Alternative Translation of the Proto-oncogene c-mycby an Internal Ribosome Entry Site*

The human proto-oncogene c-mycencodes two proteins, c-Myc1 and c-Myc2, from two initiation codons, CUG and AUG, respectively. It is also transcribed from four alternative promoters (P0, P1, P2, and P3), giving rise to different RNA 5′-leader sequences, the long sizes of which suggest that they must be inefficiently translated by the classical ribosome scanning mechanism. Here we have examined the influence of three c-myc mRNA 5′-leaders on the translation of chimeric myc-CAT mRNAs. We observed that in the reticulocyte rabbit lysate, these 5′-leaders lead to cap-independent translation initiation. To determine whether this kind of initiation resulted from the presence of an internal ribosome entry site (IRES), COS-7 cells were transfected with bicistronic vectors containing the different c-myc5′-leaders in the intercistronic region. An IRES was identified, requiring elements located within the P2 leader, between nucleotides −363 and −94 upstream from the CUG start codon. This is the first demonstration of the existence of IRES-dependent translation for a proto-oncogene. This IRES could be a translation enhancer, allowing activation of c-myc expression under the control of trans-acting factors and in response to specific cell stimuli.

The c-myc proto-oncogene has a fundamental role in various cellular events, including proliferation, differentiation, and apoptosis (1). The human c-myc gene is transcribed from four alternative promoters. In normal cells, most of the transcripts start at the P1 and P2 promoters, the latter accounting for 75-90% of the c-myc mRNA. In Burkitt's lymphoma cells that have a c-myc chromosomal translocation, the P1 promoter is preferentially used. In addition, two additional promoters, P0 and P3, are located 600 nucleotides upstream from the P1 promoter and in the first intron of the gene, respectively (2,3).
In Burkitt's lymphoma cell lines, P0 mRNA can reach levels of Ͼ10% of the total c-myc mRNA (4).
Interestingly, the coding capacity of the four c-myc mRNAs is different. P1 and P2 mRNAs encode two proteins of 64 and 67 kDa, designated as c-Myc1 and c-Myc2, respectively, by a process of alternative initiation of translation starting at two inframe codons, CUG and AUG (5). c-Myc1 and c-Myc2 proteins, whose ratio is regulated in response to methionine starvation, are transcription factors with distinct DNA targets (6,7). P0 mRNA is a natural polycistronic mRNA containing three open reading frames (ORFs). 1 The 3Ј-ORF codes for c-Myc1 and c-Myc2 proteins, whereas the middle and 5Ј-ORFs, which have initiation codons upstream from the P1 promoter, code for proteins of 188 and 114 amino acids, respectively (2,8). Only the 188-amino acid product has been characterized in HeLa cells. The function of this MycHEX1 protein remains unknown, however (9). In contrast to the coding capacity of P0, P1, and P2 mRNAs, that of P3 mRNA is restricted to the 64-kDa c-Myc2 protein.
The transcription starting points of the P0, P1, and P2 c-myc mRNAs are located 1172, 524, and 363 nt upstream from the CUG initiation codon, respectively. According to the classical cap-dependent scanning model, such long leader sequences are expected to impair translation initiation by preventing the ribosome scanning from the capped mRNA 5Ј-end (10). It has indeed been reported that the c-myc P1 leader has an inhibitory effect on translation of the c-myc mRNA in rabbit reticulocyte lysate (RRL) and in Xenopus oocytes (11,12). However, in several mRNAs, mainly of viral origin, the structure of long leader sequences has been shown to constitute an internal ribosome entry site (IRES) (13,14), allowing translation initiation to be cap-independent. The internal entry process, first shown for picornaviruses (15,16), has also been reported for a few cellular mRNAs, but never for a proto-oncogene (17)(18)(19)(20)(21). In picornaviruses, the internal entry process has been shown to require cellular trans-acting factors. The best characterized of these factors is the polypyrimidine tract-binding protein, also known as a splicing factor (22)(23)(24). Due to their positive regulation by cellular factors, IRESs can be considered as translation enhancers, allowing activation of tightly controlled genes in response to specific stimuli (25). In this report, we demonstrate the presence of an IRES in P0, P1, and P2 c-myc mRNAs, implying an element located between nucleotides Ϫ363 and Ϫ95 upstream from the CUG start codon.

MATERIALS AND METHODS
Plasmid Construction-The c-myc P1 cDNA was obtained by reverse transcription and polymerase chain reaction (PCR) amplification of total mRNA from human tumorigenic cells (kindly provided by D. Morello) using Superscript reverse transcriptase (Life Technologies, Inc.) and Goldstar Taq DNA polymerase (Eurogentec). The 5Ј-and  3Ј-primers used for this amplification, 5Ј-TCGCTCTAGATGAGGAC-CCCCGAG-3Ј and 5Ј-TTTGGCGCCGATATCCTCGCTGGGCGC-3Ј, respectively, gave a cDNA fragment extending from positions 1 to 712 from the P1 leader 5Ј-end (positions 1 and 712 are 1361 nt downstream from the P0 leader 5Ј-end and 143 nt downstream from the AUG codon, respectively). This fragment was inserted into the plasmid pFC1, derived from the pSCT vector, upstream from the CAT coding sequence and under the control of cytomegalovirus and T7 promoters (26). The chimeric construct, pSCT-MyCAT-P1, was expected to encode two Myc-CAT proteins of 29 and 31 kDa, respectively.
The c-myc P2 cDNA was obtained from the construct MyCAT-P1 by PCR amplification and cloning of the fragment extending from the P2 5Ј-end (610 nt downstream from the P0 5Ј-end) to 3Ј of CAT using oligonucleotides 5Ј-AGAGTCTAGAACTCGCTGTAGTAATTCCAGC-3Ј and 5Ј-AATCTCGAGTCTAGAGTATTCGCCCCGCCCTGCCA-3Ј (CA-Trev) as 5Ј-and 3Ј-primers, respectively. The plasmid pSCT-MyCAT-P2 was constructed by insertion of the Myc-CAT-P2 fragment into the pSCT vector.
The c-myc P0 cDNA was obtained by PCR amplification of the P0 cDNA leader from the plasmid pHLM-P1 containing the c-myc genomic sequence (kindly provided by D. Morello) with oligonucleotides 5Ј-AG-AGTCTAGAAACAAATGCAATGGGAGTTTATTCATAACGCGC-3Ј (5ЈP0) and 5Ј-CCCGCCAAGCCTCTGAGAAGCCCTGCCC-3Ј as 5Ј-and 3Ј-primers, respectively. This fragment, corresponding to nt 1-750 of the P0 cDNA, was inserted in the MyCAT-P1 construct to recreate the complete P0 leader. The new plasmid was called pSCT-MyCAT-P0.
The construct pSCT-MyCAT-⌬1, with a deleted leader, was constructed by PCR amplification from the plasmid pSCT-MyCAT-P0 of the fragment extending from nt 1158 (counting from the P0 5Ј-end) to 3Ј of the CAT sequence using oligonucleotides 5Ј-AGAGTCTAGAG-CAGCTGCTTAGACGCTGGAT-3Ј and CATrev as 5Ј-and 3Ј-primers, respectively. The PCR fragment was introduced into the pSCT vector. The construct pSCT-MyCAT-⌬2 was obtained using the same strategy as that used for MyCAT-⌬1, but with the 5Ј-primer 5Ј-AAAACTAGTC-GACGCGGGGAGGCTATT-3Ј (with a 5Ј-end at position 1078 from the P0 5Ј-end).
To mutate the P1 promoter, two PCR fragments were amplified from pSCT-MyCAT-P0 DNA. The first fragment, obtained with oligonucleotides 5ЈP0 and 5Ј-TTTCATATGAAAGGGCCGGTGGGCG-3Ј as 5Ј-and 3Ј-primers, respectively, extended from nt 1 (from the P0 5Ј-end) to nt 627 (position of the P1 TATA box), with an NdeI site at the 3Ј-end in the P1 TATA box; the second fragment, obtained with oligonucleotides 5Ј-TTTCATATGCGAGGGTCTGGACGGC-3Ј and CATrev as 5Ј-and 3Јprimers, respectively, extended from nt 622 of the P0 leader (P1 TATA box) to 3Ј of CAT, with an NdeI site at the 5Ј-end. The two fragments were cloned in the pSCT vector, and the new plasmid was called pSCT-MyCAT-P0(P1*).
The P2 promoter was mutated using the same strategy as that used for the P1 promoter. Two PCR fragments were amplified from the construct MyCAT-P0(P1*). The first, obtained with oligonucleotides 5ЈP0 and 5Ј-GGGATGCATACTCAGCGCGATCCCTC-3Ј as the 5Ј-and 3Ј-primers, respectively, extended from nt 1 (from the P0 5Ј-end) to nt 786 (position of the P2 TATA box), with an NsiI site at the 3Ј-end; the second fragment, obtained with oligonucleotides 5Ј-GGGATGCATGC-CGGTTTTCGGGGCTT-3Ј and CATrev as the 5Ј-and 3Ј-primers, respectively, extended from nt 781 of the P0 leader (position of the P2 TATA box) to 3Ј of CAT, with an NsiI site at the 5Ј-end. The two fragments were cloned in the pSCT vector, and the new plasmid, with the two mutated promoters, was called pSCT-MyCAT-P0*.
The two series of bicistronic vectors, pSCT-HP-MyCAT and BI-My-CAT (with or without a 5Ј-hairpin, respectively), were constructed from the bicistronic vectors pHP-FC1 and pBI-FC1 with an FGF-CAT fusion (20). The Myc-CAT fusions with different leaders were introduced downstream from CAT, in place of the FGF-CAT fusion. The nucleotide sequences of the different DNA constructs were verified using an automatic DNA sequencer (Applied Biosystems).
In Vitro Transcription and Translation-DNAs were linearized, and transcription was performed with T7 RNA polymerase as described previously (20). RNA transcripts were quantitated by absorbance at 260 nm and ethidium bromide staining on agarose gel, and their integrity was verified. Translation was carried out in rabbit reticulocyte lysate (Promega) (20). The amounts of mRNA present in the translation assays were checked by Northern blotting. The translation products were analyzed by electrophoresis on a 12.5% polyacrylamide gel, followed by autoradiography and quantitation on a PhosphorImager (Molecular Dynamics, Inc.).
Cell Transfections and Western Immunoblotting-COS-7 monkey cells were transfected with 1 g/ml DNA by the DEAE-dextran method (20). Cell lysates were prepared 48 h later. Total proteins were prepared, quantified, and analyzed by Western immunoblotting (5 g of proteins from each cell lysate) as described previously (20). CAT proteins were immunodetected using rabbit polyclonal anti-CAT antibodies prepared in our laboratory (1:20,000 dilution).
Cellular RNA Purification and Northern Blotting-Total cellular RNA was prepared by the Trizol method (Life Technologies, Inc.) from pellets containing 5 ϫ 10 6 transfected scraped cells. Northern blotting was performed as described previously (20). DNA probes were labeled with [ 32 P]dATP using a random priming kit (Promega). Total cellular RNA (1 g/lane) was subjected to electrophoresis through 1.2% formaldehyde-agarose gels, electrotransferred to nylon membrane, and hybridized under the conditions described previously (20). The MyCAT-P0 construct contains the P1 and P2 TATA boxes. To distinguish between the transcription and translation capacities of this plasmid in transfected cells, we abolished the putative activity of these internal promoters by mutating the two P1 and P2 TATA boxes in the MyCAT-P0 construct (as described under "Materials and Methods"); the new construct was designated MyCAT-P0* (Fig. 1A).

Cap-independent Translation Is Conferred by c-myc 5Ј-Leaders-To
Capped and uncapped mRNAs were transcribed in vitro and translated in RRL in the presence of [ 35 S]methionine. Their translation efficiencies were compared (Fig. 1, B and C). The translational levels of capped P0, P1, P2, and ⌬2 mRNAs were approximately 2, 10, 50, and 100% of that of the capped ⌬1 mRNA devoid of leader sequence, respectively (Fig. 1, B (lanes  2, 6, 8, 10, and 12) and C). P0* mRNA, with mutations of the P1 and P2 TATA boxes, was translated at the same level as P0 mRNA (Fig. 1B, lanes 2 and 4).
The translation efficiencies of the capped and uncapped mRNAs were compared to determine the cap dependence of translation initiation. The expression of the Myc-CAT proteins using P0 and P0* mRNAs was cap-independent, whereas the expression of 5Ј-ORF1 was increased in the presence of the cap ( Fig. 1, B (lanes 2-5) and C). Translation of Myc-CAT proteins with P1 mRNA was little influenced by the cap, whereas the expression of the MycHEX1 protein, hardly detectable in the P0 construct, was cap-dependent (Fig. 1B, lanes 6 and 7). Finally, Myc-CAT protein expression was weakly cap-dependent from P2 mRNA, but strongly cap-dependent from the deleted ⌬2 and ⌬1 mRNAs (Fig. 1B, lanes 6 -13).
In conclusion, these experiments clearly show that cap-independent expression of CUG-and AUG-initiated Myc-CAT proteins is conferred by the three P0, P1, and P2 c-myc 5Ј-leaders. However, a weak cap-dependent translation initiation was also observed with the P1 and P2 5Ј-leaders, whereas synthesis of the 5Ј-ORFs was cap-dependent in P0 and P1 mRNAs.
An IRES Controls the Synthesis of c-Myc1 and c-Myc2 Proteins-To determine whether the cap-independent translation initiation, observed in the RRL system for the P0 leader and also for the P1 and P2 leaders, occurred by a process of ribosome entry, we looked for the presence of an IRES in the c-myc mRNA. The same bicistronic vector assay as that described for the identification of FGF-2 and murine leukemia virus IRESs in previous reports (20,27) was used. This strategy is based on the principle that, according to the cap-dependent ribosome scanning model (10), the second ORF of a bicistronic mRNA will not be expressed unless it is preceded by an IRES. Two series of bicistronic vectors were constructed; the first one had as the first ORF the CAT coding sequence and as the second ORF the above-described Myc-CAT fusion sequence (Fig. 2, BI-MyCAT). The presence of two tandem CAT genes encoding proteins of different sizes allowed comparison of the expression of the two cistrons (20,27). The different c-myc leaders described in Fig. 1A were inserted in the intercistronic region. The second series of bicistronic vectors was derived from the first by the introduction, upstream from the first ORFs, of a stable hairpin intended to inhibit the cap-dependent, but not the IRES-mediated, initiation of translation (Fig. 2, HP-My-CAT) (17).
The different bicistronic DNAs, with or without a hairpin, were used to transiently transfect COS-7 cells. The natural with a shorter exposure of the gel for  tricistronic vector MyCAT-P0* (encoding ORF1, MycHEX1, and Myc-CAT) was also used in these transfection experiments. Northern experiments enabled us to check that the bicistronic mRNAs had the expected size (Fig. 3A). The absence of any additional band allowed us to rule out the presence of an active internal promoter.
The expression of CAT and Myc-CAT proteins was analyzed by Western immunoblotting with an anti-CAT antibody. The results showed that in COS cells, the Myc-CAT proteins were efficiently expressed, compared with CAT first cistron, from the bicistronic vectors BI-MyCAT-P0* and HP-MyCAT-P0*: both were expressed as efficiently as the MyCAT-P0* construct, which lacks the CAT first cistron (Fig. 3B, lanes 2-4). The bicistronic P2 construct also expressed the Myc-CAT cistron (Fig. 3B, lanes 5 and 6). The presence of the 5Ј-hairpin downregulated the CAT cistron expression from these vectors, but had no effect on the expression of the Myc-CAT cistron (indicating that Myc-CAT was expressed in a cap-independent manner). In contrast, Myc-CAT expression was barely detectable with the bicistronic ⌬2 and ⌬1 vectors (Fig. 3B, lanes 7-10).
We conclude from these data that an IRES is present in both the P0 and P2 5Ј-leaders of c-myc mRNA. This structure therefore involves elements located between nucleotides Ϫ363 and Ϫ94 upstream from the CUG initiation codon. Furthermore, the IRES allows a higher Myc-CAT translational level compared with the CAT first cistron, indicating that it is very efficient.
Sequence Alignment of the c-myc IRES with the FGF-2 IRES and Secondary Structure Prediction- Fig. 4A shows a sequence alignment of c-myc P2 leader sequence with nucleotides 189 -487 of human FGF-2 mRNA containing the IRES. Although there is no striking sequence homology, several conserved GCrich stretches are apparent. According to a recent report, several of these blocks are involved in stems of the E1 and E2 motifs predicted to form the FGF-2 IRES structure (28): in particular, the a/aЈ, b/bЈ, and d/dЈ elements of the E2 motif (Fig.  4A). The c-myc P2 RNA structure predicted by the Zuker method (35), with a⌬G of Ϫ122 kcal/mol, divides the leader into three major domains (Fig. 4B). The large D1 domain alone has a ⌬G of Ϫ78 kcal/mol. The D1 and D2 domains, extending from nucleotides 1 to 205 and from nucleotides 206 to 293, respectively, could participate in the IRES as they are not present in the IRES-minus ⌬2 construct (starting at nucleotide 269 from the P2 5Ј-end).

DISCUSSION
In this study, we demonstrate the presence of an IRES in the c-myc mRNA. So far, it is the first IRES to be characterized in the mRNA of a proto-oncogene. Most of the few cellular mRNAs described up to now as containing IRESs code for proteins involved in the control of cell proliferation and differentiation and require stringent regulation, as do c-myc mRNAs; these genes need to be expressed at very specific stages of cell life and/or in response to different stimuli (17)(18)(19)(20)(21). The IRES, being a recognition site for translational regulatory factors, allows translational activation of the expression of such messengers presenting a low or even non-existent level of cap-dependent translation; thus, internal ribosome entry constitutes a novel mechanism of gene expression regulation. This has been shown in the case of FGF-2, whose CUG-initiated isoforms are translationally activated in response to stress (25). Furthermore, the constitutive expression of these FGF-2 isoforms has been observed in transformed cells (25). 2 Such observations suggest that, in the case of c-myc, a relationship may exist between IRES activation and cell transformation.
To corroborate this hypothesis, it has been recently reported that c-myc expression is activated at the translational level in cell lines derived from patients with the cancer-prone disorder Bloom's syndrome (29). This activation does not appear to be mediated by the cap-binding eukaryotic initiation factor 4E, suggesting that it occurs through a cap-independent pathway. In the light of our results, one can postulate that in these cells, c-myc is activated by an IRES-dependent translation pathway. This activation may be triggered by the increased level of DNA strand breaks that occurs in Bloom's syndrome cells, as previously proposed (29). This supports our hypothesis, already suggested by the stress activation of FGF-2 isoforms (25), that the internal ribosome entry process is a cellular regulation pathway allowing a rapid response to disorder-generating stimuli.
The c-myc mRNA is the third example of an mRNA with alternative initiation codons to contain an IRES (20,27,30). However, in contrast to Moloney murine leukemia virus retroviral mRNA, showing an IRES between its CUG and AUG initiation codons and allowing translation of the AUG-initiated protein exclusively, the c-myc mRNA has an IRES located 2 B. Galy, unpublished results. upstream from the two initiation codons. Thus, this IRES controls the expression of both c-Myc1 and c-Myc2 proteins and is similar in this respect to that found in the FGF-2 mRNA. The sequence alignment of c-myc and FGF-2 IRESs does not show a strong sequence homology (Fig. 4A); however, several conserved GC-rich stretches are apparent that could lead to a structural homology (28). The c-myc IRES secondary structure prediction shows three major domains (Fig. 4B), whose function will be investigated.
The possible similarity between FGF-2 and c-myc IRESs mentioned above may be related to the fact that both genes play a role in the control of cell proliferation and differentiation. Indeed, the translation efficiency of oncogene expression emerges more and more in the literature as an important regulatory target in malignant conversion. This translation efficiency can be regulated by at least two pathways. The first pathway is cap-dependent; oncogene messengers with highly structured untranslated regions can be up-regulated by overexpression of eukaryotic initiation factor 4E, characterized as a proto-oncogene (31,32). We show here that a second possible pathway involving the IRESs should be considered. These specific structures are under the control of internal entry factors. Work is in progress to identify the factors involved in this regulation mechanism. Furthermore, both processes are related to one another in a regulatory loop: eukaryotic initiation factor 4E is indeed activated by growth factors (33), and its expression is induced by several oncogenes, including c-myc (34). Thus, the disregulation of the c-myc IRES will have a FIG. 4. The c-myc IRES: sequence alignment with the FGF-2 IRES and Zuker secondary structure prediction. A, a sequence alignment of the human c-myc P2 leader (nt 1-366) and part of the FGF-2 leader (nt 189 -487) containing c-myc and FGF-2 IRESs, respectively, was obtained with Geneworks software. Homologous nucleotides are boxed. D1, D2, and D3 correspond to the c-myc RNA domains of the secondary structure shown in B. E1 and E2 correspond to the two predicted IRES structural motifs of FGF-2 RNA proposed in the RNA structure studies of Le and Maizel (28). The elements a, aЈ, b, bЈ, c, cЈ, d, and dЈ form stems in each E1 and E2 motif, as reported by these authors. The 5Ј-end of the ⌬2 mutant is indicated with an arrow at position 269 of the c-myc sequence. B, the secondary structure of the c-myc P2 RNA leader (366 nt) was predicted with the Zuker folding program (35). ⌬G ϭ Ϫ122 kcal/mol. The P2 RNA 5Ј-end and the CUG initiation codon are indicated, as well as the position of the ⌬2 deletion. The three structural domains are shown (D1, D2, and D3). D1 alone has a ⌬G of Ϫ78 kcal/mol. cascade effect that is likely to have drastic consequences on the control of cell growth. Finally, since this IRES is remarkably efficient, it is also particularly interesting in a biotechnological point of view: it can serve for the design of polycistronic vectors expressing several proteins from the same mRNA.