveness. Endoreduplication disrupts the alternation of DNA synthesis and mitosis that maintains euploid DNA content during proliferation. To determine how the mammalian endocycle is regulated, we examined the expression of the cyclins and cyclin-dependent kinases during the transition from replication to endoreduplication in the Rcho-1 rat choriocarcinoma cell line. We cultured these cells under conditions that gave relatively synchronous endoreduplication. This allowed us to study the events that occur during the transition from the mitotic cycle to the first endocycle. With giant cell differentiation, the cells switched cyclin D isoform expression from D3 to D1 and altered several checkpoint functions, acquiring a relative insensitivity to DNA-damaging agents and a coincident serum independence. The initiation of S phase during endocycles appeared to involve cycles of synthesis of cyclins E and A, and termination of S was associated with abrupt loss of cyclin A and E. Both cyclins were absent from gap phase cells, suggesting that their degradation may be necessary to allow reinitiation of the endocycle. The arrest of the mitotic cycle at the onset of endoreduplication was associated with a failure to assemble cyclin B/p34cdk1 complexes during the first endocycle. In subsequent endocycles, cyclin B expression was suppressed. Together these data suggest several points at which cell cycle regulation could be targeted to shift cells from a mitotic to an endoreduplicative cycle.
Progression of the cell cycle requires precise replication of genomic DNA once during each cycle, and its subsequent accurate segregation during mitosis to maintain a euploid DNA content. Aberrations that disrupt this process can lead to aneuploidy and may predispose the cells to oncogenic transformation (Pathak et al., 1994 ). A variety of models have been proposed to explain the necessary interdependence of mitosis and DNA synthesis, but the molecular mechanisms that regulate the fundamental process of replication and division are only just beginning to emerge. Replication origins in yeast can be primed only during early G1, when cyclin expression and cyclin-dependent kinase (cdk) activity is low (Cocker et al., 1996 ). Once the cell is committed to cell cycle by passage through Start, the resultant expression of the cyclins that regulate progression through the remainder of the cell cycle prevents assembly of new replication origins until their degradation during mitosis, limiting the cell to one round of DNA replication per cycle (Basco et al., 1995 ; Piatti et al., 1996 ). The destruction of cyclins precipitated by passage through mitosis seems to be the event that separates one cycle from the next and enforces the strict alternation of S and M phases (Piatti et al., 1996 ).
Normal cell cycle checkpoints are uncoupled in some cell types during development. For example, endoreduplication is a naturally occurring disruption of the mitotic cycle in which rounds of DNA synthesis occur in the absence of mitosis, resulting in polyploid cells. Endoreduplication occurs in a wide variety of embryonic and adult cell types in both animals and plants, including rodent trophoblast (Zybina, 1970 ; Malinowski and Maszewski, 1994 ; Hartman and Southern, 1995 ; Datta et al., 1996 ; Lilly and Spradling, 1996 ; Zhang et al., 1996 ). Both mural trophectoderm and derivatives of polar trophectoderm in rodents are transformed into so-called trophoblast giant cells that have DNA contents up to 1000 times the haploid content as a result of endoreduplication (Zybina, 1970 ; Zybina and Grishchenko, 1970 ; Barlow and Sherman, 1972 ; Hoffman and Wooding, 1993 ). The mechanism of commitment to giant cell differentiation is poorly characterized, although down-regulation of Id1 and Id2, and up-regulation of the bHLH factor Hxt appear to play important roles (Cross et al., 1995 ). Overexpression of Id keeps the trophoblast cells as stem cells, whereas overexpression of Hxt drives differentiation to giant cells. Concomitant with this commitment the cells increase their adhesion, a necessary aspect of the implantation process that ensues in vivo (Cross et al., 1994 , 1995 ; Rinkenberger et al., 1997 ). How these changes are translated into endoreduplication and differentiation remains unclear, because the molecular pathways that produce the phenotypes are not understood. Certainly, differentiation and endoreduplication are very closely linked in trophoblast. Although it has proven to be possible to separate the two events experimentally (Gardner and Davies, 1993 ), the onset of differentiation is linked to the cells becoming postmitotic, a necessary condition for endoreduplication.
Endoreduplication has been best studied in Drosophila melanogaster. It proceeds in cycles with defined S phases separated by gap (G) phases, suggesting that progression through the endocycle is as carefully regulated as progression through the mitotic cycle. Cyclin E appears to play a central role in regulating the initiation of S phase during the endocycle, as it does in the mitotic cycle (Knoblich et al., 1994 ; Sauer et al., 1995 ; Lilly and Spradling, 1996 ), and, in the absence of cyclin A and B, mitosis is not initiated. It is not clear how the transition from the mitotic cell cycle to the endocycle is regulated even in Drosophila. For example, it is not clear whether the absence of cyclins A and B is sufficient to uncouple the reinitiation of DNA synthesis from prior passage through M phase, or whether an additional change in regulatory mechanisms is required during the transition from the mitotic cycle to the endocycle.
The control of endoreduplication has been poorly studied in other systems. One of the major hurdles to understanding trophoblast differentiation has been the paucity of material with which to study the early steps of differentiation. However, a rat choriocarcinoma-derived cell line, Rcho-1, is a good candidate in which to address trophoblast differentiation in vitro (Faria and Soares, 1991 ; Cross et al., 1995 ). Under the appropriate conditions, Rcho-1 cells undergo a transition from proliferating to postmitotic cells that eventually express trophoblast giant cell-specific markers and also continue DNA synthesis (Shida et al., 1993 ; Yamamoto et al., 1994 ; Cross et al., 1995 ; Hamlin and Soares, 1995 ; our current research). In this study we have used Rcho-1 cells to examine the regulation of trophoblast giant cell differentiation and the transition in cell cycle structure. We describe the first characterization of the regulation of a mammalian endocycle and propose a mechanism for the abrogation of mitosis and the reinitiation of rounds of DNA synthesis during endoreduplication.
The Rcho-1 cell line was derived from a rat choriocarcinoma (Faria and Soares, 1991 ) and was obtained from M. Soares. The cells were maintained in NCTC-135 supplemented with 0.1 mg/ml sodium pyruvate, 2 mM glutamine, penicillin, streptomycin (NCTC-135+), and 20% fetal bovine serum (FBS; Hyclone, Logan, UT). To set up cultures for differentiation, the cells were grown to high density and then replated into the final dishes at high density (roughly a 1:1 transfer) and maintained for 2 d in the NCTC-135+ with 20% FBS. The cells were then treated with trypsin and washed twice with phosphate-buffered saline (PBS) to remove nonadherent cells (Cross et al., 1995 ). The cells that remained were cultured in NCTC-135+ supplemented with 10% horse serum.
Rcho-1 stem cells, plated for 24 h, or differentiating giant cell cultures, maintained in NCTC+/20% FBS, were fixed in methanol and stained with bisbenzimide (Sigma Chemical, St. Louis, MO). Nuclear DNA content was estimated by fluorescence intensity on 50–100 cells for each population using an imaging system from Compix Inc. Imaging Systems (Cranberry Township, PA). Fluorescent intensity was compared with serum-starved, quiescent HeLa cells shown to be diploid by fluorescence-activated cell sorter analysis.
DNA synthesis was estimated by measuring the incorporation of [3H]thymidine into trichloroacetic acid (TCA)-insoluble material. [3H]thymidine was added directly to the medium to a final concentration of 1 μCi/ml and the cells were incubated for another 4 h. The unincorporated [3H]thymidine was extracted with three washes of 10% TCA and two washes of 95% ethanol, all at 4°C. The TCA-insoluble material was collected with 0.1 N NaOH and the [3H]thymidine incorporation was measured by liquid scintillation spectrometry. The mimosine arrest was achieved by adding mimosine to the medium to 100 μM 3 d after removing the proliferating cells. The block was reversed by washing the cells once in preequilibrated medium and then continuing the culturing in NCTC+ with 10% horse serum.
Total RNA was isolated from proliferating or differentiating Rcho-1 cells using the acid-phenol reagent RNAzol B (Biotecx, Houston, TX). The RNA concentration was calculated from the absorption at 260 nm, and 10 μg from each sample were fractionated on a 1.2% agarose-formaldehyde gel. The RNA was transferred to a nylon membrane, Duralon-UV (Stratagene, La Jolla, CA), and fixed in place by UV cross-linking. The membranes were then hybridized to probes derived from random-primed synthesis using the Rediprime kit and [α-32P]dCTP (Amersham, Arlington Heights, IL). The hybridization was performed using the Quikhyb solution (Stratagene) following the manufacturer’s protocol and washing in 2× SSC with 0.1% SDS at 60°C. The blot was then exposed to film (Kodak X-OMAT). The cDNA probes used were as follows: cyclin A1 and A2 were derived by polymerase chain reaction amplification using published sequence information (Sweeney et al., 1996 ) and a d 7.5 embryonic mouse cDNA library (Stratagene) and cloned in pCRII (Invitrogen, San Diego, CA); mouse cyclin B1 and B2 were a kind gift from D. Wolgemuth (Chapman and Wolgemuth, 1992 , 1993 ); mouse cyclins D1, D2, and D3 were a kind gift from C. Sherr (Matsushime et al., 1991 ); mouse cyclin E was kindly supplied by J. DeLoia (Damjanov et al., 1994 ); Hxt (Cross et al., 1995 ); PL-I (Faria et al., 1991 ; Faria and Soares, 1991 ); MMP-9 (Reponen et al., 1994 ); α1 integrin was a gift from A. Sutherland (Sutherland et al., 1993 ).
Protein-containing lysates were prepared from proliferating or differentiating Rcho-1 cells by washing the cells with PBS and then scraping them into a modified RIPA buffer that contained 1% Nonidet P-40, 0.5% sodium deoxycholate, 0.1% SDS, 1 mM phenylmethylsulfonyl fluoride, 1 μg/ml aprotinin, 1 mM β-glycerol phosphate, and 0.1 mM Na3VO4. The lysate was clarified by centrifugation at 20,000 × g, a sample was removed for determination of protein concentration, and the remainder was frozen at −80°C until use. Using the MicroBCA protein assay (Pierce Chemical, Rockford, IL), equal amounts of protein from each lysate were separated by SDS-PAGE using a 10% separating gel and were transferred to Immobilon-P (Millipore, Bedford, MA). The membrane was treated with PBS containing 0.05% Tween 20 and 10% nonfat milk powder and then it was incubated with the primary antibody indicated for 1 h at room temperature. Antibodies for cyclin A were obtained from Santa Cruz Biotechnology (Santa Cruz, CA), for cyclin E from J. Roberts, and for cyclin B, p34cdk1, and p33cdk2 from PharMingen (San Diego, CA). The membrane was washed with PBS containing 0.05% Tween 20, then incubated with an appropriate horseradish peroxidase-conjugated secondary antibody (Amersham), washed again, and the immunocomplexes were revealed using the enhanced chemiluminescence reagent (Amersham).
The immunoprecipitations were performed using the same primary antibodies and lysates that were used for the immunoblots. Approximately 1 μg of the primary antibody was added to the lysate and incubated for 1 h on ice, a rabbit anti-mouse IgG antibody was added (ICN, Costa Mesa, CA), and the immunoprecipitates were then collected on protein A-agarose. The pellets were washed three times in RIPA buffer, twice with a washing buffer (50 mM HEPES, pH 7.4, 10 mM MgCl2, 0.5% Nonidet P-40), and then, for histone H1 kinase assays, resuspended in 20 μl of the washing buffer lacking detergent, containing 10 mCi of [γ-32P]-ATP and 1 μg of histone H1 (Life Technologies, Gaithersburg, MD), and incubated at 30°C for 5 min. An equal volume of 2× concentrated SDS sample loading buffer was added and the samples were then fractionated on a 12.5% polyacrylamide gel. After electrophoresis, the gel was stained with Coomassie brilliant blue to reveal the histone H1, dried, and exposed to film (Kodak X-OMAT). Anticyclin immunoprecipitates were also subjected to immunoblotting and probed with antibodies against p34cdk1 or p33cdk2.
Differentiating Rcho-1 cells were fixed in Carnoy’s solution (methanol:chloroform:glacial acetic acid = 6:3:1) for 15 min or incubated with 50 μM bromodeoxyuridine (BrdU) for 2 h and then BrdU-free medium for 5 min before fixation. The samples were incubated with PBS containing 0.1% Tween 20 (PBS.1T) and 10% goat serum and then exposed to one of the anti-cyclin A or anti-cyclin E antibodies (both from Santa Cruz Biotechnology) at 1/100 dilutions in PBS.1T, or the anti-BrdU (Becton Dickinson, San Jose, CA) at 1/1000, or a mixture of the anti-BrdU and one of the anti-cyclin antibodies for 1 h at room temperature. The samples were washed with PBS.1T and then incubated for 30 min with the appropriate secondary antibodies conjugated to either Texas red or fluorescein isothiocyanate to reveal the retained primary antibodies. The samples were washed again with PBS.1T and then counterstained with bisbenzimide.
Rcho-1 cells differentiate in vitro into cells that express several markers characteristic of terminally differentiated trophoblast giant cells including P450 (Yamamoto et al., 1994 ) and members of the placental lactogen/prolactin (PL) family (Shida et al., 1993 ; Hamlin et al., 1994 ). We found previously that an early event in the commitment to giant cell differentiation was a change in cell adhesiveness, so that brief trypsinization allowed removal of the proliferating stem cells and gave a highly purified population of cells that were committed to giant cell differentiation (Cross et al., 1995 ). We were interested in whether these cells underwent synchronized cell differentiation. To study this, proliferating Rcho-1 stem cells were grown for 2 d to reach confluence and the population was then trypsinized to remove the stem cells, leaving only cells that had committed to the giant cell fate. The trypsin-resistant cells assumed a morphology typical of giant cells with enlarged nuclei that expanded with time (Figure (Figure1A-C).1A-C). Designating the day of trypsinization as d 0, the expression of PL-I mRNA was detectable as early as d 1, peaked at d 4, and declined thereafter (Figure (Figure1D).1D). The transient expression of PL-I in the differentiating Rcho-1 cells recapitulates its developmental expression in vivo, which is also transient (Faria et al., 1991 ; Faria and Soares, 1991 ; Carney et al., 1993 ; Hamlin et al., 1994 ). Hxt (Cross et al., 1995 ), gelatinase B/MMP-9 (Alexander et al., 1996 ), and α1 integrin (Sutherland, et al., 1993 ), genes that are expressed in trophoblast giant cells in vivo, were also developmentally regulated (Figure (Figure1D).1D).
After selection, the trypsin-resistant Rcho-1 cells showed no discernible increase in cell number during the first 4–6 d of differentiation, and a gradual decline at later times (Figure (Figure2A).2A). The cells continued to synthesize DNA at levels comparable to those seen in proliferating Rcho-1 stem cells, as revealed by [3H]thymidine incorporation (Figure (Figure2B,2B, and up to d 12 in other experiments). The continued synthesis of DNA in the absence of mitosis suggested that the Rcho-1 cells were undergoing endoreduplication. The Rcho-1 giant cells were analyzed for DNA content as a more direct assay for the increase in ploidy that should accompany endoreduplication. By flow cytometry, the Rcho-1 stem cells had 4–8 N DNA contents, indicating that they are tetraploid (Nakayama, Scott, and Cross, unpublished data). This was confirmed by karyotype analysis. Nuclei of Rcho-1 giant cells appeared to be too fragile to prepare for flow cytometry, since only fragments with DNA contents of up to 8–10 N could be isolated. To avoid subjecting the nuclei to manipulation and the attendant fragmentation, the DNA content of Rcho-1 giant cells was determined by scanning laser fluorimetry of cells that were fixed in situ. Confluent, serum-starved HeLa cells were used as a standard for diploid DNA content. Using this technique, the Rcho-1 stem cells showed DNA contents from 4 N to 8 N (mean 4.29 N ± 0.25), as would be expected for the G1, S, and G2 contents of a tetraploid cell line (Figures (Figures2C2C and and3).3). Analysis of the committed giant cells at d 0 showed an 8 N or G2 peak (Figures (Figures2C2C and and3).3). Cells with DNA contents equivalent to 16 N or greater appeared by the second day of differentiation, DNA contents of up to 32 N between the third and fourth day, and cells with up to 64 N appeared by the sixth day (Figures (Figures2C2C and and3).3). Interpolation with these data showed that the Rcho-1 cells were capable of at least three rounds of endoreduplication during the 6 d of analysis, suggesting an estimate of 2 d for the length of the endocycle.
The endocycles in the Rcho-1 cells appeared to consist of alternating S and G phases, based on the clustering of DNA content around twofold increases. In addition, after a brief incubation with [3H]thymidine and autoradiography, both labeled and unlabeled nuclei were apparent. In d 0 giant cells, the labeling index was 60–70% after a 1-h [3H]thymidine pulse, indicating that the cells proceeded through the first round of endoreduplication relatively synchronously. The synchrony of the population decayed over time until the labeling index reached 18–20% on d 4 of differentiation and thereafter. To further examine the structure of the endocycle, a partial synchronization was achieved by incubating d 3 giant cells in mimosine for 16 h to prevent progression through S phase (Lalande, 1990 ; Watson et al., 1991 ). The mimosine was removed and DNA synthesis was determined by measuring [3H]thymidine incorporation at later times. Incorporation of [3H]thymidine was detectable 1 and 2 h after removing mimosine and remained elevated relative to controls for 8 to 10 h, providing an estimate of S phase of 8–10 h. Combining the length of S phase with the short-term labeling index of 20% gives an estimated endocycle length of 40–50 h, similar to the estimate derived from the DNA content analysis.
The transition from a mitotic cycle to an endocycle requires changes in the function of cell cycle checkpoints, for example, dissociation of the initiation of DNA synthesis from prior passage through mitosis. To assess the status of other cell cycle checkpoints during endoreduplication in trophoblast, the effect of treatments that impede the progression of cells through the mitotic cell cycle was compared in proliferating and endoreduplicating giant cells (Table (Table1).1). DNA synthesis in Rcho-1 giant cells was relatively resistant to ionizing irradiation when compared with the proliferating cells (Table (Table1).1). Likewise, incorporation of [3H]thymidine by Rcho-1 giant cells was considerably less sensitive to treatment with the DNA alkylating agent, mitomycin C, than that of the Rcho-1 stem cells (ED50 = 160 versus 31 ng/ml, respectively). These results suggest that the checkpoint that normally inhibits DNA synthesis in response to DNA damage is suppressed in the Rcho-1 giant cells compared with the stem cells.
In replicating cells, the G1 transition is sensitive to the presence of growth factors, which induce the transcription of D-type cyclins. We tested whether this was true during the Rcho-1 endocycle as well. Incubation of stem cells in serum-free medium for 16 h reduced the incorporation of [3H]thymidine to background levels, and addition of medium containing 20% FBS stimulated incorporation approximately 70-fold by 20 h after serum addition (Table (Table1).1). In contrast, [3H]thymidine incorporation by Rcho-1 giant cells was largely independent of the presence of serum, decreasing only twofold in the absence of serum (Table (Table1).1). We next examined the expression of cyclin D mRNAs (Figure (Figure4A).4A). Cyclin D2 mRNA was expressed at relatively low levels in Rcho-1 cells at all time points. The expression of cyclin D3 was high in proliferating Rcho-1 stem cells, but significantly decreased with the onset of differentiation (Figure (Figure4A).4A). In contrast, cyclin D1 mRNA was expressed only at low levels in the proliferating Rcho-1 cells, but was induced with the onset of differentiation. Interestingly, the cyclin D1 mRNA levels did not fluctuate during the subsequent endocycles (Figure (Figure4A).4A). In combination, these data indicate that several G1-S checkpoint functions are altered with commitment to giant cell differentiation and onset of endoreduplication.
The transition from the mitotic cycle to an endocycle was accompanied by other changes in the expression of the G1-S phase cyclins and their associated kinase activities. The mRNA for cyclin A1 was not detectable at any point in the analysis. Cyclins A2 and E mRNAs were readily detectable in Rcho-1 stem cells (Figure (Figure4A).4A). They were also detectable in d 0 giant cells and both were down-regulated by d 1, suggesting a rapid inhibition of the expression of these genes. The expression of cyclin A2 and cyclin E mRNAs remained low, though detectable, relative to the expression levels in Rcho-1 stem cells. Despite the significant reductions in abundance of the mRNAs, both proteins were readily detectable by Western blotting in Rcho-1 giant cells on d 0 and d 1 (Figure (Figure4B).4B). The cyclin E protein was present at all points during differentiation and, although some fluctuation in the amount of cyclin E was seen during the time course, it followed no obvious pattern in different experiments. The cyclin A2 protein showed two peaks in abundance on d 1 and d 3 of differentiation, with subsequent declines to barely detectable levels (Figure (Figure4B).4B). A similar pattern of accumulation and degradation over a 2-d time course was seen in other experiments. The level of expression of the cyclin A2 and E proteins were not closely linked to the expression of their respective mRNAs, suggesting that the turnover of these proteins differs between the mitotic and endoreduplicative cycles. Cyclin immunoprecipitates were next assayed for cyclin A2- or cyclin E-associated H1 kinase activity (Figure (Figure4C).4C). The fluctuations in associated kinase activity during giant cell differentiation followed the changes in abundance of the respective cyclin as determined by immunoblot analysis. The cyclin E-associated kinase activity remained relatively constant throughout the differentiation time course. The cyclin A-associated kinase activity showed two peaks that coincided with the peaks of cyclin A protein expression.
Despite the continued abundance of the cyclin A and cyclin E proteins, it was notable that the kinase activity associated with both cyclins dropped significantly after the onset of differentiation compared with stem cells. This reduction was not the result of down-regulation of either p34cdk1 or p33cdk2 (Figure (Figure7A).7A). Coimmunoprecipitation of p34cdk1 and p33cdk2 by the anti-cyclin A or E antibodies revealed that these two cdks were associated with the cyclins in both the endoreduplicative (Figure (Figure4D)4D) and mitotic cycles in several independent experiments. The continued association suggested that both enzymes contributed to the cyclin-associated kinase activity detected in the differentiating Rcho-1 cells. However, the amount of p34cdk1 associated with cyclin A showed a sharp decrease at the onset of differentiation, but there was no notable, similar decrease in the cyclin E-associated p34cdk1. The amount of p33cdk2 associated with the two cyclins did not change significantly with the onset of endoreduplication (Figure (Figure44D).
By indirect immunofluorescence, both cyclin A and cyclin E were expressed in only a fraction of giant cells, indicating that the expression of these proteins did change during the endocycle. The expression of cyclin A and E relative to the timing of DNA synthesis was analyzed in d 0 giant cells that were pulse labeled with BrdU. Cyclin A was located predominantly in the nucleus (Figure (Figure5),5), although a few cells also displayed cyclin A in the cytoplasm. Colocalization of cyclin A and BrdU showed that cyclin A was only detectable in S phase cells; however, 14% of the BrdU-positive cells were cyclin A negative, indicating that cyclin A was not expressed at all times of the S phase. Analysis of cyclin E expression in a parallel sample of cells revealed that only 65% of the cyclin E-positive cells were colabeled for BrdU, but essentially all (94%) of BrdU-positive cells were also positive for cyclin E. These data indicate that cyclin E appears in S phase but also during a portion of the G phase (Figure (Figure5).5). In BrdU-positive cells cyclin E was found exclusively in the nucleus, but in BrdU-negative cells it was found either in the cytoplasm or in both the nucleus and cytoplasm. Together these data suggest that cyclin E initially accumulated in the cytoplasm and was translocated to the nucleus before the onset of DNA synthesis, where it remained through the rest of S phase.
Endoreduplication occurs, by definition, in the absence of mitosis; therefore, we examined the fate of B-type cyclins after the transition to the endocycle during Rcho-1 differentiation. The mRNAs for the cyclins B1 and B2 were readily detectable in the Rcho-1 stem cells, but were undetectable by d 2 of differentiation (Figure (Figure6A).6A). The mRNA expression of the two B-type cyclins then remained undetectable throughout differentiation, indicating that transcription was suppressed. The cyclin B1 protein product was readily detectable in the proliferating Rcho-1 stem cells by immunoblot analysis. Curiously, it was also detectable in giant cells at d 0 and d 1 of differentiation despite the loss of mRNA, but disappeared between d 1 and d 2 (Figure (Figure6B)6B) and remained essentially undetectable thereafter.
The cyclin B1-associated kinase activity was analyzed in immunocomplex histone H1 kinase assays and was readily detectable in the proliferating Rcho-1 cells (Figure (Figure6C).6C). Surprisingly, despite the persistence of the cyclin B1 protein, very little cyclin B1-associated kinase activity was detectable in the Rcho-1 giant cells on d 0 and d 1. The down-regulation of cyclin B-associated kinase activity was not achieved by elimination of its potential kinase partner, because the abundance of p34cdk1, the kinase most strongly associated with the B-type cyclins, did not change significantly during differentiation of the Rcho-1 cells (Figure (Figure77 A). However, the amount of p34cdk1 that was associated with cyclin B1 was significantly reduced with the transition to the endocycle (Figure (Figure7B).7B). This was confirmed in experiments in which the anti-cyclin B1 immunoprecipitates were formed in antigen excess (Figure (Figure7B).7B). Under these conditions, a small amount of p34cdk1 that was complexed with cyclin B1 was detectable in lysates from d 1 giant cells despite the low kinase activity. These data indicate that after the transition to the first endocycle, cyclin B-associated CDK activity remains low due primarily to reduced p34cdk1 association, but also to a decrease in specific activity.
The intimate interdependence of S and M phases is essential for maintaining the normal euploid state of a cell, and its disruption appears to be an early event leading to the aneuploidy associated with tumorigenesis (Pathak et al., 1994 ). Understanding endoreduplication should help to define the possible routes by which such a disruption might occur, because it represents a natural dissociation of S from M. The cessation of mitosis and the initiation of endoreduplication in Rcho-1 trophoblast cells were closely linked to the onset of differentiation. The cell cycle changes were detectable from the earliest time at which differentiated cells could be isolated and they preceded the changes in expression of four molecular markers of trophoblast giant cell differentiation. The first indication of commitment to giant cell differentiation is markedly increased adhesion, which we used to obtain synchronized cell populations (Cross et al., 1995 ). Although trophoblast differentiation appears to be positively regulated by Hxt and negatively regulated by Id1 and Id2 (Cross et al., 1995 ), the role of these factors in the cell cycle changes, as opposed to the differentiation changes, has yet to be determined. Indeed, Hxt increases more closely paralleled differentiation than the exit from the mitotic cycle. Trophoblast differentiation and endoreduplication can be dissociated (Gardner and Davies, 1993 ), and it seems likely that the temporal separation distinguishable in the Rcho-1 cells reflects the normal regulatory pathways seen in trophoblast. We found that the transition from the mitotic cell cycle to the endocycle involved several changes in cell cycle control, including altered expression of cyclins and alterations in checkpoint controls. Endoreduplication represents an interesting conflict for a cell as it is a terminally differentiated phenotype that retains key features of proliferation, events that appear to be mutually exclusive in other cell types. It will be of interest to discover how the regulation of differentiation is insulated from active cyclin A and E complexes. Endoreduplication may allow the growth of cells beyond the limit defined by the nuclear/cytoplasmic ratio that normally restricts the size of diploid cells, and would thus allow considerable growth in postmitotic cells. Other variant cycles in which dissociation of growth from proliferation occurs are the endomitotic cycles of megakaryocytes (Datta et al., 1996 ; Zhang et al., 1996 ) and the formation of syncytia in muscle and trophoblast of several species, including humans. The diversity of variant cycles during development suggests the complexity of the repertoire of cell cycle regulation on which the developing organism can draw and the versatility of growth regulation as a tool in morphogenesis.
The induction of trophoblast giant cell markers in differentiating Rcho-1 cells has been previously reported, and it was also suggested that the Rcho-1 cells were capable of endoreduplication (Shida et al., 1993 ; Yamamoto et al., 1994 ; Cross et al., 1995 ; Hamlin and Soares, 1995 ). The lack of synchrony within the cell population prevented more detailed analyses of the regulatory mechanisms. The ability to select a relatively synchronous cell population allowed us to analyze the temporal aspects of the regulation of the commitment to differentiation and endoreduplication. We demonstrated that the Rcho-1 cells were able to achieve DNA contents of up to 64 N by d 6 of differentiation. DNA synthesis continued past d 12 of differentiation, suggesting that the Rcho-1 cells are capable of reaching DNA contents of up to 128–256 N, close to the ploidies of 500-1000 N that have been reported for murine trophoblast in vivo (Zybina, 1970 ; Zybina and Grishchenko, 1970 ; Barlow and Sherman, 1972 ). Endoreduplication appeared to proceed in cycles that contained gap phases separating rounds of DNA synthesis in which DNA was replicated to completion, revealing an underlying regulation of initiation and progression.
Our studies revealed several changes in cell cycle control after the transition to the endocycle occurs. The loss of the G1-S checkpoints does not appear to be an oddity of the immortalized state of the Rcho-1 cell line, because DNA synthesis in trophoblast derived directly from mouse embryos is also relatively resistant to inhibition following X-irradiation (Goldstein et al., 1975 ) and occurs in the absence of growth factor stimulation (Newman-Smith et al., 1997 ). The molecular basis of the failure to inhibit DNA synthesis in response to environmental insult is not clear, but reprogramming of the cell cycle by regulators of differentiation may override the checkpoints that normally control cell cycle progression.
With the onset of Rcho-1 differentiation, the expression of the D-type cyclins changes significantly. The commitment of megakaryocytes to the endomitotic pathway requires cyclin D function and is associated with changes in expression of D-type cyclins as well (Wang et al., 1995 ; Wilhide et al., 1995 ), suggesting a role for D-type cyclins in regulating the onset of both endomitotic and endoreduplicative cycles. Rcho-1 stem cells expressed cyclin D3, but its expression was suppressed with the onset of differentiation. Cyclin D3 is the only D-type cyclin expressed in the yolk sac and ectoplacental cone of the d 8 mouse embryo (MacAuley, unpublished observation), suggesting that giant cells may replace cyclin D3 during the endocycle. The up-regulation of cyclin D1 in differentiating Rcho-1 giant cells suggested that it may function during the endocycle. The cyclin D1 knockout mouse has no obvious defect in the trophoblastic lineage (Fantl et al., 1995 ; Sicinski et al., 1995 ). However, if constitutive cyclin D1 expression in giant cells simply renders DNA synthesis independent of serum growth factors or provides a mechanism for overriding inhibitory signals that result from DNA damage, these functions would not necessarily be essential during normal mouse development.
It was surprising that the mRNAs for both the S phase cyclins, A and E, were down-regulated in the differentiated Rcho-1 cells (though still detectable) after d 1 of differentiation. The abundance of the cyclin A and E proteins did not decline as rapidly as the respective mRNAs, but it is not clear whether the slower rate of turnover is related to the lengthening of cell cycle phases during endoreduplication or a change in the regulation of protein turnover. For example, cyclin A degradation normally occurs during mitosis, but in the absence of any mitotic events a new signal must initiate cyclin A destruction at the end of the S phase (Pines and Hunter, 1991 ; Hunt et al., 1992 ). The cyclin A- and E-associated kinase activities remained readily detectable during these early stages of differentiation, although the specific activity did not appear to be as high in the differentiating cells as in the proliferating cells. Several potential positive and negative regulatory mechanisms have been described in the mitotic cell cycle (reviewed in Morgan, 1995 ), but the basis for the lower specific activity of the cyclin/cdk complexes is not clear.
The periodic cycles of accumulation of cyclin A over a time frame similar to that of the doubling of DNA content suggests that cyclin A plays a regulatory role during the endocycle. Similar cycles were not observed for cyclin E, but the shorter gap in cyclin E expression may not have been detectable with the limited degree of synchronization obtainable in these experiments. The colocalization of cyclin A and E to essentially all BrdU-positive cells suggests that both cyclins are required for S phase progression. Cyclin E was found in the cytoplasm of some BrdU-negative cells, suggesting that it was either retained briefly in the cytoplasm after the initiation of synthesis or degraded if it was translocated to the nucleus. It seems likely that the cells that were cyclin E-positive/BrdU-negative cells were before S phase, because this is similar to regulation of the mitotic cycle (Koff et al., 1992 ). These results indicate that both cyclins play a role in S phase and that, although cyclin E is required for the initiation of S phase, cyclin A is required for its completion. Another conclusion that can be drawn from the colocalization data is that G phase cells lack both cyclin E and cyclin A, supporting the idea that a period of the endocycle must lack cyclin-associated kinase activity to allow the cycle to be reset and the next round of DNA synthesis to occur. It is also apparent from the cyclin/BrdU colocalization data that the cyclin A and E proteins must be rapidly degraded at the end of S phase.
The data presented here outline a pathway that may explain the shift from the mitotic cycle to the endocycle during trophoblast differentiation (model summarized in Figure Figure8).8). With commitment to giant cell differentiation, Rcho-1 cells institute a program that prevents activation of mitosis by inhibiting the p34cdk1-dependent kinase activity resulting in cells arrested in G2. Subsequently, both cyclin E and A/p33cdk2 kinase activities are reactivated and DNA synthesis is reinitiated. At the end of the first endocycle S phase, cyclins A, E, and B are degraded. Initiation of a new cycle of endoreduplication occurs during a gap phase in which the cyclin A- and E-associated kinase activity are absent. By analogy to the mitotic cell cycle, cyclin degradation presumably allows the reactivation of the origins of replication (Cocker et al., 1996 ). Expression of cyclin E is activated first, leading to its accumulation in the cytoplasm and subsequently in the nucleus. The initiation of DNA synthesis and cyclin A expression appear later. With the completion of DNA synthesis, both cyclin A and E are degraded, signaling the end of one cycle and allowing initiation of the next. Whether the movement of cyclin E into the nucleus is the event that initiates S phase and the expression of cyclin A is unclear, but analogy to the mitotic cycle suggests that this is a likely mechanism.
The outline of the trophoblast endocycle differs somewhat from that described in other systems. In Drosophila, entry into each endocycle is driven by a new wave of cyclin E expression in the absence of both cyclin A and B (Knoblich et al., 1994 ; Duronio and O’Farrell, 1995 ; Sauer et al., 1995 ). The expression of cyclin A in the Rcho-1 cells during endoreduplication may help drive DNA synthesis closer to completion than Drosophila, where it appears that as much as 15–20% of the genome is unreplicated during each endocycle, perhaps reflecting limitations in the cdk activity (Lilly et al., 1996 ). Two apparent similarities between the Rcho-1 trophoblast and Drosophila are the necessity for a period of little or no cdk activity between rounds of DNA synthesis, and that the B-type cyclins are suppressed during endoreduplication. The regulation of the initiation of the Drosophila endocycle by developmental factors may render initiation of each cycle independent of the preceding cycle, so that the termination of one cycle does not act as a signal to initiate the next cycle (Knoblich et al., 1994 ; Sauer et al., 1995 ). Models for the regulation of the endocycle have not been well described in any other system. In corn endosperm, the total cdk activity decreases significantly with the onset of endoreduplication, although there is relatively little decrease in the abundance of the cdk proteins detectable. The loss of cdk activity appears to be an active process (Grafi and Larkins, 1995 ), perhaps similar to the Rcho-1 endocycle. The level of the S phase-associated cdk activity rises in endoreduplicating tissue, but it is not clear which cyclins are associated with this activity, nor how progression through the endocycle is regulated. Megakaryocytes also become polyploid, but through an endomitotic process in which cyclin B is expressed and is associated with kinase activity (Datta et al., 1996 ; Zhang et al., 1996 ). The expression of the endomitotic events seems to separate rounds of DNA synthesis in a way quite similar to M phase in the mitotic cycle.
The Rcho-1 cell system has allowed us to study events that occur at the transition between the mitotic cell cycle and the first endocycle, a step not clearly addressed in other systems. We found that the cyclin B1 protein remained abundant during the first endocycle. The fact that cyclin B was expressed at all during the first endocycle, but not in later endocycles, implies that this first cycle begins in G2 of a mitotic cycle. The fact that it persists until the end of the first endocycle is presumably due to the failure to activate its rapid degradation, an event usually associated with mitosis (King et al., 1994 ). After the first endocycle, cyclin B was no longer expressed and cyclin B mRNA was not detectable after commitment to giant cell differentiation at d 0, preventing reinitiation of mitosis.
In the mitotic cell cycle, cyclin B transcription is cell cycle regulated, with mRNA induction occurring in G2. The fact that cyclin B transcription is not induced during the endocycle implies a fundamental difference in the G2 phase of the first endocycle. Because cyclin B induction may depend on G1-S cyclin/cdk activity, it is possible that the abrupt termination of cyclin A- and E-associated kinase activity at the end of endocycle S phase preempts this induction. The failure of the d 0 giant cells to progress into mitosis, despite the presence of cyclin B protein, appeared to derive largely from the inability to form stable, active cyclin B/p34cdk1 complexes, as a consequence of a reduction in both the association of cyclin B and p34cdk1 and the specific activity of the complexes that did form. The latter effect could be related to p34cdk1 phosphorylation, a mechanism well described during the mitotic cycle (Morgan, 1995 ). The lack of cyclin B association with p34cdk1 was, however, somewhat surprising. The amount of p34cdk1 associated with cyclin A was also decreased early in differentiation. This suggests that p34cdk1 is the target of the inhibition, as might be expected of the cdk most closely associated with regulation of mitosis. In conclusion, it appears that one of the first differentiated functions expressed by trophoblast cells that have committed to giant cell fate is to suppress the regulatory activities associated with mitosis, first by direct inhibition of the kinase activity, and then by inhibition of expression of mitotic cyclins. Although mechanisms involved in cyclin B/p34cdk1 complex formation are unclear at present, Hsp70–2 acts as a molecular chaperone that is required for cyclin B1/CDC2 complex formation and kinase activity during meiosis I of mouse spermatocytes (Zhu et al., 1997 ). It is possible that the kinase activity of the cyclin B complex is inhibited by regulation of a chaperone or by a peptide inhibitor that also targets the complex for disassembly. Such novel mechanisms are currently under investigation.
Our sincere thanks to the people who provided reagents, particularly Drs. J. DeLoia, C. Sherr, and D. Wolgemuth for providing plasmids of various cyclins, and Dr. J. Roberts for the anti-cyclin E antibody. We thank O. Behrendtsen for his help with the figures. The work was supported by grants from the National Institutes of Health (HD-26732 and CA-75072) to Z.W. and the Medical Research Council of Canada to J.C.C. J.C.C. is a Scholar of the Medical Research Council of Canada.