Elimination of Cyclin D1 in Vertebrate Cells Leads to an Altered Cell Cycle Phenotype, Which Is Rescued by Overexpression of Murine Cyclins D1, D2, or D3 but Not by a Mutant Cyclin D1*

DT40 lymphoma B-cells normally express cyclins D1 and D2 but not D3. When cyclin D1 expression was extinguished in these cells by gene knockout, specific al- terations in their ability to transit the cell cycle were observed. These changes are exemplified by a delay of approximately 2 h in their progression through a normal 14-h cell cycle. This delay results in an increase in the number of cells in the G 2 /M phase population, most likely due to triggering of checkpoints in G 2 /M, inability to enter G 1 normally, and/or alterations of crucial event(s) in early G 1 . The defect(s) in the cell cycle of these D1 “knockout” cells can be rescued by overexpres- sion of any normal mouse D-type cyclin but not by a mutant mouse cyclin D1 protein that lacks the L X C X E motif at its amino terminus. These data suggest that the cell cycle alterations observed in the D1 (cid:50) / (cid:50) cells are a direct effect of the absence of the cyclin D1 protein and support the hypothesis that the D-type cyclins have separate, but overlapping, functions. Elimination of cyclin D1 also resulted in enhanced sensitivity to radiation, resulting in a significant increase in apoptotic cells. Ex- pression of any normal murine D-type cyclin in the D1 (cid:50) / (cid:50) cells reversed this phenotype. Intriguingly, expression of the mutant cyclin D1 in the D1 (cid:50) / (cid:50) cells partially restored resistance to radiation-induced apo-

Unlike other G 1 cyclins described thus far (e.g. cyclin C and cyclin E), the D-type cyclins are highly homologous to one another and their expression overlaps during G 1 phase, suggesting that this group of cyclins may be somewhat redundant in their function. However, it has also been noted that not all D-type cyclins are expressed in each tissue, suggesting that their function may be linked to the specific tissues in which they are expressed. This particular cell cycle pathway is subject to a number of alterations during tumorigenesis, presumably due to its importance in response to mitogenic stimulation (31,(35)(36)(37)(38)(39)(40)(41)(42)(43).
The biologic role of multiple D-type cyclins has not yet been resolved. It has been established that each of the D-type cyclins contain a functional pRb-binding motif and that these cyclins are induced in response to mitogens in a cell lineage-specific manner (3,4). However, cyclins D2 and D3 bind preferentially to pRb, whereas cyclin D1⅐pRb complexes are less stable (44,45). Microinjection of cyclin D1 antibodies and/or antisense oligonucleotides into normal human diploid fibroblasts, NIH3T3, and Rat-2 cells revealed that a sub-group of these cells could not enter S phase (46,47). The ability of these microinjected cells to pass the restriction point and begin DNA replication was directly related to the time that had elapsed between the readdition of serum and microinjection of cyclin D1 antibodies or antisense oligonucleotides. Thus, by decreasing the level of functional cyclin D1 protein prior to the restriction point in late G 1 phase, the ability of these cells to progress through the cell cycle was impaired. These agents have no apparent effect on cell cycle once cells have passed the restriction point late in G 1 phase of the cell cycle. Finally, it has been shown that ectopic expression of cyclin D1 inhibits MyoD and myogenin-mediated skeletal muscle differentiation indirectly (48). To determine the role of cyclin D1 in muscle differentiation, experiments were designed to determine whether MyoD⅐myogenin and/or pRb were targets of cyclin D1⅐CDK phosphorylation. Mutation of CDK phosphorylation sites in myogenin had no effect on the ability of cyclin D1 to inhibit differentiation. Ectopic MyoD expression in fibroblasts can induce muscle-specific gene expression, as long as wild-type pRb is also expressed (48). In pRb Ϫ/Ϫ fibroblasts expressing MyoD and a mutated pRb (which cannot be hyperphosphorylated), ectopic cyclin D1 expression continues to inhibit muscle-specific gene expression, whereas ectopically expressed cyclins A and E have no effect. These data suggest that the initiation of muscle-specific gene expression can be blocked by two distinct pathways, one of which is dependent on pRb hyperphosphorylation and the other is not.
Several models of D-type cyclin function have been proposed based on the available data, but they all envision these cyclins as growth sensors (3,4,45,46,49,50). The first model suggests that the D-type cyclins are tissue-specific regulatory subunits that are merely redundant in their function. A second model suggests that the functions of D-type cyclins are related to their sequential expression pattern during the cell cycle. More recent studies with cyclin D1 gene "knockouts" in mice have supported the former possibility (49,50). Specifically, proper development of the retina and breast ductal epithelium was perturbed, suggesting that their normal maturation depends upon the presence of cyclin D1. This mouse model has provided significant insight into the requirements of cyclin D1 for normal development, but a specific requirement for this cyclin during cell cycle progression per se, has not been established.
We now report the successful utilization of the chicken lymphoma B-cell line, DT40, to eliminate the expression of specific G 1 cyclins. This particular cell line undergoes gene-specific recombination at frequencies that are ϳ100-fold higher than those in mouse embryonic stem cells (51). DT40 cells provide a unique reagent to easily and quickly examine the necessity for each of these cyclins individually and collectively, as well as interactions between the various cyclins and other cell cycle proteins. These lymphoma B-cells normally express cyclin D1 and D2 but not cyclin D3. Elimination of cyclin D1 in DT40 cells is not lethal; however, ablation of cyclin D1 gene expression results in marked alterations in the ability of these cells to transit the cell cycle which, in turn, affects their normal growth rate. Furthermore, loss of cyclin D1 did not result in a compensatory increase in the levels of cyclin D2 in these cells. Overexpression of normal mouse cyclin D1, D2, or D3 in the D1 Ϫ/Ϫ cells restored normal growth and cell cycle progression. However, a mutant murine cyclin D1 lacking the LXCXE motif (the so-called retinoblastoma protein (pRb)-binding motif that enhances D-type cyclin interaction with pRb) did not rescue the D1 Ϫ/Ϫ phenotype (44,45).
Here we show that elimination of cyclin D1 leads to a specific cell cycle phenotype and that these cells can be rescued by expression of any one of the three murine D-type cyclins, but not by a murine cyclin D1 mutant missing the LXCXE motif. The rescue of the D1 Ϫ/Ϫ DT40 cells by murine cyclin D3 is somewhat intriguing, since avians do not express a cyclin D3 homologue in any cell line or tissue examined thus far (52,53, and this study). Elimination of cyclin D1 in DT40 cells also resulted in a significant increase in apoptotic death following either UV or ␥ irradiation. Finally, expression of either murine cyclin D1, D2, or D3 in the DT40 D1Ϫ/Ϫ cells resulted in decreased sensitivity to radiation-induced apoptosis, indistinguishable to what was observed in the parental DT40 cell line. Expression of the mutant cyclin D1 partially rescued cells from radiation-induced apoptosis, suggesting that the mechanism(s) involving cyclin D1 in this pathway is distinct from its role in normal cell cycle progression. Thus, in avian B-lymphoma cells cyclin D1 appears to fully suppress apoptosis, whereas a mutant cyclin D1 partially rescued these cells, suggesting that the targets of cyclin D1⅐CDK activity in apoptotic cells may be distinct from those in proliferating cells.

EXPERIMENTAL PROCEDURES
Materials-Agarose was purchased from FMC Bioproducts (Rockland, ME). Restriction and modification enzymes were obtained from New England Biolabs (Beverly, MA) and Boehringer Mannheim. Urea and cell culture reagents, including chicken serum, were purchased from Life Technologies, Inc. Sequenase and DNA sequencing reagents were purchased from U. S. Biochemical Corp. All radioisotopes were purchased from DuPont NEN. All chemicals were purchased from either Sigma or Fisher. The chicken cosmid genomic library was obtained from Clonetech (Palo Alto, CA). The PrimeIt labeling kit and Duralose membranes were purchased from Stratagene (La Jolla, CA). Selection drugs were purchased from Life Technologies, Inc., Sigma, or Boehringer Mannheim. Fluorescein-conjugated goat anti-mouse and goat anti-rabbit antibodies were obtained from Boehringer Mannheim or Southern Biotechnology. Rabbit polyclonal anti-cyclin D1 antibody was kindly provided by Dr. Charles Sherr (see Ref. 2), and anti-BudR antibodies were obtained from Boehringer Mannheim. The DNA-specific dye 4Ј,6-diamidino-2-phenylindole was obtained from Boehringer Mannheim. Oligonucleotides for DNA sequencing were synthesized on an Applied Biosystems 394 DNA/RNA Synthesizer by the Molecular Resources Facility of St. Jude Children's Research Hospital.
Cloning and Sequencing of the Avian Cyclin D1 cDNA, Its Corresponding Gene, and Localization of the Chromosomal Gene-Cyclin D1 cDNAs were isolated from a UG9 T-cell cDNA library (55) by low stringency hybridization with a human cyclin D1 cDNA, kindly provided by Dr. Steve Reed (56). The nucleotide sequence of these cDNAs was determined as described previously (57). Oligonucleotides were spaced approximately 80 -100 base pairs apart spanning the entire cDNA. A full-length cyclin D1 cDNA was transcribed and translated in vitro, producing a 36-kDa protein that could be immunoprecipitated with the mouse polyclonal cyclin D1 antibody (see Fig. 4, panel A). The corresponding cyclin D1 (Gallus gallus ccnd1) gene was isolated by screening a Cornish White Rock chicken cosmid library (Stratagene) with the full-length cyclin D1 cDNA as described previously (58). A 13-kb BamHI restriction fragment containing exons 1-3 of the cyclin D1 gene was subcloned into the pKS plasmid and either this DNA, or cosmid DNA, was used for double-strand DNA sequence analysis of the gene. The same oligonucleotides used for establishing the cDNA sequence were used to establish the molecular organization (intron/exon boundaries, exons) of the gene by sequencing the DNA strands in both directions. Genomic DNA was isolated as described previously (59).
Gene Constructs-The constructs containing the inserted drug selection cassettes were made by digestion of the 13-kb BamHI subclone that contained exons 1-3 of the Ggccnd1 gene, with the enzyme BstI that opened this fragment at a single site in exon 2 ( Fig. 2, panel A). This exon encodes a major portion of the cyclin box region. Approximately 5.5 kb of genomic DNA were left intact on the 5Ј side of this insertion, and approximately 7.5 kb of genomic DNA remained on the 3Ј side. Restriction fragments containing either the neo or hyg gene under the control of a ␤-actin gene promoter (kindly provided by Dr. C. Thompson) were cloned into this BstI site by blunt-end ligation using Klenow enzyme. Further details concerning the selectable marker genes can be found elsewhere (60,61). The integrity of the constructs was confirmed by DNA sequencing, demonstrating that exon 2 was disrupted and that the remainder of the gene was intact. Homologous recombination of these constructs into the Gccnd1 locus of the DT40 cells was demonstrated by the increase in size of a 15-kb XbaI fragment that contains exons 1 and 2 of the gene but that also contains a unique 5Ј region (5Ј-flanking probe) not found in the 13-kb BamHI subclone used for generating the disruption constructs (Fig. 2, panel A). This probe could then be used to analyze genomic DNA from single cell clones after XbaI digestion. Insertion of the neo cassette (2 kb in size) generated a 17-kb XbaI hybridizing fragment (Fig. 2, panels C and D), and insertion of the hyg gene cassette (3 kb in size) generated an 18-kb XbaI hybridizing fragment (Fig. 2, panels C and D).
Targeted Disruption and Analysis of Cyclin D1 Expression-DT40 lymphoma B cells (kindly provided by Dr. C-L. Chen) were grown in suspension cell culture in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum, 1% chicken serum, penicillin, streptomycin, and glutamine. pCycD1-neo and pCycD1-hyg were linearized and transfected into DT40 cells by electroporation (550 V, 25 microfarads). Twenty-four hours after DNA transfection, the appropriate concentration of selection drug (2 mg/ml G418 or 1.5 mg/ml hygromycin) was added to the culture medium, and the cells were selected for ϳ14 days. Single cells, isolated by flow sorting, were expanded into individual clones. Genomic DNA was isolated from multiple single cell-derived clones, digested with the appropriate restriction enzymes and hybridized with the cyclin D1 5Ј-flanking genomic probe (Fig. 2, panel A) to screen for homologous recombinants. Disruption of one, heterozygote (ϩ/Ϫ), or both, homozygote (Ϫ/Ϫ), cyclin D1 (Ggccnd1) alleles was observed. For all experiments, multiple single cell clones of Ϫ/ϩ and Ϫ/Ϫ cyclin D1 disruptions were examined by both Southern and Northern blotting. RNA was isolated for Northern blotting, and the blots were hybridized to the cyclin D1 cDNA probe as described previously (57,62). Equal loading of RNA samples was verified by reprobing these blots with a human ␤-actin cDNA probe. The probes were prepared using a random labeling kit (Stratagene) as specified by the manufacturer. The blots were visualized by using a Molecular Dynamics 400A Phosphor-Imager. Exposure times were as follows: 12 h for the cyclin D1 cDNA, 2 h for the cyclin B2 cDNA, and 20 min for the ␤-actin cDNA. Cyclin D1 protein expression was examined by immunoprecipitation of [ 35 S]methionine-labeled cell lysates using either mouse polyclonal cyclin D1, D2, or D3 antiserum as described previously by others (28). As controls for these experiments, the mouse preimmune sera were used in all experiments, and an appropriate unlabeled cyclin D-GST fusion protein was used as a competitor to demonstrate specificity of the various D-type cyclin immunoprecipitations in DT40 cells (Fig. 4, panel A).
Cell Growth Analysis, BudR Incorporation, Mitotic Indices, and Flow Cytometric Analysis-Cells were selected in 2 mg/ml G418 or 1.5 mg/ml hygromycin and then maintained in cell culture under continuous drug selection at the same concentrations. For the growth curves, 2 ϫ 10 5 cells were plated in triplicate and grown in the presence of 10% serum. Each point represents the average of three determinations Ϯ the standard error of the mean. Growth curves were obtained for each of the cell lines for 7 days at 1-day intervals. DT40 cells were synchronized by first using centrifugal elutriation to enrich the G 1 phase population of cells. These cells were then blocked at the G 1 /S boundary using 25 mg/ml aphidicolin for 10 h. These synchronized cells were removed from the drug and then placed into culture media containing BudR, harvested at 2-h intervals, and replicating (BudR ϩ) cells stained and counted. Cytospin slides were prepared and stained with an anti-BudR antibody and a fluorescein labeled secondary antibody according to the protocol from the manufacturer (Boehringer Mannheim). The cells were also stained with 4Ј,6-diamidino-2-phenylindole, and the number of metaphase cells was determined as well. For the BudR incorporation and mitotic index experiments, two separate clonal D1 Ϫ/Ϫ cell lines were examined (clones 1 and 3). For cell cycle analysis, asynchronously growing cells were stained with propidium iodide, and their DNA content was determined using the fluorescence activated cell sorter. This analysis was performed as described previously (62) using asynchronously growing cultures by analyzing approximately 50,000 cells from polyclonal, or single cell, populations for each sample. The percentage of cells in each phase of the cell cycle was determined from the histograms using the MODFIT cell cycle analysis program.
Rescue of Cyclin D1 "Knockouts" by Expression of Either Normal Murine Cyclin D1, D2, D3 or a Mutant Murine Cyclin D1 cDNA-Rescue of the cyclin D1 Ϫ/Ϫ DT40 cells was accomplished by transfecting a clonal D1 Ϫ/Ϫ cell line (clone 1) with either a RSV-mouse CycD1 cDNA, RSV-mouse CycD2, RSV-mouse CycD3, or the RSVmutant mouse CycD1 cDNA, which have been reported previously (28,30,47). These cDNAs were co-transfected with a puromycin (pur) selectable drug resistance marker to allow selection of stable integrants. Since all clonal D1 Ϫ/Ϫ DT40 cell lines had similar, if not identical, cell cycle alterations, only one was selected for transfection of the two mouse cDNAs (clone 1). However, multiple single cell clones or pooled clone samples were subsequently isolated from each of these transfections and selections (clones 1/56 and 1/57 for normal mouse cyclin D1, a pool of positive clones for normal mouse cyclin D2, normal mouse cyclin D3, and for the mutant mouse cyclin D1) for further analysis. Two of the original cyclin D1 Ϫ/Ϫ cell clones, as well as two clones from the same D1 Ϫ/Ϫ clone into which the murine D1 cDNA expression construct was introduced, were analyzed for cyclin D1 protein expression (Fig. 4, panel C). Similarly, clonal cell lines of the same D1 Ϫ/Ϫ cells containing either the murine cyclin D2 or D3 expression construct were analyzed for their expression of these respective proteins (Fig. 4, panel D). Two of the original DT40 D1 Ϫ/Ϫ clonal cell lines, as well as one expressing either normal murine cyclin D1, D2, or D3 protein, were analyzed further by flow cytometric analysis. The mutant mouse cyclin D1 protein migrates as a smaller molecular weight protein on SDSpolyacrylamide gel electrophoresis due to the deletion of its amino terminus, which includes the pRb-binding motif LXCXE (Fig. 4, panel C (30)).

Quantitation of Cells Undergoing Programmed Cell Death Due to DNA Damage Induced by UV or ␥ Irradiation-Normal and D1
Ϫ/Ϫ cells were exposed to varying levels of UV or ␥ irradiation to assess whether there are differences in the level of programmed cell death between these different cell lines. Others have previously shown (63,64) that DT40 cells are a radiation-sensitive lymphoma B cell line that requires PLC-␥2 activation, via the combined activities of the Bruton's tyrosine kinase and the SYK protein kinase. Apoptosis was induced by exposing the DT40, D1Ϫ/Ϫ, D1Ϫ/Ϫ[ϩ murine D1], D1Ϫ/Ϫ[ϩ murine D2], or D1Ϫ/Ϫ[ϩ murine D3] cell lines to increasing amounts of either UV or ␥ irradiation (63)(64)(65). Three different single cell clones were used for these analyses. For the UV radiation studies, DT40 cells were removed from culture (growing at a density of ϳ10 6 /ml), plated on polylysine-coated coverslips (25 g/ml), and allowed to attach for 30 min to 1 h, and washed twice with phosphate-buffered saline. These cells were then irradiated with varying amounts of either UV (UVC; 500, 1000, 1500, and 2000 J/m 2 ). For ␥ irradiation (4-Gy or 8-Gy ␥-rays), the cells were treated in culture at a density of 5 ϫ 10 5 /ml, as described by others (65,66). Both the UV and ␥ irradiated cells were returned to the incubator following addition of fresh media. Cells were then harvested 4 -8 h following exposure to radiation, and the percentage of apoptotic cells was determined by TUNEL analysis and/or Hoescht staining of nuclei.

Isolation and Characterization of the Normal Avian Cyclin D1 cDNA and Gene from Avians, and the Construction of Insertion Mutants-
The cyclin D1 cDNA was isolated by screening a UG-9 T-cell cDNA library using the human cyclin D1 cDNA as a probe (56). This particular T-cell line expresses cyclin D1 and cyclin D2 but not cyclin D3. 2 Similarly, we found that the DT40 B-cell line expressed cyclins D1 and D2 but not D3 (data not shown; Fig. 3). Mullner and colleagues (53) have suggested that chickens may lack the cyclin D3 gene and the absence of an avian cyclin D3 homologue may indicate an evolutionary divergence in the tissue specificity of D-type cyclin expression. Our results appear to support this hypothesis. Ten positive cyclin D1 cDNA clones were isolated and purified. Six of these cDNAs had inserts ϳ1.3 kb and were used for subsequent determination of the nucleotide sequence and predicted open reading frame (Fig. 1). Comparison of the avian, human, and murine cyclin D1 open reading frames demonstrates a high level of sequence conservation (86% identity with human and mouse) between the proteins, particularly in the conserved "cyclin box" region (1). The corresponding gene was isolated from a chicken cosmid library, and its structure was determined in its entirety by restriction endonuclease and DNA sequence analysis (Fig. 2, panel A). The size, ϳ10 kb, and organization of the avian cyclin D1 gene (Ggccnd1) is identical to the human (CCND1) and mouse (ccnd1) homologues (67,68).
A major portion of the Ggccnd1 gene was subcloned for further analysis and construction of insertion mutants to eliminate expression of this gene in DT40 cells. A 13-kb BamHI fragment containing exons 1-3 of the Ggccnd1 gene was used for generating disruption constructs, and a single copy ϳ3-kb BamHI fragment located 5Ј of the larger BamHI fragment, but outside the region used to disrupt this allele by homologous recombination, was used for genomic DNA analysis (Fig. 2, panel A). We took advantage of a unique BstI restriction site found only in exon 2 of the cyclin D1 gene to generate insertion mutants, containing either neomycin (neo) or hygromycin (hyg) selectable marker genes that inactivate the normal gene (Fig.  2, panel B). Insertion of either of these selectable marker gene cassettes into the BstI restriction site would alter the normal mobility of a 15-kb XbaI genomic fragment that is detected by the 5Ј-flanking probe (Fig. 2, panel A). Introduction of the neo gene cassette by homologous recombination into this region of the cyclin D1 gene generates a 17-kb XbaI fragment when hybridized with the 5Ј-flanking probe, while the insertion of the hyg gene cassette generates an 18-kb XbaI fragment using the same probe (Fig. 2, panel C). As shown in Fig. 2, panel D, the wild-type DT40 cells (ϩ/ϩ) contain two copies of the normal 15-kb XbaI hybridizing fragment, whereas cells containing one normal allele and one allele disrupted by the introduction of the neo gene (ϩ/Ϫ, heterozygote DT40 cells) into the cyclin D1 gene contain 15-and 17-kb XbaI hybridizing fragments, as expected. In DT40 cells containing both the neo and hyg gene cassettes homologously recombined into the cyclin D1 locus (Ϫ/Ϫ, homozygotes), the wild-type 15-kb XbaI hybridizing fragment is completely absent, and it has been replaced by the 17-and 18-kb XbaI fragments, respectively (Fig. 2, panel D). Additional proof of cyclin D1 elimination is provided by virtue of Northern blotting (Fig. 3) and protein analysis (Fig. 4).
Expression of Cyclin D1 mRNA and Protein in ϩ/ϩ, ϩ/Ϫ, and Ϫ/Ϫ DT40 Cells-Complete elimination of the 15-kb XbaI restriction fragment containing the normal Ggccnd1 gene was easily obtained in a number of single cell DT40 Ϫ/Ϫ clones (Fig.  2, panel D, lanes 4 -9). When expression of the cognate cyclin D1 mRNA (ϳ1.3 kb in size) was examined by Northern blotting using these same single cell clones, as well as a number of additional clones (15 total), expression of the mRNA was undetectable in Ϫ/Ϫ homozygotes and barely detectable in ϩ/Ϫ heterozygotes (Fig. 3). However, a cross-hybridizing cyclin D2 mRNA (ϳ6 kb) was readily detected in all cell lines (54,58). Consistent with the results of others, we were not able to detect an avian cyclin D3 mRNA species, or a corresponding gene, in DT40 cells (data not shown) (52,53). As with human and mouse, there is a relatively high degree of sequence homology between avian cyclin D1 and cyclin D2 (Fig. 1) (58). Concomitantly, expression of the normal cyclin D1 protein was also extinguished when examined by immunoprecipitation (IP) of metabolically labeled cell lysates (Fig. 4). In a control experiment, the mouse polyclonal cyclin D1 antibody was shown to IP the chicken in vitro transcribed and translated (IVTT) protein, as well as the [ 35 S]methionine-labeled cyclin D1 protein from the DT40 cells (Fig. 4, panel A). This ability to IP the cyclin D1 protein from DT40 cells was directly competed by preincubation with cold competitor mouse cyclin D1 protein (Fig. 4, panel  A), demonstrating specificity. A comparison of the preimmune serum control with the mouse polyclonal D1 antisera using the same [ 35 S]methioninie-labeled ϩ/ϩ wild-type DT40 cell lysates is also shown (Fig. 4, panel B, DT40). Analysis of two of the Ϫ/Ϫ cyclin D1 single cell clones shown above (clones 1 and 2) by metabolic labeling followed by cyclin D1 IP, demonstrated that neither of these cell lines expressed this protein (Fig. 4, panel  B). However, expression of either the wild-type or mutant murine cyclin D1 cDNAs in these same Ϫ/Ϫ cells (the mutant protein is truncated and no longer contains the pRb-binding motif LXCXE (44,45)) can easily be detected using this strategy (Fig. 4, panel C, normal murine cyclin D1, clones 1/56 and 1/57; mutant murine cyclin D1, clones 1/2 and 1/17).
Additional experiments involving the expression of either the murine cyclin D2 or D3 cDNA, and its corresponding protein, in the same D1 Ϫ/Ϫ DT40 cells yielded similar results (Fig. 4, panel D). The murine polyclonal antisera for cyclin D2 was capable of immunoprecipitating both the IVTT-labeled avian cyclin D2 and the metabolically labeled murine cyclin D2 overexpressed in the D1 Ϫ/Ϫ cells (Fig. 4, panel D). When excess cold murine cyclin D2-GST fusion protein was added to the cyclin D2 antibody, the immunoprecipitation of the [ 35 S]methionine-labeled IVTT protein was effectively competed (data not shown). Since an avian cyclin D3 homologue has not been isolated, a cyclin D3 IVTT control was not available for the corresponding immunoprecipitation of the metabolically labeled murine cyclin D3 protein expressed in the D1 Ϫ/Ϫ cells (Fig. 4, panel D). Steady-state levels of the various murine cyclins (D1, D2, and D3) were comparable in single cell clones, as judged by [ 35 S]methionine labeling ( Fig. 4 and data not shown). These metabolic labeling and immunoprecipitation experiments demonstrate that the cyclin D1 protein is absent in the D1 Ϫ/Ϫ DT40 cells, as well as confirm the expression of the normal murine cyclin D1, D2, D3, and mutant cyclin D1 in the same D1 Ϫ/Ϫ DT40 cells.
Alterations in the Normal Growth of the D1 Ϫ/Ϫ Cells and Changes in Their Ability to Traverse the Cell Cycle-Cyclin D1 elimination is not lethal, presumably because cyclin D2 continues to be expressed in these cells. This is consistent with results reported in homozygous cyclin D1 Ϫ/Ϫ mice, which are also viable (49,50). However, alterations in the cellular growth characteristics (Fig. 5) and normal progression through the cell cycle (Table I) were noted by flow cytometric analysis of asynchronously growing cells. The cell cycle phenotype of these cyclin D1 Ϫ/Ϫ cells, as might be expected, is distinct from that reported for modest overexpression of cyclin D1 in fibroblasts (47,69,70). Specifically, the total length of the cell cycle in these DT40 D1 Ϫ/Ϫ cells is increased by 2-3 h in two separate clones (1 and 3), increasing the cell doubling time from 14 to 16 -17 h (Table II and Fig. 5). Additionally, the percentage of asynchronously growing cells in G 2 /M phase is increased, and the percentage of cells in S phase is concomitantly decreased (Table I). In one of the mouse cyclin D1 knockout studies, fibroblasts derived from the D1 Ϫ/Ϫ mice, but not normal controls, were also delayed in G 2 /M phase when quiescent D1 Ϫ/Ϫ cells were restimulated by the addition of serum growth factors (50). Even though the effects of cyclin D1 elimination can be easily seen in asynchronously growing cell populations, they are more dramatic using synchronized cyclin D1 Ϫ/Ϫ DT40 cells (Table II; Table II). Similarly, when the MI for each cell line was monitored, the wild-type DT40 cells reached a peak in their MI at 10 h post-G 1 /S release, whereas both Ϫ/Ϫ clones had a much broader peak in their MI (spanning ϳ6 h), which peaked at 12 h post-G 1 /S release (Table II). These results provide two distinct parameters of cell cycle progression demonstrating that the growth of the D1 Ϫ/Ϫ DT40 cells was delayed by ϳ2-3 h.
Rescue of the Cyclin D1 Ϫ/Ϫ Cell Cycle Phenotype by Expression of Normal Mouse Cyclin D1, D2, or D3, but Not by a Mutant Mouse Cyclin D1 Lacking the LXCXE Motif-Having generated viable clonal cell lines lacking normal cyclin D1, but which continue to express normal levels of cyclin D2 (confirmed by IP of endogenous cyclin D2 from metabolically labeled wildtype and D1 Ϫ/Ϫ DT40 cells; data not shown), we reasoned that the observed changes in cell cycle progression could be reversed by reinstating cyclin D1 expression in these cells. We also reasoned that if D-type cyclin function were redundant, overexpression of all three D-type cyclins might rescue the observed phenotype of the D1 Ϫ/Ϫ DT40 cells. Therefore, we introduced either normal murine cyclin D1, D2, and D3 cDNAs or a D1 mutant cDNA (lacking the LXCXE motif (44,45)) into the D1 Ϫ/Ϫ DT40 cells (clone 1). Expression of these cDNAs was driven by an RSV promoter, as described previously (25). These investigators demonstrated that this particular promoter generated the highest level of "tolerable" cyclin D1 expression in murine fibroblasts and that exogenously expressed cyclin D1 could shorten the duration of G 1 phase specifically. Expression of murine cyclin D1, D2, D3, or the mutant D1 was confirmed by analysis of protein expression in single cell clones (Fig. 4, panels C and D). Normal cell cycle synchrony was restored in the D1Ϫ/Ϫ cells (i.e. no apparent block in G 2 /M) by all three D-type cyclins but not by expression of the mutant mouse cyclin D1 cDNA (in which the amino-terminal LXCXE pRb-binding sequence has been deleted (45) (Table I)). This mutant cyclin D1 cDNA construct also failed to alter normal cell cycle progression when it was overexpressed in rodent fibroblasts (47). Of particular significance is the fact that expression of any one of the murine D-type cyclin proteins was capable of rescuing cells from the apparent G 2 /M delay and not merely shortening the time spent in G 1 phase (Table I). Expression of the three murine D-type cyclin proteins was comparable in all of the single cell clones isolated and examined (Fig. 4, panels C and   D). This suggests that the characteristics of the cyclin D1 Ϫ/Ϫ phenotype are due to absence of cyclin D1 specifically but that this phenotype can be rescued by expression of any of the normal murine D-type cyclin proteins. It should be noted that cyclins D2 and D3 interact with pRb more efficiently than cyclin D1 (10,44,45), suggesting that the LXCXE motif in cyclin D1 may mediate the binding of pRb-related proteins (e.g. p107, p130) or some unknown protein. Apparently normal pRb was detected in DT40 cells; Western blotting of DT40 lysates with a human pRb antibody demonstrated a broad band at ϳ104 kDa, the reported molecular mass for avian pRb (data not shown) (71,72). The inability of the mutant murine cyclin D1, but not the normal murine cyclins D1, D2, and D3, to rescue these D1 Ϫ/Ϫ cells suggests that whatever interacts with the LXCXE motif of cyclin D1, its loss results in an altered cell cycle. These results are also consistent with the studies of Skapek et al. (48) demonstrating that cyclins A and E block muscle gene expression via phosphorylation of pRb but cyclin D1 does not; one interpretation of these data is that a pRbindependent pathway of skeletal muscle differentiation is regulated by cyclin D1. These results are consistent with what we have observed in the DT40 D1 Ϫ/Ϫ cells. Further biochemical studies of various cell cycle components (i.e. cyclins, CDKs, and CKIs) will require reagents, such as antibodies, that react with the avian proteins, since these proteins do not routinely crossreact with the available human and mouse antisera (data not shown) (52).
Radiation-induced Apoptosis in Normal and D1Ϫ/Ϫ, D1Ϫ/Ϫ ϩ Murine D1, D1Ϫ/Ϫ ϩ Murine D2, and D1Ϫ/Ϫ ϩ Murine D3 DT40 Cells-Considerable controversy exists regarding the possible role(s) of the D-type cyclins, particularly cyclin D1, in programmed cell death (73)(74)(75)(76)(77)(78). Evidence from several different cell lines in which cyclin D1 levels are manipulated, either by overexpression or antisense (including rodent neuronal, mouse mammary epithelial, NIH 3T3 mouse fibroblasts, p53 Ϫ/Ϫ mouse embryo fibroblasts, rat fibroblasts, and human HCE7 esophageal carcinoma), suggests that cyclin D1 expression is required for apoptosis (73,(75)(76)(77)(78). Conversely, others (32) have shown that the D-type cyclins function as growth factor sensors that can partially suppress apoptosis in myeloid, interleukin-3-dependent blood cells. Selective tissue atrophy and loss of proliferative capacity have also been observed in the retina and mammary epithelium of cyclin D1 Ϫ/Ϫ mice, consistent with the hypothesis that cyclin D1 acts as a tissue-specific growth factor sensor (49,50). In addition, overexpression of cyclin D1 in rodent fibroblasts, or even concomitant expression of both cyclins D1 and E, has not been associated with increased levels of programmed cell death (47,68,70). In a separate study, Sofer-Levi and Resnitzky (78) used a tetracycline-responsive promoter to direct expression of cyclin D1 in serum-starved rat fibroblasts. Induction of cyclin D1 protein expression in these serum-starved fibroblasts resulted in significant apoptosis. However, in the presence of normal serum, induction of cyclin D1 did not result in a significant increase in apoptotic cells. The availability of avian D1 Ϫ/Ϫ B-cells allowed us to examine this question directly.
UV and ␥ radiation-induced apoptosis occurred in a dose-dependent manner in both DT40 control and DT40 D1 Ϫ/Ϫ single cell clones (Table III). The percentage of cells undergoing programmed cell death resulting from 500 to 1,500 J/m 2 of UVC or ␥ radiation was substantially increased in DT40 D1 Ϫ/Ϫ cells (10 -15% increase after UVC exposure and a 15-25% increase after ␥ radiation; Table III). To determine whether this increased sensitivity to radiation, reflected as increased programmed cell death, was linked to the expression of D-type cyclins, we examined the effect(s) of ␥ radiation on D1 Ϫ/Ϫ cells FIG. 3. Northern blot analysis of cyclin D1 expression in DT40 cells and DT40 cyclin D1 ؉/؊ or ؊/؊ cells. Total RNA (50 g) from wild-type and cyclin D1 ϩ/Ϫ or Ϫ/Ϫ cells was electrophoresed on an agarose gel, transferred to nitrocellulose, and probed with a 32 P-labeled avian cyclin D1 probe. The 1.3-kb band in lanes 1, 2, 14, and 15 represents the avian cyclin D1 transcript. This band is reduced in intensity in clonal cell lines in which one allele of the cyclin D1 locus has been disrupted (lanes 3 and 16), and it is absent in clones that contain two disrupted cyclin D1 alleles (lanes 4 -13 and 17-21). The larger 6.0-kb transcript may represent cross-hybridization of the probe to the avian cyclin D2 mRNA.
in which murine D-type cyclins were expressed. These experiments involved the same D1Ϫ/Ϫ ϩD-type cyclin(s) cell lines described above, in which cell cycle abnormalities resulting from ablation of cyclin D1 were rescued by expression of any murine D-type cyclin. Expression of mouse cyclin D1 in DT40 Ϫ/Ϫ cells (Fig. 4, panel C, D1 Ϫ/Ϫ clone 1/57) eliminated the increased sensitivity to ␥ irradiation exhibited by the D1Ϫ/Ϫ cells (Table III). In fact, the percentage of apoptotic cells in the D1 Ϫ/Ϫ ϩ murine D1 cells was nearly identical to the parental cells. Similarly, when D1 Ϫ/Ϫ ϩ murine D2 or D1 Ϫ/Ϫ ϩ murine D3 cells were used (Fig. 4, panel D, D1 Ϫ/Ϫ clone 1 ϩ D2/D3), enhanced sensitivity to ␥ irradiation, as reflected by the number of apoptotic cells, was apparently eliminated (Table III). Intriguingly, expression of the mouse mutant cyclin D1 (missing the LXCXE motif) partially, but not completely, restored cellular resistance to radiation-induced apoptosis (Table  III). Since the increased sensitivity of the D1 Ϫ/Ϫ cells to radiation-induced apoptosis can be effectively reversed by expression of any normal D-type cyclin, it is likely that these proteins do, in fact, suppress apoptosis in DT40 cells. Furthermore, the ability of the mutant cyclin D1 protein to partially suppress radiation-induced apoptosis but not cell cycle abnormalities (Tables I and III) suggests that proliferating and apoptotic cells may contain distinct cyclin D1 complexes. Whether these complexes are composed of different polypeptides and/or distinct post-translational modifications is not known at this time. Further studies with the D1Ϫ/Ϫ DT40 cell lines will help to elucidate whether the observed defect in G 1 checkpoint control following radiation-induced DNA damage is due entirely to the disruption of normal p21 waf1/cip1 cyclin D1 interactions or inappropriate interactions between p21 waf1/cip1 and other cell cycle machinery in these cells. However, this response is consistent with data from p21 waf1/cip1 Ϫ/Ϫ mice, demonstrating that these mice are defective in G 1 checkpoint control (79), as well as the developmental abnormalities observed in D1 Ϫ/Ϫ mice (49,50) and the inactivation of cdks following UV irradiation (80). DISCUSSION The specific biologic functions of the three D-type cyclins remain somewhat of an enigma. Mice lacking cyclin D1 by gene targeting in embryonic stem cells have recently been generated, but the results of this study reflect a more generalized requirement of this particular cyclin during development than a specific requirement during the cell cycle (49,50). Furthermore, interpretation of the effect(s) of gene "knockouts" in animal models can be complicated by the expression of functionally redundant proteins that may mask the phenotype in certain tissues (81). We have now generated avian B-cell lines that lack cyclin D1 but retain normal cyclin D2 expression, which exhibit defects in normal cell cycle progression. The growth of the D1 Ϫ/Ϫ cells is slightly retarded (ϳ20% slower than wild-type DT40 cells), presumably due to a delay in G 2 /M phase. A similar delay in G 2 /M phase was observed when mouse embryo fibroblasts derived from the D1 Ϫ/Ϫ mice were analyzed (50). The cell cycle "defects" observed in the DT40 D1 Ϫ/Ϫ cells can be specifically rescued by the expression of normal murine cyclin D1, D2, or D3 but not by a mutant cyclin D1 protein that lacks the LXCXE pRb-binding motif (44). The fact that overexpression of the murine cyclin D3 protein, for which there is no apparent avian homologue, rescues the D1 Ϫ/Ϫ phenotype in DT40 cells strongly supports the hypothesis that the D-type cyclins have overlapping functions (3,4,82). Our results are also consistent with a specific requirement for cyclin D1 for normal transit through the cell cycle and that this requirement cannot be compensated for by the endogenous cyclin D2 protein. A normal cell cycle can only be acquired by exogenous expression of any one of the D-type cyclins. The molecular basis of this rescue is unknown at this time. One can speculate that the observed effect(s) on the cell cycle are due to inappropriate level(s) and/or association(s) between D-type cyclin interactors and other components of the cell cycle machinery (e.g. inappropriate levels of p21 waf1/cip1 , p27 Kip1 , INK4 or cdk4/6 proteins may affect the activity of other cyclins, such as cyclin E). Further biochemical analyses of these cells will require the production of appropriate antibody reagents for avian cyclins, cdks, and ckis.
We would suggest that the cell cycle perturbations observed in the cyclin D1 Ϫ/Ϫ cells might be due to 1) an insufficient "threshold" level of cyclin D1 required for entry into G 1 phase from mitosis or 2) defects in replication resulting from the absence of cyclin D1 that might trigger G 2 phase checkpoints and delay cells at this point until such damage is repaired. These two possible mechanisms are suggested since the most obvious outcome of cyclin D1 elimination, lengthening of G 1 phase, was not observed. The second possible explanation is supported by data from this study in which D1 Ϫ/Ϫ cells, blocked at the G 1 /S boundary by centrifugal elutriation and treatment with aphidicolin, are subject to a 2-3-h delay in progression through the cell cycle when released from this cycle block (Table II and Fig. 6). A delay in G 2 phase due to the activation of checkpoints before mitosis is triggered, possibly involving inappropriately associated cyclin-dependent protein kinase inhibitory (CKI) subunits, could potentially explain these observations. Whether the observed cell cycle phenotype and enhanced programmed cell death associated with these D1 Ϫ/Ϫ B-lymphoma cells will be found in other cell types, or is specific for this cell lineage, is not known at this time. Further study of these cell lines, as well as the murine cyclin D1 knockout models (49,50), should provide answers to these questions. The ability of any of the D-type cyclins to "rescue" D1 Ϫ/Ϫ DT40 cells suggests that these cyclins are somewhat redundant in function. However, we cannot completely rule out FIG. 5. Analysis of the growth rates of wild-type and cyclin D1 ؊/؊ clones. Equivalent numbers of cells (20,000) from wild-type DT40 (solid black box), a single cyclin D1 allele disruption, ϩ/Ϫ (top crosshatched box which is superimposed on the solid black box), and three independent cyclin D1 Ϫ/Ϫ clones (three progressively less shaded boxes) were seeded into 12-well cluster dishes. The number of viable cells was determined by using a hemacytometer at 24-h intervals. Each point represents the data averaged from three independent dishes. the possibility that individual D-type cyclins have specific function(s). Additional analysis of the cyclin D1 Ϫ/Ϫcell lines may help to resolve the exact relationship(s) between the various D-type cyclins and the proteins they are associated with in vivo (e.g. the CDKs, CDKIs, pRb/p107/p130 and c-Myc (29, 83, 84)). A separate, but perhaps equally important issue, has arisen regarding the possible role of the D-type cyclins in apoptosis. The availability of D1 Ϫ/Ϫ cells has allowed us to examine the possible relationship between cyclin D1 expression and the induction of apoptosis. Several studies have linked induction of cyclin D1 gene expression in post-mitotic neurons and senescent fibroblasts to programmed cell death (73,75). In addition, cyclin D1 overexpression in specific tumor cell lines results in enhanced apoptosis by selective agents (77). Finally, two different studies have been published concerning cyclin D1-induced apoptosis in cultured cells (75,78). In cultured neuronal cells, cyclin D1 overexpression, in the presence of normal growth factors (10% serum), was capable of inducing apoptosis (75). However, in cultured fibroblasts, cyclin D1 overexpression induced apoptosis only in the absence of growth factors (0.1% serum) (78). It should be noted that the human cyclin D1 gene, CCND1, was originally identified by one group as the PRAD1 oncogene residing at the breakpoint of a chromosome 11 inversion in parathyroid adenomas (85). As a consequence, these tumor cells express very high levels of apparently normal cyclin D1 mRNA and protein. Thus, it is somewhat paradoxical that elevated expression of cyclin D1 can be so well tolerated in these tumor cells, whereas its overexpression in some transformed cell lines enhances apoptosis (73,75,85). In fact, these results parallel what has been observed with c-myc, an oncogene that also induces apoptosis when overexpressed in serumdeprived cells but not when normal growth factor levels are maintained (85,86).
Our results strongly support the role of cyclin D1 as a positive growth regulator that suppresses apoptosis in proliferating cells but whose inappropriate expression in factor-deprived cells may trigger cell death by default. Independent studies of the effects of DNA damage (e.g. UV and/or ␥ irradiation) on the cell cycle machinery support the notion that D-type cyclins normally suppress apoptosis (80). In response to UV irradiation, the activity of various CDKs was inhibited in a p53-dependent manner by enhanced expression of CDK inhibitors (e.g. p21 Waf1/Cip1 ), as well as phosphorylation of Thr-14 and Tyr-15 of p34 cdc2 . Inhibition of CDK activities in response to radiation apparently allows cells to either repair their DNA damage, by suspending progress through the cell cycle until the damage is repaired, or, alternatively, leads to their elimination by apoptosis if the damage is too extensive. The enhanced level of programmed cell death observed in the DT40 D1 Ϫ/Ϫ cell lines in response to radiation-induced DNA damage, as well as the ability of exogenously expressed murine D-type cyclins to desensitize these cells, suggests that cyclin D1 may suppress   apoptosis. Once again, we do not know whether this is a cell type-specific effect. Further studies of the DT40 D1Ϫ/Ϫ cells could provide valuable insight into how cyclin D1 functions as a mediator of apoptosis and help determine whether this reflects changes in cyclin D1 complexes with other proteins, such as CDKs, CDKIs, and pRb/p107/p130, as cells transit the cell cycle. In addition, it will now be of interest to determine whether elimination of cyclin D2, on its own or concomitantly with cyclin D1, has a similar effect(s) on the cell cycle and/or programmed cell death.
Finally, these studies demonstrate, persuasively, that the DT40 cell line is amenable to easy and rapid manipulation of specific genes that regulate the cell cycle and/or programmed cell death and that the resulting cell lines can be valuable reagents for the analysis of the function of these genes.