Inhibition of CCAAT/Enhancer-binding Protein a and b Translation by Upstream Open Reading Frames*

CCAAT/enhancer-binding protein (C/EBP) a is a bZIP transcription factor whose expression is restricted to specific cell types. Analysis of C/EBP a mRNA and protein levels in various mammalian cells indicates that expression of this gene is controlled both transcription-ally and post-transcriptionally. We report here that C/EBP a translation is repressed in several cell lines by an evolutionarily conserved upstream open reading frame (uORF), which acts in cis to inhibit C/EBP a translation. Mutations that disrupt the uORF completely abolished translational repression of C/EBP a . The related c/ebp b gene also contains an uORF that suppresses translation. The length of the spacer sequence between the uORF terminator and the ORF initiator codon (7 bases in all c/ebp a genes and 4 bases in c/ebp b homologs) is precisely conserved. The effects of insertions, deletions, and base substitutions in the C/EBP a spacer showed that both the length and nucleotide sequence of the spacer are important for efficient translational repression. Our data indicate that the uORFs regulate translation of full-length C/EBP a and C/EBP b and do not play a role in generating truncated forms of these proteins, as has been suggested by start site multiplicity models.

The transcription factor CCAAT/enhancer-binding protein (C/EBP) 1 ␣, a member of the bZIP (basic region leucine zipper) class of DNA-binding proteins (1), has been a useful model for investigating the role of transcription factors in cellular differentiation. C/EBP␣ is primarily expressed in terminally differentiated cells such as hepatocytes and adipocytes, where it activates transcription of differentiation-specific target genes (reviewed in Refs. 2 and 3). Several studies demonstrated that C/EBP␣ plays an essential role in the conversion of adipoblasts to mature adipocytes. For example, overexpression of C/EBP␣ antisense RNA in the preadipocytic cell line 3T3-F442A inhib-ited hormonally induced differentiation of these cells (4). Similarly, Lin and Lane (5) showed that expression of C/EBP␣ antisense RNA in 3T3-L1 preadipocytes prevents their conversion to adipocytes and the subsequent expression of fat-specific genes such as aP2, GLUT4, and SCD1. In the converse type of experiment, expression of an inducible C/EBP␣-estrogen receptor fusion protein in 3T3-L1 adipoblasts resulted in estrogendependent cell growth arrest (6). In addition, 3T3-L1 cells stably transfected with a C/EBP␣ expression vector exhibited reduced proliferative potential and displayed the differentiated adipocyte morphology at high frequency (7). Forced expression of C/EBP␣ also promoted the adipogenic program in fibroblastic cell lines such as NIH 3T3, which normally are unable to differentiate into mature adipocytes (8 -10). These studies demonstrated that C/EBP␣ is a critical regulator of the adipogenic program and led to the proposal that C/EBP␣ plays a role in controlling the balance between cellular growth and differentiation (6).
In view of the important role of C/EBP␣ in cellular differentiation, the mechanisms that regulate the tissue specificity and developmental timing of C/EBP␣ expression are of considerable interest. Surveys of C/EBP␣ expression show that whereas C/EBP␣ transcripts are present at varying levels in many mammalian tissues and cell lines (Refs. 11 and 12; see also Fig. 1), C/EBP␣ protein occurs in only a subset of these cell types. Cells in which the C/EBP␣ protein has been detected include differentiated hepatocytes (1,12), adipocytes (13), intestinal epithelial cells (14), myelomonocytic progenitor cells (15), ovarian follicles (16), and type II cells of the lung (17). The disproportionate levels of C/EBP␣ mRNA and protein observed in certain cells indicates that post-transcriptional regulation plays a role in restricting C/EBP␣ expression.
In this study, we have investigated the molecular mechanism underlying post-transcriptional control of C/EBP␣ expression. We demonstrate that C/EBP␣ translation is inhibited in transfected cell lines by a short, evolutionarily conserved upstream open reading frame (uORF) located 7 bases upstream of the C/EBP␣ ORF (18). uORF-mediated repression was also observed for the related C/EBP family member C/EBP␤. The uORF repressed C/EBP␣ translation by a cis-acting mechanism, and inhibition was overcome by mutations that inhibit translation of the uORF. Certain changes in the length and sequence of the uORF-ORF spacer region also caused translational derepression. Interestingly, we did not detect truncated C/EBP␣ and C/EBP␤ isoforms in these experiments. These findings may necessitate a re-evaluation of models proposing that initiation at internal start sites produces truncated forms of C/EBP proteins (19 -22). vine serum (Hyclone Laboratories) in the presence of kanamycin, streptomycin, and penicillin. Differentiation of 3T3-L1 cells was performed as described previously (13). HepG2 cells were cultured as described (12). HeLa and 293 cells (30 -40% confluent) were transiently transfected in 10-cm plates by a standard calcium phosphate coprecipitation procedure (23). HeLa cells were transfected with 7.5 g of test plasmid, 7.5 g of internal control plasmid, and 5 g of carrier plasmid (pMEX) for a total of 20 g of DNA/plate. The cells were incubated in precipitate for 18 -24 h, washed twice with unsupplemented Dulbecco's modified Eagle's medium, and then incubated in complete medium for an additional 18 -24 h prior to harvesting. 293 cells were transfected with 2.5 g of test plasmid, 2.5 g of internal control plasmid, and 15 g of pMEX plasmid. The cells were incubated in DNA precipitate for 12-15 h, washed twice with unsupplemented Dulbecco's modified Eagle's medium, and incubated in complete medium for an additional 12-15 h prior to harvesting. For the experiment of Fig. 7, 10-cm plates of HeLa cells were transfected using 2.5 g of DNA and 10 l of DMRIE-C reagent (Life Technologies, Inc.) according to the supplier's recommendations. Duplicate plates were harvested after 48 h for protein and mRNA, respectively. For all transfection assays, at least two independent isolates of each recombinant plasmid were tested, and the transfections were repeated independently at least three times.
For stably transfected HeLa cells, transfection conditions were identical to those described above, except that 20 g of test plasmid were used in addition to 2.5 g of the selectable marker (pMEX.neo). After ϳ48 h, cells were split into fresh 10-cm plates and allowed to recover for an additional 48 h. Cells were then fed with complete medium supplemented with G418 (0.3-0.5 units/ml; Life Technologies, Inc.). Pools of ϳ20 -50 independent neo r transfectants were obtained and analyzed.
Wild-type and mutant C/EBP␤ PCR products were cloned into pMEX-C/EBP␤ using a SalI site in the pMEX polylinker at the 5Ј-end and an MscI site within the C/EBP␤ open reading frame at the 3Ј-end. The PCR protocol described above was used to generate the wild-type and ATG2*-C/EBP␤ constructs using the following oligonucleotides: 5Ј-oligonucleotide for both the wild type and ATG2*, 5Ј-GACGGCGTC-GACTCCGAGCCGCGCACG; wt 3Ј, 5Ј-GTAACGTGGCCACTTCCAT-GGGTC; and ATG2* 3Ј, 5Ј-GTAACGTGGCCACTTCCATGGGTCTAA-AGGCGGCGGGCGGCGGCGGGAGGTGTGCTGCGTC. The mutant constructs AUG1* and AUG1*/2* were made using the four-primer PCR mutagenesis procedure (24,25). Oligonucleotides used to generate these C/EBP␤ mutants were the following: ATG1* 5Ј: 5Ј-CGCGTTCA-CACACCGCCTG; and ATG1* 3Ј, 5ЈGGCGGTGTGTGAACGCG. For the ATG1* mutant, the first-round PCR primers were wt 5Ј and ATG1* 3Ј in one reaction and ATG1* 5Ј and wt 3Ј in another reaction. The resulting PCR products were then annealed and amplified in a second round of PCR. The ATG1*/2* mutant was generated identically to the ATG1* mutant, except that the ATG2* 3Ј-oligonucleotide was used in place of the wild-type 3Ј-oligonucleotide.
The pMEX-uORF plasmid used in Fig. 5 (i.e. a C/EBP␣ construct containing the uORF and the 5Ј-leader, but lacking the C/EBP␣ ORF) was made by digesting wt uORF-C/EBP␣ with NcoI, removing most of the C/EBP␣ ORF. The 4.6-kilobase vector fragment was then isolated and religated to produce pMEX-uORF.
The spacer ϩ200 construct was made as follows. The uORFϩStu-C/ EBP␣-F construct was digested with StuI, and the linear vector fragment was treated with calf intestinal phosphatase (Boehringer Mannheim) to prevent religation. A 217-base pair fragment was obtained from an MspI digestion of the plasmid pBR322. The ends were made blunt using T4 DNA polymerase (Boehringer Mannheim), and the fragment was ligated to the calf intestinal phosphatase-treated vector. Positive isolates were sequenced to identify plasmids with the insert in either orientation.
FLAG-tagged constructs were generated as follows. First, the leucine zippers of C/EBP␣ and C/EBP␤ were removed by digesting the constructs pMEX-C/EBP:G LZ and pMEX-CRP2:G LZ (26), in which the leucine zippers were replaced by the GCN4 zipper, with XhoI and HindIII. A double-stranded oligonucleotide encoding the FLAG epitope was then ligated into the vector fragments. The sequence of the FLAG epitope double-stranded oligonucleotide was as follows (Sequence 1).

RNA Extraction and Northern
Blot Analysis-Total cellular RNA was isolated from transfected cells using RNA STAT-60 (TEL-TEST, Inc.) or Trizol reagent (Life Technologies, Inc.). RNA (15 g) was separated by electrophoresis using formaldehyde-containing 1.2% agarose gels and then transferred to GeneScreen membrane (DuPont) according to the manufacturer's instructions. Membranes were hybridized overnight at 42°C with [␣-32 P]dCTP-labeled DNA probes. The hybridization solution consisted of 50% formamide, 5ϫ Denhardt's solution, 5ϫ SSPE (750 mM NaCl, 50 mM NaH 2 PO 4 ⅐H 2 O, and 5 mM EDTA), 1% SDS, and 0.1 mg/ml salmon sperm DNA (27). Following incubation, blots were washed with 0.1% SDS and 0.5ϫ SSPE at 65°C and then exposed to x-ray film (Eastman Kodak Co.). DNA probes for analysis of RNA from transfected cells by Northern blotting were prepared by isolating the following fragments: c/ebp␣, 350-base pair PstI-SacI fragment from the 3Ј-end of the coding region of the murine c/ebp␣ gene; FLAG, 170-base pair XhoI-HpaI fragment from pMEX-C/EBP␣-F (contains the FLAG epitope and 3Ј-untranslated sequences from pMEX); ␤-actin, 2-kilobase HindIII fragment excised from the plasmid ␤2000 (28); and cyclophilin, as described by Danielson et al. (29). DNA fragments were labeled to high specific activity with [␣-32 P]dCTP using the Prime-It II kit (Stratagene).
Cell Extracts and Western Blotting-Whole cell extracts were prepared by lysing cells directly in radioimmune precipitation assay buffer (30). Nuclear extracts were prepared as described by Lee et al. (31), except that Buffer A also contained 0.1% Nonidet P-40 (Sigma) and 5 g/ml leupeptin (Boehringer Mannheim), and nuclear extracts were not dialyzed against Buffer D. Nuclear extracts from tissue samples were prepared according to the method of Gorski et al. (32). Protein concentrations were determined either by the Bio-Rad protein assay or by estimating protein following SDS-polyacrylamide gel electrophoresis and Coomassie Blue staining. Samples were denatured in sample buffer at 85°C and then electrophoresed on 10% SDS-polyacrylamide gels (33). Molecular weights of the proteins were determined by comparison to Rainbow molecular weight markers (Amersham Pharmacia Biotech). SDS-polyacrylamide gels were transferred to nitrocellulose (Schleicher & Schuell) by electroblotting. Blots were blocked in 5% nonfat dry milk and Tris-buffered saline, pH 7.6, for 2-6 h at room temperature. Blots were incubated with primary antibody for 4 -16 h at room temperature. The antigen-antibody complex was visualized using the Amersham Pharmacia Biotech enhanced chemiluminescence kit. Pan-CRP antiserum was used as described (26). The FLAG epitope was detected using the anti-FLAG M2 monoclonal antibody (Eastman Kodak Co.). Fig. 1 shows a survey of C/EBP␣ mRNA and protein expression in several mammalian cell lines and murine tissues. C/EBP␣ transcript levels were highest in liver and differentiated 3T3-L1 adipocytes, two cell types in which the C/EBP␣ protein is abundant. HeLa and HepG2 cells and kidney tissue contained lower but significant levels of C/EBP␣ mRNA, whereas they lacked detectable C/EBP␣ protein. Liver contained 5-10fold more C/EBP␣ mRNA than kidney (Fig. 1A), but at least 20-fold more C/EBP␣ protein, as estimated by serial dilution of the protein extract (Fig. 1C). The difference in protein levels is a minimum estimate since the C/EBP␣ signal in kidney extracts was below the level of detection. This disparity between mRNA and protein levels suggests that a post-transcriptional mechanism contributes to the regulation of C/EBP␣ expression.

Post-transcriptional Control of C/EBP␣ Expression-
Inhibition of C/EBP␣ Translation by a Conserved Upstream Open Reading Frame-A comparison of the 5Ј-leaders of rat, mouse, human (34), bovine (35), chicken, and Xenopus laevis c/ebp␣ genes revealed the presence of a short (5-amino acid) conserved ORF ( Fig. 2A) (18). This uORF is located precisely 7 nucleotides upstream of the C/EBP␣ initiation codon in all c/ebp␣ homologs. The evolutionary conservation of the uORF element and its location in the transcript suggest an important function in regulating C/EBP␣ expression. We suspected that the uORF might suppress translation of the C/EBP␣ ORF since many eukaryotic mRNAs lack upstream AUG codons, and the insertion of a single functional initiation codon upstream can severely inhibit translation from downstream AUG codons (reviewed in Refs. 36 -39).
To investigate the potential role of the uORF in regulating C/EBP␣ translation, we constructed rat c/ebp␣ genes containing either wild-type or mutant uORF sequences (Fig. 2B) and inserted these into the eukaryotic expression vector pMEX. The mutations included alterations of the uORF initiator (ATG*) and terminator (TAA*) sequences and precise deletion of the uORF (⌬uORF). The mutant genes were tested by transient transfection into HeLa cells, a cell line in which expression of the endogenous C/EBP␣ protein is repressed (Fig. 1). As an internal control, each construct was cotransfected with a constitutive C/EBP␤ expression vector (pMEX-C/EBP␤). Trans-fected cell extracts were analyzed for C/EBP␣ and C/EBP␤ expression by Western blotting using an antibody (pan-CRP) that recognizes both C/EBP proteins (26).
As shown in Fig. 2C (lane 3), a gene containing optimal translation initiation signals in place of the C/EBP␣ 5Ј-leader (pMEX-C/EBP␣) was expressed at high levels in HeLa cells. In contrast, a construct bearing the wild-type 5Ј-leader (wt uORF-C/EBP␣; lane 4) produced very low amounts of C/EBP␣ protein. However, the mutants uORF(ATG*) (lane 5) and ⌬uORF (lane 7) exhibited high levels of C/EBP␣ expression, comparable to pMEX-C/EBP␣, whereas the uORF termination codon mutant (uORF(TAA*); lane 6) did not express C/EBP␣. The same set of mutations introduced into the murine c/ebp␣ gene gave identical results (data not shown). These findings indicate that translation of the uORF severely inhibits C/EBP␣ expression in HeLa cells.
To test whether translation initiates at the uORF AUG codon, we generated a construct in which the uORF was fused in frame to the C/EBP␣ ORF. This vector expressed high levels of a protein that was slightly larger than C/EBP␣ (Fig. 2C, lane  8). The fact that the fusion protein was strongly expressed from the uORF-ORF construct demonstrates that the uORF initiation codon is utilized efficiently in HeLa cells. Accordingly, the uORF initiation codon is in a favorable context for efficient initiation (40) in all species examined ( Fig. 2A). Collectively, the data of Fig. 2C are consistent with a mechanism in which translation of the uORF cistron represses initiation from the downstream C/EBP␣ AUG codon. Several other eukaryotic genes are also translationally repressed by uORFs (reviewed in Ref. 37), including the gene encoding the Saccharomyces cerevisiae bZIP protein GCN4 (41).
To further establish that the uORF suppresses C/EBP␣ translation, we examined the expression of wild-type and mutant uORF constructs after stable transfection into HeLa cells. This approach enabled us to readily compare mRNA and protein expression from the engineered constructs. Because forced C/EBP␣ expression can cause cell growth arrest, we generated a functionally inactive protein by deleting the leucine zipper, thus preventing dimerization and DNA binding. To facilitate detection of the protein, we fused the FLAG epitope (42) to the C terminus, producing a hybrid designated C/EBP␣-F. The levels were compared in several mammalian tissues and cell lines. A, total RNA (10 g) from the indicated adult mouse tissues or cell lines was analyzed by Northern blotting. The blot was probed using a labeled DNA fragment from the mouse c/ebp␣ gene, stripped, and reprobed for ␤-actin mRNA. B, nuclear extracts prepared from the same cell lines and tissues were analyzed for C/EBP␣ protein expression. Protein concentrations were estimated by Coomassie Blue-stained SDS-polyacrylamide gel electrophoresis analysis, and equivalent amounts of protein were assayed by Western blotting using a peptide antibody that recognizes an internal C/EBP␣ epitope (1). p42 ␣ is the full-length (42-kDa) C/EBP␣ protein, and p30 ␣ is a 30-kDa product proposed to arise from an internal initiation codon within the C/EBP␣ transcript (20,21). C, shown is the relative C/EBP␣ protein expression in liver and kidney tissue. Serial dilutions of liver nuclear extract (lanes 1-4) were analyzed by Western blotting and compared with undiluted kidney extract (lane 5). Diff., differentiated; Non-Diff., nondifferentiated.
C/EBP␣-F plasmids were cotransfected with a selectable marker, and pools of transfectants were selected and analyzed for C/EBP␣-F mRNA and protein expression. Again, the wildtype 5Ј-leader was found to severely suppress C/EBP␣ protein expression, and mutation of the uORF AUG codon eliminated this inhibitory effect (Fig. 3). In addition, the uORF-ORF fusion gene was efficiently translated. The differences in protein levels were not due to effects on mRNA expression since similar levels of C/EBP␣-F mRNA were detected in each of the transfectant pools. The results further support the conclusion that translational repression of C/EBP␣ in HeLa cells is controlled by the uORF and confirm that transient transfection is a valid assay for assessing C/EBP␣ translational control.
To test whether uORF-mediated inhibition occurs in other cell types, we introduced the C/EBP␣-F constructs into 293 cells (human embryonic kidney) by transient transfection. As shown in Fig. 4A, the wild-type uORF sequence again strongly reduced C/EBP␣ expression. Mutations that eliminate uORF translation (uORF(ATG*) and ⌬uORF; lanes 5 and 7) reversed this inhibitory effect. The uORF terminator mutant, uORF-(TAA*) (lane 6), was completely repressed, whereas the uORF-ORF fusion construct was efficiently translated (lane 8). Equivalent levels of C/EBP␣-F transcripts were produced in each transfection (Fig. 4B). These data corroborate the results obtained using HeLa cells and, together with similar results from HepG2 hepatoma cells (data not shown), demonstrate that uORF-dependent repression operates in a variety of cell types.

FIG. 2. An evolutionarily conserved uORF inhibits C/EBP␣ translation.
A, sequence comparison of the proximal 5Јleaders of C/EBP␣ transcripts from several species. A schematic diagram of the C/EBP␣ transcript is depicted at the top, showing the relative size and proximity of the uORF to the ORF. The uORF region is expanded to show the sequence of the uORF and the conserved spacing between the uORF and ORF. B, schematic diagram of uORF mutations. C, C/EBP␣ expression from wild-type and mutant uORF constructs in transiently transfected HeLa cells. HeLa cells were transiently transfected with the C/EBP␣ constructs shown in B. A constitutive C/EBP␤ expression vector, pMEX-C/ EBP␤, was included in each transfection as a control for transfection efficiency. Nuclear extracts were analyzed by Western blotting using a peptide antibody (pan-CRP) that recognizes an epitope common to all C/EBP proteins. UT, untranslated; Bact, bacterial.
cis-We next addressed the possibility that the uORF encodes a peptide product that inhibits C/EBP␣ translation in trans. The uORF(ATG*) mutant was transfected into HeLa cells together with a construct that contains the uORF but lacks the C/EBP␣ ORF. Transfecting increasing amounts of the uORF expression plasmid did not diminish expression either from uORF(ATG*)-C/EBP␣-F or from the control, pMEX-C/EBP␣-F (Fig. 5). Although we cannot directly assess synthesis of the uORF-encoded peptide in the cell, the high expression seen from the uORF-ORF fusion construct (Fig. 2C, lane 8) demonstrates that translation initiates efficiently at the uORF in HeLa cells. Thus, the inability of the uORF to repress C/EBP␣ translation when expressed from a separate transcript suggests that the uORF inhibits C/EBP␣ expression via a cisacting mechanism.
The uORF-ORF Spacer Length and Nucleotide Sequence Are Critical for Translational Repression-The spacing between the uORF termination codon and the C/EBP␣ start site is exactly 7 bases in all c/ebp␣ homologs characterized ( Fig. 2A), suggesting that the spacer length is an important feature for C/EBP␣ regulation. We examined whether mutations that increase or decrease the spacer length (Fig. 6A) affect translational repression in 293 cells. Insertion of 1-3 cytosines in the spacer did not abolish suppression of C/EBP␣ translation by the uORF (Fig. 6B, lanes 5-7). In contrast, removing a single cytosine caused a large increase in C/EBP␣ translation (Fig.  6C, lane 5), and deleting 2 or 3 of the 5 cytosines preceding the ORF also resulted in significant derepression (Fig. 6C, lanes 6  and 7). Thus, decreasing the spacing by 1 nucleotide renders the uORF unable to inhibit C/EBP␣ translation, demonstrating that the length and/or sequence of the spacer is critical for uORF-mediated repression.
Since small increases in spacer length did not interfere with translational repression (Fig. 6B), we next tested the effects of larger insertions in the spacer element. Spacing was increased by inserting 10 bases (Spϩ10) or 10 bases containing a StuI restriction site (SpϩStu) (Fig. 6A). C/EBP␣ expression from these constructs remained repressed (Fig. 6C, lanes 8 and 9). In addition, inserting ϳ200 bases into the spacer (in both orientations) did not disrupt the repressive activity of the uORF (Fig. 6C, lanes 10 and 11). The lack of derepression for the larger insertions was unexpected since, in an independent study, reinitiation efficiency at downstream initiation codons was found to improve as the intercistronic distance was increased (43). While the repression observed with the Spϩ200 mutant could be due to unforeseen inhibitory sequences within these DNA inserts, it is unlikely that such sequences would occur in both orientations of the insert.
We further investigated whether altering the nucleotide sequence of the spacer would modify the efficiency of repression. We first inspected the C/EBP␣ spacer for the presence of conserved nucleotides located upstream of the C/EBP␣ initiation (Kozak) sequence. A cytosine residue at nucleotide Ϫ4 relative to the ORF AUG codon was the only conserved base (Fig. 2A). We changed this nucleotide to adenine (Sp C*(Ϫ4)) ( Fig. 6A) as well as constructed a mutant in which nucleotide Ϫ4 and its 2 flanking bases were altered (Sp C*(Ϫ33Ϫ5)). Both of these mutations caused derepression of C/EBP␣ expression (Fig. 6C,  lanes 12 and 13). Collectively, the results of Fig. 6 indicate that the uORF-ORF spacer length and nucleotide sequence are critical for full translational repression.
A Conserved uORF Regulates C/EBP␤ Translation-Expression of another C/EBP family member, C/EBP␤, is also regulated post-transcriptionally. Descombes et al. (44) reported that C/EBP␤ (or LAP (liver-enriched transcriptional activator protein) transcripts occur in liver, lung, spleen, kidney, brain, and testis, whereas the protein was detected only in liver. These results imply that a translational control mechanism suppresses C/EBP␤ protein expression in certain cell types. Inspection of the 5Ј-leader of the c/ebp␤ gene revealed a conserved uORF located 4 bases upstream of the ORF (Fig. 7A) (22). As depicted in Fig. 7A, the C/EBP␤ transcript contains four potential translation start sites (19). The first precedes the uORF and is in frame with the C/EBP␤ ORF, the second corresponds to the uORF start site, and the third represents the C/EBP␤ initiation codon. A fourth AUG codon forms a potential internal start site that may produce a truncated protein. Mutation of the first, third, and fourth initiation codons indicated that each potential in-frame start site could generate a polypeptide, which were designated FL-LAP (p38), LAP (p34), and LIP (p20), respectively (19). LAP (p34 ␤ ) is the most common C/EBP␤ isoform expressed in vivo (19).
Because of the evidence that C/EBP␤ is translationally regulated, we investigated the possibility that the C/EBP␤ uORF modulates p34 ␤ translation. We mutated the first in-frame AUG codon (ATG1*) and the uORF initiation codon (ATG2*), both individually and in combination (ATG1*/2*). The leucine zipper region was also replaced by the FLAG epitope to facilitate detection of the expressed proteins. Expression was analyzed in transiently transfected HeLa cells and was compared with constructs containing either the wild-type 5Ј-leader (wt C/EBP␤-F) or a canonical Kozak sequence in place of the 5Јleader region (pMEX-C/EBP␤-F). These experiments showed that the C/EBP␤ 5Ј-leader inhibits expression of p34 ␤ -F (Fig.  7B, lane 3). Mutating the first AUG codon (ATG1*) had only minor effects on p34 ␤ expression (lane 4), and p38 ␤ was not significantly expressed from any of the constructs. However, altering the uORF initiation codon (ATG2*) caused a large increase in C/EBP␤ expression (lane 5), and the double mutant (ATG1*/2*) also expressed high levels of p34 ␤ -F (lane 6). These data demonstrate that the C/EBP␤ uORF exerts an inhibitory effect on translation of the ORF located immediately downstream.
Surprisingly, in these experiments, we did not detect LIP (p20 ␤ ), which was predicted to be expressed from constructs containing the wild-type C/EBP␤ 5Ј-leader (19). In addition, p30 ␣ , a putative alternative translation product of C/EBP␣ (20 -22), was not produced by any of the C/EBP␣ constructs transfected into HeLa or 293 cells (Figs. 3 and 4). These observations, which are discussed in greater detail below, suggest that the 5Ј-leader regions do not necessarily regulate translation initiation at internal AUG codons.

DISCUSSION
In this report, we have examined the molecular basis for post-transcriptional regulation of C/EBP␣ and C/EBP␤ expression. Mutational analysis revealed that the uORF is a potent inhibitor of C/EBP␣ translation. The uORF initiation codon was readily recognized by the translational machinery, and mutations that abolished uORF translation disrupted repression of the C/EBP␣ ORF. The repressive effect of the uORF on C/EBP␣ translation was not cell-specific since the same results were observed in several cell lines using both transient and stable transfection assays. Although repression was seen in multiple cell lines, there must be a mechanism to overcome repression imposed by the uORF in cells that express the C/EBP␣ protein abundantly, such as terminally differentiated hepatocytes and adipocytes. C/EBP␤ also contains a conserved uORF that represses translation (Fig. 7). This mechanism could control the induction of p34 ␤ translation in response to cell-specific or physiological cues. For example, the human homolog of C/EBP␤, NF-IL6 (45), was shown to be translationally regulated in the pulmonary alveolar epithelial cell line A549 (46). Protein expression from the NF-IL6 transcript was induced upon infection of the cells with respiratory syncytial virus. Enhanced protein expression occurred without detectable changes in NF-IL6 transcript abundance or size, indicating the existence of a translational control mechanism. Since C/EBP␤ is implicated in transcription of cytokine genes and other inflammatory mediators in monocytic and epithelial cells (47), up-regulation of C/EBP␤ translation may be an important component of the cellular response to infectious or toxic agents. C/EBP␤ protein expression, but not mRNA levels, was also found to be upregulated during terminal differentiation of primary keratinocytes in culture 2 and differentiation of rat Leydig cells at puberty (49). We suggest that induction of C/EBP␤ expression in each of these cases involves derepression of uORF-mediated translational inhibition.
The truncated forms of C/EBP␣ and C/EBP␤ (p30 ␣ and p20 ␤ / LIP, respectively) have been hypothesized to result from alternative translation initiation (19 -22). It has been proposed that internal initiation within the C/EBP␣ or C/EBP␤ ORFs occurs by a mechanism that requires the 5Ј-leader region. The resulting truncated proteins contain a DNA-binding domain, but lack an N-terminal activation domain. LIP, which lacks activating sequences altogether, has been proposed to function as an inhibitory isoform of C/EBP␤. Internal initiation in the C/EBP␣ and C/EBP␤ mRNAs is believed to arise either from "leaky scanning" (19 -21) or from termination after uORFs and reinitiation at downstream AUG codons (22,50).
Our results do not support the hypothesis that C/EBP␣ and C/EBP␤ translation reinitiates at downstream AUG codons within the ORF. The presence of the wild-type C/EBP␣ 5Јleader diminished translation of p42 ␣ (Figs. 3-5), but no trun-cated proteins such as p30 ␣ were produced from any of the C/EBP␣ constructs. The absence of truncated proteins was not a function of the cell type since these proteins were not observed in any cell line tested, including HeLa, 293, HepG2, and undifferentiated 3T3-L1 cells (data not shown). Similarly, truncated proteins were not expressed from the C/EBP␤ constructs (Fig. 7). These results were consistently observed with two different antibodies (pan-CRP and anti-FLAG M2), indicating that antibody specificity does not account for the inability to detect p30 ␣ and p20 ␤ /LIP. Rather, we have detected only p30 ␣ or p20 ␤ /LIP in nuclear cell extracts from transfected cells under certain cell lysis conditions. 3 Specifically, the addition of detergent (Nonidet P-40) and the protease inhibitor leupeptin to the lysis buffer prevented the appearance of truncated proteins. The level of endogenous p30 ␣ protein observed in differentiated 3T3-L1 cells (Fig. 1B, lane 4) could also be diminished or eliminated by altering the cell lysis protocol. In addition, we have found that a calpain protease cleaves recombinant C/EBP␤ to generate a product that is indistinguishable in size from LIP. 4 Collectively, our observations suggest that truncated forms of C/EBP␣ and C/EBP␤ may arise in many cases from proteolysis rather than translation initiation. We propose that the function of C/EBP␣ and C/EBP␤ uORFs is to regulate initiation at the ORFs immediately downstream and not to generate truncated proteins by internal initiation.
Mutations designed to test the mechanism of uORF repres- FIG. 7. The c/ebp␤ gene transcript contains a conserved uORF that represses C/EBP␤ translation in HeLa cells. A, sequence comparison of 5Ј-leader regions from several c/ebp␤ homologs. A schematic diagram of the C/EBP␤ transcript is shown at the top, depicting the relative size and proximity of the uORF to the ORF. The two possible activator proteins resulting from alternative initiation codons (19) are designated p38 ␤ and p34 ␤ . The putative start site for LIP (p20 ␤ ) is also indicated. References for the sequences are as follows: rat (12), mouse (62), human (45), chicken (63), and bovine (48). B, C/EBP␤ protein expression from wild-type and mutant C/EBP␤ constructs in HeLa cells. FLAG-tagged constructs were transiently transfected into HeLa cells, and whole cell extracts were analyzed by Western blotting using the anti-FLAG M2 monoclonal antibody. C, Northern blot analysis of C/EBP␤-F mRNA expression. Total RNA (15 g) was analyzed by hybridization with a FLAG-specific probe and then with a cyclophilin probe to control for RNA loading. sion showed that the uORF-ORF spacer is critical for inhibition of C/EBP␣ translation. Removal of 1-3 cytosine residues significantly decreased translational repression, especially when a single cytosine was deleted. One explanation for this result is that the C/EBP␣ spacer sequence dictates the efficiency of ribosome release after uORF termination. Support for this notion comes from studies of the yeast gcn4 gene, which encodes a transcriptional activator that regulates the synthesis of several biosynthetic genes in response to amino acid starvation. Gcn4 translation is controlled by four uORFs within the 5Ј-leader and is regulated primarily by the first and fourth uORFs (41). uORF1 alone inhibits GCN4 translation by ϳ50% and is also required for translational derepression in response to amino acid starvation, whereas uORF4 constitutively represses translation of GCN4. The sequences surrounding the termination codons of both uORFs are critical for regulation of GCN4 translation (51). Sequences downstream of the uORF1 terminator are required for reinitiation at uORF4 when amino acids are plentiful or at GCN4 under starvation conditions. A rare proline codon at the 3Ј-end of uORF4 and sequences just downstream of the terminator appear to be necessary for efficient ribosome release following termination, thereby preventing reinitiation at downstream AUG codons. Thus, sequences 3Ј of the uORF1 terminator favor reinitiation downstream, whereas those 3Ј of uORF4 favor ribosome release.
C/EBP␣ translational repression could also involve ribosome release after uORF termination, resulting in weak reinitiation at the C/EBP␣ AUG codon. Spacer deletions and nucleotide substitution mutants that up-regulate C/EBP␣ expression may function by decreasing the efficiency of uORF termination/ release. An alternative possibility is that these mutations increase initiation efficiency at the C/EBP␣ start site. However, this explanation seems less likely since the sequence context of the initiation codon should be the predominant factor dictating the efficiency of initiation (40,52,53), and the mutations were designed to minimize disruption of the C/EBP␣ Kozak sequence. In addition, mutants in which uORF translation is disrupted express high levels of C/EBP␣, showing that the ORF AUG codon is in a favorable sequence context for initiation. The spacer length may also affect the efficiency of reinitiation, although the derepression seen with spacer deletion mutants could be explained by the removal of nucleotides required for efficient ribosome release. Nevertheless, the stringent conservation of uORF-ORF spacing in c/ebp␣ and c/ebp␤ genes argues that spacer length per se is an important feature of translational regulation.
In some uORF-containing genes, repression of downstream ORFs is dependent on the peptide translated from the uORF. Examples of this include yeast CPA-1 (54), the human cytomegalovirus gpUL4 (gp48) gene (55,56), and AdoMetDC (57,58). Experimental data suggest that the nascent polypeptide translated from these uORFs in some way stalls the translational machinery, possibly during termination of uORF translation. The peptides encoded by these various uORFs do not contain any apparent sequence homologies and are variable in size. Like the C/EBP␣ uORF, all appear to function only in cis. Further mutagenesis experiments should establish whether the uORF amino acid sequence is important for repression of C/EBP␣ and C/EBP␤ translation.
We speculate that uORF-mediated inhibition of C/EBP␣ translation may operate to restrict C/EBP␣ expression to cells that are undergoing terminal differentiation. C/EBP␣ possesses potent differentiative and antimitotic activities (6,8,59), and its expression may not be tolerated by most proliferating cells. Overexpression of C/EBP␤ was also found to cause growth arrest in hepatoma cells (59). C/EBP␣ activates expres-sion of two growth-arrest associated genes: p21 (WAF-1/CIP-1/SDI-1), a member of the cyclin-dependent kinase inhibitor family that blocks G 1 -S cell cycle progression (60), and gadd145, a growth arrest-inducible gene (61). Thus, stringent regulation of C/EBP␣ and C/EBP␤ expression may be necessary to prevent premature mitotic arrest, differentiation, and activation of target genes. uORF-mediated translational regulation may contribute to the control of C/EBP␣ and C/EBP␤ expression during development and could also modulate C/EBP␤ protein synthesis in response to cell stress or viral infection.