C/EBPβ (NF-M) Is Essential for Activation of the p20K Lipocalin Gene in Growth-Arrested Chicken Embryo Fibroblasts

Abstract

The p20K gene is induced in conditions of reversible growth arrest in chicken embryo fibroblasts (CEF). This expression is dependent on transcriptional activation and on a region of the promoter designated the quiescence-responsive unit (QRU). In this report, we describe the regulatory elements of the QRU responsible for activation in resting cells and characterize the trans-acting proteins interacting with these elements. We show that the QRU consists of functionally distinct domains including quiescence-specific and weak proliferation-responsive elements. The quiescence responsiveness of the QRU was mapped to two C/EBP binding sites, and the activity of the p20K promoter and its QRU was inhibited by the expression of a dominant negative mutant of C/EBPβ in nondividing cells. The activation of QRU in response to serum starvation and contact inhibition correlated with the presence of a growth arrest-specific complex in electrophoretic mobility shift assays. This complex was supershifted by antibody for C/EBPβ. C/EBPβ accumulated in conditions of contact inhibition as a result of transcriptional activation. Therefore, C/EBPβ was itself regulated as a growth arrest-specific gene in CEF. Finally, we show that the expression of p20K is regulated by linoleic acid, an essential fatty acid binding to p20K. The addition of linoleic acid to contact-inhibited CEF markedly repressed the synthesis of p20K without inducing mitogenesis. The activity of the QRU was inhibited by linoleic acid or the peroxisome proliferator-activated receptor PPARγ2 in transient expression assays. Therefore, we have identified C/EBPβ as a key activator of a growth arrest-specific gene in CEF and implicated an essential fatty acid, linoleic acid, in regulation of the QRU and the p20K lipocalin gene.

Mitogenic stimulation is characterized by extensive changes in gene expression. Gene induction occurs within minutes of the addition of a mitogen or growth factor, often independently of de novo protein synthesis (3, 21, 38, 41). Many of the immediate-early genes expressed at the G₀/G₁ transition code for a transcription factor which has a viral or cellular oncogenic counterpart. Growth arrest caused by serum starvation or contact inhibition is also characterized by changes in gene expression. While the gene products up-regulated during mitogenesis are repressed in nondividing cells, a different set of genes, referred to as growth arrest-specific (GAS), growth arrest- and DNA damage-inducible (GADD), or quiescence-specific genes (8, 9, 22, 31, 77), are also induced in conditions of reversible growth arrest or G₀. The function of their gene product is generally poorly understood, but proteins of the extracellular matrix (18, 22, 58), proteins with high affinity for lipids (9, 13, 41, 73), proteins acting as negative regulators of cell proliferation (4, 13, 24, 34) or survival factors (35, 60), or proteins capable of enhancing the response of quiescent cells to mitogens (51) are part of the growth arrest-specific program of gene expression.

Despite the recent efforts in the cloning and characterization of the GAS genes, little is known about the regulatory mechanisms controlling their expression. Many of the GAS genes are regulated at the posttranscriptional level in serum-starved cells and may depend on transcript stabilization for expression (31, 80). Others are regulated at multiple levels including transcriptional activation (25, 31, 56, 58). The p53 tumor suppressor induces growth arrest in cells irradiated with γ rays and activates the transcription of the p21^waf1 and gadd45 genes (28, 43). However, these genes are also induced in response to serum starvation even in cells devoid of functional p53 (95). Therefore, multiple mechanisms, signaling pathways and trans-acting factors may control the expression of p21^waf1, gadd45, and other GAS genes. Thus far, the trans-acting factors and regulatory mechanisms controlling the expression of the GAS genes in G₀ have remained elusive, yet their characterization is likely to be important for our understanding of growth control and reversible growth arrest, in particular.

We have characterized the expression of a secretory protein, p20K, expressed by quiescent chicken embryo fibroblasts (CEF) and chicken heart mesenchymal (CHM) cells. While the function of p20K in quiescent cells is unknown, it is a member of the lipocalin family of lipid binding proteins (9, 30). The expression of p20K is markedly elevated in quiescent CHM cells and is rapidly inhibited in response to mitogenic stimulation. p20K is also synthesized by serum-starved or density-arrested CEF. In contrast, compounds which inhibit DNA synthesis, such as hydroxyurea, are poor inducers of p20K synthesis, suggesting that growth arrest in G₀ but not G₁/S controls the expression of this gene (56). In addition, p20K is not synthesized by senescent CEF or CEF undergoing apoptosis (7, 52), indicating that p20K is a marker of reversible growth arrest, i.e., of cells capable of reentering the cell cycle following mitogenic stimulation and the establishment of favorable growth conditions (5).

The expression of p20K is regulated at the transcriptional level in CEF. We have previously identified a region of the p20K promoter capable of conferring quiescence responsiveness to a heterologous, minimal promoter. This 48-bp region, termed the quiescence-responsive unit (QRU) of the p20K gene, is both essential and sufficient for activation in conditions of serum starvation and contact inhibition in CEF and rat FR3T3 fibroblasts (7, 56). In this report, we describe the characterization of the QRU and trans-acting factors controlling its activity in quiescent cells. We identified C/EBPβ (designated NF-M [nuclear factor myeloid] for the chicken homolog) as an essential regulator of p20K expression in normal, density-arrested CEF and serum-starved Rous sarcoma virus (RSV)-transformed CEF and provide evidence that the expression of C/EBPβ is itself regulated during reversible growth arrest. We show that linoleic acid, an essential fatty acid (EFA) binding to p20K (17), inhibits the activity of the QRU and expression of p20K in contact-inhibited cells. These results establish C/EBPβ as a key regulator of a GAS gene in CEF and suggest a role for EFAs in control of the p20K gene.

MATERIALS AND METHODS

Cells and viruses.

Early passages of CEF were cultured at 41.5°C in Richter-improved minimal essential medium containing insulin and zinc (I⁺ medium; Irvine Scientific, Santa Ana, Calif.), 5% heat-inactivated newborn bovine serum (MediaserII; Montreal Biotech Inc., Kirkland, Quebec, Canada), 5% tryptose phosphate broth, glutamine, penicillin, and streptomycin. CEF were also infected with the Schmidt-Ruppin A strain of RSV (SR-A RSV) or with recombinant viruses generated with the RCASBP(B) retroviral vector. RSV-infected CEF were starved in the complete absence of serum in Dulbecco’s modified Eagle’s medium containing 10% tryptose phosphate broth. To maintain a constant pH, 50 mM HEPES (pH 7.4) was added routinely to the culture medium of RSV-transformed cells. Cells, cultured in serum-containing medium, were also treated overnight (16 h) with various concentrations (25 to 200 μM) of fatty acids dissolved in ethanol. Cells with lipid-containing droplets were fixed in 3.7% formaldehyde in phosphate-buffered saline (PBS) and then stained for 1 h with a 0.25% solution of oil red O dissolved in 40% ethanol (71).

Promoter and expression vector constructs.

The avian C/EBPβ cDNA was inserted in the Cla12 adapter plasmid and subcloned in the unique ClaI site of the RCASBP(B) retroviral vector to generate plasmid RCASBP-C/EBPβ (16, 69). The resulting vector was transfected in CEF, and a productive infection was established by the replication-competent avian retrovirus. The same procedure was followed to generate the RCASBP-20K virus expressing the p20K gene. The p12E CAT (chloramphenicol acetyltransferase) reporter plasmid containing 2.3 kb of the 5′ flanking region of the p20K gene was described previously (56). Constructs of the QRU of the p20K promoter (corresponding to positions −217 to −169) were generated with plasmid pJFCAT-TATA, which includes a minimal promoter consisting of a TATAAAA box and the initiation start site of the human β-globin gene. To generate these minimal promoter constructs, synthetic double-stranded oligonucleotides representing various regions of the QRU were multimerized and inserted in the HindIII site of plasmid pJFCAT-TATA. The intact QRU and several mutant derivatives of this 48-bp DNA fragment were also synthesized chemically and inserted in the SalI site of the same plasmid. These mutations are shown in Fig. 1.

FIG. 1.

Schematic representation of the QRU. Potential binding sites for the C/EBP and Ets families of transcription factors as well as the position of a palindrome are indicated. The QRU was arbitrarily divided into three overlapping regions designated regions A, B, and C and analyzed in transient expression assays. Mutations generated in the QRU are indicated by lowercase letters.

Transient expression assays and conditions of quiescence.

All transfections were done by calcium phosphate precipitation as described previously (36, 56). Briefly, dense cultures of normal or RSV-transformed CEF grown in 100-mm-diameter dishes were transfected with a total of 30 μg of DNA consisting of 10 μg of test reporter plasmid, 2 μg of the LacZ-containing plasmid pCH110, and 18 μg of salmon sperm carrier DNA. In experiments with expression plasmids, 2 to 10 μg of plasmid pCDM8 encoding avian C/EBPα or -β or the peroxisome proliferator-activated receptor (PPAR) expression plasmid pCMV2-PPARγ2 were included in the transfection mixture. The plasmid encoding the dominant negative mutant of C/EBPβ (Δ184) was described by Kowenz-Leutz et al. (48); 10 μg of the Δ184 expression plasmid was cotransfected as described above. For all transient expression assays, the activity of the CAT reporter enzyme was determined in lysates representing equal levels of β-galactosidase activity. All constructs were analyzed in duplicate or triplicate in at least two separate experiments. The conversion of chloramphenicol into its acetylated forms was quantitated in an InstantImager (Packard/Canberra). The results of previous studies indicated that the simian virus 40 enhancer/promoter of plasmid pCH110 is not regulated in our conditions of proliferation and quiescence and therefore can serve as an internal standard to control for differences in transfection efficiency.

The transfection of serum-starved or density-arrested CEF resulted in a poor expression of any of the promoter constructs investigated thus far, including reporter genes driven by strong, constitutive promoter/enhancers used as internal standards (55). Therefore, two different protocols were developed to induce quiescence in actively dividing CEF transfected with p20K promoter constructs. Our original procedure was based on the transfection of dense RSV-transformed CEF (56). In this protocol, quiescence is obtained by resuspending the cells in serum-free medium the day after transfection and preparing the cell lysate 48 h after transfection. Dense cultures of RSV-transformed CEF deplete the medium of essential nutrients and growth factors more rapidly than their normal counterparts and do not need to be starved for more than 24 h to become quiescent and express p20K. A second protocol was developed to study the activation of the p20K promoter in normal, density-arrested CEF. In this protocol, cultures of normal CEF grown at approximately 80% confluence were transfected by the procedure described above. Calcium phosphate precipitation is essential in this protocol because calcium acts as a strong mitogen for the cells, which must reach confluence and become contact inhibited before preparation of the lysate. In these conditions, the normal CEF reach confluence the day after transfection, at which time the cells are refed with serum-containing medium to maximize contact inhibition at the time of cell lysis (48 h after transfection). Actively dividing normal or RSV-transformed CEF were obtained by trypsinization of the culture the day after transfection and replating of the cells at one-sixth of the original density in fresh serum-containing medium. Both protocols yielded the same results unless otherwise specified in the text. Results obtained with the starvation of RSV-transformed CEF or contact inhibition of normal CEF are presented to emphasize specific points or differences relevant to the activation of the QRU.

Electrophoretic mobility shift assay (EMSA).

Nuclear extracts were prepared by a modification of the methods of Briggs et al. (14) and Dignam et al. (27). Routinely, 10⁷ cells were washed once with cold PBS, collected for 5 min in a microcentrifuge at 1,500 × g, and resuspended in 500 μl of buffer A (10 mM HEPES [pH 7.9], 10 mM KCl, 0.5 mM dithiothreitol [DTT], 1.5 mM MgCl₂, cocktail of protease and phosphatase inhibitors consisting of 0.5 mM phenylmethylsulfonyl fluoride, 0.3 μg of leupeptin per ml, 0.3 μg of antipain per ml, 0.5 μg of aprotinin per ml, 1 mM sodium fluoride, and 1 mM sodium vanadate). Cells were lysed with a glass Dounce homogenizer (B pestle), and the nuclei were pelleted by centrifugation at 12,000 × g for 20 min. The nuclear pellet was extracted with 1 ml of buffer C (20 mM HEPES [pH 7.9], 25% glycerol, 1.5 mM MgCl₂, 0.42 mM KCl, 0.5 mM DTT, 1 mM EDTA, protease and phosphatase inhibitors as mentioned above) for 20 min at 4°C with continuous agitation. At the end of the incubation, the reaction mixture was centrifuged at 12,000 × g for 20 min, and the supernatant was dialyzed against buffer D (20 mM HEPES [pH 7.9], 20% glycerol, 0.1 M KCl, 0.5 mM DTT, 0.2 mM EDTA, protease and phosphatase inhibitors). Alternatively, the nuclear lysate was partially purified by ammonium sulfate precipitation. In this case, the nuclear lysate was centrifuged at 80,000 × g for 1 h at 4°C, and 0.33 g of ammonium sulfate was added slowly per ml of the resulting supernatant. The mixture was then incubated for 30 min at 4°C with gentle agitation and finally centrifuged at 21,000 × g for 20 min. The supernatant was discarded, and the pellet was resuspended in buffer D and dialyzed against the same buffer for 5 h at 4°C. Both methods gave the same results. The dialysate from the purified nuclear extract was aliquoted and stored at −80°C until needed. Protein concentration was determined by the Bradford assay using bovine serum albumin as a standard.

Sequences of synthetic DNA oligomers used for DNA binding analysis are shown below in Table 1; mutations of the QRU are described in Fig. 1. The NF-M oligonucleotide is a composite element derived from the two NF-M (C/EBPβ) binding sites of the chicken myelomonocytic growth factor (cMGF) promoter (82). All probes were generated by filling in the ends with Klenow enzyme, using [³²P]dATP, [³²P]dCTP, and unlabeled dGTP and TTP. Binding reactions were performed in a total volume of 20 μl of 1× binding buffer (20 mM Tris-HCl [pH 7.5], 1 mM MgCl₂, 60 mM NaCl, 5% glycerol) for 30 min at room temperature. For EMSAs, the ³²P-end-labeled oligonucleotides (approximately 10,000 cpm/0.1 ng) were incubated with 2 to 5 μg of nuclear extracts in the presence of 2 μg of poly(dI-dC). Competition binding reactions were performed by preincubating the nuclear extract with an excess of unlabeled, filled-in oligonucleotides for 15 min at room temperature. This was followed by the addition of labeled oligonucleotides and incubation for an additional 30 min. The DNA-protein complexes were resolved on 4.8% nondenaturing polyacrylamide gels in 0.5× Tris-borate-EDTA buffer and visualized by autoradiography.

TABLE 1.

Synthetic DNA oligomers used

Oligomer	Sequencea
QRU (−217 to −169)	5′AGCTTCTCCTCAGGGCTTGCAACACTTTCCTCTTTCCGTAAGCGTCTGTTTACA3′
	AGAGGAGTCCCGAACGTTGTGAAAGGAGAAAGGCATTCGCAGACAAATGTTCGA
(−217 to −198)
Region A	5′AGCTTCTCCTCAGGGCTTGCAACA3′
	AGAGGAGTCCCGAACGTTGTTCGA
(−200 to −179)
Region B	5′AGCTTACACTTTCCTCTTTCCGTA3′
	ATGTGAAAGGAGAAAGGCATTCGA
(−188 to −169)
Region C	5′AGCTTCCGTAAGCGTCTGTTTACA3′
	AGGCATTCGCAGACAAATGTTCGA
(−217 to −198)
μpalA	5′AGCTTCTggTCAaaGCTTGCAACA3′
	AGAccAGTttCGAACGTTGTTCGA
μB-Ets (−200 to −179)	5′AGCTTACACTTTggTCTTTggGTA3′
	ATGTGAAAccAGAAAccCATTCGA
NF-Mb	5′AGCTTCACAATGAGGCAACA3′
	AGTGTTACTCCGTTGTTCGA

^aLowercase letters indicate mutations introduced in the sequence. 

^bComposite consisting of two NF-M binding sites of the cMGF promoter (82). 

Immunoprecipitation, Western blotting analysis, and production of C/EBPβ antiserum.

The synthesis of p20K was investigated by immunoprecipitation analysis of proteins metabolically labeled with [³⁵S]methionine followed by separation on sodium dodecyl sulfate (SDS)-polyacrylamide gels and fluorography as described previously (56). For Western blotting analyses, 40 μg of total cell protein extract prepared in SDS sample buffer was subjected to SDS-polyacrylamide gel electrophoresis and blotted on nitrocellulose (BA85; Schleicher & Schuell). A polyclonal antibody for avian C/EBPβ was kindly provided by K.-H. Klempnauer and used at a dilution of 1:2,000 in a 0.1% solution of milk dissolved in PBS (42). This was followed by incubation with a peroxidase-conjugated secondary antibody (Pierce, Rockford, Ill.) and detection with an enhanced chemiluminescent substrate according to protocols provided by the manufacturer (Amersham). Antibodies for avian C/EBPβ were also generously provided by A. Leutz or generated in a rabbit by using a recombinant avian C/EBPβ protein as the antigen and established immunization protocols (8). Antibodies for ERK-1 and ERK-2 (SC-93 and SC-154, respectively) were obtained from Santa Cruz Biotechnology (Santa Cruz, Calif.).

Northern blotting and transcription run-on analyses.

RNA was purified by high-salt urea precipitation and analyzed on Northern blots as described by Mao et al. (56). Run-on transcription assays were performed as described by Dehbi et al. (23). In both analyses, glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used as an internal standard to control for loading or hybridization. The intensity of radioactive bands on blots was quantitated with an InstantImager (Packard/Canberra).

Screening of cDNA library.

A contact-inhibited CEF cDNA library (9) was screened with the SacII restriction fragment of the chicken C/EBPα gene corresponding to the DNA binding and leucine zipper domains of the protein. A similar screen was done with a DNA fragment of the murine CHOP-10/gadd153 cDNA (also corresponding to the DNA binding domain of the protein; 75). The final wash of nitrocellulose filters harboring phage DNAs was performed with 0.5× SSC (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate)–0.1% SDS at 55°C (C/EBPα) or with 2× SSC–0.1% SDS at room temperature (CHOP-10). Following secondary and tertiary screens of positive phages, excision of phagemids was accomplished with the ExAssist helper phage according to protocols provided by the manufacturer (Stratagene, La Jolla, Calif.). Following excision, clones were analyzed by a combination of restriction enzyme and sequencing analyses followed by FastA database analysis (68).

RESULTS

Characterization of functional domains of the QRU.

The region of the promoter required for activation of the p20K gene in quiescent cells was termed the QRU. The 48-bp QRU confers quiescence responsiveness to an heterologous promoter in conditions of serum starvation or contact inhibition and is necessary and sufficient for induction in growth-arrested CEF and rat FR3T3 cells (7, 56). EMSAs indicated that numerous DNA binding proteins interact with the 48-bp QRU in conditions of quiescence and logarithmic proliferation (55). To begin the characterization of the QRU, we identified potential binding sites for known transcription factors by using the program Signal Scan (72) and performed a series of transient expression assays with individual segments of the QRU. The analysis of potential binding sites is presented in Fig. 1 and 2. Two consensus elements for the C/EBP and Ets families are present within the QRU. A sequence located at the 3′ end of the QRU (GCGTCTG) shows similarity to the core sequence recognized by the Maf family of transcription factors (44, 70). A palindrome (CCTCAGG) with no extensive similarity to known regulatory elements is also found at the 5′ end of the 48-bp fragment. To identify functional domains within the QRU, we then arbitrarily defined three partially overlapping segments designated regions A, B, and C and analyzed their activity in transient expression assays. Region A includes the first C/EBP binding site and the palindrome located at the 5′ end of the QRU. Region B contains the two potential Ets response elements and the second C/EBP binding site. Region C includes the second C/EBP binding site and the potential Maf recognition element. Synthetic double-stranded oligonucleotides corresponding to each region were multimerized, inserted in proximity to a minimal promoter, and investigated in actively dividing or serum-starved RSV-transformed CEF. A construct containing two copies of region A was strongly activated in quiescent cells, with a 40-fold induction over the level observed in actively dividing cells (construct A-2 in Fig. 3). In contrast, constructs containing two or three copies of region B or C did not respond to growth arrest and were in fact more active in proliferating cells. This proliferation-dependent activity was more obvious with region B but remained modest when compared to the strong activity of region A in nondividing cells. Therefore, region A contains a potent quiescence-responsive element, but the QRU is composed of multiple, functionally distinct domains.

FIG. 2.

Sequence similarity between the QRU and the binding sites of known transcription factors. The consensus binding sites of known transcription factors were obtained from reference 29 and other references indicated in the text. Numbers indicate positions and strands of the elements of the QRU (e.g., −199 to −207 refers to the minus strand, whereas −189 to −181 indicates the position of the element on the plus strand). N represents any nucleotide.

FIG. 3.

Characterization of functional elements of the QRU. Double-stranded oligonucleotides corresponding to various regions of the QRU were cloned in the minimal promoter plasmid pJFCAT-TATA and investigated in transient expression assays. For each construct, the number of inserted copies of region A, B, or C or μB-Ets is indicated by a numeral and the orientation is indicated by − or +. Normalized values of CAT activity expressed as the means of duplicate samples are indicated for serum-starved CEF (0% serum) and actively dividing RSV-transformed CEF (5% serum).

Since several members of the Ets family have been implicated in gene activation in response to mitogens (12, 90), we synthesized a modified version of the region B oligonucleotide containing a mutation in both potential Ets binding sites but affecting noncritical residues of the putative C/EBP binding site (oligonucleotide μB-Ets in Fig. 1 and constructs μB+2 and μB+3 in Fig. 3) (33, 66, 84). In transient expression assays, mutation of the potential Ets binding sites reduced the activity of region B in actively dividing cells, in agreement with the notion that one or multiple proliferation-responsive elements are also present in the QRU. The activity of regions B and C was observed in RSV-transformed CEF but not in normal, actively dividing cells (46). Therefore, the proliferation-specific activity of regions B and C may be restricted to v-src-transformed cells. This activity was not characterized further.

C/EBP is required for activation of the QRU.

The overexpression of several members of the Maf (c-MAFI, c-MAFII, MAF B, MAF F, MAF G, and MAF K) and Ets (PEA3 and c-Ets-1) families did not affect the activity of the QRU in CEF (46). In contrast, members of the C/EBP family were potent activators of the QRU. Results shown in Fig. 4 demonstrate the action of C/EBPα and -β on the p20K promoter (p12E), on the 48-bp QRU, and on individual regions of the QRU. The QRU constructs studied in this analysis are depicted in Fig. 4A. C/EBPα and -β were strong activators of p12E, the QRU, and a multimer of region A (construct 3XA in Fig. 4). Likewise, regions B and C (constructs 3XB and 3XC) were also activated by C/EBP, but to a lesser extent (35- to 60-fold activation for region A, versus 3- to 12-fold for regions B and C). pJFCAT-TATA, the parental promoter construct, did not respond to the overexpression of C/EBPα or -β. The replacement of two nucleotides known to be critical for DNA binding by C/EBP markedly impaired the activation of a construct of region A by C/EBPα or C/EBPβ (construct 2XμA in Fig. 4). Constructs μB+2 and μB+3, harboring mutations in the two potential Ets binding sites of region B, were still activated by the overexpression of C/EBPα or -β (7). We also studied the effect of a mutation in one or both of the potential C/EBP binding sites of the QRU. As expected, the mutation of both sites nearly abolished the activation of the QRU by C/EBP (QRU-μ2C/EBP in Fig. 4). The single mutants were still well activated, suggesting that both sites are effective targets of C/EBPα or -β, at least in conditions where these factors are overexpressed (QRU-μA and QRU-μB). The mutation of the palindrome of region A was also activated efficiently (QRU-μpalA). This is not surprising, considering that the palindrome does not overlap with the core sequence of the C/EBP binding site of region A (Fig. 1). Therefore, we conclude that the QRU contains two C/EBP-responsive elements and is strongly transactivated by the α and β members of this family. In contrast, none of the QRU constructs were affected by CHOP-10/Gadd153 when this member of the C/EBP family was expressed alone or in combination with C/EBPα or -β (46).

FIG. 4.

Activation of the p20K promoter and QRU by C/EBP. (A) Schematic representation of promoter constructs studied in this analysis. p12E is a construct of the p20K promoter comprising the −2289 to +42 region of p20K genomic DNA. All other constructs were generated in the minimal promoter plasmid pJFCAT-TATA. (B) Effects of C/EBP on the p20K promoter (p12E), the QRU, and various regions of the QRU, determined by transient expression assays in actively dividing CEF. Constructs of the QRU consist of one copy of the wt QRU or a mutant derivative containing a mutation in one (QRU-μA or QRU-μB) or both (QRU-μ2C/EBP) C/EBP binding sites. QRU-μpalA is a derivative of the QRU containing nucleotide changes in the palindrome identified in region A. Constructs 3XA, 3XB, and 3XC contain three copies of regions A, B and C, respectively, inserted in the minimal promoter plasmid pJFCAT-TATA. A similar construct was generated with a derivative of region A containing a substitution of two critical residues of the C/EBP binding site (2XμA). pJFCAT-TATA is the parental CAT plasmid devoid of any element of the QRU. Cotransfection was done with 10 μg of test plasmid, 10 μg of effector plasmid, and 2 μg of the LacZ-containing plasmid pCH110 in a 100-mm-diameter dish. The activity of CAT was determined in duplicate samples for CEF transfected with an expression vector for C/EBPα or C/EBPβ or with the parental expression vector, pCDM8.

To confirm the role of C/EBP in activation of the QRU, we studied the effect of a mutation in a single or both C/EBP binding sites of the QRU in actively dividing and contact-inhibited CEF. As shown in Fig. 5A, the quiescence-dependent activation of the QRU was completely eliminated by the mutation of either one of the two C/EBP binding sites. Interestingly, the site located in region B or C was essential when studied in the context of the QRU since a mutation in this element abolished the activation by contact inhibition. In contrast, the mutation of the palindrome of region A had no effect on the activation by quiescence (construct QRU-μpalA in Fig. 5A).

FIG. 5.

(A) Role of the C/EBP binding sites in the activity of the QRU. Activities of QRU constructs containing a mutation in one (QRU-μA or QRU-μB) or both (QRU-μ2C/EBP) C/EBP binding sites or in the palindrome of region A (QRU-μpalA) were analyzed by transient expression assays in actively dividing and contact- inhibited CEF. The activation is indicated as the ratio of the activity in quiescent over actively dividing cells (defined as 1). (B) Effect of a dominant negative mutant of C/EBPβ on the activity of the p20K promoter. Plasmids p12E, QRU, and AP1+3 were transfected alone, with a plasmid encoding a dominant negative mutant of C/EBPβ (D184 for Δ184), or with the parental vector (pCDM8) and analyzed by transient expression assays in contact-inhibited CEF. Plasmid p12E is a construct of the p20K promoter. Plasmid QRU contains a single copy of the QRU in the minimal promoter vector pJFCAT-TATA; plasmid AP1+3 contains three copies of the tetradecanoyl phorbol acetate-responsive element inserted in the same vector. The results represent the average of duplicate samples.

The activity of the QRU was then analyzed in the presence of a dominant negative mutant of C/EBPβ (Δ184). This mutant, constructed by Kowenz-Leutz and coworkers, encodes a C/EBPβ protein lacking a transactivating region (48). The expression of the deleted form of C/EBPβ markedly reduced the activity of a construct of the QRU or p20K promoter (p12E) in contact-inhibited cells (Fig. 5B). In contrast, plasmid Δ184 did not affect the activity of a construct controlled by three copies of the tetradecanoyl phorbol acetate-responsive element (AP1+3). Likewise, the overexpression of a dominant negative mutant of c-Jun (Tam-67 [2]) had no effect on the activity of the QRU in contact-inhibited cells (46). Taken together, these results indicate that the activation of the QRU is dependent on a member of the C/EBP family in growth-arrested CEF.

C/EBPβ (NF-M) binds to the QRU in quiescent CEF.

To determine if a member of the C/EBP family interacts with the QRU, we performed a series of EMSAs with regions of the QRU used as probes. Nuclear extracts were prepared from normal actively dividing and density-arrested CEF and were incubated with double-stranded radiolabeled oligonucleotides. A quiescence-specific complex, designated C1 in Fig. 6A (lanes 1 and 2), was observed with region A. Complexes migrating approximately at the same position were also apparent when regions B and C were incubated with the quiescent cell nuclear extract (lanes 3 to 6). A series of fast-migrating complexes, referred to collectively as C2, were also obtained with all three regions of the QRU. However, complexes C2 were not observed with region A containing a mutation in the palindrome, indicating that their formation required sequences adjacent to the C/EBP binding site (Fig. 6B, lane 8). Since complexes C2 were obtained with all three regions of the QRU, it is probable that their formation depends on the putative Ets binding sites (TCCT or TTCC) disrupted in the μpalA mutant. Complex C1 comigrated with a complex obtained with an unrelated C/EBP binding site, namely, the NF-M oligonucleotide, while complexes C2 were not formed with this probe (lane 9). Complex C1 was also observed when a nuclear extract was prepared from serum-starved but not actively dividing SR-A RSV-transformed CEF (lanes 10 to 12).

FIG. 6.

Analysis of nuclear proteins binding to the QRU by EMSA. (A) Analysis of nucleoprotein complexes obtained with regions A, B, and C incubated with a nuclear extract from actively proliferating CEF (P) or quiescent, density-arrested CEF (Q). The positions of complexes C1 and C2 are indicated. (B) Patterns of nucleoprotein complexes obtained with region A (lane 7) or a derivative of region A containing mutations disrupting a palindromic sequence identified in this region (lane 8; μpalA). A radiolabeled, double-stranded oligonucleotide corresponding to the C/EBP element designated NF-M was used as a probe in lane 9. (C) Nucleoprotein complexes obtained when region A is incubated with a nuclear extract from SR-A RSV-transformed CEF cultured in the presence (5%) and absence (0%) of serum (lanes 10 and 11, respectively). The pattern obtained with a nuclear extract prepared from normal, density-arrested CEF (P) is shown in lane 12. (D) Supershift of complex C1 with (+) and without (−) a C/EBPβ antibody (ab; arrowhead in lanes 15 and 16). As a control, the preimmune serum was used in lanes 17 and 18.

The specificity of the quiescence-specific complex C1 was determined by competition EMSA (Fig. 7). A 50-fold excess of the region A oligonucleotide competed for the formation of complex C1 (lane 12 in Fig. 7). In contrast, similar oligonucleotides differing by two nucleotides critical for C/EBP binding did not compete complex C1 even when present at a 50-fold molar excess (oligonucleotide μA [lanes 10 and 11]). A 50-fold excess of region B and NF-M oligonucleotides also could compete for formation of complex C1 (lanes 15 and 19, respectively). However, region C was a poor competitor of complex C1 even at a 50-fold molar excess (lane 17). Similar competition assays were performed with oligonucleotides corresponding to the QRU or derivatives containing a mutation in one or both C/EBP binding sites. Complex C1 was competed by a 50-fold excess of the wild-type (wt) QRU and the QRU containing a mutation in either one of the two C/EBP binding sites (lanes 3, 5, and 7). In contrast, a QRU fragment containing a mutation in both C/EBP binding sites did not eliminate complex C1 (lane 9). Taken together, these results indicate that the formation of the quiescence-specific complex C1 is dependent on nucleotides that are also critical for binding by C/EBP.

FIG. 7.

Complex C1 is competed by oligonucleotides containing a C/EBP binding site. Region A was used as a probe in EMSA and incubated with a nuclear extract from density-arrested CEF. The specificity of the nucleoprotein complexes was assessed by preincubating the extract with a 10- or 50-fold molar excess of unlabeled oligonucleotides corresponding to arbitrarily defined regions of the QRU (A, B, and C), a derivative of region A containing a mutation in the C/EBP binding site (μA), or the C/EBP element designated NF-M. In lanes 2 to 9, competition is performed with the wt QRU or a derivative of the QRU containing a mutation of the C/EBP element in region A (QRU-μA), a mutation of the C/EBP element in region B (QRU-μB), or a mutation in both C/EBP binding sites (QRU-μ2C/EBP).

To identify a relevant activator(s) of the QRU, we screened a contact-inhibited CEF cDNA library with a DNA fragment corresponding to the conserved DNA binding domain of avian C/EBPα. More than 15 independent clones were isolated and then characterized by restriction digest and sequencing analyses. The member most frequently isolated was NF-M, i.e., the avian C/EBPβ (82). A minority of the clones encoded C/EBPα, which was the only other member of the family isolated in this screen. A similar screen was performed with the DNA binding region of murine CHOP-10 but no C/EBP-related clones, including one encoding CHOP-10, were isolated, in agreement with the notion that the expression of this factor is induced predominantly in conditions of glucose depletion and stress (89).

To determine if C/EBPβ binds to the QRU in quiescent cells, a nuclear extract prepared from contact-inhibited CEF was preincubated with a C/EBPβ antibody or with the preimmune serum and analyzed by EMSA. As shown in Fig. 6D, the majority of complex C1 was supershifted by the C/EBPβ antibody but not by the preimmune serum, indicating that C/EBPβ is a component of C1. The results of other assays not shown in this report confirmed that the entire C1 complex can be supershifted or eliminated by antibodies for C/EBPβ. A less abundant C/EBPβ-containing complex was also supershifted by the antibody in the sample corresponding to actively dividing CEF (arrowhead in Fig. 6D). Complex C1 was also supershifted or eliminated by C/EBPβ antibodies when a nuclear extract of serum-starved RSV-transformed CEF was incubated with region A (7). Therefore, C/EBPβ is a component of the quiescence-specific complex C1 of normal density-arrested and serum-starved RSV-transformed CEF.

p20K and the QRU are regulated by linoleic acid in CEF.

To determine if C/EBPβ can regulate the expression of the resident p20K gene and not only that of promoter constructs activated transiently, we constructed a replication-competent retrovirus capable of infecting CEF and overexpressing C/EBPβ ectopically. CEF infected with the parental virus, RCASBP, were used as a control in this experiment. The overexpression of C/EBPβ was verified on a Western blot (Fig. 8A), while its nuclear localization was confirmed by indirect immunofluorescence analysis (46). Total RNA was purified from CEF infected with RCASBP-C/EBPβ or RCASBP or from uninfected, contact-inhibited CEF. Subconfluent cells infected with a retrovirus were investigated in serum-containing medium, i.e., in conditions where p20K is normally not expressed. As shown in Fig. 8B, CEF infected with RCASBP-C/EBPβ but not RCASBP expressed elevated levels of the p20K mRNA. This level was comparable to that found in contact-inhibited CEF. Likewise, the synthesis of p20K was markedly stimulated in CEF overexpressing C/EBPβ but not in CEF infected with the parental virus, RCASBP (Fig. 8C). CEF infected with a virus encoding p20K (RCASBP-20K) express high levels of the protein and thus provided a positive control in this experiment.

FIG. 8.

Analysis of p20K expression in CEF infected with a retrovirus encoding C/EBPβ (RCASBP-C/EBPβ). (A) Western blotting analysis of C/EBPβ in CEF infected with RCASBP-C/EBPβ or the parental vector RCASBP. (B) Northern blotting analysis of p20K mRNA in CEF infected with RCASBP or RCASBP-C/EBPβ or in uninfected but density-arrested CEF (quiescent CEF). RNA loading was controlled by probing the blot with a GAPDH cDNA. (C) Synthesis of the p20K protein examined by metabolic labeling with [³⁵S]methionine followed by immunoprecipitation, separation on a polyacrylamide gel, and fluorography. As a positive control, p20K was immunoprecipitated from CEF infected with a retrovirus expressing the p20K cDNA (RCASBP-20K).

CEF infected with RCASBP-C/EBPβ rapidly developed vesicles reminiscent of the lipid droplets found in adipocytes (37). Staining of the cells with oil red O confirmed that these vesicles contained lipids (Fig. 9B). Lipid-containing vesicles were not observed in serum-starved or density-arrested CEF (Fig. 9C or D, respectively). The same was true for actively dividing or serum-starved RSV-transformed CEF (7). However, since p20K is a lipid binding protein of the lipocalin family (9, 30), it is possible that its expression reflects the differentiation of CEF into adipocytes and not reversible growth arrest. To resolve this issue, we treated CEF with increasing concentrations of different lipids and examined the synthesis of p20K in these conditions. The addition of a high concentration (200 μM) of the polyunsaturated fatty acid linoleic acid induced the formation of lipid droplets as efficiently as overexpressed C/EBPβ (Fig. 9H). Smaller vesicles appeared with concentrations of linoleic acid as low as 25 μM in cells treated for 16 h. In contrast, the addition of palmitic or oleic acid induced the formation of only very small vesicles even at a concentration of 200 μM (Fig. 9F or G, respectively). CEF treated with lipids were metabolically labeled with [³⁵S]methionine, and the synthesis of p20K in actively dividing or density-arrested CEF was analyzed by immunoprecipitation. No synthesis of p20K was detected when lipid-treated CEF were actively dividing (Fig. 10A, lanes 1 to 4). Moreover, the synthesis of p20K was markedly reduced in density-arrested cells treated with linoleic but not palmitic acid (lanes 8 and 9). Linoleic acid did not induce DNA synthesis in growth-arrested cells, indicating that the repression of p20K synthesis was not the result of mitogenic stimulation (Table 2). Likewise, the addition of linoleic acid, in the experimental conditions described for Fig. 10, did not induce apoptosis (46). Therefore, we conclude that linoleic acid, a known ligand of p20K (17), is a negative regulator of p20K expression. These results imply that the expression of p20K is not the result of growth-arrested CEF being committed to adipogenesis.

FIG. 9.

The overexpression of C/EBPβ or the addition of linoleic acid induces the formation of lipid droplets in CEF. CEF infected with RCASBP (A) or RCASBP-C/EBPβ (B) were stained for lipid-containing vesicles with the lipophilic dye oil red O. (C) Oil red O staining was also performed on CEF transferred to serum-free medium and starved for 2 days. The formation of lipid-containing vesicles was also investigated in density-arrested CEF (D); in this sample, confluent CEF were treated with fresh serum-containing medium and stained with oil red O the day after. Staining was also performed on uninfected CEF treated with 200 μM palmitic, oleic, or linoleic acid for 16 h (F, G, or H, respectively) or treated with the diluent for the same period (0.1% ethanol) (E).

FIG. 10.

Regulation of p20K expression by linoleic acid and PPARγ2. (A) The synthesis of p20K was examined by metabolic labeling with [³⁵S]methionine and immunoprecipitation analysis in CEF treated for 16 h with 200 μM palmitic acid or linoleic acid in conditions of logarithmic proliferation (lanes 2 and 3) or contact inhibition (lanes 8 and 9). As a control, the synthesis of p20K was also examined in actively dividing CEF infected with RCASBP or RCASBP-C/EBPβ (lanes 5 and 6). CEF treated with the diluent, 0.1% ethanol, were also analyzed (lanes 1 and 7) and compared with untreated CEF (lanes 4 and 10). (B) The activity of a QRU construct was examined by transient expression assays in growth-arrested (quiescent [Q]) CEF treated with a low concentration of linoleic acid (LA; 25 μM) for 16 h and/or cotransfected with 2 μg of an expression vector for PPARγ2. (C) The activation of the QRU was examined upon ectopic expression of C/EBPα or -β in cells cotransfected or not with an equal amount (2 μg) of an expression vector for PPARγ2.

TABLE 2.

DNA synthesis in density-arrested CEF treated with lipids^a

Treatment	[3H]thymidine incorporation (103 cpm)/μg of protein ± SE [n = 4])
0.1% ethanol	1.29 ± 0.04
200 μM palmitic acid	1.09 ± 0.06
200 μM oleic acid	1.18 ± 0.21
200 μM linoleic acid	1.24 ± 0.05
Change of medium	1.82 ± 0.12

^aConfluent, density-arrested CEF were treated with different fatty acids at a concentration of 200 μM or with the diluent (0.1% ethanol) for 20 h. [³H]thymidine (1 μCi/ml) was added at the same time to measure DNA synthesis. In parallel, [³H]thymidine uptake was measured in density-arrested CEF transferred to fresh serum-containing medium for the same period of time; a modest increase in DNA synthesis was observed in these conditions. 

To investigate the mechanism of action of linoleic acid, we performed a series of transient expression assays with the QRU construct transfected in CEF treated with this lipid or the diluent alone (Fig. 10B). No stimulation of the QRU was observed at any of the added concentrations of linoleic acid (7). A suboptimal concentration of linoleic acid (25 μM) reduced the activity of the QRU in density-arrested CEF, suggesting that linoleic acid regulates the expression of p20K at least in part at the transcriptional level. Two transcription factors, C/EBPα and PPARγ2, promote adipogenesis in model cell systems of adipogenesis (53). Since the former is a potent activator of the QRU (Fig. 4B), we examined the effect of the overexpression of PPARγ2 in density-arrested CEF. As shown in Fig. 10B, PPARγ2 considerably reduced the activity of the QRU in quiescent cells. This potent activity of PPARγ2 did not require the addition of linoleic acid to the culture medium. However, PPARγ2 did not significantly decrease the activation of the QRU when C/EBPα or -β was overexpressed in transient expression assays (Fig. 10C). Taken together, these results indicate that p20K and the QRU are regulated negatively by factors promoting adipogenesis, namely, linoleic acid and PPARγ2.

The expression of C/EBPβ is induced by contact inhibition.

CEF express nearly undetectable levels of C/EBPβ (NF-M) compared to cells of the myelomonocytic lineage (45). Therefore, we examined the expression of C/EBPβ in actively dividing and density-arrested CEF. As determined by Western blotting analysis (Fig. 11B), a significant accumulation of the 44-kDa C/EBPβ protein was observed in contact-inhibited cells. No other forms of C/EBPβ were detected in this analysis using different antisera, including one recognizing the conserved C terminus of the protein (corresponding to the dimerization and DNA binding domains [32]). Thus, a low-molecular-weight protein corresponding to the liver-enriched inhibitory protein form of C/EBPβ was not detected (26). As expected, p20K was also more abundant in conditions of growth arrest whereas the level of ERK-1 and -2 was not regulated by the proliferative state of the cell. The transcription of the C/EBPβ gene was also more active in contact-inhibited CEF, indicating that the accumulation of the protein depends at least in part on transcriptional activation of the gene (Fig. 11A). Following quantitation in an InstantImager, we determined that contact inhibition caused a sixfold activation of transcription of the C/EBPβ gene. Hence, C/EBPβ was itself regulated as a GAS gene in CEF. The accumulation of C/EBPβ is thus one of the mechanisms controlling the induction of p20K in quiescent CEF.

FIG. 11.

Activation of C/EBPβ expression by contact inhibition. (A) Transcription of the p20K, C/EBPβ, and GAPDH genes was examined by run-on transcription assays performed with nuclei isolated from contact-inhibited (Q [quiescent]) or actively proliferating (P) CEF. (B) The abundance of p20K and C/EBPβ was determined by Western blotting analysis of total cell protein extract of density-arrested (Q) or actively proliferating (P) CEF. Protein loading was monitored by incubating the blot with antibodies for ERK-1 and -2.

DISCUSSION

Activation of the p20K gene is dependent on C/EBP.

Gene activation has been intensively investigated in cells submitted to ionizing radiations. In these conditions, the p53 tumor suppressor is essential for the induction of p21^waf1, gadd45, and gadd153 (28, 43, 95). However, in response to other stresses and conditions causing growth arrest, these genes remain at least partly inducible in the absence of functional p53, suggesting that they are regulated by multiple transcription factors (43, 95). Little is known about the trans-acting proteins and regulatory mechanisms of the GAS genes in conditions of reversible growth arrest. Promoter fragments activated by confluence or serum starvation have been described for the gas1 and decorin genes, but a full characterization of relevant transcription factors and their cognate elements has not been reported (25, 58). A member of the Maf family has been implicated in the activation of the retina-specific QR1 gene in the chicken and appears to play an essential role for expression of this gene in resting neuroretinal cells (70). Interestingly, these authors outlined the sequence similarity of the QR1 and p20K QRUs and provided evidence that the Maf-like factor responsible for the activation of QR1 is competed by the p20K QRU in EMSA (70). However, despite these observations, we failed to show a role for Maf in activation of the p20K promoter.

We now report that C/EBP is responsible for the activation of p20K in quiescent CEF. The overexpression of a dominant negative mutant of C/EBPβ reduced markedly the activity of the QRU and p20K promoter in growth-arrested CEF (Fig. 5B). The mutation of two C/EBP binding sites of the QRU abolished its activation by quiescence (Fig. 5A). It is not absolutely clear why the C/EBP element of region A was the only one capable of conferring quiescence responsiveness to a heterologous, minimal promoter (Fig. 3). It is possible that the two C/EBP binding sites are qualitatively different. By EMSA, we identified a quiescence-specific complex designated C1 when region A was used as a probe (Fig. 6). A similar but more diffuse complex was also observed with region B. While these complexes were both supershifted or eliminated by a C/EBPβ antibody (Fig. 6D and reference 7), they may represent a different C/EBPβ dimer. It is possible that essential nucleotides located outside the known consensus binding site of C/EBP were missing in the arbitrarily defined regions B and C, rendering these elements nonfunctional in transient expression assays. We also investigated the role of nucleotides diverging between the C/EBP binding sites of the QRU or found in other promoters such as that of the cMGF cytokine gene (82). We observed that differences at the 5′ end did not significantly affect the activity of these elements in transient expression assays. In contrast, the presence of a C residue at the extreme 3′ end of the element (thus diverging from the T or G residue of the established C/EBP consensus binding site) generates an element more responsive to C/EBPβ but not C/EBPα (7). Thus, while region A harbors a classical C/EBP binding site on the minus strand, it also contains the more C/EBPβ-responsive sequence on the plus strand. We showed that the expression of C/EBPβ is induced at contact inhibition (Fig. 11), but even in these conditions, this factor is rare in CEF. Thus, given the limiting amounts of C/EBPβ, differences in the sequence of C/EBP binding sites are likely to be translated into differences in the activity of the elements. That being said, the mutation analysis of the QRU indicated that both C/EBP elements are required for activation by quiescence, implying that factors binding to these elements cooperate in the activation of the QRU. In this respect, the C/EBP binding site of region B behaved as a typical type “B enhanson,” as defined originally by Fromental et al. (31a).

The results of EMSAs indicated that C/EBPβ is a component of the quiescence-specific complex C1. However, it is not known if complex C1 corresponds to the homodimer of C/EBPβ or includes a second member of the C/EBP family. A potential partner of C/EBPβ would be the structurally related CHOP-10 protein, itself the product of a GAS gene known as gadd153 (31). The C/EBP binding site of region A fits approximately the extended consensus sequence defined for the CHOP-10/C/EBPβ heterodimer (PuPuPuTGCAAT[A/C]CCC [87]). However, the overexpression of CHOP-10 alone or in combination with C/EBPα or C/EBPβ did not alter the activity of the QRU in transient expression assays. Moreover, our cDNA library screen for C/EBP-related factors did not yield any clone for CHOP-10 even when a DNA fragment corresponding to the conserved DNA binding domain of the murine CHOP-10 protein was used as a probe. Since CHOP-10 is thought to be induced principally by stress to the endoplasmic reticulum, it may not be expressed in our conditions of quiescence (89). Likewise, a clone for C/EBPδ, a factor induced by serum starvation in mammary epithelial cells (67), was not isolated in our screen. Using a probe for avian C/EBPδ and Northern blotting analysis, we were also unable to detect any transcript for this C/EBP family member expressed in quiescent CEF (46). Since contact inhibition induced the expression of C/EBPβ and this protein is a component of the quiescence-specific C1 complex (Fig. 6 and 11), our results point to this member of the C/EBP family as the key activator of p20K expression in quiescent CEF.

Multiple pathways of p20K regulation.

During development, C/EBPβ is involved in the commitment of the hematopoietic cell lineage and is a critical regulator of macrophage and helper T-cell functions (20, 62, 78, 83). It is also essential for the acute-phase response of the liver, and it participates with C/EBPα in the control of energy metabolism and lipid homeostasis (6, 53, 88). It is thus perhaps not surprising that p20K, a lipid binding protein, is regulated by C/EBPβ in growth-arrested CEF and by linoleic acid, an EFA. Thus, our results support and expand the role of C/EBPβ as a switch factor for the expression of differentiation-specific and now growth arrest-regulated genes.

Using a retroviral construct (RCASBP-20K), we observed that the formation of lipid droplets by CEF treated with increasing amounts of linoleic acid is enhanced by the ectopic overexpression of p20K (7). Cancedda and collaborators showed that p20K has high affinity for polyunsaturated fatty acids such as linoleic acid and therefore is thought to act as a transporter for this class of molecules (17). Our preliminary results thus suggest that p20K promotes the uptake of linoleic acid and therefore functions as a modulator of EFA metabolism in CEF. An intriguing possibility is that EFAs become limiting in serum-starved or density-arrested CEF and that the induction of p20K represents an adaptive response to deleterious growth conditions. Interestingly, Iyer et al. reported recently that the genes for several enzymes involved in cholesterol biosynthesis are most dramatically regulated by serum starvation in human fibroblasts (41). These investigators speculated that the repression of these genes by serum stimulation was the consequence of the addition of lipoproteins provided by the serum. In agreement with this hypothesis, we observed that the addition of linoleic acid to quiescent CEF inhibited the expression of p20K (Fig. 10), suggesting the existence of a negative feedback mechanism involving linoleic acid. Several predictions can be made from this model and are presently being tested in our laboratory.

Lu et al. showed previously that several nonsteroidal anti-inflammatory drugs induce apoptosis in RSV-transformed CEF and markedly inhibit the expression of p20K (52). Since fatty acid methyl esters and several nonsteroidal anti-inflammatory drugs activate members of the PPAR family (50, 76), we sought to determine if PPARγ2, a nuclear receptor induced by the expression of C/EBPβ and a key regulator of adipogenesis in preadipocyte cell systems (53, 85, 93), controls the activity of the QRU. As shown in Fig. 10B, ectopically expressed PPARγ2 was indeed a potent inhibitor of a QRU construct in density-arrested CEF. Thus, it is possible that a member of the PPAR family is part of a regulatory network of proteins controlling the expression of p20K. The QRU does not harbor any consensus binding site for the nuclear receptor family. Moreover, we found that PPARγ2 did not diminish the activation of the QRU by ectopically expressed C/EBPβ, an observation which may explain why CEF infected with RCASBP-C/EBPβ express elevated levels of p20K and form lipid vesicles. Whether PPARγ2 can regulate the activity or expression of C/EBPβ under physiological conditions remains to be investigated. Finally, it is not known if the reciprocal is true, i.e., that the ectopic expression of C/EBPβ stimulates the expression of a member of the PPAR family and PPARγ2 in particular in CEF, as reported for other fibroblasts (53). These questions are important for our understanding of p20K regulation in growth-arrested cells and are presently under investigation.

The regulation of p20K by EFAs may also explain the high level of p20K expression in quiescent CHM cells since these cells were made quiescent by culture in lipid-poor plasma (8). However, the study of CHM cells also suggests that p20K is not regulated solely by the availability of lipids in the medium. Indeed, the addition of a wide variety of growth factors and mitogens was sufficient to induce mitogenesis of CHM cells in plasma and invariably cause the repression of p20K synthesis. In these conditions, CHM cells underwent several rounds of cell division and were clearly not starved for lipids or other nutrients. Second, we showed that CHM cells transformed by a temperature-sensitive mutant of RSV stop dividing in plasma when transferred to the nonpermissive temperature (9). Following temperature shift, the p20K mRNA accumulated in quiescent CHM cells within a few hours but not in their transformed counterparts, which continued to proliferate at the permissive temperature. Third, the synthesis of p20K was repressed in serum-starved CEF by the addition of fresh serum-free medium (56); fourth, our conditions promoting contact inhibition in CEF include the addition of serum-containing medium, which would ensure a fresh supply of nutrients to the cell (Materials and Methods). Thus, p20K is likely to be regulated by the products of immediate-early genes and/or components of the cell cycle machinery as cells exit G₀. c-Myc is a good candidate for this role since it inhibited the expression of the QRU in transient expression assays (54). A role for c-Myc in the inhibition of C/EBPβ-dependent gene expression has been described by Mink et al. (61). In contrast, Rb and the structurally related p107 protein interact with and antagonize the action of c-Myc and promote gene activation by C/EBPβ (10, 19, 40). Therefore, the activation of the p20K promoter may depend on multiple mechanisms and trans-acting factors controlling the expression or activity of C/EBPβ. In this respect, the relationship between progression through the cell cycle, lipid metabolism, and expression of p20K remains to be defined.

Regulation of C/EBPβ in quiescent CEF.

Unlike cells of the myelomonocytic lineage, CEF express very low levels of C/EBPβ. However, results presented in Fig. 11 indicated that transcription of the C/EBPβ gene is stimulated by contact inhibition, a process providing a mechanism of induction of the p20K gene. Hence, C/EBPβ is regulated as a GAS gene in CEF. Therefore, it is of interest to characterize the transcriptional activation of C/EBPβ in growth-arrested CEF and to identify the trans-acting proteins responsible for this process.

The activity of C/EBPβ is known to be controlled by conserved negative regulatory domains adjacent to and interacting with the trans-activating region of the protein. According to models proposed by two independent groups, derepression of the concealed transactivating region involves the phosphorylation of critical residues of the regulatory domains or association with unrelated trans-acting factors resulting in the unfolding of the C/EBPβ protein (48, 92). Experimental evidence supports the existence of these regulatory processes as C/EBPβ is phosphorylated and activated by several kinases (63, 86, 91) and cooperates with an increasing number of transcription factors, including Myb AP-1, NF-κB, Pu.1, serum response factor, the glucocorticoid receptor, and SP1, in gene activation (11, 15, 47, 49, 57, 64, 65, 74, 79, 81). It is therefore probable that the activity of C/EBPβ is also regulated at growth arrest in CEF. The characterization of these regulatory mechanisms is likely to be important for our understanding of GAS gene expression and G₀.