Two different E2F6 proteins generated by alternative splicing and internal translation initiation Tillman Dahme 1 , Jason Wood 1 , David M. Livingston 1 and Stefan Gaubatz 2 1 Dana Farber Cancer Institute and Harvard Medical School, Boston, MA, USA; 2 Institute for Molecular Biology and Tumor Research (IMT), Philipps-University Marburg, Germany E2F transcription factors play an important role in the regulation of cell cycle progression. E2F6, the most recently identified member of the E2F family, is a retinoblastoma- protein-independent transcriptional repressor that is required for developmental patterning of the axial skeleton. It has recently been shown that the E2f6 locus produces two different mRNAs, E2F6 and E2F6b. The E2F6b mRNA contains an additional exon that is inserted by alternative splicing. This exon contains an in-frame stop-codon and an in-frame translation initiation codon. However, whether a protein is translated from the E2F6b mRNA has not yet been addressed. We now show that internal translation initiation gives rise to E2F6b, an amino-terminal truncated E2F6 protein. We also show that E2F6 and E2F6b mRNAs are ubiquitously expressed in primary mouse tissues. During the cell cycle, the highest expression of both forms is found at the G1 to S transition. The 5¢ untranslated regions of E2F6 and E2F6b are unusually long, and they contain several upstream AUG codons followed by short reading frames. Our results suggest that translation of E2F6b is initiated by internal ribosome entry. We propose that regulated trans- lation initiation can produce distinct E2F6 isoforms under different physiological conditions. Keywords:E2F;transcriptionfactor;cellcycle;alternative splicing; internal ribosomal entry. E2F transcription factors have been intensively studied for their ability to control cellular proliferation. E2F-responsive elements have been identified in genes that play key roles in cell cycle progression, synthesis of nucleotides, and DNA replication (reviewed in [1]). Consequently, deregulation of E2F can promote tumorigenic transformation (reviewed in [2]). In normal cells, E2F activity is regulated by the binding to pRB, the product of the retinoblastoma gene, and by bind- ing to two related Ôpocket proteinsÕ, p107 and p130. Recent studies suggest that E2F proteins have a dual function in cell cycle control. First, ÔfreeÕ, uncomplexed E2F is a transcrip- tional activator with growth promoting activities. Secondly, complexes between E2F and pocket proteins act as tran- scriptional repressors and growth inhibitors [1]. E2F is a heterodimeric complex containing an E2F- and a related DP- subunit. So far, six E2F proteins (E2F1 through E2F6) and two DP proteins (DP-1 and DP-2) have been identified. Conserved domains mediate DNA-binding, dimerization, transactivation and pocket protein binding. In contrast to the other E2F proteins, E2F6, the most recently identified E2F protein, lacks a transactivation domain, and is a pocket protein independent transcriptional repressor [3–6]. We have recently shown that, in mice, E2F6 is required for developmental patterning of the axial skeleton [7]. E2f6 deficient animals display homeotic transformations of the skeleton that are strikingly similar to the transforma- tions of certain polycomb deficient mice [7]. These observa- tions are consistent with the recent finding that E2F6 associates with members of the mammalian polycomb complex [8,9]. Taken together, these observations suggest that one function of E2F6 is to recruit polycomb multipro- tein complexes to target promoters during development. It has been recently shown, that the E2f6 locus produces two distinct mRNAs, E2F6 and E2F6b [10]. The E2F6b mRNA contains the newly discovered exon 2. It has been noted that this exon introduces an in frame termination codon as well as an AUG codon on its 3¢ extremity, which could potentially serve as a translation initiation codon. However, no evidence for the generation of a protein has been given. In this study, we now demonstrate that an amino-terminal truncated E2F6 protein is generated by internal translation initiation of E2F6b. In addition, we show that the 5¢ untransl- ated region of the E2F6 mRNA is unusually long, and that they contain several upstream AUG codons followed by short reading frames, features that impair normal CAP- dependent translation initiation. E2F6 and E2F6b mRNAs are widely expressed in primary mouse tissues. We propose that regulated translation initiation can produce distinct E2F6 isoforms under different physiological conditions. MATERIALS AND METHODS Cell culture Cells were cultivated at 37 °Cina10%CO 2 -containing atmosphere. Cells were maintained in Dulbecco’s modified Eagle’s medium (Cellgrow) supplemented with 10% fetal bovine serum (HyClone). For the cell cycle experiment, MEF cells were serum starved in Dulbecco’s modified Correspondence to S. Gaubatz, Institute for Molecular Biology and Tumor Research (IMT), Philipps-University Marburg, 35037 Marburg, Germany. Fax: + 49 6421286 5196, Tel.: +49 6421286 6240, E-mail: gaubatz@imt.uni-marburg.de (Received 25 June 2002, revised 22 August 2002, accepted 28 August 2002) Eur. J. Biochem. 269, 5030–5036 (2002) Ó FEBS 2002 doi:10.1046/j.1432-1033.2002.03210.x Eagle’s medium/0.1% fetal bovine serum for 48 h, then released from starvation by addition of Dulbecco’s modified Eagle’s medium/10% fetal bovine serum. The cells were harvested at different time points after serum addition. Synchronization was confirmed by FACS analysis. RNA isolation Total RNA was isolated from primary MEFs with the RNeasy kit (Qiagen) according to the manufacturer’s instructions. To isolate RNA from primary mouse tissues, an adult female C57/BL6 mouse was sacrificed, and then the organs were harvested and quickly frozen in liquid nitrogen. Total RNA was isolated with TRIZOL (Invitrogen) according to the manufacturer’s instructions. Luciferase fusion constructs We first modified the sequence around the translation initiation of the luciferase cDNA in pGL2-basic (Promega) by introducing an AvrII-site and thereby abolishing the initiating ATG, generating pGL2-Avr. A fragment enco- ding for the first eight codons of E2F6 and partial 5¢UTR was amplified by PCR with primers SG108 and SG92 using a genomic E2f6 clone as a template. The resulting product was digested with BglII and XbaI and inserted into pGL2- Avr digested with BglII and AvrII. In a second step, a 2.4-kb genomic SacI–BglII fragment containing the remaining 5¢UTR and 2.2 kb of the E2F6 promoter was inserted to generate Exon1-luc. A similar strategy was used to generate Exon3-luc and Exon2-luc. We first used RT-PCR of total RNA with primers SG92 and SG122 to amplify part of the coding regions and 5¢ UTRs of E2F6 and E2F6b mRNAs. PCR-products were digested with BglII and XbaIand inserted into pGL2-Avr. Secondly, the 2.4 kb genomic SacI–BglII fragment was inserted to generate Exon2-luc and Exon3-luc. In Exon2-mut-luc, the translation initiation codon was modified by PCR with primers SG92 and SG142 and Exon2-luc as a template. The resulting PCR product was digested with BglII and XbaI and inserted into pGL2- Avr. Finally, the 2.4 kb genomic promoter fragment was inserted as described above. An expression plasmid for E2F6b was generated by PCR with primers SG109 and SG35. The PCR product was digested with BamHI and EcoRI and inserted into pcDNA3. All constructs were confirmed by DNA sequencing. Other plasmids have been described previously: E1B-luc [11], E2F1-luc [12]. Polyclonal antibodies and immunoprecipitations A polyclonal serum was raised in rabbits with KLH-coupled peptide N-14 (CMSQQRTARRLPSLL). The antiserum was affinity purified on columns with immobilized peptide (Sulfolink, Pierce) and eluted in 100 m M glycine (pH 2.5). The C-10 antiserum has been described previously [7]. For immunoprecipitation, MEFs were metabolically labeled for12–14hwith1mCi[ 35 S]methionine and then lysed in Tris/NaCl/NP-40 (50 m M Tris, pH 7.4, 150 m M NaCl, 0.5% NP-40, 1 · complete protease inhibitors (Roche), 1m M dithiothreitol). Lysates were incubated with 1 lg affinity-purified antiserum. Proteins were collected on Protein-A Sepharose beads, washed five times in Tris/ NaCl/NP-40 and then separated by SDS/PAGE. Oligonucleotides SG24: 5¢-GGGAATTCTCAGATAGAGTCTTCTCTGG GAGC-3¢ SG50: 5¢-GGGGACCAGGGTGACCGCGG-3¢ SG92: 5¢-GCGCGGGAGATCTAACGGACGG-3¢ SG108: 5¢-CCTCTAGACGCGCCGTCCGGTGCTGAC TCAT-3¢ SG109: 5¢-GGGGATCCATGCCATCAAAAATAAGGA TTAAT-3¢ SG142: 5¢-CCTCTAGATTAATCCTTATTTTTGATGG CCCCCTTCTGTCTCTGCCTCCCAAGGACTGGC-3¢ SG35: 5¢-CGGAATTCCCCGTGCTGGAGGCGACT CG-3¢ SG122: 5¢-CGGACGGCGCGGAGAC 3¢ b-actin fw: 5¢-TGTGATGGTGGGAATGGGTCAG-3¢ b-actin bw: 5¢-TTTGATGTCACGCACGATTTCC-3¢ RT-PCR RT-PCR was performed with the Superscript One-Step RT- PCR kit (Invitrogen) with 100 ng of total RNA. For the detection of E2F6 mRNA with primers SG122 and SG24, the following conditions were used: 1 cycle: 30 min, 50 °C; 1 cycle: 2 min, 94 °C; 35 cycles: 30 s, 94 °C, 30 s, 55 °C; 1min,72°C, followed by 1 cycle for 10 min at 72 °C. For the detection of b-actin, the number of cycles was reduced to 32. Products were separated on 1.4% agarose gels. Transient transfections and reporter assays Cells (2 · 10 4 ) were plated per each well of a 24 well cell culture dish. 24 h later, cells were transfected in triplicate using the indicated amount of luciferase fusion construct or empty vector and 3–5 lL of Fugene (Roche) diluted in 100 lL Dulbecco’s modified Eagle’s medium per triplicate reaction. To analyze the transcriptional properties of E2F6b, U2-OS cells were transfected with 0.200 lgof E2F-dependet luciferase reporter construct (E2F1-luc or E1B-luc), 0.050 lgofCMV-bGal (to monitor transfection efficiency), and 0.200 lg pCDNA3-mE2F6, pCDNA3- mE2F6b or pcDNA3-DP2 expression plasmids. Cells were harvested 48 h after transfection and lysed in 1 · passive lysis buffer (Promega). Luciferase assays were performed on a Dynex Luminometer. Luciferase activity was normalized to b-Gal activity to correct for differences in transfection efficiency. Primer extension Total RNA was extracted from mouse embryonic fibro- blasts using the RNeasy mini kit (Qiagen). SG50 primer was end labeled with [c- 32 P]ATP and T4 polynucleotide kinase. For the primer extension reaction, 10 lgofRNAwas hybridized to radiolabelled SG50 at 65 °Cfor90minin hybridization buffer (0.15 M KCl/0.01 M Tris/Cl, pH 8.3/ 1m M EDTA). Hybridized RNA was then extended for 1 h at 42 °C by addition of 30 lL reaction mix [1 lL1 M Tris, pH 8.3, 1 lL0.5 M MgCl 2 ,1lL0.25 M dithiothreitol, 3.33 lL2m M dNTP mixture, 22.67 lLH 2 O, 1 lL1:20 diluted Superscript II reverse transcriptase (Invitrogen)]. RNA was subsequently degraded by incubating with 105 lLRNase(100lgÆmL )1 sonicatedsalmonsperm Ó FEBS 2002 E2F6b, an alternatively spliced E2F6 isoform (Eur. J. Biochem. 269) 5031 DNA/20 lgÆmL )1 RNaseA)for15minat37°C. A sequencing reaction was performed with the same SG50 32 P-end-labeled primer using Sequenase 2.0 (United States Biochemical) and the pBS-Not/Xho4.5 template. The sequencing products and the primer extension product were separated on a QuickPoint gel (Novex) and visualized by autoradiography. RESULTS Two distinct mRNAs are produced from the E2f6 locus It has recently been shown that the E2f6 locus produces two alternatively spliced isoforms [10] (see Fig. 1). The larger splice variant is generated by insertion of an additional exon 2. Exon 2 contains a stop codon, in-frame with the upstream E2F6 sequence. The last triplet of exon 2, however, is a potential translation initiation codon (AUG), immediately upstream and in-frame with exon 3. Thus, it is possible that the E2F6b mRNA gives rise to an amino-terminal truncated protein, which is initiated from the internal AUG in exon 2. A second possibility is that a protein is synthesized by a read-through of the stop codon in exon 2. The latter would give rise to a protein of slightly higher molecular mass containing 22 additional amino acids. The E2F6 5¢ untranslated region impairs ribosomal scanning Translation of eukaryotic mRNAs is usually initiated at the first AUG triplet downstream of the 5¢-terminal CAP, and it is usually separated from it by 50–100 nucleotides [13]. Nevertheless, translation initiation from internal start sites by internal ribosome entry is not uncommon. Certain features, which are predicted to impair ribosome recruit- ment and linear scanning, have been found in the 5¢ untranslated regions (5¢UTRs) of mRNAs favouring inter- nal ribosome entry. These features include a long 5¢UTR, stable secondary structures, and potential upstream initi- ation codons [14]. To investigate whether E2F6 has a 5¢UTR that could promote protein synthesis from an internal AUG, we determined the site of transcription initiation by primer extension analysis (Fig. 2). A fragment containing 3 kb of genomic sequences upstream of the E2F6 coding sequence was subcloned and partially sequenced. Mouse E2F6 EST clones start approximately Fig. 1. Genomic organization of mouse E2f6 and the E2F6b splice variant. Black boxes represent coding exons. The open box represents the 5¢ untranslated region of exon 1 (see Fig. 3). The sequence of exon 2 is shown below. Translation termination and initiation codons are indicated. Fig. 2. Mouse E2F6 promoter and 5¢ untranslated region (5¢UTR). (A) Identification of the E2F6 transcriptional start site by primer extension analysis. Primer extension analysis was performed with radiolabelled primer SG50 (right lane). A genomic clone was sequenced with the same labelled primer, and the reaction was resolved on the same gel together with the primer extension reaction (left lanes). For primer location see (B). (B) Sequence of the mouse E2F6 promoter and 5¢UTR. The 650 bp nucleotide sequence 5¢ of the transcriptional start site (+1) is shown. The translation initiation codon is at +457. Upstream AUG codons are boxed. The location of the SG50 primer used in the primer extension reaction is shown. (C) Schematic representation of upstream initiation codons and of corresponding open reading frames in E2F6 and E2F6b. Black lines represent open reading frames with AUG codons in a context that favors translation initiation. Grey lines represent other open reading frames. Translation initiation of E2F6 is at +457. Translation of E2F6b is initiated at +628. Numbers represent location of initiation and termination codons relative to the transcription initiation site. 5032 T. Dahme et al. (Eur. J. Biochem. 269) Ó FEBS 2002 200 base pairs upstream of the E2F6 open reading frame, suggesting a minimum length of 200 nucleotides for the 5¢UTR of E2F6 (not shown). For primer extension analysis, we chose a backward primer in the untranslated region that anneals close to the beginning of the known EST clones (see Fig. 2B, primer SG50). Total RNA was isolated from primary MEFs, then radiolabelled primer SG50 was hybridized to the RNA and extended with reverse tran- scriptase. Sequencing reactions with the same radiolabelled primer and the subcloned genomic template were resolved on the same gel with the primer extension reaction. We observed a single band in the primer extension reaction corresponding to a transcriptional start site about 457 base pairs upstream of the translation start in Exon 1 (Fig. 2A). The presence of a single primer extension product indicates that transcription of E2F6 and E2F6b is initiated at the same position. Therefore, the length of the E2F6 5¢UTR is about 457 nucleotides, and that of the leader of the E2F6b mRNA up to the potential translation initiation codon in exon 2 is about 628 nt. The sequence of the mouse E2F6 5¢UTR with the transcriptional and translational start sites, as well as the partial promoter sequence, are shown in Fig. 2B. We noted that there are three upstream AUG triplets in the 5¢UTR of E2F6, and five upstream AUGs in E2F6b (Fig. 2C). None of the start codons have an optimal Kozak sequence. The third AUG at position +224, however, is in a context, which is favourable for translation initiation. However, it would only allow the generation of a short polypeptide terminated by in-frame stop codons. Furthermore, the 5¢UTRs of E2F6 and E2F6b mRNAs are predicted to form extensive secondary structures by the M - FOLD prediction software [15,16], as expressed in DGs of )225.7 kcalÆmol )1 for E2F6¢sand)270.9 kcalÆmol )1 for E2F6b’s 5¢UTRs. Taken together, our findings strongly oppose efficient translation initiation by ribosomal scanning for both E2F6b and E2F6, and suggest that their translation may be initiated internally. Two different E2F6 proteins accumulate in cells If translation of E2F6b is indeed initiated at the internal AUG in exon 2, it will give rise to an amino-terminal truncated protein lacking the first 36 amino acids of E2F6. This protein would be predicted to be of a smaller molecular mass than the previously described E2F6 protein. To identify a potential N-terminal truncated E2F6 protein, we used two different polyclonal antisera directed against oligopeptides derived from the C-terminus (C10) [7] and from the amino-terminus (N14) of murine E2F6 (see Fig. 3A). C10 is predicted to recognize both forms of E2F6, while N14 will only recognize the full-length E2F6 protein. The specificity of C10 and N14 antisera was confirmed by immunoprecipitation-Western experiments with in vitro translated proteins (Fig. 3B). Interestingly, lysates of MEFs derived from E2F6 deficient mice lacked two E2F6 specific bands, as compared with lysates of wildtype MEFs in a Western blot probed with C10 antiserum (data not shown, but see [7]). In these experi- ments, a second, slightly faster migrating protein was detected in wildtype MEFs but not in E2F6 deficient MEFs [7]. Unfortunately, neither Western blots, nor immuno- precipitation-Western blots of MEF lysates, probed with N14 antiserum, revealed any E2F6 specific signal that could be unambiguously distinguished from background (data not shown). We therefore employed an approach that turned out to be of higher sensitivity and specificity, and immunoprecipitated E2F6 from lysates of metabolically labelled MEFs. In these experiments, a common band corresponding in size to E2F6 was detected by the C10 and N14 antisera (Fig. 3C). Importantly, an additional, faster migrating protein was immunoprecipitated by C10, but not by the N14 antiserum, which is specific for the full-length E2F6 variant. The two proteins immunoprecipitated from cellular lysates by C10 antiserum correspond in size to in vitro translated E2F6 and E2F6b proteins when com- pared to molecular mass standards (compare Fig. 3B,C). Because of the lower affinity of the N14 antiserum, we cannot completely exclude the possibility that the band in the C10 immunoprecipitation corresponding in size to E2F6b is a proteolytic breakdown product. To address this possibility it will be necessary to generate a higher affinity antiserum that is specific for the E2F6 protein. Translation of E2F6b is initiated at the internal AUG codon To verify that the smaller E2F6b protein was the predicted E2F6b, we generated a reporter construct in which the luciferase coding sequence was fused in frame to the E2F6b coding sequence (Exon2-luc, see Fig. 4A). In this construct, the E2F6b N-terminal coding sequence replaces the initi- ation codon of the luciferase. Thus, luciferase enzyme activity can only be generated if translation is initiated within the E2F6b sequence. The mRNA transcribed from Fig. 3. E2F6b gives rise to an amino-terminal truncated protein that is initiated from an internal initiation codon. (A) Schematic representation of the E2F6 and E2F6b protein structure. The location of the peptides used to generate polyclonal antisera (C10 and N14) is schematically indicated (black lines). (B) Characterization of the C10 and N14 antisera. E2F6 and E2F6b were in vitro translated, and then subjected to immunoprecipitation and Western-blotting with the indicated antisera. Note that the C10 antiserum recognizes both E2F6 forms, whereas the N14 antiserum is specific for the full length E2F6 protein. When E2F6b was in vitro translated, two E2F6b bands of slightly different mobility were observed. (C) E2F6 was immunoprecipitated from lysates of metabolically labelled MEFs with the affinity purified E2F6 specific antisera C10 and N14, as indicated. C10 precipitates two proteins that correspond in size to E2F6 and E2F6b, while N14 recognizes only one protein that corresponds to E2F6. Because of the lower affinity of the N14 antiserum, we cannot completely exclude the possibility that the band in the C10 immunoprecipitation that corres- ponds in size to E2F6b is a proteolytic breakdown product. Ó FEBS 2002 E2F6b, an alternatively spliced E2F6 isoform (Eur. J. Biochem. 269) 5033 this construct contains exon 1, exon 2, the first seven triplets of exon 3, and thereafter the luciferase coding sequence lacking only the initiating AUG. Two constructs that lack exon 2 served as controls. In Exon3-luc, the luciferase coding sequence was introduced after the seventh codon of exon 3, which is the same fusion point as in Exon2-luc. Since there is the possibility that this rather long amino- terminal E2F6 sequence fused to the luciferase will influence its activity, a second control that only contained the first seven codons of exon 1 (Exon1-luc) was used. Importantly, all luciferase constructs contain the complete E2F6 5¢UTR, and their transcription is driven by a 2.2-kb fragment of the E2F6 promoter (see Fig. 4A). Transient transfection assays in U2-OS cells revealed dose-dependent activity of Exon2-luc which was two to three times lower than that of Exon1-luc and Exon3-luc (Fig. 4B). However, activity of Exon2-luc was still up to more than 1000 times higher than the activity of the empty vector, pGL2-Avr (Fig. 4B, right lanes). Similar results were found in NIH-3T3 cells (data not shown). We therefore concluded that a protein is expressed from the E2F6b mRNA. Our results suggest that translation of Exon2-luc is initiated internally. Alternatively, it is possible that a protein is generated by a read-through of the stop codon in exon 2. To examine which mechanism is involved in the synthesis of this protein, we mutated the internal AUG in exon 2 to GGG (to generate Exon2-mut-luc), and compared its activity to the activity of Exon2-luc in transient transfection assays. The mutation led to complete loss of activity in these assays, indicating that the integrity of the AUG is crucial for translation of E2F6b (Fig. 4C). Taken together, this strongly suggests that translation can be initiated at the internal AUG in exon 2 of the E2F6b mRNA. E2F6 and E2F6b are ubiquitously expressed It has previously been reported that E2F6 is ubiquitously expressed in mouse tissues [10]. To analyze the relative expression of E2F6 and E2F6b in a larger panel of mouse tissues, we established a semiquantitative RT-PCR strategy. Total RNA was isolated from multiple tissues of an adult C57/Black6 mouse and subjected to reverse transcription followed by PCR amplification with the primer set SG122 and SG24. E2F6 and E2F6b mRNA expression is shown in Fig. 5A (top). Beta-actin specific primers were used as a control in a separate RT-PCR reaction (Fig. 5A, bottom). Both E2F6 and E2F6b mRNA were detected in all tissues examined. The highest expression levels of E2F6 and E2F6b were found in heart and skeletal muscle. In most tissues, E2F6 levels were higher or equal to E2F6b. However, one exception was skeletal muscle, where E2F6b was several fold more abundant than E2F6 (Fig. 5A, right lane). To analyze whether the expression of E2F6 and E2F6b is cell cycle dependent, MEFs were serum starved for 48 h, and then released from starvation by the addition of serum. Total RNA was isolated at three-hour intervals. In addition, we isolated RNA from asynchronously growing cells, and from confluent, contact inhibited cells. E2F6 and E2F6b expression was again analyzed by RT-PCR with Fig. 5. Expression of E2F6 and E2F6b. (A)ExpressionofE2F6and E2F6b in primary mouse tissues was analyzed by RT-PCR with primers SG122 and SG24 (top). RT-PCR with b-actin specific primers was used as a control. (B) Expression of E2F6 and E2F6b during the cell cycle. Mouse primary fibroblasts were brought to quiescence by incubationfor48hinserumfreemedium,andthenreleasedintothe cell cycle by the addition of 10% serum. RNA was isolated at the indicated times after the addition of serum, and E2F6 and E2F6b expression was analyzed by RT-PCR with primers SG122 and SG24. RT-PCR with b-actin specific primers was used as a control. Expres- sion in confluent (confl.) and asynchronously growing cells (asynchr.) was also analyzed. The percentage of cells in the G0/G1, S, and G2/M phases of the cell cycle at each time point was determined by FACS analysis and is shown at the bottom. Fig. 4. Translation of E2F6b is initiated at an internal initiation codon. (A) Schematic representation of the E2F6-luciferase fusion constructs. The SacIsiteat)2.2 kb was used for cloning of the E2F6 promoter (see Fig. 1). Transcription initiation (+ 1) is indicated by a right arrow (see Fig. 2). (B) Activity of Exon2-luc (Ex2-luc), Exon1-luc (Ex1-luc), and Exon3-luc (Ex3-luc), compared to the activity of the empty vector (pGL2-Avr). Plasmids were transiently expressed in U2-OS cells, and luciferase activity was determined. Plasmid inputs (ng DNA) are indicated. Activity of pGL2-Avr at 100 ng was set to 1. (C) Activity of Exon2-mut-luc compared to Exon2-luc after transient expression in U2-OS cells. Activity of Exon2-luc at 100 ng was set to 1. 5034 T. Dahme et al. (Eur. J. Biochem. 269) Ó FEBS 2002 the primer set SG122 and SG24 (Fig. 5B, top). Cell cycle synchronization was monitored by FACScan analysis of parallel samples (Fig. 5B, bottom). E2F6 and E2F6b mRNA levels were lowest in serum-starved cells. During re-entry into the cell cycle, E2F6 and E2F6b transcription increased with peak levels in late G1/early S. Moderate levels of the two transcripts were found in confluent and asynchronously growing cells. The relative abundance of the two E2F6 mRNAs did not change significantly during the cell cycle. E2F6b is a transcriptional repressor Although E2F6b lacks the amino-terminus of E2F6, it is otherwise identical in sequence to E2F6, except for the initiating methionine derived from exon 2 (see Fig. 3A). Notably, E2F6b contains the DNA-binding and dimeriza- tion domains, suggesting that E2F6b will bind to DNA and dimerize with DP proteins like the previously described E2F6 protein. Previous work has shown that E2F6 is a pocket protein independent transcriptional repressor [3–6]. The repression domain of human E2F6 is localized in the C-terminus of the protein [5]. In contrast, in mouse E2F6 (also termed EMA), a repression domain has been identified in the amino-terminus [3]. Since E2F6b lacks the first 36 amino-terminal amino acids, we wanted to determine the properties of E2F6b as a transcriptional regulator. To address this issue, we performed transient transfection assays. Two different reporter plasmids were utilized, in which luciferase expression is under the control of the p14 ARF and the E2F1 promoters. Both promoters contain E2F sites and have previously been shown to be regulated in an E2F-dependent manner [11,12]. p14 ARF and E2F1 luciferase reporter plasmids were cotransfected with E2F6 and E2F6b expression plasmids. As expected, E2F6 reduced activity of both promoters by about 50% to 60% (Fig. 6A,B). Surprisingly, E2F6b also reduced the activity of these promoters, although it was slightly less efficient than E2F6. To rule out the possibility that sequestration of the dimerization partner DP is responsible for the reduction in luciferase activity, we coexpressed DP2 together with E2F6 and E2F6b (Fig. 6A,B, + DP2). Coexpression of DP2 resulted in even further reduction of reporter activity, indicating that repression by E2F6b is not a result of titration of endogenous DP proteins. Taken together, these results show that E2F6b is also a transcriptional repressor. DISCUSSION E2F6, the most recently identified E2F protein, is a retinoblastoma-protein independent transcriptional repres- sor. In mice, E2F6 is required for developmental patterning of the axial skeleton [7]. Together with the recent finding that E2F6 associates with polycomb proteins [8,9], these observations suggest that E2F6 recruits polycomb com- plexes to certain target promoters during development. It has recently been reported that the E2f6 locus generates two differentmRNAs,E2F6andE2F6b[10].However,no evidence for the generation of a protein from the E2F6b mRNA was given. Interestingly, exon 2 in E2F6b contains a stop codon, in-frame with the upstream E2F6 sequence. mRNA species that contain premature termination codons are often destroyed by nonsense mediated decay so that only full-length proteins are produced [17]. However, not all mRNAs that contain premature termination codons are targeted for destruction. At the moment it is not clear how the E2F6b splice variant escapes nonsense mediated decay. In this study, we present evidence that a truncated E2F6 protein, E2F6b, is produced from the alternatively spliced mRNA by internal initiation of translation. First, we show that an amino-terminal truncated E2F6 protein is present in cellular lysates of mouse embryonic fibroblasts. Secondly, we found the following structural features in the 5¢ leaders of E2F6b that make translation initiation by a ribosome- scanning mechanism unlikely: (a) a long 5¢ untranslated region, (b) stable secondary structure, and (c) potential upstream initiation codons. The presence of these features in the E2F6b mRNA strongly suggests that translation of this protein is not initiated by a normal CAP-dependent mechanism. Finally, with a set of luciferase-reporter constructs, we demonstrate that an internal initiation codon is required for translation of E2F6b. Translation initiation from internal start codons is usually not compatible with normal CAP-dependent ribosomal scanning. In conclusion, these data strongly suggest that E2F6b is initiated by internal ribosome entry. We noted that in an earlier study, the transcription start site of E2F6 was mapped to 256 and 241 nucleotides upstream of the E2F6 AUG codon [10]. In contrast, we found a single start at 457 nucleotides upstream of the AUG. It is possible that the RNase-protection approach used by Kherrouche et al. is more sensitive to extensive mRNA secondary structure than our primer extension strategy. Alternatively, it is possible that the transcription start site of E2F6 is tissue dependent. In agreement with an earlier report, we found that the E2F6b mRNA is ubiquitously expressed in a wide variety of tissues. The highest expression levels were found in heart and skeletal muscle. In most tissues, except for skeletal muscle and heart, E2F6 was more abundant than E2F6b. However, in another study, for some of those tissues reverse ratios between E2F6 and E2F6b were found [10]. It is possible that this discrepancy is due to the different primers used for the reverse transcription reaction. While we used a Fig. 6. E2F6b is a transcriptional repressor. U2-OS cells were trans- fected with 200 ng E2F-dependent reporter plasmids E2F1-luc (A) [12], or with E1B-luc (with the p14 ARF promoter) [11] (B), and with 200 ng E2F6, E2F6b, or DP2 expression plasmids, as indicated. 50 ng CMV-b gal was cotransfected, and luciferase activity was normalized to b-galactosidase activity. Basal activity of the reporter plasmid in presence of empty expression vector was set to 1. Ó FEBS 2002 E2F6b, an alternatively spliced E2F6 isoform (Eur. J. Biochem. 269) 5035 gene specific primer, an oligo(dT) primer was used in the previous study. We also show that the ratio between the two E2F6 mRNAs does not change significantly during the cell cycle. The E2F6b protein lacks the amino-terminal 36 amino acids derived from exon 1. It shares with E2F6 the DNA- binding and dimerization domain, as well as the C-terminus. Consequently, E2F6b is predicted to bind to DNA, and to dimerize with DP proteins in a manner similar to E2F6. Surprisingly, E2F6b, like E2F6, is a repressor of E2F-site dependent transcription (Fig. 6), despite the fact that a repression domain has been assigned to the amino terminus of mouse E2F6 [3]. It is possible that E2F6b specifically represses some, but not other E2F-dependent promoters in vivo. Further experiments will be necessary to address this possibility. Internal ribosome entry is known to still be functional under conditions in which the usual CAP-dependent mechanism is inactive [18–20]. Examples include translation of c-myc, which is mediated by an internal ribosome entry site during apoptosis [18], and efficient translation of vascular endothelial growth factor from an internal ribosome entry site during hypoxia [19]. It is worth noting that upstream open reading frames and CAP-independent translation are commonly found in genes whose products play roles in the regulation of cell growth and in the cellular response to stress. Interestingly, deregulation of translation initiation is a common feature of tumorigenesis. For example, increased expression of the eukaryotic translation factor eIF4E has been reported in a number of different cancers [21]. It has been speculated that overproduction of eIF4E in tumors promotes translation of mRNAs with long and complex 5¢UTRs. Indeed, sequence analysis demonstrated that mRNAs with complex 5¢UTRs often encode for proto- oncogenes [21]. It remains to be shown whether E2F6 and/ or E2F6b play a role in tumorigenesis. ACKNOWLEDGEMENTS We wish to thank Stefanie Hauser and our laboratory and divisional colleagues for many helpful conversations. We thank Kelly Farrenkopf for proofreading and for helpful comments. We also thank Gordon Peters and William Kaelin for the E1B-luc and E2F1-luc constructs, respectively. This work was supported by fellowships from the Leukemia and Lymphoma Society and the Volkswagenstiftung to S. G. REFERENCES 1. Trimarchi, J.M. & Lees, J.A. (2002) Sibling rivalry in the E2F family. Nat.Rev.MolCellBiol.3, 11–20. 2. Adams, P.D. & Kaelin, W.G.J. (1996) The cellular effects of E2F overexpression. Curr. Top. Microbiol. Immunol. 208, 79–93. 3. Morkel, M., Wenkel, J., Bannister, A.J., Kouzarides, T. & Hagemeier, C. (1997) An E2F-like repressor of transcription. Nature. 390, 567–568. 4. Cartwright, P., Muller, H., Wagener, C., Holm, K. & Helin, K. (1998) E2F-6: a novel member of the E2F family is an inhibitor of E2F-dependent transcription. Oncogene 17, 611–623. 5. Gaubatz, S., Wood, J.G. & Livingston, D.M. (1998) Unusual proliferation arrest and transcriptional control properties of a newly discovered E2F family member, E2F-6. Proc.NatlAcad. Sci. USA 95, 9190–9195. 6. Trimarchi, J.M., Fairchild, B., Verona, R., Moberg, K., Andon, N. & Lees, J.A. (1998) E2F-6, a member of the E2F family that can behave as transcriptional repressor. Proc. Natl. Acad. Sci. USA 95, 2850–2855. 7. Storre, J., Elsaesser, H P., Fuchs, M., Ullmann, D., Livingston, D.M. & Gaubatz, S. (2002) Homeotic transformations of the axial skeleton that accompany a targeted deletion of E2F6. EMBO Reports 3, 695–700. 8. Trimarchi, J.M., Fairchild, B., Wen, J. & Lees, J.A. (2001) The E2F6 transcription factor is a component of the mammalian Bmi1-containing polycomb complex. Proc. Natl Acad. Sci. USA 98, 1519–1524. 9. Ogawa, H., Ishiguro, K., Gaubatz, S., Livingston, D.M. & Nakatani, Y. (2002) A complex with chromatin modifiers that occupies E2F- and Myc-responsive genes in G0 cells. Science. 296, 1132–1136. 10. Kherrouche, Z., Begue, A., Stehelin, D. & Monte, D. (2001) Molecular cloning and characterization of the mouse E2F6 gene. Biochem. Biophys. Res. Commun. 288, 22–33. 11. Bates, S., Phillips, A.C., Clark, P.A., Stott, F., Peters, G., Ludwig, R.L. & Vousden, K.H. (1998) p14ARF links the tumour sup- pressors RB and p53. Nature 395, 124–125. 12. Neuman, E., Flemington, E.K., Sellers, W.R. & Kaelin, W.G. (1994) Transcription of the E2F-1 gene is rendered cell cycle dependent by E2F DNA binding sites within its promoter. Mol. Cell. Biol. 14, 6607–6615. 13. Pestova, T.V., Kolupaeva, V.G., Lomakin, I.B., Pilipenko, E.V., Shatsky, I.N., Agol, V.I. & Hellen, C.U. (2001) Molecular mechanisms of translation initiation in eukaryotes. Proc. Natl Acad. Sci. USA 98, 7029–7036. 14. Vagner, S., Galy, B. & Pyronnet, S. (2001) Irresistible IRES. Attracting the translation machinery to internal ribosome entry sites. EMBO Report 2, 893–898. 15. Mathews,D.H.,Sabina,J.,Zuker,M.&Turner,D.H.(1999) Expanded sequence dependence of thermodynamic parameters improves prediction of RNA secondary structure. J. Mol. Biol. 288, 911–940. 16. Zuker, M., Mathews, D.H. & Turner, D.H. (1999) Algorithms and Thermodynamics for RNA Secondary Structure Prediction: A Practical Guide. RNA Biochemistry and Biotechnology (Barcisz- weki, J. & Clark, B.F.C., eds), pp. 11–43, Kluwer Academic Publishers, Dordrecht. 17. Maquat, L.E. (2002) Nonsense-mediated mRNA decay. Curr. Biol. 12, R196–R197. 18. Stoneley, M., Chappell, S.A., Jopling, C.L., Dickens, M., MacFarlane, M. & Willis, A.E. (2000) c-Myc protein synthesis is initiated from the internal ribosome entry segment during apop- tosis. Mol. Cell. Biol. 20, 1162–1169. 19. Stein,I.,Itin,A.,Einat,P.,Skaliter,R.,Grossman,Z.&Keshet, E. (1998) Translation of vascular endothelial growth factor mRNA by internal ribosome entry: implications for translation under hypoxia. Mol. Cell. Biol. 18, 3112–3119. 20. Harding, H.P., Novoa, I., Zhang, Y., Zeng, H., Wek, R., Schapira, M. & Ron, D. (2000) Regulated translation initiation controls stress-induced gene expression in mammalian cells. Mol. Cell. 6, 1099–1108. 21. Watkins, S.J. & Norbury, C.J. (2002) Translation initiation and its deregulation during tumorigenesis. Br. J. Cancer 86, 1023–1027. 5036 T. Dahme et al. (Eur. J. Biochem. 269) Ó FEBS 2002 . both E2F6b and E2F6, and suggest that their translation may be initiated internally. Two different E2F6 proteins accumulate in cells If translation of E2F6b. Two different E2F6 proteins generated by alternative splicing and internal translation initiation Tillman Dahme 1 , Jason