Translation initiation of the insulin-like growth factor I receptor mRNA is mediated by an internal ribosome entry site.

The insulin-like growth factor I receptor (IGF-IR) is a heterotetrameric receptor mediating the effects of insulin-like growth I and other growth factors. This receptor is encoded by an mRNA containing an unusually long, G-C-rich, and highly structured 5' untranslated region. Using bicistronic constructs, we demonstrated here that the 5' untranslated region of the IGF-IR allows translation initiation by internal ribosome entry and therefore constitutes an internal ribosome entry site. In vitro cross-linking revealed that this internal ribosome entry site binds a protein of 57 kDa. Immunoprecipitation of UV cross-linked proteins proved that this protein was the polypyrimidine tract-binding protein, a well known regulator of picornavirus mRNA translation. The efficiency of translation of the endogenous IGF-IR mRNA is not affected by rapamycin, which is a potent inhibitor of cap-dependent translation. This result provides evidence that the endogenous IGF-IR mRNA is translated, at least in part, through a cap-independent mechanism. This is the first report of a growth factor receptor containing sequence elements that allow translation initiation to occur by internal initiation. Because the IGF-IR has a pivotal function in the cell cycle, this mechanism of translation regulation could play a crucial role in the control of cell proliferation and differentiation.

The insulin-like growth factor I receptor (IGF-IR) is a heterotetrameric receptor mediating the effects of insulin-like growth I and other growth factors. This receptor is encoded by an mRNA containing an unusually long, G-C-rich, and highly structured 5 untranslated region. Using bicistronic constructs, we demonstrated here that the 5 untranslated region of the IGF-IR allows translation initiation by internal ribosome entry and therefore constitutes an internal ribosome entry site. In vitro cross-linking revealed that this internal ribosome entry site binds a protein of 57 kDa. Immunoprecipitation of UV cross-linked proteins proved that this protein was the polypyrimidine tract-binding protein, a well known regulator of picornavirus mRNA translation. The efficiency of translation of the endogenous IGF-IR mRNA is not affected by rapamycin, which is a potent inhibitor of cap-dependent translation. This result provides evidence that the endogenous IGF-IR mRNA is translated, at least in part, through a cap-independent mechanism. This is the first report of a growth factor receptor containing sequence elements that allow translation initiation to occur by internal initiation. Because the IGF-IR has a pivotal function in the cell cycle, this mechanism of translation regulation could play a crucial role in the control of cell proliferation and differentiation.
Insulin-like growth factor I (IGF-I) 1 is a ubiquitous peptide that has a fundamental role in both prenatal and postnatal development (1,2). The effects of IGF-I are mediated through the IGF-I receptor (IGF-IR),which resembles the insulin receptor in primary and tertiary structure (3). The IGF-IR plays a central role during the cell cycle, as demonstrated by the fact that overexpression of IGF-IR abrogates requirements for exogenous growth factors (4). In vascular smooth muscle cells (VSMCs),the IGF-IR mediates the mitogenic effects of various growth factors such as platelet-derived growth factor, endothelial growth factor, fibroblast growth factor ␤, thrombin, and angiotensin II, consistent with an important role in cardiovascular growth responses (5)(6)(7)(8). A variety of studies have shown that relatively small changes in IGF-IR expression have important proliferative or antiproliferative effects (9 -11). However, studies characterizing growth factor and tumor suppressor gene regulation of the IGF-IR have focused on transcriptional mechanisms (12)(13)(14)(15)(16).
Some features of the IGF-IR gene suggest that it could be regulated at the posttranscriptional level. Like housekeeping genes, which appear to be transitionally regulated, the IGF-IR promoter gene is devoid of TATA and CAAT elements and is also very GC-rich. Moreover, the human and rat 5Ј untranslated regions (UTRs) are unusually long, 1038 and 943 bp, respectively, and the rat 5Ј UTR has been reported to possess two additional AUGs upstream in frame with the initiator AUG (17).
In eukaryotic cells, different modes of translation initiation are used depending on the nature of the mRNA to be translated and on the physiological state of the cell. The most frequently used are the "scanning mechanism" and "internal initiation." In the scanning mechanism, the initiation of translation requires the formation of a "43 S complex," which binds to the 5Ј-m 7 G cap structure of the mRNA and scans along the 5Ј UTR up to the initiator AUG. Then, the 60 S subunit attaches to this complex and translation is initiated (18). However, the presence of highly structured 5Ј UTR will cause the repression of this cap-dependent mechanism, and many mRNAs with structured 5Ј UTRs are poorly translated using this mechanism (19).
A cap-independent mechanism called internal initiation was first demonstrated in picornaviruses, which lack a 5Ј-m 7 G cap and have long structured 5Ј UTRs in their RNA. The presence of an internal ribosome entry site (IRES) has been shown in different picornaviruses, such as encephalomyocarditis virus (EMCV), human rhinoviruses, and hepatitis A virus (see Ref. 20 for review). This mechanism requires secondary structures that allow ribosomes to bind directly next to the initiator AUG and permit translation to start without previous scanning. Regulation of IRESs sometimes also involves cellular factors such as the polypyrimidine tract-binding protein (PTB), a member of the heterogeneous nuclear ribonucleoprotein family involved in nuclear splicing regulation (21).
In this study, we investigated whether IGF-IR mRNA translation could be driven by an internal initiation mechanism. We provide evidence for the first time that the mRNA encoding a growth factor receptor (IGF-IR) is translated by an internal initiation mechanism. Moreover, we show that the 5Ј UTR of the IGF-IR mRNA can be considered to possess an IRES that binds PTB. These findings have important implications for understanding mechanisms of control of IGF-IR expression and hence mechanisms of cellular growth control.

EXPERIMENTAL PROCEDURES
Cell Culture-Vascular smooth muscle cells were isolated from rat thoracic aorta as described previously (31). They were grown in DMEM supplemented with 10% FCS, 2 mM glutamine, 100 units/ml penicillin, and 100 g/ml streptomycin. Human MCF7 cells were grown in DMEM without phenol red, supplemented with 2 mM glutamine, 10% fetal bovine serum, and antibiotics. Rat2 (ATCC CRL-1764, embryo fetus fibroblasts) were grown in DMEM supplemented with 5% fetal bovine serum, 2 mM glutamine, and antibiotics. COS-1 monkey cells were grown in DMEM supplemented with 10% fetal bovine serum, 2 mM glutamine, and antibiotics.
Plasmid Constructions-The rat IGF-I receptor coding sequence and partial 5Ј UTR was previously cloned in our laboratory (10). The DNA fragment corresponding to the other part of the 5Ј UTR (ϩ1/ϩ640) was kindly provided by Dr. H. Werner (plasmid -2350/ϩ640). The plasmid pIGFIR-Trunc was constructed by cloning the BamHI/EcoRI fragment purified from pTrunc-smp8 (a kinase-deficient rat IGF-IR cDNA construct, kindly provided by Dr. J. Du, Emory University) into the corresponding sites of plasmid pBSK-(Stratagene).
The pBSK-430 plasmid was obtained by inserting the SacI/PvuII Klenow-filled fragment of plasmid p611 (10) into the SmaI site of the pBSK-vector. A BamHI fragment was then removed, and the resulting clone contained 430 bp (ϩ513/ϩ943) of the sequence coding for the 5Ј UTR plus 130 bp of the adjacent coding sequence. Plasmid pBSK-943 was obtained by inserting the Klenow-filled fragment AflIII of plasmid -2350/ϩ640 into the SmaI site of the 5Ј UTR (position ϩ513). This clone contains the sequence coding for the entire 5Ј UTR (943 bp) and 130 bp of the coding sequence.
Construction of the bicistronic vector: The Renilla luciferase (RLuc) reporter gene from plasmid pRL-TK (Promega) was extracted by NheI/ XbaI digestion and inserted into the Klenow-filled XhoI site of plasmid pCI (Promega) giving the pC-RL plasmid. Firefly Luciferase reporter gene (FLuc) was then subcloned from the pGL3 plasmid (Promega) into the SmaI site of pC-RL. In the resulting plasmid, called pBiC, the transcription of the first cistron corresponding to RLuc is under the control of the cytomegalovirus promoter.
The sequence coding for the 5Ј UTR of the IGF-IR mRNA was amplified by PCR using plasmid pBSK-943 as template and oligonucleotides 5Ј-AAAGAATTCAGTGTGTGGCGGCGGCGG-3Ј and 5Ј-AAAG-TCGACTCCTTTTATTTGGGACGA-3Ј, which contained an EcoRI site and a SalI site, respectively. The PCR product was then inserted in the pGEM-T easy vector (Promega), giving the pGEM-943 plasmid, and sequenced. The EcoRI/SalI fragment containing the sequence coding for the 5Ј UTR was then introduced in the EcoRI and SalI sites of pBiC vector giving the pBiC-943 vector. A stable hairpin (⌬G ϭ -46.2 kcal⅐mol Ϫ1 , calculated by using Zucker's RNA folding software) was created by introduction of a double-stranded autocomplementarity oligonucleotide 5Ј-GCTAGCTGAACTGGGAGTGGACACCTGCTAGC-3Ј and 5Ј-GCTAGCAGGTGTCCACTCCCAGTTCAGCTAGC-3Ј in the NheI site of plasmid pBiC-943 upstream of the Renilla luciferase sequence. The resulting plasmid was called pHBiC-943. Plasmid pBiC-⌬, containing 413 bp of the sequence coding for the 5Ј UTR, was constructed by extraction of the SmaI/SalI fragment of pGEM-943 and ligation in the Klenow-filled EcoRI/XbaI site of pBiC vector.
Dual Luciferase Assay-COS-1 cells were plated in 24-well plates and transfected with 1 g of bicistronic plasmid per well with Lipo-fectAMINE reagent. After 24 h, cells were washed and lysed according to the manufacturer's instructions. FLuc and RLuc activities were measured by using the dual-luciferase reporter assay system (Promega) in a EG&G Berthold luminometer (Bad Wildbad, Germany). For rapamycin treatment, 80% confluent COS-1 cells were serum-starved for 48 h and then stimulated with DMEM containing 10% FCS in the presence or absence of various concentrations of rapamycin.
UV Cross-linking Assay-UV cross-linking experiments were done according to Huez et al. (22). Cytoplasmic extracts from MCF-7 and rat VSMCs were prepared as follows: subconfluent cells were scraped in phosphate-buffered saline and pelleted by centrifugation. The cell pellet was resuspended in 500 l of lysis buffer (10 mM NaCl, 10 mM Tris-HCl, pH 7.4, 0.5 mM phenylmethylsulfonyl fluoride, 0.5 mM dithiothreitol) and frozen-thawed three times. The extract was centrifuged at 12,000 ϫ g for 10 min, and the supernatant (S10) was brought to 5% (v/v) glycerol, aliquoted, and stored at Ϫ80°C.
Probe A was synthesized and 32 P-labeled by T3 polymerase from plasmid pBSK-943 linearized at the EcoRI site. Probes B, C, and D were synthesized and 32 P-labeled by T3 polymerase from plasmid pBSK-430 linearized at the EcoRI, BbsI, and AccIII sites, respectively. Probe EMCV was transcribed and 32 P-labeled by T7 polymerase from plasmid pIRES (Promega) linearized at the XbaI site. 32 P-labeled RNA probes (10 6 cpm) were incubated with 6 g of S10 extract, preincubated with 2.5 g of yeast tRNA (Ambion) for 15 min at 30°C, in buffer containing 5 mM Hepes (pH 5.2), 25 mM KCl, 2 mM MgCl 2 , 3.8% glycerol, 0.2 mM dithiothreitol, and 1.5 mM ATP in a final volume of 10 l. Samples were then transferred to ice and irradiated with UV Stratalinker (Stratagene) at 10 cm from the bulbs and routinely with 400,000 J/cm 2 at 254 nm for 4 min. The samples were then treated with a mixture of 5 units of RNase One (Promega) and 3 g of RNase A at 37°C for 30 min and, when indicated, with 1 mg/ml of proteinase K (Ambion) at 37°C for 20 min. Laemmli buffer was added, and the samples were heated for 2 min at 95°C and loaded on a 12.5% polyacrylamide denaturing gel. The gel was fixed in 40% methanol-10% acetic acid for 30 min, dried, and submitted to autoradiography.
Immunoprepitation of UV Cross-linked Proteins-After UV crosslinking as described above, the samples (10 l) were diluted to 450 l with NETS buffer (50 mM Tris-HCl, pH 7.4, 5 mM EDTA, 1 mM dithiothreitol, 100 mM NaCl, and 0.05% Nonidet P-40) and mixed with either 5 l of monoclonal antibody directed against PTB (generously provided by Dr. E. Wimmer, Stony Brook, NY) or preimmune serum. After a 2-h incubation at 4°C, the immunocomplexes were immobilized on protein G-Sepharose (Amersham Pharmacia Biotech). The unbound materials were washed five times with the same buffer. Bound proteins were analyzed by SDS-polyacrylamide gel electrophoresis and identified by autoradiography.
Distribution of Specific mRNAs among Polyribosomal Subfractions-Polyribosomes were prepared from Rat2 cells. 80% confluent cells were serum-starved for 48 h and then stimulated with DMEM containing 10% FCS in the presence or absence of 50 ng/ml of rapamycin. Sucrose gradient fractionation of ribosomes was carried out on postmitochondrial supernatants as described previously (32). Cells were washed in cold phosphate-buffered saline in the presence of cycloheximide (100 g/ml), scraped off the plates, and collected in phosphate-buffered saline. After centrifugation, cells were resuspended in cold buffer containing 250 mM sucrose, 50 mM Tris-HCl, pH 7.4, 25 mM KCl, 5 mM MgCl 2 , 100 g/ml cycloheximide and lysed by addition of Nonidet P-40 to a final concentration of 0.07%. Nuclei and mitochondria were pelleted by two successive centrifugations at 750 and 12,000 ϫ g. The postmitochondrial supernatant was layered onto a 10 -50% linear sucrose gradient made in the same buffer containing 2 mM dithioerythritol, and centrifuged in a Beckman SW41Ti rotor at 40,000 rpm for 90 min at 4°C. Twenty fractions of identical volume were collected from the top of the gradient. Each fraction was treated with 100 mg/ml proteinase K and 1% SDS for 30 min before phenol-chloroform extraction and ethanol precipitation of the RNA. RNA from each fraction was resuspended in 10 mM Tris-HCl, pH 7.4, 1 mM EDTA, 1 unit of RNAsin. The quantitative distribution of the various mRNAs among the different fractions was then analyzed by slot-blot analysis. The specific IGF-IR and ␤-actin probes were the 1768-bp BamHI/NcoI fragment of plasmid pIGFIR-Trunc and the 1100-bp PstI fragment of plasmid pAL41-␤ (33), respectively. The probes were gel-purified and labeled by random primer DNA labeling. The blots were prehybridized for 1 h at 65°C and hybridized over night. After autoradiography each fraction was quantified using a PhosphorImager system (Molecular Dynamics).

RESULTS
Characteristics of the IGF-IR mRNA 5Ј UTR-The features of the IGF-IR mRNA 5Ј UTR prompted us to examine whether this region had the potential to adopt a highly structured secondary conformation. Predictions of the rat and human IGF-IR 5Ј UTR secondary structures were performed using Zucker's algorithm, and the structures were folded with the ESSA software ( Fig. 1) (34). Extended base pairing is apparent Internal Ribosome Entry Site in IGF-IR mRNA all along this 5Ј UTR, which has a ⌬G of -465.7 kcal⅐mol Ϫ1 and several hairpin structures. Remarkably, a long polypyrimidine (poly-U) stem-loop located just upstream of the initiator AUG (between position ϩ850 and ϩ950) was identified. Sequence comparison of the rat and human 5Ј UTR showed the same stem-loop at nearly the same location. Analysis of this stemloop sequence revealed a typical binding site for PTB. Moreover, the poly-U sequence is embedded in a Y-shaped structure (ϩ780/ϩ960), which is an important element of picornavirus IRES (35). These features led us to investigate whether the rat IGF-IR mRNA can be translated by internal ribosome entry.
The 5Ј UTR of the IGF-IR mRNA Contains an IRES-To determine the presence of an IRES in the 5Ј UTR of the IGF-IR mRNA, we used a bicistronic vector strategy in which the first cistron is translated by a cap-dependent scanning mechanism and the second requires translation by internal entry (36). The basic bicistronic vector was constructed by cloning the RLuc gene as the first cistron under the control of the cytomegalovirus promotor and the FLuc gene as the second cistron ( Fig.  2A, pBiC). The sequence coding for the entire or the partial 5Ј UTR of the IGF-IR mRNA (pBiC-943 or pBiC-⌬), as well as that coding for the encephalomyocarditis virus (EMCV) IRES as a positive control (pBiCV) were introduced between the RLuc and FLuc genes. These plasmids were transfected into COS-1 cells and tested for the two luciferase enzyme activities. The ratio between the two activities was then calculated, and the results were normalized using the pBiC FLuc/RLuc as the reference. The FLuc/RLuc ratio was 8-fold higher in cells transfected with pBiCV in which the FLuc was under the control of the EMCV IRES, than in cells transfected with the pBiC control plasmid. In cells transfected with pBiC-943, the ratio was 18-fold higher than that of the control plasmid pBiC. We also tested whether the 5Ј UTR of the rat IGF-IR mRNA can drive internal initiation in human cells (MCF-7) and similar results were obtained (data not shown). In cells transfected with pBiC-⌬, which contains only a part of the sequence coding for the 5Ј UTR of the IGF-IR (region ϩ513/ϩ943), translation of the FLuc gene was reduced because the ratio was only 3-fold higher than that obtained with the pBiC vector. We also constructed the pBiC-⌬inv, which contains the last 430 nt (ϩ513/ ϩ943) of the 5Ј UTR, and 100 bp of the coding sequence cloned in the opposite orientation. This construct was used to prove that the IRES activity was specific for the 5Ј UTR sequence. RNA synthesized from this plasmid was clearly incapable of allowing translation of FLuc compared with the pBiC-⌬ plasmid.
To prove that the IRES activity was not due to translation initiation from the first cistron or by a mechanism of reinitiation, we introduced a hairpin (⌬G ϭ -46.2) just upstream of the RLuc gene (pHBiC-943). The relative activities measured for each enzyme are presented in Fig. 2B. Activity of RLuc encoded by the first cistron of plasmid pBiC-943 was higher than that of the FLuc coded by the second cistron. In contrast, when a hairpin was inserted upstream of the RLuc gene, activity of the RLuc enzyme was strongly decreased, as expected, whereas that of the second enzyme FLuc persisted and was even increased. These results showed that translation of the second cistron was totally independent of that of the first one. Thus, synthesis of the FLuc enzyme was initiated from the IRES contained in the rat 5Ј UTR IGF-IR.
Rapamycin Treatment Does Not Affect the Efficiency of Translation of the IGF-IR IRES in Vitro-To demonstrate further that the IGF-IR 5Ј UTR contained an IRES, we investigated its ability to drive translation in the presence of rapamycin which was known to specifically reduce cap-dependent translation both in vivo and in vitro (37). To this end bicistronic mRNAs containing either the EMCV IRES or the IGF-IR 5Ј UTR were expressed in COS-1 cells (Fig. 3). In the pBiCV construct the translation of RLuc was cap-dependent, whereas the translation of FLuc under the control of the EMCV IRES was cap-independent (Fig. 3A). Rapamycin caused a dose-dependent inhibition (with a maximum of 55% at 100 ng/ml) of the cap-dependent translation of RLuc, whereas the cap-independent translation was not significantly affected. Similar re- sults were obtained with the pBiC-943 construct, in which the translation of FLuc was under the control of the IGF-IR 5Ј UTR (Fig. 3B). A dose-dependent inhibition of RLuc was observed with a maximum of 35% at 100 ng/ml of rapamycin. Interestingly, the IGF-IR 5Ј UTR retained the ability to drive translation in a cap-independent manner as the level of FLuc activity did not significantly decrease. Thus, these results proved that the IGF-IR 5Ј UTR could initiate translation through a capindependent mechanism.
Distribution of the Endogenous IGF-R mRNA among Polyribosomes Is Not Affected by Rapamycin Treatment-The effect of rapamycin on the efficiency of translation of the endogenous IGF-IR mRNA was evaluated by analyzing the distribution of IGF-IR mRNA among polyribosomes before and after rapamycin treatment. For these experiments, a transformed fibroblast cell line (Rat2) was used because it contains sufficient IGF-IR mRNA for detection by slot blot hybridization. ␤-Actin mRNA was used as a control representing mRNA initiated by cap-dependent translation. Exponentially growing cells were treated for 3 h with or without rapamycin, and the distribution of IGF-IR and ␤-actin mRNAs was determined among polyribosomal subfractions (Fig. 4). Distribution of IGF-IR mRNA from control and rapamycin-treated cells among the different fractions was very similar (Fig. 4, top panel). 83 and 87% of total IGF-IR mRNA was associated with polyribosomes containing at least two ribosomes (fractions 5-20) in control and rapamycin-treated cells. Thus, rapamycin did not induced a significant change in the IGF-IR mRNA distribution. In contrast, there was less ␤-actin mRNA associated with polyribosomes (fractions 5-20) in cells treated with rapamycin than in control cells (Fig. 4, bottom panel, 70% versus 81%). An increase in the amount of ␤-actin mRNA in fractions containing less than two ribosomes (fractions 1-4) was observed. Each ␤-actin mRNA molecule can theoretically bind 10 ribosomes; thus, the 11% shift in ␤-actin mRNA distribution from fractions containing polysomes to fractions corresponding to monosomes (fractions 1-4) corresponds to a significant decrease of the efficiency of translation. In contrast, rapamycin treatment did not modify the efficiency of translation of the endogenous IGF-IR mRNA.
PTB Binds to the 5Ј UTR of the rat IGF-IR mRNA-Several RNA-binding proteins are known to interact with the IRES of the EMCV, poliovirus, and mammalian mRNAs. To determine whether the IGF-IR 5Ј UTR was able to bind particular proteins, we used rat VSMCs and performed UV cross-linking experiments. Specifically, we tested whether the part of the UTR containing the typical PTB binding site (see Fig. 1) was able to bind proteins present in VSMCs. The 32 P-labeled B RNA probe (Fig. 5A) corresponding to the last 430 nt of the 5Ј UTR, containing the poly-U tract and the first 123 nt of the adjacent coding sequence, was incubated with S10 VSMC extract and then UV-irradiated. Analysis of proteins cross-linked to the labeled RNA was carried out by SDS-polyacrylamide gel electrophoresis (Fig. 5B, lane 1). Two major cross-linked complexes of 60 and 35 kDa were detected, and both complexes disappeared after proteinase K treatment (Fig. 5B, lane 2), confirming the proteic nature of these signals. The specificity of the interaction was demonstrated by competition experiments.
In the presence of a 50-or 150-fold molar excess of unlabeled B probe (Fig. 5B, lanes 3 and 4), both signals were strongly and progressively decreased.
We then investigated the possibility that the 60-kDa protein could correspond to PTB. We used a 32 P-labeled RNA probe corresponding to the EMCV IRES as a positive control for PTB binding. The 60-kDa protein bound to the B probe comigrated with a factor bound to the EMCV IRES (Fig. 5B, lanes 5 and 6). Moreover, labeling of the 60-kDa protein was specifically and progressively abolished when a 150-or 300-fold molar excess of unlabeled EMCV IRES was used as competitor (Fig. 5B, lanes  7 and 8). By contrast, an unknown protein with an apparent molecular mass of 35 kDa was not linearly competed by this probe. Further, no competition was observed when an unlabeled nonspecific RNA probe was used (Fig. 5B, lanes 9 and 10). To localize, within the 5Ј UTR, the site of interaction with the 60-kDa protein, we carried out complementary experiments using the different RNA probes described in Fig. 4A. Probe C contained the poly-U tract but not the coding sequence. In probe D, the poly-U tract was removed. The 60-kDa protein was able to bind to both probes B and C (Fig. 5C, lanes 1 and 2). However, this protein did not interact at all with the probe D (lane 3). This indicated that the presence of the poly-U tract is absolutely necessary for binding of the 60-kDa protein. These results are consistent with the hypothesis that the 60-kDa protein could be PTB and with the predicted secondary structure of the region of the RNA recognized by this protein (Fig. 1). As the poly-U tract is also present in the 5Ј UTR of the human IGF-IR mRNA, UV cross-linking assays were performed with S10 extract of MCF-7 human cells. Again, two proteins of 60 and 35 kDa were revealed with probes A, which corresponds to the entire rat IGF-IR 5Ј UTR. These two complexes were formed with probes B and C (Fig. 5C, lanes 4 -6) whereas only that at 35 kDa was found for probe D (Fig. 5C, lane 7). These results revealed that the same proteins might be present in rat and human extracts and that they were able to bind to the rat IGF-IR 5Ј UTR.
To demonstrate the identity of the 60-kDa protein, UV crosslinked complexes obtained with MCF-7 and VSMC extracts and probe B were immunoprecipitated using a monoclonal anti-PTB antibody (Fig. 6A). The 60-kDa complex was immunoprecipitated with both the EMCV probe and the probe containing the last half part of the IGF-IR 5Ј UTR mRNA (probe B) (Fig.  6A, lanes 1-4). The 35-kDa protein was also detected after immunoprecipitation of UV cross-linked complexes from human MCF-7 S10 extracts incubated with probe B but was barely detectable in immunoprecipitates from VSMCs. The detection of this protein after immunoprecipitation might result either from interaction of this protein with PTB or from residual background. Immunoprecipitation performed with protein G alone and protein G plus preimmune serum (Fig. 6A, lanes 5  and 6) did not reveal these proteins. Finally, to verify that the interaction between probe B and PTB was direct, the immunoprecipitated complex (Fig. 6B, top panel) was then analyzed by Western blot, using the same anti-PTB antibody. This analysis (Fig. 6B, bottom panel) revealed a unique immunoreactive band in both cell types, which migrated at around 60 kDa. This result demonstrated that the PTB was indeed present in the UV cross-linked complex. DISCUSSION IGF-I receptor gene expression is controlled by a number of physiological conditions and cell metabolic states. In this study, we demonstrate that initiation of translation of this growth factor receptor mRNA can occur by internal ribosome entry, providing a potential additional level of control of its expression. To our knowledge, this is the first IRES to be characterized in the mRNA of a growth factor receptor.
Because the rat IGF-IR 5Ј UTR is long and displayed a high G-C content, the question about the ability of ribosomes to linearly scan along the IGF-IR 5Ј UTR mRNA according to the conventional scanning model was raised. Internal ribosome entry was first described for picornavirus (38),and different elements involved in the internal ribosome entry process are now well known. The rat IGR-IR 5Ј UTR mRNA does not present the conventional motifs of picornavirus IRES, such as a GNRA or CAAA loop (39). Nonetheless, our prediction of the secondary structure revealed that this 5Ј UTR mRNA presents a conserved polypyrimidine-rich stem-loop structure composed of two different loops. The first contains two repeated UUUC motifs described to be the consensus binding site for PTB (40) and a stretch of 15 uridines. Interestingly the second loop contains a third UUUC sequence localized 21 nt upstream of the AUG, as is the case of a number of picornavirus IRESs (41).
Moreover, this stem-loop structure is embedded in a Y-shaped conformation described for other cellular mRNAs, such as vascular endothelial growth factor and c-myc (22,42). The predicted secondary structure of the 5Ј UTR of the human IGF-IR mRNA revealed some differences between rat and human 5Ј UTRs. First, the human 5Ј UTR is longer than the rat and contains a GNRA loop localized 625 nt upstream of the initiator AUG. Moreover, the two upstream in-frame AUGs described for the rat IGF-IR mRNA are not found in the human IGF-IR mRNA.
We initially investigated the ability of the rat IGF-IR 5Ј UTR to direct internal translation initiation. For this purpose, we used the bicistronic strategy (36). The translation efficiency of the second cistron was clearly increased when the IGF-IR 5Ј UTR mRNA was cloned in the intercistronic region, and the efficiency of the IGF-IR IRES was higher than that of the control EMCV IRES. Furthermore, we tested whether this IRES was also functional after inhibition of cellular cap-dependent translation by rapamycin treatment. These experiments were first conducted in vitro and demonstrated that the IGF-IR 5Ј UTR conserved the ability to initiate translation when the cap-dependent translation machinery is blocked. This result was also confirmed in vivo as the endogenous IGF-IR mRNA was still engaged into polyribosomes under rapamycin treatment.
The limits of the IGF-IR IRES remain to be defined, although an important element is contained in the first 400 nt of the 5Ј UTR, and it will be of primary interest to determine whether this IRES is composed of short modules as described recently for the Gtx homeodomain protein mRNA (43).
UV cross-linking experiments performed with rat vascular smooth muscle cell and human MCF-7 cells revealed that at least two proteins can bind to the rat IGF-IR 5Ј UTR mRNA and that the patterns using rat and human cell extracts were very similar. Moreover, these proteins were detected both with the EMCV IRES and the rat IGF-IR 5Ј UTR mRNA. The apparent 35-kDa protein could correspond to the previously described p36 protein known to interact with EMCV IRES (44). We demonstrated by immunoprecipitation experiments that the 60-kDa protein corresponds to the PTB splicing factor. Based on the predicted secondary structure, the potential bind-ing site should be localized within the stem-loop structure described above. This hypothesis was confirmed by the fact that the removal of this structure completely abolished the binding of the PTB in human and rat cellular extracts. PTB, also known as hnRNP 1, is a predominantly nuclear protein, present in lower abundance in the cytoplasm of many cells, which binds to pre-mRNA elements (45). The role of this factor both in the splicing mechanism and in IRES-directed translation initiation remains unclear. Nonetheless, it has been shown recently that this protein enhances the translational activity of different viral IRESs (46).
The biological significance of IRESs has only recently been appreciated. There is evidence that posttranscriptional regulation involving IRES-dependent translation plays a critical role in expression of several proto-oncogenes and homeodomain protein genes. The proto-oncogene c-myc is a critical regulator of cell proliferation, differentiation, and apoptosis, and recently it was demonstrated that the c-myc IRES was active during apoptosis (47). Moreover, IRESs were found in mRNA that direct the translation of important mammalian regulating proteins, such as fibroblast growth factor-2, IGF-II, and plateletderived growth factor. A translational mechanism mediated by an IRES confers clear advantages to the mRNA bearing such an element. The vascular endothelial growth factor IRES is active during hypoxia when protein synthesis is inhibited (48), likely playing an important role in angiogenesis. The plateletderived growth factor IRES is more active during cell differentiation, in which protein synthesis rates are also reduced (26). Thus, IRES-dependent translational regulation allows translation under conditions that are not favorable for cap-dependent translation, such as apoptosis, mitosis, development, or viral FIG. 5. UV cross-linking of rat and human cellular factors to the IGF-IR 5 UTR mRNA and EMCV IRES. A, schematic representation of the different 32 P-labeled rat RNA probes, obtained from T3 in vitro transcription and corresponding to the complete or partial IGF-IR 5Ј UTR mRNA. B, S10 extracts from rat VSMCs were incubated with 10 6 cpm of probe B or EMCV probe (lane 5). Competition experiments were carried out by addition of unlabeled probe B at a molar excess of 50 -150-fold (lanes 3 and 4), unlabeled EMCV probe (lanes 7 and 9), and unlabeled nonspecific RNA (lanes 9 and 10) at a molar excess of 150 -300-fold. UV irradiation was performed as described under "Experimental Procedures," and samples were treated with a mixture of RNase One and RNase A before analysis by SDS-polyacrylamide gel electrophoresis. Lane 2 corresponds to sample treated with proteinase K. C, the different 32 P-labeled probes indicated were cross-linked with proteins extracted from S10 VSMCs and MCF-7 cells. Proteins are revealed by autoradiography of the dried gels. Molecular mass markers, in kDa, are shown in B and C.  1-4) was used to immunoprecipitate human (lanes 1 and 2) and rat (lanes 3 and 4) proteins cross-linked to the EMCV IRES (lanes 1 and 3) or the rat IGF-IR 5Ј UTR fragment ϩ513/ϩ1066 (see probe B in Fig. 4A) (lanes 2 and 4). Lane 5 and 6 represent UV cross-linked proteins incubated with protein G and protein G plus preimmune serum, respectively. B, the proteins cross-linked to the IGF-IR 5Ј UTR (probe B) were immunoprecipitated using an anti-PTB antibody. After SDS-polyacrylamide gel electrophoresis, the proteins were transferred to polyvinylidene difluoride membrane and revealed either by autoradiography of the membrane (top panel) or by Western blot using the same antibody (bottom panel).

infection.
Accumulating evidence has shown that the IGF-IR is at a convergence point in the control of cell growth and that a functional IGFI/IGF-IR autocrine loop is required for the mitogenic effects of a variety of growth factors, including plateletderived growth factor (5), endothelial growth factor (6), angiotensin II, and thrombin (8,49). Deletion of the IGF-IR is lethal, and overexpression of the receptor has transforming potential. Thus, IGF-IR gene expression must be tightly regulated to guarantee correct protein levels at the appropriate time and location. In quiescent cells, cap-dependent translation is strongly inhibited, and proteins of the cap-dependent machinery are poorly expressed. We can postulate that the internal ribosome entry process allow the translation of key proteins such as growth factor receptors to permit the cell to engage in mitosis. In this respect, it is interesting to note that a recent study reported that in human kidney cells, the level of IGF-IR remained unchanged after peroxide injury (50). This is consistent with the mechanism of cap-independent translation initiation of the IGF-IR mRNA, as has been hypothesized recently for the insulin receptor based on microarray analysis (51). Additionally, we have recently shown that native low density lipoprotein up-regulates IGF-IR on VSMCs without affecting IGF-IR mRNA levels, suggesting that translational control mechanisms could be important (52).
In summary, we have shown that the 5Ј UTR of the IGF-IR mRNA contains an IRES that binds PTB and allows cap-independent translation. Because of the critical role played by the IGF-IR in cell cycle control and development, our findings have important implications for understanding mechanisms of normal and abnormal cell proliferation and mechanisms for developmental control of gene expression.