Opposing Roles of C/EBPβ and AP-1 in the Control of Fibroblast Proliferation and Growth Arrest-specific Gene Expression*

Chicken embryo fibroblasts (CEF) express several growth arrest-specific (GAS) gene products in G0. In contact-inhibited cells, the expression of the most abundant of these proteins, the p20K lipocalin, is activated at the transcriptional level by C/EBPβ. In this report, we describe the role of C/EBPβ in CEF proliferation. We show that the expression of a dominant negative mutant of C/EBPβ (designated Δ184-C/EBPβ) completely inhibited p20K expression at confluence and stimulated the proliferation of CEF without inducing transformation. Mouse embryo fibroblasts nullizygous for C/EBPβ had a proliferative advantage over cells with one or two functional copies of this gene. C/EBP inhibition enhanced the expression of the three major components of AP-1 in cycling CEF, namely c-Jun, JunD, and Fra-2, and stimulated AP-1 activity. In contrast, the over-expression of C/EBPβ caused a dramatic reduction in the levels of AP-1 proteins. Therefore, C/EBPβ is a negative regulator of AP-1 expression and activity in CEF. The expression of cyclin D1 and cell proliferation were stimulated by the dominant negative mutant of C/EBPβ but not in the presence of TAM67, a dominant negative mutant of c-Jun and AP-1. CEF over-expressing c-Jun, and to a lesser extent JunD and Fra-2, did not growth arrest at high cell density and did not express p20K. Therefore, AP-1 interfered with the action of C/EBPβ at high cell density, indicating that these factors play opposing roles in the control of GAS gene expression and CEF proliferation.

The entry into G 0 is characterized by the activation of a specific program of gene expression. Growth arrest-specific (GAS) genes encode proteins fulfilling roles in lipid metabolism (5,6,20,31), cellular architecture (1-3), cell proliferation (4,5) the control of gene expression (6 -8), and the response to growth factors such as platelet-derived growth factor (9). Al-though the activity and function of several GAS genes remain poorly characterized, gene products such as GAS6 and the secreted apoptosis-related proteins (SARPs) promote the survival of growth-arrested cells (10,11). Therefore, a component of the G 0 program of gene repression may represent an adaptive response of cells that are poised to re-enter the cell cycle but facing deleterious growth conditions.
Chicken embryo fibroblasts (CEF) 1 reaching confluence express a novel set of proteins secreted into the culture medium (12). The most abundant of these quiescence-specific proteins is a member of the lipocalin family of lipid-binding proteins, designated p20K, chicken p21, or extracellular fatty acid-binding protein (EX-FABP) (13,14). Maximal expression of p20K is observed in contact-inhibited CEF and in chicken heart mesenchymal cells rendered quiescent by culture in plasma-containing rather than serum-containing medium (12). Serum starvation causes a less important increase in the synthesis of p20K, whereas agents arresting cells at the G 1 /S border have no effect on the expression of this gene (15). The induction of p20K is regulated at the transcriptional level and depends on a 48-bp region of the promoter termed the quiescence-responsive unit or QRU (15). The QRU is both necessary and sufficient for the transcriptional activation of p20K in contact inhibited CEF. The QRU includes two C/EBP binding sites that are essential for induction in quiescent CEF. C/EBP␤, the predominant member of the C/EBP family expressed in contact-inhibited CEF, binds to both C/EBP binding sites of the QRU and is capable of inducing the expression of p20K when over-expressed in cycling CEF.
We showed recently that C/EBP␤ is also induced in CEF transformed by the Rous sarcoma virus (RSV) (16). In these cells, C/EBP␤ cooperates with AP-1 and NF-B in the constitutive induction of the CEF-4 chemokine, a marker of v-Src transformation in CEF (16 -18). The activity and expression of C/EBP␤ is thus controlled by conditions associated with both the stimulation and inhibition of cell proliferation. However, the expression of a dominant negative mutant of C/EBP␤, termed ⌬184-C/EBP␤, accentuated the phenotype of v-Srctransformed cells, suggesting that C/EBP␤ acts mainly as a growth inhibitor in CEF cultured in vitro. In this report, we describe the proliferation of CEF expressing the dominant negative mutant ⌬184-C/EBP␤. We confirm that the inhibition of C/EBP enhances proliferation in CEF, an effect dependent on the stimulation of AP-1. Thus although C/EBP␤ and AP-1 cooperate in RSV-transformed cells and at the G 0 /G 1 transition in non-transformed cells, they play opposing roles in growtharrested CEF. This interaction regulates the expression of the growth arrest specific p20K gene and governs in part the proliferation of chicken embryo fibroblasts.

EXPERIMENTAL PROCEDURES
Cells and Viruses-Early passages of CEF were cultured at 41.5°C in Richter-improved minimal medium containing insulin and zinc (I ϩ medium, Irvine Scientific, Santa Ana, CA), 5% heat-inactivated newborn serum (MediaserII, Montreal Biotech Inc., Kirkland, Quebec, Canada), 5% tryptose phosphate broth, glutamine, penicillin, and streptomycin. CEFs were infected with recombinant viruses generated with the RCASBP(A) or RCASBP(B) retroviral vectors. Recombinant retroviruses used in this study have been described elsewhere (16) except for RCASBP-TAM67. The TAM67 cDNA was excised from pCMV-TAM67 by BamHI restriction endonuclease digestion and subcloned into the BamHI site of the Cla12 adaptor plasmid to obtain convenient ClaI restriction sites at both the 5Ј and 3Ј end of the insert. The TAM67 cDNA was subsequently excised by ClaI digestion and cloned into RCASBP(A) or RCASBP(B) retroviral vectors (19). RCASBP vectors were transfected by the calcium phosphate method (20), initiating the infection of the entire CEF population. For co-expression of two different transgenes, CEF transfected with a RCASBP vector were superinfected with a second retrovirus of a different genetic subgroup (A or B). Mouse embryo fibroblasts (MEF) isolated from 15-day-old embryos were cultured in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 10% fetal calf serum and glutamine, penicillin, and streptomycin at 37°C. MEF with different C/EBP␤ genetic complements were isolated from littermate embryos and compared in parallel in proliferation assays.
Antibodies and Western Blotting Analysis-Total cellular protein extract (30 g) was prepared in SDS sample buffer, subjected to SDSpolyacrylamide gel electrophoresis, and blotted on nitrocellulose (Schleicher and Schuell, BA85). The membranes were blocked in a 5% solution of milk powder dissolved in Tris-buffered saline. The p20K and C/EBP␤ antiserum generated from immunized rabbits was used at a dilution of 1:1000. This was followed by incubation with a peroxidaseconjugated secondary antibody (Pierce) and detection with a chemiluminescent substrate according to protocols provided by the manufacturer (ECL, Amersham Biosciences). Antibodies for the AP-1 family members c-Jun (SC-45X or SC-44X for TAM67), JunD (SC-74X), Fra-2 (SC-604X), and p21 CIP (SC-393G), as well as ERK-1 (SC-93), were obtained from Santa Cruz Biotechnology (Santa Cruz, CA) and used according to the specifications provided by the manufacturers. The c-Fos antibody was purchased from Geneka Biotechnology Inc. (catalog no. 2000752; Montréal, Québec). The p27 KIP (no. 558902) and cyclin D1 (no. 556470) antibodies were obtained from BD Pharmingen and the PhosphoPlus® Rb Antibody Kit (no. 9300) was obtained from Cell Signaling Technologies and used according to the manufacturers' guidelines.
Transient Expression Assays-Transient expression assays were performed on subconfluent cells by the DEAE-dextran method as described previously (17) using a total of 20 g of DNA consisting of 5 g of a TPA-responsive element (TRE)-controlled chloramphenicol acetyltransferase (CAT) reporter plasmid, 2.5 g of a ␤-galactosidase expression plasmid (pCH110), and 12.5 g of salmon sperm carrier DNA. For all transient expression assays, the activity of the CAT reporter enzyme was determined in lysates representing equal levels of ␤-galactosidase activity. All transfections were performed in duplicates or triplicates in at least two separate experiments. [ 14 C]Chloramphenicol and its acetylated derivatives were separated by thin layer chromatography and quantitated using and InstantImager (Packard/Canberra). Plasmid pJFCAT/TATA is a minimal promoter construct consisting of a TATA-AAA box, the initiation start site of the human ␤-globin gene, and the CAT reporter gene. Plasmids 3x TRE-pJFCAT/TATA, 3x TRE-pJF-CAT/TATA, and 3x PRDII pJFCAT/TATA are derivative of pJFCAT/ TATA and include three copies of a TRE, mutant TRE, and PRDII domain, respectively (17,21).
Electrophoretic Mobility Shift Assays (EMSA)-Nuclear extracts were prepared as described previously (16). The sequence of the synthetic DNA oligomers corresponding to the TRE are (ϩ)5Ј-AGCTTGT-GACTCATTT-3Ј and (Ϫ)5Ј-GATCAAATGAGTCACA-3Ј, whereas the TRE oligomers include three nucleotide substitutions within the AP-1 binding site (17). The radiolabeled probe was generated by end-filling reactions with the Klenow fragment of DNA polymerase using [␣-32 P]dCTP and unlabeled dGTP, dATP, and dTTP. Binding reactions were performed in a total volume of 20 l of 1ϫ binding buffer (20 mM Tris-HCl pH 7.5, 1 mM MgCl 2 , 60 mM NaCl, and 5% glycerol) for 30 min at room temperature. For mobility shift assays, the radiolabeled oligonucleotides (ϳ10,000 cpm/0.1 ng) were incubated with 2 g of nuclear extract in the presence of 2 g of poly(dI-dC). Competition binding reactions were performed by pre-incubating the nuclear extract with an excess of unlabeled, end-filled wild-type TRE or mutant TRE oligonucleotides for 15 min at room temperature followed by an additional 30-min incubation with the probe. Supershift analysis of AP-1 complexes was performed by incubating AP-1-specific antibodies with the radiolabeled probe-nuclear extract binding reaction for 1 h at 4°C. The resulting DNA-protein complexes were resolved on 4.8% nondenaturing polyacrylamide gels in 0.5ϫ TBE buffer and visualized by autoradiography.
Proliferation Assays-Proliferation was analyzed for CEFs infected with RCASBP(B), RCASBP-⌬184-C/EBP␤, RCAS-c-Jun, RCAS-JunD, and RCAS-Fra-2 and the dominant negative c-Jun mutant (TAM67). Cells were split from confluent 100-mm plates into 35-mm plates and incubated for the specified number of days without change of medium. Cells were counted each day in triplicates using a Coulter counter (model Z2, Coulter Corp., Miami, FL). Analysis of MEF proliferation was performed essentially as described for CEFs.
Northern Blotting Analysis-Northern blotting analysis was performed as described previously (22). Briefly, 8 g of total RNA were loaded per lane and separated on a 1% agarose gel containing formaldehyde. The RNA was transferred onto Nytran membranes and probed with a radiolabeled cDNA for genes of interest. A glyceraldehyde-3phosphate dehydrogenase probe was used as control to monitor RNA loading.
Genomic DNA Isolation and DNA Fragmentation Analysis-Cells scraped from tissue culture dishes and floating in the medium were collected together from 2-4 100-mm dishes and used for genomic DNA isolation. Cells were spun down at 6000 ϫ g for 5 min. and lysed in 500 l of TTE/RNase (10 mM Tris, pH 8, 0.25%(v/v) Triton X 100, 1 mM EDTA, 0.1 mg/ml RNase)/pellet at 37°C for 1.5 h. This solution was then treated with proteinase K/SDS (to give 0.2 mg/ml proteinase K and 1% SDS in a 500 l volume) for 2 h at 50°C. This solution was then extracted twice with phenol/chloroform and once with chloroform and precipitated with sodium acetate and ethanol. Final pellets were typically dissolved in 50 l Tris-EDTA, pH 8, and stored at 4°C. Typically 2 l of the genomic DNA was run on a 1% Separide gel (Invitrogen) to assess the degree of DNA fragmentation.

C/EBP␤ Is Growth-inhibitory in Cycling Embryonic
Fibroblasts-To characterize the role of C/EBP in the control of CEF proliferation and quiescence-specific gene expression, we expressed the dominant negative mutant ⌬184-C/EBP␤ with the RCASBP retroviral vector in CEF (16). ⌬184-C/EBP␤ encodes a protein devoid of trans-activating regions but capable of homoand heterodimerization with members of the C/EBP family (23). Using this approach, we showed previously that ⌬184-C/ EBP␤ is over-expressed in all cells of the monolayer, is localized to the nucleus, and binds to C/EBP elements (16). CEFs infected with the ⌬184-C/EBP␤ or control RCASBP virus were grown to confluence, and p20K expression was examined by Western blotting analysis of total cell extracts ( Fig. 1). Control CEFs expressed p20K at confluence (reached after day 3 in this experiment) and continued to accumulate the protein at contact inhibition (Fig. 1, lanes 3-5). In contrast, p20K was not expressed by dense CEFs infected with RCASBP-⌬184-C/EBP␤ even when the cells became growth-arrested ( Fig. 1, lanes 6 -10, and see below). Likewise, the p20K mRNA did not accumulate in CEF expressing ⌬184-C/EBP␤ in response to confluence (Fig. 1B, lanes 4 and 10). Starvation of subconfluent CEF in serum-free medium was inefficient in the induction of the p20K mRNA (lanes 1-3). Moreover, substituting serum-containing for serum-free medium led to a decrease of p20K mRNA levels in density-arrested cells (lane 5). Under these conditions, CEF re-enter the cell cycle, undergo one round of cell division, and do not express p20K for a 24-h period (15). Serum starva-tion had no effect on the expression of p20K in CEF infected with the RCASBP-⌬184-C/EBP␤ virus. In contrast, the overexpression of wild-type C/EBP␤ caused a marked accumulation of the p20K mRNA even when cells were actively dividing; under these conditions, the p20K mRNA was induced to levels well above those observed at contact inhibition (Fig. 1B, lane  13). Therefore the expression of p20K was strictly dependent on C/EBP in CEF. These results also demonstrate the efficacy of the ⌬184-C/EBP␤ dominant negative mutant in density-arrested CEF.
The proliferation of CEF infected with the ⌬184-C/EBP␤ or control RCASBP(B) virus was examined at low cell density in medium containing 5% serum ( Fig. 2A). The expression of the dominant negative mutant strongly stimulated CEF proliferation. Moreover, cells infected with RCASBP-⌬184-C/EBP␤ were more refractile and reached a higher cell density. However, ⌬184-C/EBP␤ did not promote the formation of foci in monolayers or the growth of colonies in soft agar (data not shown). Therefore the inhibition of C/EBP enhanced proliferation but did not transform CEF, which were still contact-inhibited despite reaching a higher saturation density.
To determine whether C/EBP␤, the predominant C/EBP member expressed by CEF (6), is responsible for the inhibition of proliferation, MEFs with a disrupted C/EBP␤ gene were compared with MEF harboring one or two functional copies of the gene. In proliferation assays, nullizygous C/EBP␤ (Ϫ/Ϫ) MEF had a proliferative advantage over MEF expressing C/EBP␤. Nullizygous (Ϫ/Ϫ) and MEF with a single functional copy of the gene (ϩ/Ϫ) continued to proliferate at a higher rate even as the primary fibroblasts with two functional copies of the C/EBP␤ gene approached replicative senescence in culture (Fig. 2B). Therefore C/EBP␤ acts as a negative regulator of embryonic fibroblast proliferation.
C/EBP␤ Controls CEF Proliferation through AP-1-The growth-promoting effect of ⌬184-C/EBP␤ is consistent with our previous observation that blocking the activity of C/EBP␤ enhances CEF transformation by the Rous sarcoma virus (16). C/EBP␤ is one of the transcription factors activated by pp60 v-src in CEF. However, unlike AP-1, it is dispensable for the in vitro transformation of these cells by RSV. Furthermore, the v-srcdependent activation of AP-1 and C/EBP␤ was affected differently by the expression of the dominant negative mutant. Although a minimal promoter controlled by C/EBP binding sites was completely inhibited, a TRE-controlled promoter was stimulated by the expression of ⌬184-C/EBP␤ in RSV-transformed CEF. Therefore, we sought to determine whether the growth stimulatory effect of ⌬184-C/EBP␤ was mediated by the induction of AP-1. Transient expression assays performed on CEF infected with the ⌬184-C/EBP␤ or parental RCASBP(B) virus confirmed the activation of the TRE-controlled promoter in the presence of the dominant negative mutant (Fig. 3A). In contrast, the minimal promoter controlled by multiple copies of PRDII, an element recognized by p50 NF-B1 in cycling CEF (18), was not activated by the dominant negative mutant of C/EBP␤. Finally, the minimal promoter construct pJFCAT/ TATA or a derivative including multiple copies of a mutant TRE did not respond to the action of ⌬184-C/EBP␤. This was true for cells grown in medium containing 5 or 1% serum (data not shown). Therefore the stimulation of AP-1 by ⌬184-C/EBP␤ was specific and not dependent on pp60 v-src . The level of TRE binding activity (AP-1) was then examined by EMSA. As shown in Fig. 3B, CEFs expressing the dominant negative mutant exhibited higher levels of AP-1 activity than control cells infected with the parental RCASBP virus. Competition assays with an excess of the homologous TRE oligonucleotide or a mutant form of this element confirmed the specificity of the TRE nucleoprotein complex.
The components of AP-1 were identified by EMSAs using antibodies for various Jun or Fos family members. The appearance of supershifted complexes confirmed that dimers of Fra-2/JunD, and, to a lesser extent, Fra-2/c-Jun account for the majority of, if not all, AP-1 activity of cycling CEF. c-Fos was not detected in this complex (Fig. 4A). We then examined the level of these proteins in cells expressing ⌬184-C/EBP␤ or infected with the control virus (Fig. 4B). All three proteins were expressed at higher levels in CEF expressing the dominant negative mutant (Fig. 4B, lane 1 versus lane 4). In contrast, the steady state level of all three components of AP-1 was markedly reduced in CEF over-expressing C/EBP␤, whereas the expression of p20K was strongly stimulated in these cells (Fig. 4B,  lanes 7 and 8). Therefore these data suggest that C/EBP␤ is a negative regulator of the expression of c-Jun, JunD, and Fra-2 in CEF.
The activity of C/EBP␤ is induced in contact inhibited CEF where it drives the expression of quiescence-specific genes such as p20K (6). Therefore one would expect that the expression of c-Jun, JunD, and Fra-2 be reduced in density-arrested cells as a result of C/EBP␤ activation. To address this question, we compared the levels of all three components of AP-1 in actively dividing (G), contact-inhibited (CI), and serum-starved (S) CEF; as shown in Fig. 4B, the level of c-Jun decreased at contact inhibition but not markedly in CEF over-expressing ⌬184-C/EBP␤ (lane 2 versus 5). Contact inhibition did not cause a reduction in the expression of JunD but altered the relative abundance of JunD isofoms. The accumulation of a faster migrating JunD protein suggests that quiescent CEFs express predominantly the truncated and less potent Jun D isoform described by other investigators (24). These changes were partly antagonized by the expression of the dominant negative mutant ⌬184-C/EBP␤ although less drastically for the faster migrating JunD isoform. Fra-2 expression was nearly abolished at contact inhibition in the absence or presence of ⌬184-C/EBP␤ expression. Similar results were observed for Fra-2 in response to serum starvation (Fig. 4B, lanes 3 and  6). Interestingly, c-Jun was repressed at contact inhibition but not by serum starvation (Fig. 4B, lane 3).
To characterize the induction of p20K in conditions of AP-1 down-regulation, we compared the levels of p20K and c-Jun in cycling CEF (Fig. 4C, lanes 1 and 4), in CEFs reaching confluence (lanes 2 and 5), and in contact-inhibited CEFs 2 days after reaching confluence (lanes 3 and 6). This was done for cells infected with the RCASBP control virus or RCASBP-⌬184-C/ EBP␤ (Fig. 4C, lanes 1-3 and 4 -6, respectively). High levels of c-Jun correlated with the absence of p20K in CEF expressing ⌬184-C/EBP␤. However, p20K was maximally induced in control CEF reaching confluence, i.e. in conditions where c-Jun was not fully down-regulated. Using transient expression as-says, we examined the activity of a TRE-controlled promoter in actively dividing, contact-inhibited, and serum-starved CEF (Fig. 4D). This activity decreased in growth-arrested CEF even in the presence of the dominant negative mutant ⌬184-C/ EBP␤, which was growth-stimulatory mainly in actively dividing CEF. Therefore the expression and activity of AP-1 is controlled by C/EBP in cycling CEF, but other factors cooperate to inhibit AP-1 and particularly Fra-2 expression during growth arrest.
To assess the role of AP-1 in the proliferation of ⌬184-C/ EBP␤ infected CEF, we over-expressed the TAM67 deletion mutant of c-Jun with the RCASBP(A) vector. TAM67 encodes the rat c-Jun protein devoid of trans-activation domains but capable of homo-and heterodimerization with members of the Jun and Fos families (25). Stable expression of the c-Jun deletion mutant was confirmed by Western blotting analysis (Fig.   FIG. 2. Inhibition or  5A). TAM67 markedly inhibited the proliferation of CEF expressing ⌬184-C/EBP␤. This inhibition was comparable with that of CEF expressing TAM67 alone, suggesting that the proliferation of control CEF and CEF over-expressing ⌬184-C/ EBP␤ is critically dependent on AP-1.
TAM67 is known to induce growth arrest by enhancing p53 and p21 Waf1 expression (26). However the level of these proteins was not altered by the over-expression of TAM 67 or ⌬184-C/EBP␤ in CEF (data not shown). In contrast cyclin D1, a known target of AP-1 (27,28), was induced in cells expressing ⌬184-C/EBP␤ (Fig. 5B). To determine the role of AP-1 in the accumulation of cyclin D1, we co-infected CEF with viruses expressing the dominant negative mutant of C/EBP␤ (⌬184-C/ EBP␤) and c-Jun (TAM67). Cyclin D1 levels were reduced in control and ⌬184-C/EBP␤ CEF expressing TAM67 (Fig. 5C). Moreover, the dominant negative mutant of c-Jun reduced significantly the level of Rb phosphorylation on Ser 807/881 and Ser 780 , a residue phosphorylated in vivo in a cyclin D1-depend-ent manner (29,30). Using the same phospho-Rb-specific antibodies, we also observed that CEF expressing ⌬184-C/EBP␤ showed enhanced phosphorylation on Ser 780 and Ser 807/811 (Fig. 5C, lane Aϩ⌬184). However hyperphosphorylation was reduced by the co-expression of TAM67, which was equally potent in promoting Rb de-phosphorylation in control and RCASBP-⌬184-C/EBP␤-infected CEF (Fig. 5C). These results suggest that ⌬184-C/EBP␤ stimulates proliferation by enhancing AP-1 activity which controls cyclin D1 expression and Rb phosphorylation. Thus C/EBP␤ controls CEF proliferation through AP-1.
AP-1 Interferes with Density-dependent Growth Arrest and p20K Expression-The inhibition of AP-1 and CEF proliferation by C/EBP␤ is consistent with a role for this factor in growth arrest. However it is not clear that this function is related to the induction of GAS genes such as p20K by C/EBP␤. To address this question, we first examined the activation of p20K in CEF over-expressing individual components of AP-1. CEFs were infected with RCAS viruses encoding c-Jun, JunD, or Fra-2 and analyzed in proliferation assays. When expressed individually, all three components of AP-1 stimulated CEF FIG. 3. The effect of ⌬184-C/EBP␤ over-expression on TRE activity. A, CEF infected with the RCASBP-⌬184-C/EBP␤ or control RCASBP retrovirus were transfected with a CAT reporter plasmid including a minimal promoter fused to three copies of the TRE (3x TRE). Control vectors consisting of the parental minimal promoter construct (pJFCAT/TATA), the same vector fused to three copies of a mutant form of the TRE (3x TRE), or a distinct element (PRDII in 3x PRDII) were included to establish the specificity of the TRE activation. B, EMSA performed with nuclear extracts of RCASBP-or RCASBP-⌬184-C/EBP␤-infected CEF and a radiolabeled oligonucleotide probe corresponding to the TRE. Competition experiments were performed with a 50ϫ excess of unlabeled wild-type TRE oligonucleotide (wt TRE) or a mutant form of the TRE unable to bind to AP-1 ( TRE). Higher levels of TRE binding activity are observed in CEF expressing the dominant negative mutant ⌬184-C/EBP␤.  7 and 8). Protein loading was monitored by probing the blots with ERK-1 antibody. C, the expression of p20K and c-Jun was compared by Western blotting analysis in cycling CEF (lanes 1 and 4), CEF reaching confluence (lanes 2 and 5), or density-arrested CEF (lanes 3 and 6) in the presence or absence of the dominant negative mutant ⌬184-C/ EBP␤. D, effect of A retrovirus over-expressing ⌬184-C/EBP␤ on TRE activity. CEF infected with RCASBP(B)-⌬184-C/EBP␤ or the control virus, RCASBP(B), were transfected with the 3XTRE/CAT reporter plasmid, and reporter activity was determined under conditions of proliferation/growth, contact inhibition (confluence), and serum starvation. proliferation (Fig. 6A). This stimulatory effect was even more potent when Fra-2 was co-expressed with c-Jun or JunD (data not shown). There were also marked differences in the fate of these cells at high density. Although control CEF became contact inhibited at confluence, CEF expressing c-Jun, JunD, or Fra-2 began to die with morphological features of apoptotic cells. This was confirmed by the analysis of genomic DNA, which showed the typical fragmentation pattern of apoptotic cells (Fig. 6B). CEF over-expressing c-Jun, JunD, and Fra-2 did not show any signs of DNA fragmentation at low density but entered apoptosis as they reached confluence (day 4 in Fig. 6B). This was most obvious in cells over-expressing c-Jun. Control CEF did not show DNA fragmentation at high cell density and entered apoptosis only after several days of growth arrest without medium replenishment, i.e. as a result of prolonged starvation (data not shown). Therefore CEF over-expressing individual components of AP-1 show alterations in densitydependent growth arrest.
We examined the induction of p20K in CEF over-expressing c-Jun, JunD, or Fra-2 (Fig. 7). In this experiment, control CEF infected with the RCAS virus reached high cell density at day 2 and became progressively growth-arrested at days 3 and 4. This was confirmed by the induction of the p27 Kip1 cyclindependent kinase inhibitor, the induction of p20K and the down-regulation of the endogenous c-Jun, JunD, and Fra-2 proteins, as described above. In contrast, p20K and p27 Kip1 did not accumulate in c-Jun over-expressing CEF. Albeit less dramatically, the induction of p20K was also impaired in CEF over-expressing JunD and Fra-2, whereas p27 Kip1 accumulated normally in these cells. Therefore p20K and p27 Kip1 are controlled by different mechanisms in response to cell density. Moreover these results indicate that AP-1 interferes with the C/EBP␤-dependent activation of p20K. C/EBP␤ is itself induced by contact inhibition, a process that likely contributes to the activation of p20K (6). Although C/EBP␤ levels remained unchanged in CEF over-expressing JunD, the density-dependent induction of C/EBP␤ was not affected by the over-expression of c-Jun and Fra-2, which thus interfered with p20K induction at a succeeding step (Fig. 7). Therefore the decrease of AP-1 activity, observed in response to cell confluence and controlled in part by C/EBP␤, is necessary for entry into G 0 and the induction of the growth arrest-specific p20K gene in CEF.
Forced expression of a single component of AP-1 altered the expression of its binding partners. For instance, c-Jun was not detected in CEF over-expressing JunD, and conversely, JunD levels were significantly reduced in CEF over-expressing c-Jun. In contrast, JunD and Fra-2, the major components of AP-1 in cycling CEF, exerted a mutual stimulatory effect leading to the accumulation of JunD in Fra-2 over-expressing CEF and vice versa. Whether or not this feedback regulation can account for the differences in p20K repression and growth arrest remains to be investigated.
We then examined the effect of phorbol esters (TPA), which induce AP-1 activity, on the over-expression of p20K in CEF infected with the RCASBP-C/EBP␤ virus. As shown above, high levels of C/EBP␤ caused a marked accumulation of the p20K mRNA even when cells were actively dividing (Fig. 7B,  lanes 6 -9). However, the addition of TPA reduced the induction of p20K by C/EBP␤. In contrast, the expression of the CEF-4 chemokine, which depends on C/EBP␤ and AP-1 (16, 17), was markedly stimulated by the addition of TPA in the RCASBP-C/EBP␤-infected cells. However, the over-expression of C/EBP␤ had a modest effect on the induction of the CEF-4 mRNA in the absence of TPA, which was a more potent inducer of the chemokine on its own (Fig. 7B, lanes 2 and 6). Therefore, TPA stimulation altered the pattern of C/EBP␤-controlled genes from one corresponding to G 0 (p20K) to one typical of the G 0 /G 1 transition (CEF-4).

C/EBP␤ Is a Negative Regulator of Fibroblast Proliferation-
The dominant negative mutant of C/EBP␤ (⌬184-C/EBP␤) abolished the induction of the p20K lipocalin gene in densityarrested CEF ( Fig. 1; Ref. 6). In addition, ⌬184-C/EBP␤ stimulated CEF proliferation implicating C/EBP␤ in growth control. Because ⌬184-C/EBP␤ interacts and inhibits other members of the C/EBP family, we cannot rule out the possibility that the dominant negative mutant interfered with the growth inhibitory activity of C/EBP␣. The latter interacts with Cdk2 and Cdk4 preventing binding of cyclins and promoting the degradation of Cdk4 (31). C/EBP␣ also represses E2F-dependent gene activation, a process required for the terminal differentiation of adipocytes and granulocytes (32,33). However, C/EBP␣ is rare in CEF and is unlikely to exert the activity described by others in adipocytes and granulocytes, where it is induced as part of the differentiation program. Although it may accumulate and inhibit the proliferation of CEF over-expressing C/EBP␤ (6), it is unlikely to be a key determinant of growth arrest in normal CEF. In addition, mouse embryo fibroblasts nullizygous for C/EBP␤ had a proliferative advantage in vitro over cells with a single or two functional copies of the gene (Fig.  2) supporting the notion that C/EBP␤ is growth inhibitory for embryonic fibroblasts. Interestingly, the expression of a second member of the family, C/EBP␦, is also induced in confluent epithelial cells. Down-regulation of C/EBP␦ expression in a mammary epithelial cell line expressing an antisense RNA enhanced proliferation, suggesting that C/EBP␤ and C/EBP␦ play a similar role in the proliferation of fibroblasts and epithelial cells, respectively (8,34).
C/EBP␤ is expressed in several tissues where it activates the expression of tissue-specific genes and participates in cell differentiation. However, C/EBP␤ is often required for the stimulation of cell proliferation in the same tissues. For instance, C/EBP␤ is widely expressed in quiescent hepatocytes and is capable of inhibiting the proliferation of hepatoma cells, but mice nullizygous for C/EBP␤ have impaired liver regeneration following partial hepatectomy (35)(36)(37). Likewise, C/EBP␤ is expressed in the mammary epithelium, which is composed  1-4) and a virus expressing c-Jun (lanes 5-8), JunD (lanes 9 -12), or Fra2 (lanes [13][14][15][16]. Confluence (RCASBP) or maximal cell density (c-Jun, JunD, Fra-2) was reached at day 2 in this experiment. ERK-1 was used to control for even loading, whereas the over-expression of each AP-1 component was confirmed with c-Jun, JunD, and Fra2 antibodies. B, effect of TPA on p20K and CEF-4 mRNA levels in RCASBP-or RCASBP-C/EBP␤-infected CEF. Northern blotting analysis was performed on mRNA isolated from CEF incubated for different time intervals in 800 nM TPA. Glyceraldehyde-3-phosphate dehydrogenase mRNA (GAPDH) was probed to control for RNA loading. predominantly of quiescent cells, but also in transformed, actively dividing mammary epithelial cells (38). Therefore, it is clear that the role of C/EBP␤ in growth control is both cell-and context-specific. Several mechanisms have now been proposed to account for the dual role of C/EBP␤ in quiescent and dividing cells. One of the proposed mechanisms is based on the expression of different C/EBP␤ isoforms. In several cell types, three different C/EBP␤ isoforms are generated by initiation at different in-frame AUG codons. The two longest forms, designated LAP* and LAP in rodents (liver-enriched activating protein) or C/EBP␤-1 and -2 (35,38,39), differ by the presence of an additional 21 amino acids located at the N terminus of LAP*/ C/EBP␤-1. Although both LAP* and LAP harbor trans-activation regions and are capable of gene induction, functional differences have been observed between these two isoforms. Indeed Kowenz-Leutz and Leutz (40) reported that the longer LAP* form, but not LAP, interacts with Swi/Snf chromatin remodeling complexes and cooperates with Myb in the induction of the mim-1 myeloid-specific gene in the context of chromatin. In contrast, LAP was able to stimulate the activity of a cyclin D1 promoter construct, whereas LAP* repressed the same promoter (38). More recently Bundy and Sealy (41) extended this study by demonstrating that LAP (C/EBP␤-2) but not LAP* transforms normal mammary epithelial cells. Other investigators have concluded that the expression of a third isoform, the truncated LIP (liver-enriched inhibitor protein) or C/EBP␤-3, is key to the process of mammary cell transformation. LIP/C/EBP␤-3 lacks trans-activating regions and therefore may inhibit cell differentiation by the longer C/EBP␤ isoforms (42). LIP is not expressed in contact-inhibited or v-Src transformed CEF and the longest C/EBP␤ isoform (C/EBP␤-1) remains the predominant species expressed in these cells (16). Therefore C/EBP␤ activity is not controlled through modulation of the ratio of C/EBP␤ isoforms in CEF.
A second mechanism entails the post-translational modification by kinases and in particular by RSK, which phosphorylates C/EBP␤ on threonine 217 in the mouse or serine 105 in the rat and is required for hepatocyte proliferation in response to transforming growth factor-␣ (43). Although relevant target genes remain to be identified, phosphorylated C/EBP␤ would enhance the expression of gene products promoting cell proliferation. More recently, Buck et al. (43) have uncovered a second activity of C/EBP␤ dependent on RSK by showing that phosphorylation on threonine 217 in the mouse generates a binding inhibitory site for procaspases 1 and 8. This function is essential for the hepatotoxin-induced proliferation of hepatic stellate cells, which undergo apoptosis in the absence of C/EBP␤ (43). A pro-survival role for C/EBP␤ has also been proposed in keratinocytes and the development of skin tumors. However the role of C/EBP␤ phosphorylation by RSK in this process is presently unknown (44). One of the key predictions of the post-translational mechanism is that C/EBP␤ would be functionally different in quiescent and cycling cells. Since uncovering targets of activation by C/EBP␤ in contact-inhibited and v-Src-transformed CEF (6,16), we began to address this question by studying the activity of fusion constructs of the activation and regulatory domains of avian C/EBP␤ with the DNA binding domain of GAL4. Using these constructs we observed a marked potentiation of the activation function in v-Src-transformed but not in contact-inhibited CEF. 2 Therefore, transcriptional activation by C/EBP␤ is likely to depend on different mechanisms or factors expressed in growth-arrested and v-Src-transformed cells. Our results suggest the existence of a third mechanism, based on the activity of other transcription factors such as AP-1, which modulates the expression of C/EBP␤-controlled genes in mitogenically stimulated cells (Fig. 7). Moreover, we present evidence that AP-1 is regulated by C/EBP␤, a process involved in the control of CEF proliferation.
AP-1 Is a Target of C/EBP␤-The activity of AP-1 was elevated in CEF expressing the dominant negative mutant ⌬184-C/EBP␤ (Fig. 3). The expression of c-Jun, JunD, and Fra-2, the main components of AP-1 in cycling CEF, was induced and accounts at least in part for the stimulation of AP-1 by ⌬184-C/EBP␤ (Fig. 4). Nearly complete repression of c-Jun, JunD, and Fra-2 was also observed in cells over-expressing C/EBP␤, indicating that the latter is a potent regulator, directly or indirectly, of AP-1 expression and activity in CEF. The c-Jun/ AP-1 dominant negative mutant TAM67 drastically reduced the proliferation of normal CEF and CEF expressing ⌬184-C/ EBP␤ (Fig. 5). These data suggest that the dominant negative mutant of C/EBP␤ regulates CEF proliferation through AP-1. Elevated levels of cyclin D1, a known target of AP-1 (27,28,45), and Rb hyperphosphorylation were also observed in CEF expressing ⌬184-C/EBP␤ but not in the presence of TAM67 (Fig.  5). Therefore a simple model for the action of ⌬184-C/EBP␤ is that the dominant negative mutant promotes the expression of c-Jun, JunD, and Fra-2, thus increasing AP-1 activity, which induces the expression of cyclin D1 and stimulates cell proliferation.
So what is the mechanism of action of ⌬184-C/EBP␤? Although we have not investigated the level of c-Jun, JunD, and Fra-2 regulation, the fact that the dominant negative mutant stimulates the expression of all three components of AP-1 suggests a common mechanism of action. No functional C/EBP binding sites have been described in the promoter of c-Jun, JunD, and Fra-2 (46 -49). However, C/EBP␤ interacts with c-Jun and p300/CBP co-activators such as AP-1 (50,51). Because AP-1 itself is one of the factors controlling the expression of c-Jun, JunD, and Fra-2 (Ref. 46 and Fig. 7), C/EBP␤ may disrupt the AP-1 loop of autoregulation.
The effect of the ⌬184-C/EBP␤ mutant on AP-1 was restricted to cycling cells (Fig. 4), which nevertheless became contact-inhibited at high cell density. Changes in the expression of c-Jun and JunD, observed in density-arrested CEF, were attenuated by the expression of ⌬184-C/EBP␤. However this was not true for Fra-2, which was repressed at contact inhibition or in response to serum starvation in the presence or absence of ⌬184-C/EBP␤. Because C/EBP␤ is also activated by mitogens and growth factors at the G 0 /G 1 transition and in response to transformation by v-Src and Ras, it may serve as a feedback mechanism restricting the activation of AP-1 (16, 17) (44) (52).
AP-1 Is a Negative Regulator of p20K Expression-The inhibition of AP-1 by C/EBP␤ may also contribute to the activation of GAS genes as cells exit the cell cycle. In support of this notion, we observed that the forced expression of c-Jun and, to a lesser extent, JunD/Fra-2 blocked the expression of the p20K lipocalin at contact inhibition (Fig. 7). Sustained activation of AP-1 also resulted in apoptosis and alterations in growth arrest at high cell density (Fig. 6). This is in agreement with results reported by other investigators working on c-Jun (28,53). Therefore C/EBP␤ and AP-1 play opposing roles in the control of GAS gene expression and CEF proliferation. The mechanism of action of AP-1 on p20K induction remains to be characterized but may entail the inhibition of C/EBP␤ by direct interaction or the competition for a common co-activator such as p300/CBP (50,51,54).
Independently of the mechanism of action, the interplay between AP-1 and C/EBP␤ will likely determine the pattern of gene expression in G 0 and at the G 0 /G 1 transition. TPA stim-ulation of CEF over-expressing C/EBP␤ reduced the ectopic expression of p20K despite the fact that TPA has been reported to enhance the activity of C/EBP␤ (52). In contrast, TPA combined with the over-expression of C/EBP␤ caused a marked induction of the CEF-4 chemokine gene, which is controlled by both C/EBP␤ and AP-1 (Fig. 7B) (17,16). This observation suggeststhatmitogenicstimulationaltersthepatternofC/EBP␤dependent gene expression from growth arrest-specific to that of the G 0 /G 1 transition. The interaction with other transcription factors may explain how a single transactivator, C/EBP␤, exerts a dual activity in the control of gene expression and cell proliferation.