Retinoic Acid Promotes Ubiquitination and Proteolysis of Cyclin D1 during Induced Tumor Cell Differentiation*

Mechanisms by which differentiation programs engage the cell cycle are poorly understood. This study demonstrates that retinoids promote ubiquitination and degradation of cyclin D1 during retinoid-induced differentiation of human embryonal carcinoma cells. In response to all-trans-retinoic acid (RA) treatment, the human embryonal carcinoma cell line NT2/D1 exhibits a progressive decline in cyclin D1 expression beginning when the cells are committed to differentiate, but before onset of terminal neuronal differentiation. The decrease in cyclin D1 protein is tightly associated with the accumulation of hypophosphorylated forms of the retinoblastoma protein and G1 arrest. In contrast, retinoic acid receptor γ-deficient NT2/D1-R1 cells do not growth-arrest or accumulate in G1 and have persistent cyclin D1 overexpression despite RA treatment. Notably, stable transfection of retinoic acid receptor γ restores RA-mediated growth suppression and differentiation to NT2/D1-R1 cells and restores the decline of cyclin D1. The proteasome inhibitor LLnL blocks this RA-mediated decline in cyclin D1. RA treatment markedly accelerates ubiquitination of wild-type cyclin D1, but not a cyclin D1 (T286A) mutant. Transient expression of cyclin D1 (T286A) in NT2/D1 cells blocks RA-mediated transcriptional decline of a differentiation-sensitive reporter plasmid and represses induction of immunophenotypic neuronal markers. Taken together, these findings strongly implicate RA-mediated degradation of cyclin D1 as a means of coupling induced differentiation and cell cycle control of human embryonal carcinoma cells.

Mechanisms by which differentiation programs engage the cell cycle are poorly understood. This study demonstrates that retinoids promote ubiquitination and degradation of cyclin D1 during retinoid-induced differentiation of human embryonal carcinoma cells. In response to all-trans-retinoic acid (RA) treatment, the human embryonal carcinoma cell line NT2/D1 exhibits a progressive decline in cyclin D1 expression beginning when the cells are committed to differentiate, but before onset of terminal neuronal differentiation. The decrease in cyclin D1 protein is tightly associated with the accumulation of hypophosphorylated forms of the retinoblastoma protein and G 1 arrest. In contrast, retinoic acid receptor ␥-deficient NT2/D1-R1 cells do not growtharrest or accumulate in G 1 and have persistent cyclin D1 overexpression despite RA treatment. Notably, stable transfection of retinoic acid receptor ␥ restores RA-mediated growth suppression and differentiation to NT2/ D1-R1 cells and restores the decline of cyclin D1. The proteasome inhibitor LLnL blocks this RA-mediated decline in cyclin D1. RA treatment markedly accelerates ubiquitination of wild-type cyclin D1, but not a cyclin D1 (T286A) mutant. Transient expression of cyclin D1 (T286A) in NT2/D1 cells blocks RA-mediated transcriptional decline of a differentiation-sensitive reporter plasmid and represses induction of immunophenotypic neuronal markers. Taken together, these findings strongly implicate RA-mediated degradation of cyclin D1 as a means of coupling induced differentiation and cell cycle control of human embryonal carcinoma cells.
The mechanisms coordinating the cell cycle and terminal differentiation have attracted intense interest (1, 2) because uncoupling these pathways may contribute to tumorigenesis, and pharmacological agents such as retinoids reverse or restore normal maturation control in certain premalignancies and overt malignancies (3). Mechanisms active in signaling alltrans-retinoic acid (RA) 1 effects involve the retinoid receptors: retinoic acid receptors (RARs) and retinoid X receptors (RXRs) (4,5). These ligand-dependent transcription factors alter expression of downstream species mediating retinoid biologic effects (6,7). Three RARs (RAR␣, RAR␤, and RAR␥) and three RXRs (RXR␣, RXR␤, and RXR␥) exist and have cell context-dependent expression patterns reflecting tissue-specific retinoid functions and developmental programs (8,9). The current study proposes that signaled degradation of cyclin D1 is one mechanism linking RA-induced terminal differentiation and cell cycle control in human embryonal carcinoma cells.
The predominant control of cell proliferation occurs during the G 1 phase of the cell cycle, as previously reviewed (15). Cyclin-dependent kinases (Cdks) play major roles in the regulation of eukaryotic cell cycle events, as previously reviewed (16). The cyclin D family of G 1 cyclins regulates activity of Cdk4 and Cdk6 (16). One major target of cyclin D1/Cdk complexes is the retinoblastoma (Rb) protein that is maintained in an underphosphorylated state through much of G 1 but becomes hyperphosphorylated in late G 1 , resulting in the release of sequestered transcription factors of the E2F family (15). Expression of cyclin-dependent kinase inhibitors can block S phase entry. The Cdk inhibitors are grouped into two classes (17). The Ink4 class contains p15, p16, p18, and p19, and the Kip class contains p21, p27, and p57 (17). Thus, the "Rb pathway" typified by cyclin D1/cdk4/p16/Rb constitutes a G 1 -S checkpoint that tightly controls cell cycle progression (18,19). The importance of this pathway is underscored by the clinical finding that one component of the pathway is deregulated or mutated in the majority of cancers (18).
Cyclin D1 is regulated by transcriptional and posttransla-tional mechanisms, and its degradation is required for cells to enter S phase (reviewed in Ref. 20). Cyclin D1 is targeted for rapid degradation involving the ubiquitin-dependent proteasome pathway (21)(22)(23)(24). Cyclin D1 protein rapidly degrades following RA-mediated growth suppression of human bronchial epithelial cells due to a proteasome-dependent process (23,24). This study indicates that cyclin D1 is a retinoid target during induced tumor cell growth suppression and terminal differentiation. A notable finding is that RA treatment markedly increases the formation of cyclin-D1 ubiquitin conjugates, which are substrates for proteasomal degradation. These data indicate that increased ubiquitination and instability of cyclin D1 link RA-mediated growth arrest and effective RA-mediated differentiation of NT2/D1 cells.

EXPERIMENTAL PROCEDURES
Cell Culture and Derivation of Cell Lines-The RA-sensitive human embryonal carcinoma cell NT2/D1 (a gift from Dr. Peter Andrews, Wistar Institute) is a clonal line derived from a xenograft of Tera-2 cells (25). Derivation and characterization of the NT2/D1-R1 cell line (an RA-resistant subclone of NT2/D1 cells) are described in detail elsewhere (11, 13, 14, 26 -28). NT2/D1-R1 cells stably transfected with episomal expression vectors containing RAR␣ or RAR␥ were previously described and were maintained in 200 g/ml hygromycin-containing media (13). NT2/D1 cells and clones derived and engineered from them were cultured in high glucose DME with 10% fetal bovine serum supplemented with penicillin, streptomycin, and glutamine under humidified 5% CO 2 . The RA induction protocols were as follows. On day Ϫ1, 1 ϫ 10 5 or 4 ϫ 10 5 cells were seeded into six-well plates or 10-cm dishes, respectively. On day 0, medium was replaced with medium containing 10 M RA or dimethyl sulfoxide (Me 2 SO) (1:1000 dilution). Cells could be treated for 6 days before reaching confluence. RA was stored under liquid N 2 in the dark as a 10 mM Me 2 SO stock solution.
Cell Cycle Analysis and Thymidine Incorporation Assay-For cell cycle phase analysis, cells harvested with trypsinization were stained with propidium iodide for 30 min at 37°C. The percentages of cells in G 1 , S, and G 2 /M were then determined on a Becton Dickinson FACscan flow cytometer using CellFit software and established techniques as described previously (26). Proliferation was measured by the thymidine incorporation method, as described (23,24).
Immunophenotypic Analysis-Indirect fluorescence-activated cell sorter (FACS) analysis to evaluate RA-induced neuronal differentiation of NT2/D1 cells was performed using established techniques (28). Briefly, NT2/D1 and NT2/D1-R1 cells were harvested independently by trypsinization and incubated with a monoclonal antibody to the cell surface antigen A2B5 or an isotype matched monoclonal control antibody. The A2B5 antibody recognizes a neuronal epitope in RA-treated NT2/D1 cells and was prepared from hybridoma cultures purchased from ATCC (29). Cells were indirectly assayed with fluorescein isothiocyanate-conjugated goat anti-mouse antibody, and fluorescence was measured as described (29). Mean peak fluorescent values and the percentage of positive cells were measured for the entire population. At least 2 ϫ 10 4 cells were analyzed per assay.
Western Analysis-Exponentially growing cells were lysed in a modified radioimmune precipitation buffer and analyzed by SDS-polyacrylamide gel electrophoresis as described (12,23,30). In some experiments, 5 mM N-ethylmaleimide (Sigma) was added to the lysis buffer to preserve ubiquitin conjugated proteins. Protein concentrations were determined using the Bradford technique. Cyclin D1 (M20) polyclonal antibody was purchased from Santa Cruz Biotechnology. HA antibody (16B12), a monoclonal antibody, was purchased from Babco. A monoclonal antibody specific for underphosphorylated forms of Rb (G99-549) was purchased from PharMingen.
Construction of Expression Vectors-Full-length murine cyclin D1 cDNA was excised with BamHI from vector pGEX-3x Cyl-1 and inserted into BamHI-and BglII-restricted pEGFP-C1 (CLONTECH) to create an in-frame fusion between GFP and the N terminus of cyclin D1. GFPcyclin D1 (T286A), in which threonine 286 is replaced by alanine, was constructed using a two-step polymerase chain reaction strategy with the above-described pGFP-cyclinD1 plasmid as template. The pRcCMV-cycD1-HA plasmid is an HA-tagged human cyclin D1 expression vector driven by the CMV promoter in pRC/CMV (Invitrogen) that was kindly provided by Dr. Steven Dowdy (Howard Hughes Medical Institute, Washington University School of Medicine). A CMV-driven Myc-Histagged ubiquitin expression plasmid, pCW7, was kindly provided by Dr.
Ronald Kopito (Stanford University, CA). The pRcCMV-cycD1(T286A)-HA plasmid was generated using the QuikChange site-directed mutagenesis system (Stratagene). DNA sequence analysis was used to confirm the identity of each construct.
Transient Transfection Reporter Assay-The FGF4 reporter assay was previously described (12,14). Briefly, cells were transfected with 5.0 g of DNA including 2.5 g of reporter plasmid, 1.5 g of expression plasmid, and 1.0 g of a CMV-driven ␤-galactosidase expression plasmid using a modified calcium phosphate precipitation technique. For expression of human wild-type or mutant cyclin D1, the HA-tagged expression plasmids pRcCMV-cycD1-HA and pRcCMV-cycD1(T286A)-HA were used, respectively. The FGF4-CAT construct contains a human RA-sensitive FGF4 enhancer placed within a CAT-containing reporter driven by the endogenous human FGF4 promoter (31). Cells (3 ϫ 10 5 ) were exposed to DNA precipitate for 14-16 h and then to a HEPES-buffered saline-glycerol solution for 45 s. Cells were then washed and cultured with or without RA for an additional 48 h before harvesting in isotonic buffer (150 mM NaCl, 40 mM Tris-HCl, pH 8, and 1 mM EDTA). CAT activity was measured as previously reported and was normalized to ␤-galactosidase activity (12,14).
In separate experiments, NT2/D1 cells were transfected with HAtagged plasmids encoding human wild-type cyclin D1 and mutant cyclin D1 (T286A). NT2/D1 cells were preincubated for 24 h with RA before transfection. After overnight incubation with DNA precipitate, cells were incubated with RA for an additional 24 h; during the last 16 h, cells were also exposed to calpain inhibitor I (LLnL) or vehicle at a concentration of 50 M. Cells were subsequently harvested for Western analysis using an antibody specific for HA.
GFP Fusion Experiments-Cells were transfected with 20 g of either pGFP-cyclin D1, pGFP-cyclin D1 (T286A), pEGFP-C1, or pBluescript SK, which was used as a control for autofluorescence. Cells were treated 24 h posttransfection with 10 M RA or vehicle control for an addition 4 days. Cells were then harvested and stained for expression of the immunophenotypic marker A2B5 as described above, except that goat anti-mouse R-phycoerythrin-labeled IgG (Kirkegaard & Perry Laboratories) at a dilution of 1:100 was used in place of the usual fluorescein isothiocyanate-conjugated goat anti-mouse secondary antibody. FACS analysis was performed on nonfixed cells that were dually gated for GFP and R-phycoerythrin fluorescence. Nuclei from a portion of the cells were isolated and stained with ethidium bromide for cell cycle analysis. At least 2 ϫ 10 4 cells were analyzed for each assay.

RESULTS
Terminal Differentiation in NT2/D1 Cells Causes G 1 Arrest in RA-sensitive but Not RA-resistant Cells-RA causes NT2/D1 cells to accumulate in the G 1 phase of the cell cycle with a concomitant decrease in S phase (Fig. 1A). Unlike NT2/D1 cells, NT2/D1-R1 cells failed to accumulate in G 1 after 6 days of RA treatment (Fig. 1A). To assess RA-induced changes in proliferation, NT2/D1 and NT2/D1-R1 cells were treated with 1 and 10 M RA or Me 2 SO vehicle control and assayed for thymidine incorporation (Fig. 1B). NT2/D1 cells treated with 10 M RA displayed a 60% decrease in S phase cells by day 2, as compared with vehicle control. A prominent induction of the neuroectodermal marker A2B5 appeared by day 4 of RA treatment of NT2/D1 cells but not NT2/D1-R1 cells (Fig. 1C). Complete morphologic differentiation, such as neuronal outgrowth, did not typically appear until approximately 3 weeks of RA treatment. As expected, NT2/D1-R1 cells did not display a decreased proliferation rate, despite RA treatment. The increase in thymidine incorporation over time was similar in RA and vehicle control treated NT2/D1-R1 cells (Fig. 1B). Morphologic and immunophenotypic maturation did not follow RA treatment of NT2/D1-R1 cells (Ref. 14 and Fig. 1C). These findings indicate that RA resistance in NT2/D1-R1 cells is associated with an aberrant G 1 -S cell cycle blockade.
RA Represses Cyclin D1 Expression before Onset of Terminal Differentiation of NT2/D1 but Not NT2/D1-R1 Cells-Expression of proteins known to regulate the G 1 phase of the cell cycle was examined. A marked increase in steady-state cyclin D1 levels in NT2/D1-R1 cells compared with NT2/D1 cells was noted ( Fig. 2A). The level of this overexpression as compared with parental NT2/D1 cells ranged from 3-to 7-fold in independent analysis. Fig. 2B compares the time course of cyclin D1 expression in NT2/D1 and NT2/D1-R1 cells after RA treatment. Cyclin D1 expression markedly declined in NT2/D1 cells within 2 days of 10 M RA treatment. This is well before onset of expression of early immunophenotypic differentiation markers (Fig. 1C) but within the time frame when commitment to differentiation occurred. In contrast, RA fails to signal a decline of cyclin D1 in NT2/D1-R1 cells. NT2/D1 and NT2/D1-R1 had similar basal and post-retinoid treatment levels of cyclin A, cyclin E, cyclin D2, Cdk-4, Cdk-6, Rb, p27, and p16 (data not shown). Cyclin D3 protein was not detected in NT2/D1 cells.
Down-regulation of Cyclin D1 Is Tightly Linked to RA-Induced G 1 Arrest and Hypophosphorylated Rb Accumulation in NT2/D1 Cells-NT2/D1 and NT2/D1-R1 cells were exposed to varying dosages of RA and simultaneously assayed by immunoblot and cell cycle analyses. RA treatment of NT2/D1 cells resulted in a dose-dependent decline in cyclin D1 protein (Fig.   3A). In contrast, cyclin D1 expression was maintained at much higher levels than in NT2/D1R1 cells. Even at the highest doses of RA (10 M) cyclin D1 expression in NT2/D1-R1 cells was comparable to levels detected in untreated NT2/D1 cells. Persistent cyclin D1 expression in NT2/D1-R1 cells is observed after extended (up to 4 weeks) exposure to RA. The decline in cyclin D1 expression coincided with an RA dose-dependent G 1 arrest in NT2/D1 cells (Fig. 3B). G 1 arrest was not evident in RA-treated NT2/D1-R1 cells (Fig. 3B).
RA-treated lysates were probed with an antibody recognizing hypophosphorylated Rb. The decline in cyclin D1 expression after RA treatment of NT2/D1 cells was tightly correlated with a dose-dependent accumulation of hypophosphorylated Rb. In contrast, the levels of hypophosphorylated Rb were negligible in NT2/D1-R1 cells (Fig. 3C). This is expected because NT2/D1-R1 cells can be continuously passaged in RA without induction of cell cycle arrest. This suggests that, in this cell context, cyclin D1 expression is rate-limiting in regulating Rb-specific Cdk activity.
RA-induced Decline in Cyclin D1 Is Mediated by a Ubiquitindependent Process-Treatment with the proteasome inhibitor LLnL reversed the decline in cyclin D1 protein after RA treatment of NT2/D1 cells, as shown in Fig. 5A. Diehl et al. (21) have reported that phosphorylation of threonine 286 signals ubiquitination and degradation of cyclin D1 in response to mitogen withdrawal. NT2/D1 cells were transiently transfected with an ubiquitin expression plasmid and either HA-tagged wild-type cyclin D1 or cyclin D1 in which threonine 286 was replaced with alanine (T286A). Cells were then exposed to RA in the presence or absence of LLnL. As shown in Fig. 5B, slow migrating cyclin D1 species were evident in wild-type transfectants treated with RA and LLnL, consistent with appearance of  1 and 2) and NT2/D1-R1 cells (lanes 3 and 4). B, Western analysis depicting the time course (in days) of the RA-mediated decrease in cyclin D1 expression in NT2/D1 cells and persistent overexpression of cyclin D1 in NT2/D1-R1 cells, despite 10 M RA treatment.
higher ubiquitinated forms. This effect was dependent on threonine 286 because higher molecular weight forms were not detected in cyclin D1 (T286A) mutant transfectants treated with LLnL and RA. This finding demonstrates that RA accelerates formation of ubiquitinated forms of cyclin D1 (Fig. 5B,  compare lanes 1 and 3). Other mechanisms in addition to proteolysis may contribute to regulation of cyclin D1 by RA. A slight decline in cyclin D1 mRNA was evident in NT2/D1 cells as assayed by Northern and reverse transcription-polymerase chain reaction analyses (data not shown). This occurred rela-tively late, 72 h after RA treatment, suggesting that this effect is secondary to G 1 arrest, decreased cell proliferation, and commitment to differentiation, which occur within 48 h. Interestingly, levels of cyclin D1 transcripts were lower in NT2/ D1-R1 cells compared with parental NT2/D1 cells (data not shown), despite substantial overexpression of cyclin D1 protein in NT2/D1-R1 cells. The extent to which transcriptional control contributes to RA-mediated decline of cyclin D1 protein is a subject for future work.
The Effect of Cyclin D1 on the Differentiation State-dependent FGF4 Reporter Assay-We previously reported the adaptation of a transient transfection reporter assay to rapidly assess differentiation signals in NT2/D1 cells (12,14). This assay uses a construct containing human FGF4 promoter-enhancer elements linked to the CAT reporter gene (31,32). The decline of FGF4 mRNA and protein expression reflects effective RA-induced differentiation and growth suppression of murine and human embryonal carcinoma cells and is "downstream" of direct retinoid receptor-driven transactivation in this differentiation program (10,12,27,32). Fig. 6A demonstrates that at low RA dosages, cyclin D1 transfection prevented the decline in FGF4 reporter activity, as compared with an insertless control vector. This cyclin D1 transfection effect was not observed at higher RA dosage. Because this RA dose-dependence may relate to degradation of cyclin D1 by RA, additional experiments were performed utilizing the cyclin D1 (T286A) mutant. This substitution has no effect on the ability of cyclin D1 to act as a regulatory subunit of Cdks (21). Fig. 6B demonstrates that as compared with wild-type cyclin D1 transfectants, cyclin D1 (T286A) transfectants retained FGF4-reporter activity at high RA dosages. Western analysis of the cell lysates reveals high cyclin D1 expression was achieved after RA treatment only FIG. 3. Down-regulation of cyclin D1 is tightly linked to RAinduced G 1 arrest and hypophosphorylated Rb accumulation in NT2/D1 cells. Cells were exposed to the indicated RA dosages for 4 days, harvested, and subdivided for each analysis. A, Western analysis of the dose-dependent decrease in cyclin D1 expression in NT2/D1 cells. NT2/D1-R1 cells expressed substantial levels of cyclin D1 even at the highest RA dosages. B, NT2/D1 cells exposed to RA arrest in the G 1 phase of the cell cycle in a dose-dependent fashion with a corresponding decrease in S phase cells. RA failed to induce an accumulation of the G 1 phase in NT2/D1-R1 cells, despite RA treatment. Cells treated as above were staining with propidium iodide for flow cytometry analysis. C, Western analysis of RA-treated NT2/D1 and NT2/D1-R1 cells was performed with an antibody specific for the underphosphorylated forms of Rb. NT2/D1 cells exhibit an RA dose-dependent increase in hypophosphorylated Rb. NT2/D1-R1 cells expressed little hypophosphorylated Rb, and these low levels did not appreciably change at all examined RA dosages. with the mutant cyclin D1 (T286A) (data not shown).
Effects of Transient Overexpression of Cyclin D1 on Immunophenotypic Differentiation-NT2/D1 cells were transfected with GFP vector alone, GFP-cyclin D1, GFP-cyclin D1 (T286A), or control DNA and exposed to 10 M RA for 4 days. Cells were then simultaneously analyzed for GFP fluorescence and expression of the established immunophenotypic differentiation marker A2B5 by FACS analysis. Nuclei from a portion of the samples were isolated and stained with ethidium bromide for cell cycle analysis of the entire population of cells. Results are displayed in Table I. The GFP vector alone had no observed effect on RA-mediated G 1 arrest or immunophenotypic differentiation of NT2/D1 cells when compared with control DNA or mock-transfected cells (data not shown). In the presence or absence of RA, cells transfected with wild-type cyclin D1 exhibited little GFP fluorescence. In contrast, a relatively strong GFP signal was seen in cyclin D1 (T286A) transfectants, despite RA treatment. A much lower percentage of cyclin D1 (T286A) transfectants exhibiting GFP fluorescence also expressed A2B5 after RA treatment, as compared with cells transfected with GFP alone. This suggests that forced overexpression of cyclin D1 blocks or delays RA-mediated maturation of these cells. Notably, the entire population of cyclin D1 (T286A)-transfected cells exhibited a diminished percentage of cells in the G 1 phase of the cell cycle after RA treatment as compared with GFP control cells. This is consistent with the established ability of cyclin D1 to accelerate the G 1 -S transition when overexpressed (33,34). This effect is likely greater because a mixed population of transfected and untransfected cells were analyzed. Taken together with the FGF4 reporter results, these data are consistent with the view that forced overexpression of cyclin D1 blocks typical RA-mediated differentiation responses in NT2/D1 cells.

DISCUSSION
Studies of tumor differentiation mechanisms are clinically relevant because an increased understanding of tumor differentiation programs offers strategies for improved tumor diagnosis, classification, or therapy. A common event in terminal differentiation is G 1 cell cycle arrest; however, the precise mechanisms governing this process are not fully understood (2). Recent work indicates that cell cycle components regulating the G 1 -S transition actively promote or inhibit cell contextspecific differentiation programs (reviewed in Refs. 35 and 36). This in turn implies that differentiation programs recruit specific cell cycle components. The current study proposes that one such coupling mechanism active during RA-induced terminal differentiation of human embryonal carcinoma cells is the RAmediated degradation of cyclin D1.
RA-signaled decline of cyclin D1 is an early retinoid receptormediated event during differentiation of NT2/D1 cells. As compared with NT2/D1 cells, RA-resistant NT2/D1-R1 cells overexpress cyclin D1 before and after RA treatment. These cells also fail to exhibit a decline in cell proliferation, to differentiate, or to G 1 arrest. RAR␥ activation is necessary and required for RA-mediated differentiation in NT2/D1 cells (12)(13)(14). The current finding that restoring RAR␥ expression to retinoid-resistant NT2/D1-R1 cells also restores a decline in cyclin D1 levels supports the conclusion that the decline in cyclin D1 is mediated by, and is downstream of, retinoid receptor activation of these cells.
This decline appears to be in large measure due to accelerated ubiquitin-dependent degradation of cyclin D1 by the 26 S proteasome, although transcriptional regulation may also contribute to this regulation. The levels of many G 1 regulators are controlled by ubiquitin-mediated proteolysis, including cyclin D1, cyclin E, p53, and p27 (21,22,(37)(38)(39)(40)(41)(42)(43). In all examined cases, site-specific phosphorylation targets these proteins for ubiquitination, linking protein stability to extracellular and intracellular signals (reviewed in Refs. 44 and 45). We previously reported a similar mechanism is active in human bronchial epithelial cells (23,24). We propose this as a general mechanism by which RA signals cell cycle control. It is interesting that in immortalized or transformed human bronchial epithelial cells, RA signals primarily through RAR␤, not RAR␥ (24,46). Thus, it appears that different retinoid receptors are recruited to regulate a common cell cycle component, cyclin D1. It would be interesting to ascertain whether this pathway is also active in HL60 or NB4 cells, in which RAR␣ (47) and PML-RAR␣, respectively, have predominant roles.
A notable finding was the ability of RA treatment to markedly increase the ubiquitination of cyclin D1. The precise mech-  6. Cyclin D1 antagonizes RA-mediated repression of FGF4 transcription. A, the effect of cyclin D1 on the CAT reporter activity of the FGF4 promoter-enhancer in NT2/D1 cells. An HA-tagged expression vector containing wild-type cyclin D1 or the insertless expression vector were co-transfected with the FGF4-CAT plasmid in NT2/D1 cells in the presence or absence of the indicated RA dosage. Each point is the average of three independent determinations. Error bars are S.D. B, cyclin D1 (T286A) is more effective than wild-type cyclin D1 in preventing the decline in FGF4 reporter activity. HA-tagged wild-type cyclin D1 or cyclin D1 (T286A) expression vectors were co-transfected with the FGF4-CAT plasmid in NT2/D1 cells in the presence or absence of RA at the indicated dosages. Each point is the average of three independent determinations. All results were corrected for ␤-galactosidase as described under "Experimental Procedures." anism by which retinoids promote cyclin D1 ubiquitination and degradation is not fully elucidated. The differential effects on ubiquitination and stability of the mutant cyclin D1 (T286A) suggest that this residue is important. It will be interesting to determine whether RA induces the ubiquitination of other proteins including other regulators of the G 1 -S transition or whether this effect is specific for cyclin D1. Future work will clarify how RA promotes cyclin D1 degradation and whether RA affects either threonine 286 phosphorylation or ubiquitin conjugation. Recently, glycogen synthase kinase-3␤ was reported to phosphorylate cyclin D1 at threonine 286, triggering rapid cyclin D1 turnover (22). Glycogen synthase kinase-3␤ is negatively regulated by mitogen stimulation through the RAS 3 PI3K 3 AKT pathway and may contribute to increased levels of cyclin D1 upon serum addition (48 -50). It will be interesting to determine whether RA affects glycogen synthase kinase-3␤ activity. However, because the present studies were performed under serum replete conditions and direct links between retinoid signaling and RAS signaling pathways are not yet established, the involvement of this kinase during RAmediated cyclin D1 degradation remains a speculation.
The current study demonstrates that forced cyclin D1 overexpression blocks or delays maturation responses of NT2/D1 cells. Components of the cell cycle regulated during differentiation vary in a cell context-specific manner (51,52). Manipulating G 1 cell cycle regulators alters induced differentiation programs (reviewed in Refs. 51 and 52). Cyclin D2 or cyclin D3 inhibits neutrophil differentiation in response to granulocyte colony-stimulating factor (53). Ectopic expression of cyclin D1 abrogates myogenesis mediated by MyoD (54). G 1 arrest may simply be a requirement for induced differentiation. However, regulators of the G 1 -S transition may play an active role in regulating tissue-specific gene expression. Underphosphorylated Rb interacts directly with MyoD and CCAAT/enhancerbinding protein to increase the transcription of tissue-specific genes during myoblast and adipocyte differentiation, respectively (55)(56)(57)(58). The mechanism by which Rb is recruited during myoblast and adipocyte differentiation is unclear, in part because G 1 arrest is induced by cell contact inhibition and mitogen withdrawal as part of the standard induction protocol in these systems. An important distinction in the current model is that RA induction is performed on subconfluent, log phase cells maintained in replete growth conditions. Thus, RA is able to signal cyclin D1 decline, G 1 arrest, and terminal differentiation in the presence of a full complement of growth factors and mitogens.
In summary, the findings presented here strongly implicate that RA-mediated degradation of cyclin D1 coordinates induced differentiation and cell cycle arrest of human embryonal carcinoma cells and perhaps other tumor cells. This mechanism depends on ubiquitination and is downstream of retinoid receptor activation. This may represent a general mechanism by which retinoids signal cell cycle control that may have therapeutic implications in the pharmacologic triggering of growth suppression and maturation of tumor cells.