CCAAT/Enhancer-binding Protein- a Cooperates with p21 to Inhibit Cyclin-dependent Kinase-2 Activity and Induces Growth Arrest Independent of DNA Binding*

CCAAT/enhancer-binding protein- a (C/EBP a ) is a basic leucine zipper protein that controls transcription of genes important for liver function, white adipose tissue development, and granulocyte differentiation. In addition to its function in controlling gene expression in differentiated tissues, C/EBP a is also associated with an antimitotic activity. We have previously demonstrated that C/EBP a interacts with p21, a cyclin-dependent kinase (CDK) inhibitor, and that C/EBP a inhibits proliferation when expressed in several different cell types (Timchenko, N. A., Harris, T. E., Wilde, M., Bilyeu, T. A., Burgess-Beusse, B. L., Finegold, M. J., and Darlington, G. J. (1997) Mol. Cell. Biol. 17, 7353–7361). Here we define the regions of C/EBP a required for interaction with p21 and demonstrate that CDK2 also interacts with C/EBP a . We show that C/EBP a can cooperate with p21 to inhibit CDK2 activity in vitro . The effect of C/EBP Cell Culture, and Recombinant Proteins— The full-length human C/EBP a coding sequence was cloned into pGEX-4T-1, producing an amino-terminal GST tag and a carboxyl-terminal HA tag. GST*

CCAAT/enhancer-binding protein-␣ (C/EBP␣) 1 is a member of the basic leucine zipper protein family of transcription factors and is expressed in several terminally differentiated tissue types such as liver, fat, and lung and in certain hematopoietic cells (1). C/EBP␣ regulates the expression of genes involved in gluconeogenesis, including phosphoenolpyruvate carboxykinase and glucose 6-phosphatase, in the liver (2). It is also involved in the regulation of uncoupling protein-1 in brown adipose tissue (3) and of the granulocyte colony-stimulating factor receptor during granulocyte differentiation (4). C/EBP␣-deficient animals exhibit an absence of subcutaneous white adipose tissue and granulocytes and typically die within a few hours after birth due to hypoglycemia (3,4). The C/EBP␣deficient animals demonstrate that this transcription factor plays a key role in tissue-specific gene expression and differentiation.
Early attempts to study the transcriptional activity of C/EBP␣ in cell culture revealed that C/EBP␣ overexpression leads to cell cycle arrest (5,6). Subsequent investigations by several laboratories have suggested that there are multiple pathways by which C/EBP␣ causes growth arrest (6 -9). Recent work has shown that C/EBP␣ interacts directly with a number of components of the cell cycle such as Rb, p107, and p21 (8,10,11). It has been suggested that C/EBP␣ requires interaction with Rb to bring about differentiation of adipocytes (10). C/EBP␣ may also directly interact with important regulatory proteins to bring about cell cycle arrest (8,11). In particular, C/EBP␣ directly interacts with p21 WAF1/CIP1/sdi1 (8). p21, p27 KIP1 , and p57 KIP2 are potent inhibitors of cyclin-dependent kinases (CDKs), the enzymes that are crucial for the orderly progression of the cell cycle (12). Paradoxically, these CDK inhibitors have also been shown to activate CDKs by an increase in the rate of CDK-cyclin assembly (13). The carboxyl terminus of p21 directly binds to and inhibits proliferating cell nuclear antigen-dependent DNA polymerase processivity (14). This ability is not shared by p27. Finally, there are several reports that the inhibitory activity of p21 for CDK2 can be modulated by exogenous protein binding to the carboxyl terminus. These proteins include the human papilloma virus (HPV)-16 E7 gene product (15), the SET gene product (16), and TOK-1␣ (17).
In this study, we show that the carboxyl-terminal region of p21 binds to C/EBP␣ and that this site of interaction is important for a C/EBP␣-dependent enhancement of CDK2 inhibition by p21. We also show that C/EBP␣ interacts with CDK2 and that C/EBP␣ mutants incapable of enhancing p21 activity for CDK2 are also defective in growth arrest. We demonstrate that the ability of C/EBP␣ to bind DNA is not required for C/EBP␣mediated growth arrest. We propose that C/EBP␣ interacts specifically with p21 and CDK2 to directly inhibit the CDK enzymatic activity required for cell cycle progression. Our findings lend support for the role of p21 as an "adaptor" protein with its effects on CDK activity being modulated by additional proteins.
Cell Culture and Transfection Protocols-All cell lines were maintained in 3:1 minimal essential media to Waymouth's media plus 10% fetal calf serum, and HT-1 cells were induced with either 10 mM isopropyl-␤-D-thiogalactopyranoside or glucose as previously described (7). Transient transfections for growth arrest assays were performed using Fugene (Roche Molecular Biochemicals) following the manufacturer's recommendations. All growth arrest studies were performed in the 293 cell line and confirmed in both Hep3B2 cells and immortalized mouse embryo fibroblasts (22).
Isolation of Recombinant Proteins-GST-tagged, bacterially expressed proteins were purified using glutathione beads (Amersham Pharmacia Biotech) according to the manufacturer's instructions. The recombinant proteins were expressed in either BL21(DE3) LysS or TG-2 cells. The bacterial pellet was lysed in phosphate-buffered saline and 0.5% Nonidet P-40 plus protease inhibitor tablets (Roche Molecular Biochemicals) using a Branson Model 450 sonifier. The proteins were incubated with ϳ25 l of glutathione beads/100 ml of culture, washed three times with 1 M NaCl and three times with 50 mM Tris (pH 8.0), and eluted with 50 mM glutathione. The eluate was dialyzed against phosphate-buffered saline and concentrated using Centricon columns (Amicon, Inc.). Proteins were stored in 50 mM Tris (pH 8.0), 20% glycerol, 1 mM dithiothreitol, 1 mM EDTA, and protease inhibitors. Protein concentration was determined by Coomassie Blue (Sigma) staining relative to an albumin standard (Sigma).
CDK Activity Assays-For CDK activity assays performed with recombinant CDK2-cyclin E purified from Sf9 cells (23), ϳ2-4 ng of CDK2 was used per reaction. To this were added the recombinant proteins diluted in 20 l of kinase buffer (50 mM HEPES (pH 7.5), 2.5 mM EGTA, 10 mM MgCl 2 , and 1 mM dithiothreitol), and the reaction was allowed to incubate on ice for 15 min. 2.5 g of histone H1 (HH1), 1 Ci of [␥-32 P]ATP, and 10 M unlabeled ATP were added in kinase buffer, and the reaction was incubated at 37°C for 30 min. The reaction was stopped by the addition of 5 l of Laemmli sample buffer and boiling. The samples were resolved on a 10% SDS-acrylamide gel and transferred to nitrocellulose membrane. Phosphorylated HH1 was visualized by autoradiography, and the membranes were subsequently probed with antibodies. Approximately 500 g of 293 cell nuclear extract was used for the immunoprecipitation of CDK2, which was then divided into 10 reactions for individual kinase assays. 1 g of antibody and 10 l of protein A-agarose were used to immunoprecipitate CDK2. After immunoprecipitation, the complexes were washed three times and resuspended in kinase buffer. Kinase reactions were then performed as described above. All assays were phosphorimaged using the Storm Model 860 PhosphorImager (Molecular Dynamics, Inc.) and quantitated using ImageQuant.
GST Pull-down Assays and Co-immunoprecipitations-GST pulldown assays were performed by incubating bacterial lysate containing ϳ1 g of GST-tagged protein with 14 l of glutathione-Sepharose 4B (Amersham Pharmacia Biotech) in pull-down buffer (phosphate-buffered saline containing 0.5% Nonidet P-40, 1 mM dithiothreitol, and protease inhibitors). After 30 min of incubation at 4°C, the mixture was washed and resuspended in 500 l of pull-down buffer, and to this were added either bacterially expressed proteins or in vitro translated and [ 35 S]methionine-labeled proteins produced using the TNT T7-coupled reticulocyte lysate kit (Promega). After a 2-h incubation at 4°C, the Sepharose beads were washed extensively with pull-down buffer. The Sepharose beads were then resuspended in Laemmli buffer, and the proteins were resolved on an SDS-acrylamide gel and transferred to nitrocellulose membrane. Proteins interacting with the GST-tagged proteins were visualized by Western blotting or directly by autoradiography of 35 S-labeled proteins. To verify equal loading of the GST-tagged proteins in the pull-down assays, the membrane was subsequently stripped and stained with Coomassie Blue.
Co-immunoprecipitation of C/EBP␣ and CDK2 was carried out using extracts from 293 cells containing C/EBP␣-HA, expressed from the transfected pcDNA3.1C/EBP␣-HA vector. One 10-cm plate was harvested for nuclear extracts (7) and used for each immunoprecipitation reaction. 1 g of antibodies was attached to protein A-agarose beads by incubation at 4°C for 30 min. After washing, the nuclear extract plus Nonidet P-40 and phosphate-buffered saline was added to give a final concentration of 0.5% Nonidet P-40 and 140 mM NaCl in 500 l. Following a 2-h incubation at 4°C, the agarose beads were washed extensively and resuspended in Laemmli buffer.
Immunofluorescence-Cells were seeded onto four-chamber plasticcoated slides (LabTek). For staining, cells were washed and then fixed in 4% formaldehyde overnight at 4°C, permeabilized in 0.5% Triton X-100 for 30 min at room temperature, and then blocked in 5% dry milk. Cells were stained with a 1:50 dilution of mouse anti-BrdUrd antibodies (Roche Molecular Biochemicals) and 1:200 dilution of rabbit anti-C/ EBP␣ antibodies (Santa Cruz Biotechnology, Inc.) for 1 h at 37°C in the presence of 200 units/ml exonuclease III and 0.02 units/ml DNase I. Slides were then washed and incubated with a 1:500 dilution of antimouse antibodies conjugated to fluorescein isothiocyanate and anti-rabbit antibodies conjugated to tetramethylrhodamine B isothiocyanate.

Regions of Interaction between C/EBP␣ and p21-It
has been previously demonstrated by co-immunoprecipitation studies and by mammalian two-hybrid assays that C/EBP␣ interacts with p21 (8). The amino terminus of p21 contains cyclinand CDK-binding sites and is sufficient for inhibition of CDK activity (24,25). The carboxyl terminus of p21 contains a binding site for proliferating cell nuclear antigen and cyclins ( Fig.  1A) (14,21,26,27). GST fusion proteins were constructed as shown in Fig. 1A. To identify interaction sites, in vitro translated and [ 35 S]Met-labeled human C/EBP␣ was incubated with full-length GST-p21, and GST-p21 deletion mutants bound to glutathione beads (Fig. 1B). Full-length p21 was found to interact with C/EBP␣ (Fig. 1B, lane 1). However, a truncated p21 mutant lacking the carboxyl terminus, p21-(1-71), did not (lane 2). Thus, C/EBP␣ binds to the carboxyl terminus of p21. To verify that C/EBP␣ binds to p21 directly and is not mediated through an unknown protein present in the reticulocyte lysate used for the in vitro translation reactions, bacterially expressed proteins were used in a GST pull-down assay (Fig. 1C). GST (Fig. 1C, lanes 1 and 2), GST-tagged p21 (lanes 3 and 4), and GST-tagged p21 missing the last 20 amino acids (p21-(1-144)) (lanes 5 and 6) were attached to glutathione beads and incubated with His-tagged C/EBP␣ alone (lanes 1, 3, and 5) or with His-tagged CDK2 (lanes 2, 4, and 6). C/EBP␣ bound only to full-length p21 and not to GST or GST-p21-(1-144) (lanes 3 and 4). Both full-length p21 and p21-(1-144) were able to bind to CDK2 with equal efficiency (lanes 4 and 6). Thus, C/EBP␣ binds within the last 20 amino acids of p21. This experiment also shows that the interaction of C/EBP␣ with p21 is likely to be direct, due to the absence of associated cellular proteins in this assay. C/EBP␣ interacts with a region of p21 that overlaps or is in close proximity to the proliferating cell nuclear antigenbinding site of p21 (14,26,28). A weak cyclin-binding site in p21 has been described to encompass amino acids 153-159 (27). The carboxyl-terminal region of p21 has also been shown to interact with several proteins, including SET (16), calmodulin (29), c-Myc (30), TOK-1␣ (17), and the HPV16 E7 protein (15,31).
Experiments to define the regions of C/EBP␣ capable of interaction with p21 were carried out using in vitro translated p21 incubated with GST-tagged C/EBP␣ deletion mutants. Fig.  1D outlines the relevant C/EBP␣ deletion mutants used in this analysis. In Fig. 1E, both full-length C/EBP␣ (lane 1) as well as a C/EBP␣ mutant containing only the first 226 amino acids (lanes 2 and 6) were shown to interact with p21 as compared with GST alone (lanes 3 and 8) or C/EBP␣-(1-119) (lane 5). A C/EBP␣ mutant containing only the leucine zipper was also shown to interact with p21 (Fig. 1E, lane 9), suggesting the existence of two independent binding sites for p21 within C/EBP␣. One interaction site was localized to the leucine zipper region (amino acids 313-360), whereas the other site was found to reside between amino acids 119 and 226 in activation domain II (32). For both in vitro translated p21 binding to GST-C/EBP␣ and in vitro translated C/EBP␣ binding to GST-p21, the amount of protein bound typically fell within the range of 2-4%. This is similar to previously reported levels of binding between the carboxyl terminus of p21 and other proteins (15,17,31).
C/EBP␣ Interacts with CDK2-The ability of C/EBP␣ to directly interact with other cell cycle components was investigated. A panel of GST-tagged C/EBP␣ mutants were tested for their ability to bind bacterially expressed CDK2. As shown in Fig. 2A, C/EBP␣ was found to specifically interact with CDK2 through two different regions. An amino-terminal region of C/EBP␣ located between amino acids 119 and 160 as well as a site within the basic region located between amino acids 280 and 313 both demonstrated binding to CDK2 (Fig. 2B). The amino-terminal binding site for CDK2 in C/EBP␣ is contained within the region defined for p21 interaction, and further definition of interaction sites will be required to determine whether they are different for p21 and CDK2. The second site of interaction for CDK2 is clearly distinct from that for p21. The basic region of C/EBP␣ interacts with CDK2 and not with the leucine zipper, as was seen for p21, thus demonstrating that the carboxyl terminus of C/EBP␣ contains separate binding sites for CDK2 and p21.
To test the in vivo interaction of C/EBP␣ with CDK2, coimmunoprecipitation studies were performed. C/EBP␣ containing a carboxyl-terminal HA tag was transiently transfected into 293 cells. After 48 h, the cells were harvested, and CDK2 was immunoprecipitated from nuclear extracts. The associated complexes were assayed by Western blotting for the HA tag present on C/EBP␣-HA. As shown in Fig. 2C, C/EBP␣-HA was co-immunoprecipitated with CDK2 (lane 1), but not with anti-p19 antibodies (used as a negative control) (lane 2). When no C/EBP␣-HA was transfected (lane 3), no associated proteins were detected by anti-HA antiserum. These observations show that C/EBP␣ is capable of interaction with CDK2 in cells.
Inhibition of CDK2 Activity by C/EBP␣ Requires the Interaction Sites for p21-p21-binding proteins such as HPV16 E7 (15,31) and SET (16) have been shown to affect CDK2 activity by reversing the inhibitory activity of p21 for CDK2, whereas TOK-1␣ is capable of promoting the inhibitory activity of p21 for CDK2 (17). It has been previously reported that C/EBP␣ controls the activity of CDK2 in vivo (33). Because C/EBP␣ interacts with both p21 and CDK2, we tested whether or not recombinant C/EBP␣ influences p21-mediated inhibition of CDK2 activity in vitro.
Purified recombinant CDK2-cyclin E from baculovirus-infected Sf9 cells was assayed for phosphorylation of HH1 in the presence of purified p21, C/EBP␣, or both. After incubation, the reaction was run on a 12% acrylamide gel, transferred to a membrane, and probed using antibodies to the HA tag present on C/EBP␣ and p21 (Fig. 3A). As shown in Fig. 3B, CDK2 phosphorylation activity was easily detectable in the absence of p21 (lane 1; this value was arbitrarily set to 100), whereas the addition of 100 nM recombinant p21 inhibited CDK2 activity to FIG. 3. C/EBP␣ cooperates with p21 to inhibit CDK2-cyclin E phosphorylation of HH1. A, shown is a Western blot of recombinant proteins added to baculovirus-expressed and purified CDK2-cyclin E. Proteins were added to the kinase assay as detailed under "Experimental Procedures" and, after incubation, were resolved on SDS gel, transferred to a membrane, and probed with anti-HA antibodies. Lane 1 contains only CDK2-cyclin E. Lane 2 contains ϳ100 nM GST-p21-HA, and lanes 3-6 contain 5 nM GST-p21-HA. Increasing amounts of GST-C/EBP␣-HA (10, 50, and 250 nM) were added to lanes 4 -6 and 7-9, respectively. B, after blotting, the membrane was dried and exposed to film. B shows the level of HH1 phosphorylation by CDK2/E. C, the dried membrane was Coomassie Blue-stained to show equal loading of the HH1 substrate. D, shown is the phosphorimaging of five repeats. Bar 1 was arbitrarily set to 100.

FIG. 2. C/EBP␣ interacts with CDK2.
A, GST-tagged C/EBP␣ was used to test the interaction with bacterially expressed CDK2. After incubation and extensive washing, the GST-tagged proteins and associated CDK2 was resolved on SDS gel, transferred to a nylon membrane, and probed with anti-CDK2 antibodies. B, shown is a summary of the regions of C/EBP␣ determined to be critical for interaction with CDK2 measured as described for A. C, C/EBP␣ and CDK2 are able to associate in vitro. C/EBP␣-HA was overexpressed in cycling 293 cells, and either anti-CDK2 antibodies (lanes 1 and 3) 4 -6). In the absence of p21, equivalent levels of C/EBP␣ had no significant effect on CDK2 activity (lanes 7-9). Subsequent Coomassie Blue staining of the membrane showed equal loading of HH1 (Fig. 3C). Phosphorimaging of five repeat experiments is shown in Fig. 3D. These findings suggest that C/EBP␣ is capable of enhancing the inhibitory activity of p21 while not directly affecting CDK2 activity at these concentrations. Immunoprecipitation of CDK2 under conditions identical to the kinase assay did not show a specific increase in p21 bound to CDK2 complexes in the presence of C/EBP␣ (data not shown). This suggests that C/EBP␣ does not enhance the rate of p21 binding to CDK2. The interaction between C/EBP␣ and p21 may be transient, or there may be a conformational change induced upon binding that leads to p21 binding to CDK2 in a more "inhibitory" fashion. C/EBP␣ was also shown to affect CDK2 activity immunoprecipitated from cycling cells. CDK2 was immunoprecipitated from 293 cell nuclear extracts, and kinase activity was measured as described above. 293 cells do not express C/EBP␣ and contain low, but detectable endogenous levels of p21. As shown in Fig. 4A (lane 2), p21 was titrated into the reaction to 100 nM, thus strongly inhibiting CDK2 activity. Lanes 3-8 (containing 20% of this level of p21 or ϳ20 nM) showed a 50% reduction in CDK2 activity. The addition of C/EBP␣ in conjunction with low levels of p21 caused a dose-dependent decrease in CDK2 activity that was significantly greater than that seen in the absence of p21 (lane 9). The phosphorimaging data in Fig. 4B are for the experiment shown and are representative of three repeats.
It has been previously reported that p27, another member of the p21 WAF1/CIP1/sdi1 family, does not bind to C/EBP␣, thus predicting that C/EBP␣ would not affect p27 inhibition of CDK2 (8). As shown in Fig. 5A (lane 2), the addition of ϳ100 nM p27 significantly inhibited CDK2 activity. When p27 was added at 25 nM, a partial inhibition of HH1 phosphorylation was observed (lane 3). However, when C/EBP␣ was titrated into the reaction in the presence of 25 nM p27, no further reduction in CDK2 activity was seen (lanes 4 and 5). These experiments demonstrate that the ability of C/EBP␣ to influence CDK activity requires interaction specifically with p21. Furthermore, the C/EBP␣-binding site in p21 is essential for enhancement of p21-mediated repression of CDK2 activity by C/EBP␣. As shown in Fig. 5A, a p21 construct missing the last 20 amino acids was capable of inhibiting immunoprecipitated CDK2 activity at levels comparable to wild-type p21 (100 nM) (lane 6) and provided partial inhibition of CDK2 activity at 20 nM (lane 7). However, titration of C/EBP␣ into the kinase reaction showed no further decrease in CDK2 activity (lanes 8 and 9). Thus, the domain of p21 that is involved in the interaction with C/EBP␣ is required for the cooperativity between p21 and C/EBP␣ in CDK2 inhibition.
To define the regions of C/EBP␣ required for the enhancement of inhibition of CDK2 activity by p21, we tested aminoand carboxyl-terminal truncation mutants of C/EBP␣ for their ability to inhibit CDK2 activity in conjunction with p21. As shown above (Fig. 1D), p21 can interact with amino acids 119 -212 and 313-360 of C/EBP␣. N-and C-terminal truncation mutants of C/EBP␣ were incubated with CDK2 to determine whether individual p21-binding regions in C/EBP␣ are sufficient to inhibit CDK2 activity. Although full-length C/EBP␣ was capable of cooperating with p21 to inhibit CDK2 immunoprecipitated from 293 cells (Fig. 6A, lanes 3 and 4), neither of the C/EBP␣ deletion mutants was able to do so. Neither the 5Ј-region (amino acids 1-226) (Fig. 6B, bars 5 and  6) nor the 3Ј-region (amino acids 280 -360) (bars 7 and 8) had any effect on CDK2 activity. This observation suggests that both the N-and C-terminal p21-and CDK2-binding sites are required for C/EBP␣ to exert its effects on CDK2 and that a single binding site for p21 or CDK2 is not sufficient to cooperate in inhibiting CDK activity. When performing deletion analysis, it is difficult to know for certain whether the mutants are folding correctly. However, the carboxyl-terminal mutant of C/EBP␣, containing amino acids 280 -360, includes both the basic DNA-binding region as well as the complete leucine zipper and binds to DNA in an electrophoretic mobility shift assay. By this measure of functionality, the basic leucine zipper protein mutant has DNA-binding ability, but is deficient in CDK2 inhibition.
Expression of C/EBP␣ in Cells in Culture Inhibits Proliferation-Since C/EBP␣ is capable of inhibiting CDK2 activity in vitro, it was necessary to examine whether the ability of C/EBP␣ to inhibit CDK2 activity correlates with its ability to inhibit proliferation. Previous reports have indicated that overexpression of C/EBP␣ in cultured cells causes growth arrest (5,6). These studies concluded that the effect of C/EBP␣ on growth is dependent on its ability to dimerize as well as to bind DNA. Deletion of a large internal portion of the transactivation region of C/EBP␣ results in loss of growth inhibitory activity (5,6). The ability of C/EBP␣ to inhibit growth is not dependent on Rb or p53 and is not inhibited by SV40 large T-antigen (6). To better define domains of C/EBP␣ that bring about growth arrest and to determine whether this inhibitory activity correlates with p21 or CDK2 binding, we generated a panel of C/EBP␣ mutants under the direction of the cytomegalovirus promoter (Fig. 7A) for expression in mammalian cells. When the empty vector pcDNA3.1 was cotransfected with the blasticidin resistance plasmid into 293 cells, ϳ100 colonies/10-cm plate were observed after 7-14 days of selection. When C/EBP␣ was cotransfected, Ͻ10 colonies/plate were seen after selection (Fig. 7C). Transfection of LacZ also showed a partial colony inhibition, suggesting that a high level of expression of this control protein could influence growth.
Transactivation Domain I of C/EBP␣ Is Not Required for Growth Arrest-The amino-terminal portion of C/EBP␣ is responsible for transcriptional activation and has been broken down into two broad domains, a strong activation domain, activation domain I (amino acids 1-107), and a weaker transcriptional activation region, activation domain II (amino acids 171-224). It is within activation domain I that C/EBP␣ binds to the basal transcription factors TATA-binding protein and transcription factor IIB (34). This region also contains a putative binding domain for Rb family members (35). Cells that lack the expression of Rb are defective in C/EBP␣-mediated adipocyte conversion, and it has been hypothesized that the interaction with the Rb family of proteins is important for the ability of C/EBP␣ to cause growth arrest. There is potentially conflicting information in the literature as to whether the 30-kDa alter-natively translated isoform of C/EBP␣ is capable of inhibiting proliferation. It has been reported that 30-kDa human C/EBP␣, which is composed of amino acids 119 -360, can cause growth arrest (6), whereas 30-kDa rat C/EBP␣ does not (36). This is despite a sharp decrease in the transcriptional activity of the C/EBP␣ isoform (36). We therefore tested an amino-terminal deletion mutant of C/EBP␣ for the ability to inhibit proliferation. The expression of amino acids 119 -360 of human C/EBP␣ was sufficient to inhibit proliferation in all cell lines tested (Fig.  7, C-E). These data suggest that transactivation domain I of C/EBP␣ is not required to influence the cell cycle (32,37).
We next tested whether deletion of the upstream CDK2 interaction domain (previously localized to amino acids 119 -160) or the N-terminal p21-binding site (located between amino acids 119 and 212) eliminates the ability of C/EBP␣ to inhibit growth. Fig. 7C shows that deletion of amino acids 1-186 (C/EBP␣-(187-360)) eliminated the ability of C/EBP␣ to cause growth arrest, demonstrating that the regions required for binding to CDK2 and/or p21 are important for C/EBP␣-mediated growth arrest. Because the C/EBP␣ construct containing amino acids 119 -360 was fully proficient in inhibiting proliferation, the N-terminal residues critical for C/EBP␣-mediated inhibition of proliferation are located between residues 119 and 187. Although C/EBP␣-(187-360) did generate a statistically significant difference in colonies compared with full-length C/EBP␣ (p Ͻ 0.05), there was not a statistically significant difference in the number of colonies generated by the empty vector or by the lacZ plasmid.
Deletion of the Leucine Zipper Reduces the Inhibitory Effect of C/EBP␣ on Growth-C/EBP␣ deleted for the extreme carboxyl-terminal leucine zipper and containing only amino acids 1-320 was only partially growth-inhibitory (Fig. 7C). This C/EBP␣ construct lacks the entire leucine zipper and was incapable of dimerization or of binding to DNA (Fig. 7A). This finding is consistent with work that described a C/EBP␣ leucine zipper mutant that was ineffective in suppressing colony formation (5). The leucine zipper deletion construct was highly expressed compared with the other constructs (Fig. 7B,  lane 3), although comparable levels of plasmid DNA were transfected into all cells. The effect of high levels of C/EBP␣-(1-320) protein is unknown, but may contribute to partial inhibition of proliferation, as was seen for lacZ (Fig. 7C). Analysis of Ͼ10 experiments showed that the leucine zipper deletion construct was statistically different from either the empty vector (p Ͻ 0.005) or C/EBP␣ (p Ͻ 0.001). To verify that the inhibition of proliferation by C/EBP␣ is not confined to 293 cells, we tested several of the constructs in Hep3B2 cells and mouse embryo fibroblasts (Fig. 7, D and E). The same profile of inhibition was seen in these cell lines as was observed in 293 cells.
DNA Binding Is Not Required for C/EBP␣-mediated Growth Arrest-To demonstrate that C/EBP␣ does not mediate growth arrest through DNA binding, a basic domain mutation (K300E) was generated. The K300E mutation changes the charge of the amino acid at this position, but keeps a similarly sized side group, and the surrounding region maintains an ␣-helical structure as predicted by the Chou-Fasman secondary structure prediction program (38) (data not shown). DNA binding as measured by electrophoretic mobility shift assay was eliminated (Fig. 8A, lanes 7 and 8) in K300E, and there is no activation of a C/EBP␣-driven reporter gene (data not shown). However, this construct was as efficient as wild-type C/EBP␣ in inhibiting proliferation (p Ͻ 0.05) (Fig. 8B).
A mutation of the rat C/EBP␣ gene described by Landschulz et al. (19), called BR3, contains four amino acid changes in the DNA-binding domain (R299G, K300T, R302G, and K304N). These mutations eliminate DNA binding and cause an ϳ50% inhibition of colony formation compared with the empty vector. The BR3 mutations do not conserve secondary structure, as two of the amino acid changes are to glycines, amino acids known to disrupt ␣-helical structure. To compare the activity of the BR3 mutant with the K300E mutant, the former was subcloned into the pcDNA3.1 vector. BR3 was expressed at levels similar to those of the K300E mutant (data not shown). A significant decrease in colony formation was observed with the BR3 construct compared with empty vector (p Ͻ 0.0001), but not compared with the LacZ construct. However, compared with wildtype C/EBP␣ or the K300E mutant, the growth inhibitory activity of BR3 was significantly decreased (p Ͻ 0.001). Because we have demonstrated that the basic region of C/EBP␣ binds to CDK2, we tested the ability of the C/EBP␣ DNAbinding mutants to interact with CDK2. As shown in Fig. 8C  (lanes 2 and 4), the wild-type C/EBP␣ basic region as well as the basic region with the K300E mutation were capable of binding to CDK2. However, the BR3 mutations within the basic region of C/EBP␣ eliminated the ability of the basic region to interact with CDK2 (lane 5). This suggests that the lack of effect of the BR3 mutant on proliferation is due to its inability to bind to CDK2, not an inability to bind to DNA, as the K300E mutant also lacks DNA-binding activity, but is fully capable of binding to CDK2 and inhibiting proliferation. Since the K300E basic domain binds to CDK2 and the full-length K300E construct inhibits proliferation as efficiently as wild-type C/EBP␣, we were interested in determining whether the K300E construct can also inhibit CDK2 activity in vitro. In Fig. 8D, we tested whether CDK2 immunoprecipitated from cells was af- fected by the addition of GST-K300E and p21. Titration of increasing amounts of GST had no effect on kinase activity (lanes 4 -6), whereas the addition of an equivalent level of GST-K300E reduced HH1 phosphorylation in a dose-dependent manner (lanes 7-9). These data demonstrate that although the K300E mutant is defective in DNA binding, the basic region of this construct is still capable of binding to CDK2, and full-length K300E can inhibit CDK2 activity in vitro. We then tested whether the BR3 mutant, which cannot bind to CDK2 through the basic domain, and deletion mutants 187-360 and 1-320, which are missing the N-terminal CDK2 interaction domain and the C-terminal p21 interaction domain, respectively, are defective in cooperating with p21 to inhibit CDK2 activity. The BR3 mutant as well as deletion mutants 187-360 and 1-320 were tested for the ability to affect CDK2 activity in vitro (Fig. 8E). Although full-length C/EBP␣ was capable of enhancing the effect of p21 against baculovirus-expressed CDK2-cyclin E activity (lane 4), mutants 187-360 and 1-320 and the BR3 mutant were all defective in affecting kinase activity (lanes 5-7). Construct 187-360 is missing the N-terminal CDK2 interaction site and did not inhibit proliferation. Despite containing the N-and C-terminal CDK2-binding sites, construct 1-320 is missing the leucine zipper, which was identified as a p21 interaction site. Because mutant 1-320 did not cooperate with p21 to inhibit CDK2 activity, this suggests that the leucine zipper is necessary for C/EBP␣ to inhibit CDK2 activity in a p21-dependent fashion. The BR3 construct maintains an N-terminal CDK2-binding site; but as shown in Fig.  8C, the basic domain of BR3 was defective in binding to CDK2. These constructs demonstrate that loss of CDK2 inhibitory activity in vitro in all C/EBP␣ mutants tested correlates with loss of growth inhibitory activity in vivo.
DNA Binding by C/EBP␣ Is Not Required to Inhibit BrdUrd Incorporation-To ensure that the constructs tested were localizing to the nucleus as predicted and to provide another method for measuring the inhibitory effects of C/EBP␣ on growth, we examined DNA synthesis as measured by BrdUrd uptake. Cells were transiently transfected, labeled with BrdUrd on day 3, and stained for expression of C/EBP␣ and for BrdUrd incorporation. As shown in Fig. 9 (A1-D1), expression of C/EBP␣, C/EBP␣-(1-286), C/EBP␣-BR3, and C/EBP␣-K300E was detectable in a subpopulation of cells. BrdUrd uptake (shown in Fig. 9, A2-D2) was also detected in a subpopulation of cells. Fig. 9 (A3-D3) shows DNA staining by 4,6-diamidino-2-phenylindole in all cells. To measure whether C/EBP␣ or the various mutants affected DNA synthesis, we counted C/EBP␣-expressing cells and scored them as positive or negative for BrdUrd uptake. As shown in Fig. 9E, there was a 4 -5-fold difference in the level of DNA synthesis in C/EBP␣expressing cells compared with a truncated C/EBP␣ expression construct lacking the basic leucine zipper protein domain (amino acids 1-286). The DNA-binding mutant BR3, which lacks the ability to bind to CDK2 in the basic region and to inhibit CDK2 activity, was also ineffective in the inhibition of BrdUrd incorporation (Fig. 9E). However, the DNA-binding mutant K300E, which inhibited CDK2 in a p21-dependent fashion, showed an inhibition of DNA synthesis as measured by BrdUrd uptake. Together with the assays from Figs. 7 and 8, this experiment demonstrates that although the basic domain of C/EBP␣ may be mutated to eliminate DNA binding, this region must still be capable of interacting with CDK2 to inhibit BrdUrd uptake.
Our results indicate that although the ability to bind to DNA is not required for growth arrest, the leucine zipper is essential for the ability of C/EBP␣ to inhibit proliferation. Finally, the domains of C/EBP␣ involved in binding CDK2 and p21 correlate with the regions of C/EBP␣ required for maximal growth arrest.

DISCUSSION
In this report, we have defined regions of C/EBP␣ that interact with p21 and CDK2 as well as the possible functions of these interactions. We have demonstrated that DNA binding is not required for C/EBP␣-mediated growth arrest and have shown that the regions of C/EBP␣ that are important for enhancement of p21 inhibition of CDK2 are also necessary for growth arrest. This has led us to conclude that the effects of C/EBP␣ on the cell cycle are separate from its ability to activate transcription through DNA binding. Although we have implicated p21 and CDK2 as mediators of C/EBP␣-induced growth arrest, there may well be additional mechanisms by which C/EBP␣ can block proliferation.
A recent report showed that HPV16 E7 is capable of specifically reversing the inhibitory activity of p21 against CDK2 through an interaction with the carboxyl-terminal 20 amino acids of p21 (15,31). A separate publication demonstrated that the ability of C/EBP␣ to induce growth arrest could be reversed by coexpression of the HPV16 E7 protein (39). The Rb-binding region of E7 was not necessary for eliminating C/EBP␣-mediated growth arrest; rather, the domain required to block the effects of C/EBP␣ on growth was the casein kinase II phosphorylation site within HPV16 E7. We speculate that the effects of C/EBP␣ on CDK2 activity through p21 can be overridden by a competitive interaction with HPV16 E7. It was further shown that the ability of the E7 protein to prevent C/EBP␣-mediated growth arrest does not affect the transcriptional activity of C/EBP␣. This suggests that C/EBP␣-mediated inhibition of proliferation can be separated from its transcriptional abilities, as was demonstrated (39). Our data portray a scenario similar to that of HPV16 E7 with the important difference being that C/EBP␣ enhances rather than reverses the inhibitory activity of p21. An additional report has shown that yet another transcription factor known as SET can reverse the inhibitory activity of p21 through binding to the carboxyl terminus of p21 (16). A recently described novel gene called TOK-1␣ behaves like C/EBP␣ in that it enhances the inhibitory activity of p21 through interaction with the carboxyl terminus of p21 (17). Together, these findings suggest that a number of proteins capable of interacting with p21 may directly influence cell cycle components in a positive or negative fashion, perhaps through the use of p21 as an adaptor or scaffold protein.
Muller et al. (39) showed that C/EBP␣ can cause growth arrest in a p21-deficient cell line and argued that p21 is not required for growth arrest by C/EBP␣. The ability of C/EBP␣ to cause growth arrest in the absence of p21 strongly suggests that p21 is not the only pathway by which C/EBP␣ inhibits proliferation. Other mechanisms for C/EBP␣-mediated growth arrest have been implicated, including direct interaction with the pocket proteins p107 and Rb. It is known that p107 can directly bind to and inhibit CDK2 through a p21-like domain (40). C/EBP␣ can also bind to p107 (11). It may be that C/EBP␣ can cooperate with p107 to inhibit CDK activity and bring about growth arrest. In the latter instance, p107 may directly substitute for p21. A recent report suggests that C/EBP␣ inhibits E2F transcriptional activity through complex formation (9). Alternatively, high levels of C/EBP␣ may have a direct effect on CDK activity such as seen with MyoD for CDK4 (41). Any of these possibilities would provide for an alternative pathway for C/EBP␣-mediated growth arrest in p21-deficient cells.
A remaining question is whether other leucine zipper-containing transcription factors or C/EBP family members are capable of acting in a similar fashion. The leucine zipper of C/EBP␤ can interact with p21 (data not shown). When the leucine zipper from C/EBP␣ is replaced with that from C/EBP␤, the fusion protein is still capable of causing growth arrest (data not shown). C/EBP␤ and C/EBP␦ have both been shown to cause growth arrest under certain conditions, although the mechanism by which this occurs is unknown (42)(43)(44)(45). These findings lead to the possibility that there are several different leucine zipper-containing proteins capable of modulating proliferation via a direct interaction with components of the cell cycle. Although additional pathways of C/EBP␣-mediated growth arrest may exist in differing cell types, there is a strict correlation between the presence of the p21-and CDK2-binding sites and inhibition of proliferation. Furthermore, DNA-binding activity is not required for this activity, strongly suggesting that the mechanisms for C/EBP␣-mediated inhibition of proliferation involve protein/protein interactions and not transcriptional activation.