Functional Characterization of the Internal Ribosome Entry Site of eIF4G mRNA
1998; Elsevier BV; Volume: 273; Issue: 9 Linguagem: Inglês
10.1074/jbc.273.9.5006
ISSN1083-351X
AutoresWeiniu Gan, Michael La Celle, Robert E. Rhoads,
Tópico(s)RNA regulation and disease
ResumoThe eIF4 group initiation factors are required for cap-dependent translation initiation. Infection of mammalian cells by picornaviruses results in proteolytic cleavage of one of these factors, eIF4G, which severely restricts cap-dependent initiation but permits cap-independent initiation to proceed from an internal ribosome entry site (IRES) in picornaviral RNAs. The first 357 nucleotides (nt) of the 5′-untranslated region of eIF4G mRNA also contains an IRES. Using bicistronic constructs for expression in K562 cells, we have now shown that progressive deletions of the 5′-untranslated region can have either stimulatory or inhibitory effects. Furthermore, a 101-nt segment exhibits full IRES activity, and an 81-nt segment exhibits detectable IRES activity. A polypyrimidine tract (PPT) at the 3′ terminus is essential for internal initiation, a property which is characteristic of picornaviral IRESs but not the other host cellular IRESs studied to date. IRES activity does not require sequences beyond 357 nt. Out-of-frame AUGs have no effect on IRES-driven luciferase expression when introduced upstream of the PPT but markedly decrease expression when introduced at sites between the PPT and the authentic initiation codon at nt 369. These results suggest that the ribosomal subunit enters at or near the PPT and then scans downstream for the initiation codon. The eIF4 group initiation factors are required for cap-dependent translation initiation. Infection of mammalian cells by picornaviruses results in proteolytic cleavage of one of these factors, eIF4G, which severely restricts cap-dependent initiation but permits cap-independent initiation to proceed from an internal ribosome entry site (IRES) in picornaviral RNAs. The first 357 nucleotides (nt) of the 5′-untranslated region of eIF4G mRNA also contains an IRES. Using bicistronic constructs for expression in K562 cells, we have now shown that progressive deletions of the 5′-untranslated region can have either stimulatory or inhibitory effects. Furthermore, a 101-nt segment exhibits full IRES activity, and an 81-nt segment exhibits detectable IRES activity. A polypyrimidine tract (PPT) at the 3′ terminus is essential for internal initiation, a property which is characteristic of picornaviral IRESs but not the other host cellular IRESs studied to date. IRES activity does not require sequences beyond 357 nt. Out-of-frame AUGs have no effect on IRES-driven luciferase expression when introduced upstream of the PPT but markedly decrease expression when introduced at sites between the PPT and the authentic initiation codon at nt 369. These results suggest that the ribosomal subunit enters at or near the PPT and then scans downstream for the initiation codon. Initiation of nearly all eukaryotic mRNAs proceeds by a cap-dependent mechanism whereby the AUG nearest the 5′-end serves as the initiation codon (1Kozak M. Annu. Rev. Cell Biol. 1992; 8: 197-225Google Scholar). Yet other modes of initiation codon selection are used in special cases, e.g. leaky scanning, termination-reinitiation, ribosome shunting, and internal initiation (2Jackson R.J. Hershey J.W.B. Mathews M.B. Sonenberg N. Translational Control. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY1996: 71-112Google Scholar). In the latter case, ribosomes are directed to internal AUGs by an internal ribosome entry site (IRES). 1The abbreviations used are: IRES, internal ribosome entry site; 5′-UTR, 5′-untranslated region; BiP, immunoglobulin heavy chain binding protein; nt, nucleotide(s); CAT, chloramphenicol acetyltransferase; eIF, eukaryotic initiation factor; LUC, luciferase; nt, nucleotide; PCR, polymerase chain reaction; PPT, polypyrimidine tract.1The abbreviations used are: IRES, internal ribosome entry site; 5′-UTR, 5′-untranslated region; BiP, immunoglobulin heavy chain binding protein; nt, nucleotide(s); CAT, chloramphenicol acetyltransferase; eIF, eukaryotic initiation factor; LUC, luciferase; nt, nucleotide; PCR, polymerase chain reaction; PPT, polypyrimidine tract. Internal initiation has been demonstrated by both in vitro and in vivo experimentation for picornaviruses (3Ehrenfeld E. Hershey J.W.B. Mathews M.B. Sonenberg N. Translational Control. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY1996: 549-573Google Scholar), certain other viruses (4Tsukiyama-Kohara K. Iizuka N. Kohara M. Nomoto A. J. Virol. 1992; 66: 1476-1483Google Scholar, 5Liu D.X. Inglis S.C. J. Virol. 1992; 66: 6143-6154Google Scholar, 6Berlioz C. Darlix J.L. J. Virol. 1995; 69: 2214-2222Google Scholar), and a growing number of cellular mRNAs (7Macejak D.G. Sarnow P. Nature. 1991; 353: 90-94Google Scholar, 8Oh S.-K. Scott M.P. Sarnow P. Genes Dev. 1992; 6: 1643-1653Google Scholar, 9Vagner S. Gensac M.C. Maret A. Bayard F. Amalric F. Prats H. Prats A.C. Mol. Cell. Biol. 1995; 15: 35-44Google Scholar, 10Gan W. Rhoads R.E. J. Biol. Chem. 1996; 271: 623-626Google Scholar, 11Bernstein J. Sella O. Le S.-Y. Elroy-Stein O. J. Biol. Chem. 1997; 272: 9356-9362Google Scholar, 12Ye X. Fong P. Iizuka N. Choate D. Cavener D.R. Mol. Cell. Biol. 1997; 17: 1714-1721Google Scholar). Picornaviral IRESs have been the most characterized to date (3Ehrenfeld E. Hershey J.W.B. Mathews M.B. Sonenberg N. Translational Control. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY1996: 549-573Google Scholar). The minimal size of picornaviral IRESs is ∼450 nt, with further deletion into this sequence dramatically decreasing IRES activity. These IRESs have been divided into three groups: (i) entero- and rhinoviruses, (ii) cardio- and aphthoviruses, and (iii) hepatoviruses. Within each group, there is strong conservation of IRES secondary structure, somewhat less conservation of primary structure, but little conservation between groups apart from the existence of a polypyrimidine tract (PPT) located ∼25 nt from the 3′-end of the IRES and a 3′-terminal AUG. The precise sequence of the PPT is important for IRES function in entero- and rhinoviruses but not cardioviruses (2Jackson R.J. Hershey J.W.B. Mathews M.B. Sonenberg N. Translational Control. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY1996: 71-112Google Scholar). Several host cellular proteins bind to, and in some cases stimulate translation from, picornaviral IRESs, including La (13Meerovitch K. Svitkin Y.V. Lee H.S. Lejbkowicz F. Kenan D.J. Chan E.K.L. Agol V.I. Keene J.D. Sonenberg N. J. Virol. 1993; 67: 3798-3807Google Scholar, 14Svitkin Y.V. Meerovitch K. Lee H.S. Dholakia J.N. Kenan D.J. Agol V.I. Sonenberg N. J. Virol. 1994; 68: 1544-1550Google Scholar), the polypyrimidine tract-binding protein (15Hellen C.U.T. Witherell G.W. Schmid M. Shin S.H. Pestova T.V. Gil A. Wimmer E. Proc. Natl. Acad. Sci. U. S. A. 1993; 90: 7642-7646Google Scholar, 16Borovjagin A. Pestova T. Shatsky I. FEBS Lett. 1994; 351: 299-302Google Scholar, 17Kaminski A. Hunt S.L. Patton J.G. Jackson R.J. RNA. 1995; 1: 924-938Google Scholar) and eIF4G (18Pestova T.V. Shatsky I.N. Hellen C.U. Mol. Cell. Biol. 1996; 16: 6870-6878Google Scholar). The mechanisms of initiation codon selection in picornaviral IRESs are likewise divided into three models. For cardioviruses such as encephalomyocarditis virus, and perhaps hepatoviruses as well, the 40 S ribosomal subunit is thought to enter at the 3′-terminal AUG of the IRES and use it as the translation initiation codon (the "landing" model). For entero- and rhinoviruses, the ribosomal subunit appears to enter at the same AUG but then scans for the next downstream AUG (the "scanning" model). The aphthoviruses, e.g. foot and mouth disease virus, may combine landing and scanning, as translation initiates at both the 3′-terminal AUG of the IRES and also at the next downstream AUG. Another determinant of IRES utilization is the host protein synthesis machinery. Under normal conditions, most mRNAs utilize the cap-dependent translation pathway, but in cells infected with entero-, rhino-, and aphthoviruses, this pathway is shut down and viral mRNAs use instead a cap-independent mechanism (3Ehrenfeld E. Hershey J.W.B. Mathews M.B. Sonenberg N. Translational Control. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY1996: 549-573Google Scholar). The switch from cap-dependent to cap-independent initiation is mediated by the proteolytic cleavage of eIF4G (19Etchison D. Milburn S.C. Edery I. Sonenberg N. Hershey J.W.B. J. Biol. Chem. 1982; 257: 14806-14810Google Scholar, 20Etchison D. Fout S. J. Virol. 1985; 54: 634-638Google Scholar, 21Lloyd R.E. Grubman M.J. Ehrenfeld E. J. Virol. 1988; 62: 4216-4223Google Scholar). eIF4G functions as a linking protein which joins, by virtue of its binding sites for eIF4E, eIF3, eIF4A, and poly(A)-binding protein, the various factors involved in mRNA recruitment to the 40 S ribosomal subunit (22Lamphear B.J. Kirchweger R. Skern T. Rhoads R.E. J. Biol. Chem. 1995; 270: 21975-21983Google Scholar, 23Mader S. Lee H. Pause A. Sonenberg N. Mol. Cell. Biol. 1995; 15: 4990-4997Google Scholar, 24Tarun S.Z. Sachs A.B. EMBO J. 1996; 15: 7168-7177Google Scholar). The action of the 2A proteases of entero- and rhinoviruses (25Liebig H.-D. Ziegler E. Yan R. Hartmuth K. Klump H. Kowalski H. Blaas D. Sommergruber W. Frasel L. Lamphear B. Rhoads R.E. Kuechler E. Skern T. Biochemistry. 1993; 32: 7581-7588Google Scholar, 26Lamphear B.J. Yan R. Yang F. Waters D. Liebig H.-D. Klump H. Kuechler E. Skern T. Rhoads R.E. J. Biol. Chem. 1993; 268: 19200-19203Google Scholar) or the L protease of foot and mouth disease virus (27Kirchweger R. Ziegler E. Lamphear B.J. Waters D. Liebig H.-D. Sommergruber W. Sobrino F. Hohenadl C. Blaas D. Rhoads R.E. Skern T. J. Virol. 1994; 68: 5677-5684Google Scholar) releases the N-terminal portion of eIF4G, bound to the cap-binding protein eIF4E, from the initiation complex but leaves the C-terminal portion of eIF4G bound to eIF3 and eIF4A in the initiation complex (22Lamphear B.J. Kirchweger R. Skern T. Rhoads R.E. J. Biol. Chem. 1995; 270: 21975-21983Google Scholar). Such modified initiation complexes can apparently participate in internal initiation but not cap-dependent initiation, although the mechanism is not clear (18Pestova T.V. Shatsky I.N. Hellen C.U. Mol. Cell. Biol. 1996; 16: 6870-6878Google Scholar, 28Ohlmann T. Rau M. Pain V.M. Morley S.J. EMBO J. 1996; 15: 1371-1382Google Scholar, 29Borman A.M. Kirchweger R. Ziegler E. Rhoads R.E. Skern T. Kean K.M. RNA. 1997; 3: 186-196Google Scholar). In comparison with picornaviral IRESs, relatively little is known about cellular (non-viral) IRESs. The 5′-UTRs of BiP (7Macejak D.G. Sarnow P. Nature. 1991; 353: 90-94Google Scholar),Ultrabithorax (12Ye X. Fong P. Iizuka N. Choate D. Cavener D.R. Mol. Cell. Biol. 1997; 17: 1714-1721Google Scholar), and Antennapedia (8Oh S.-K. Scott M.P. Sarnow P. Genes Dev. 1992; 6: 1643-1653Google Scholar) mRNAs are, respectively, 220, 968, and 252 nt, but the locations and sizes of the IRESs within them are not known. In the case of the 318-nt 5′-UTR of fibroblast growth factor 2 mRNA, the IRES resides in a 165-nt segment (9Vagner S. Gensac M.C. Maret A. Bayard F. Amalric F. Prats H. Prats A.C. Mol. Cell. Biol. 1995; 15: 35-44Google Scholar). In the 1022-nt 5′-UTR of platelet-derived growth factor 2 mRNA, the IRES is present in a 395-nt segment (11Bernstein J. Sella O. Le S.-Y. Elroy-Stein O. J. Biol. Chem. 1997; 272: 9356-9362Google Scholar). None of these cellular mRNAs contains a picornavirus-like PPT nor is there significant homology among them or with picornaviral IRESs. The 5′-UTR of human eIF4G mRNA (30Yan R. Rychlik W. Etchison D. Rhoads R.E. J. Biol. Chem. 1992; 267: 23226-23231Google Scholar) is unusually long (368 nt), compared with typical 5′-UTRs of cellular mRNAs (31Kozak M. Nucleic Acids Res. 1987; 15: 8125-8148Google Scholar), and contains four upstream open reading frames, suggesting that its translation would be extremely inefficient if the scanning mechanism were used. Previously we demonstrated, using reporter constructs transfected into K562 cells, that the first 357 nt of the 5′-UTR of eIF4G mRNA has IRES activity (10Gan W. Rhoads R.E. J. Biol. Chem. 1996; 271: 623-626Google Scholar). In the present study we have further characterized the IRES of eIF4G with respect to size, sequence requirements, and mechanism of initiation codon selection. Knowledge of this mechanism may shed light on how internal initiation occurs in other RNAs, both viral and cellular. Because of the unique role of eIF4G in protein synthesis initiation, this may also give insight into such questions as the mechanism by which intracellular levels of eIF4G are maintained and the balance between cap-dependent and cap-independent initiation. Luciferin, restriction endonucleases, the DNA Cycle Sequencing System (fmol ®), enzymes used for in vitro transcription, and S1 nuclease were purchased from Promega (Madison, WI). RNase T1 was obtained from Life Technologies, Inc. (Gaithersburg, MD). RNase V1 was purchased from Pharmacia Biotech Inc. (Piscataway, NJ). Acetyl-coenzyme A was obtained from Sigma. Radioisotopes were purchased from ICN (Costa Mesa, CA). Silica gel thin layer chromatography plates (LK5F, 150 Å) were obtained from Whatman (Clifton, NJ). Pfu polymerase was purchased from Stratagene (La Jolla, CA). Deletions in the DNA corresponding to nt 1–357 of human eIF4G mRNA (30Yan R. Rychlik W. Etchison D. Rhoads R.E. J. Biol. Chem. 1992; 267: 23226-23231Google Scholar) were made from both 5′- and 3′-ends by PCR using Pfu polymerase. The PCR products used to make the first three 5′-deletions contained HindIII sites at the 3′ but not 5′ termini. It was therefore necessary to incorporate a 5′-HindIII site by cloning into pBluescript KS (Stratagene, La Jolla, CA) at the EcoRV site, which is adjacent to aHindIII site, and then screening for an insert with the same orientation as the T3 promoter. The fragments were excised withHindIII and inserted into pGL2/CAT/LUC (10Gan W. Rhoads R.E. J. Biol. Chem. 1996; 271: 623-626Google Scholar) at the uniqueHindIII site between coding regions for CAT and luciferase. (Henceforth, plasmids names are abbreviated, e.g.pGL2/CAT/LUC is abbreviated C/L, pGL2/CAT/4G/LUC is abbreviated C/4G/L, etc.) PCR products for the rest of the plasmids shown in Fig. 1 A and the 3′-deletion plasmids shown in Fig. 2 Acontained HindIII sites at each end. These were digested with HindIII and inserted into the HindIII site of C/L. The plasmid C/4G+/L contained DNA corresponding to the full-length 5′-UTR and the first three codons of eIF4G mRNA (nt 1–377), fused upstream and in-frame with the luciferase gene. This plasmid was made by a three-step PCR strategy. In the first step, pHFC5 (30Yan R. Rychlik W. Etchison D. Rhoads R.E. J. Biol. Chem. 1992; 267: 23226-23231Google Scholar) was used as template. The product contained sequences corresponding to nt 1–377 of eIF4G mRNA followed by nt 1–17 of the luciferase coding region. In the second step, pGL2/LUC (10Gan W. Rhoads R.E. J. Biol. Chem. 1996; 271: 623-626Google Scholar) was used as template. The product contained sequences corresponding to nt 358–377 of eIF4G mRNA followed by nt 1–55 of the luciferase coding region. In the third step, gel-purified PCR products from the first two steps were mixed and used as templates for PCR. The product, which had a HindIII site at its 5′-end and a NarI site at its 3′-end, was digested with both enzymes and inserted into C/L at the same sites. Plasmids containing mutations in the 3′-region of the 5′-UTR (Figs. 3 A, 4 A, 5 A, and6 A) were made by the same strategy but with C/4G+/L as template. Plasmid C/4G+BΔ322–363/LA (Fig. 6 A) was made by PCR with C/4G+B/LA as template. The resulting product, which had a 5′HindIII site and 3′ BclI site, was digested with the same enzymes and inserted into C/4G+B/LA at the same sites. The plasmid used for in vitro transcription, pKS4GΔMCS, was made in the following way. The PCR product corresponding to nt 1–357 of human eIF4G mRNA (10Gan W. Rhoads R.E. J. Biol. Chem. 1996; 271: 623-626Google Scholar) was inserted into pKS (Stratagene, La Jolla, CA) at the HindIII site. The polylinker upstream of the insert was removed by digesting with XbaI and SacI and re-ligating the plasmid. The structures of plasmids were confirmed by restriction digestion and sequencing (32Sanger F. Nicklen S. Coulson A.R. Proc. Natl. Acad. Sci. U. S. A. 1977; 74: 5463-5467Google Scholar) using double-stranded DNA as template.Figure 23′-Deletion analysis of the eIF4G IRES. A, construction of vectors. DNA segments corresponding to nt 1–357 of human eIF4G mRNA or the indicated 3′-deletions were inserted between the CAT and luciferase coding regions of plasmid C/L.B, relative luciferase activity was measured as in Fig. 1 B.View Large Image Figure ViewerDownload (PPT)Figure 3The effect of sequence variations in the PPT on the IRES activity. A, construction of vectors. DNA segments corresponding to the entire 5′-UTR (nt 1–368) plus the first three codons (nt 369–377) of human eIF4G mRNA, or the indicated sequence variants in the PPT (nt 304–315), were inserted between the CAT and luciferase coding regions of plasmid C/L. B, relative luciferase activity was measured as in Fig. 1 B.View Large Image Figure ViewerDownload (PPT)Figure 4Determination of the translation initiation codon used for luciferase in the bicistronic mRNA. A, construction of vectors. DNA segments corresponding to nt 1–377 of human eIF4G mRNA or the indicated mutations (lowercase) of ATG to AAG codons were inserted between CAT and luciferase coding regions of plasmid C/L. The first ATG (underlined) in plasmid C/4G+/L corresponds to the putative initiation codon for eIF4G mRNA at nt 369. The second ATG corresponds to the original initiation codon of luciferase mRNA. B, relative luciferase activity was measured as in Fig. 1 B.View Large Image Figure ViewerDownload (PPT)Figure 5The effects of out-of-frame AUGs on luciferase expression initiated from the AUG at nt 369. A, construction of vectors. DNA segments corresponding to the entire 5′-UTR plus first three codons of human eIF4G mRNA, or the indicated mutations in this sequence, were inserted between CAT and luciferase coding regions of plasmid C/L in which the original initiation codon of luciferase was mutated to AAG. Mutations (lowercase) introduced out-of-frame AUG codons (underlined) in the transcribed RNA at nt 298, 319, or 340.PPT signifies the polypyrimidine tract. The grouping of nucleotide residues into triplets corresponds to the reading frame of luciferase. B, relative luciferase activity was measured as in Fig. 1 B.View Large Image Figure ViewerDownload (PPT)Figure 6The effect of a "spacer" between the PPT and initiation codon. A, construction of vectors. A DNA segment corresponding to the entire 5′-UTR plus first three codons of human eIF4G mRNA, but with an A → T mutation at nt 363 to produce a BclI site (B), was inserted between CAT and luciferase coding regions of plasmid C/L, in which the original initiation codon of luciferase was mutated to AAG, to make C/4G+B/LA. The same sequence lacking nt 322–363 was also inserted to make C/4G+BΔ322–363/LA. B, relative luciferase activity was measured as in Fig. 1 B.View Large Image Figure ViewerDownload (PPT) K562 cells were transfected by electroporation using a Gene Pulser from Bio-Rad (Hercules, CA) as described previously (10Gan W. Rhoads R.E. J. Biol. Chem. 1996; 271: 623-626Google Scholar). Electroporation was performed in triplicate for each plasmid and utilized 30 μg of DNA unless otherwise specified. Luciferase activity was measured (33Braster A.R. Tate J.E. Habener J.F. BioTechniques. 1989; 7: 1116-1122Google Scholar) using a Monolight 2010 Luminometer from Analytical Luminescence Laboratory (San Diego, CA). CAT activity (10Gan W. Rhoads R.E. J. Biol. Chem. 1996; 271: 623-626Google Scholar) was used as an internal control to normalize luciferase activity in each sample. Corrected luciferase activity values were averaged and standard deviations determined. Total RNA was isolated from transfected K562 cells (34Chomcynski P. Ausubel F.M. Brent R. Kingston R.E. Moore D.D. Seidman J.G. Smith J.A. Struhl K. Current Protocols in Molecular Biology. Greene Publishing Associates, Inc. and John Wiley and Sons, Inc., New York1991: 4.2.4-4.2.8Google Scholar) and poly(A)-containing mRNA was purified (35Kingston R.E. Sheen J. Ausubel F.M. Brent R. Kingston R.E. Moore D.D. Seidman J.G. Smith J.A. Struhl K. Current Protocols in Molecular Biology. Greene Publishing Associates, Inc. and John Wiley and Sons, Inc., New York1994: 9.6.2-9.6.9Google Scholar). Poly(A)-containing mRNA from each sample (∼1 μg) was used for electrophoresis and Northern blot analysis (36Selden R.F. Ausubel F.M. Brent R. Kingston R.E. Moore D.D. Seidman J.G. Smith J.A. Struhl K. Current Protocols in Molecular Biology. Greene Publishing Associates, Inc. and John Wiley and Sons, Inc., New York1994: 4.9.1-4.9.7Google Scholar). The hybridization probe was synthesized as described previously (10Gan W. Rhoads R.E. J. Biol. Chem. 1996; 271: 623-626Google Scholar) and consisted of RNA complementary to nt 47–547 of luciferase mRNA (where the initiating AUG is nt 1). A 5′-end labeled form of the eIF4G 5′-UTR (nt 1–357) was produced by in vitro transcription of HindIII-linearized pKS4GΔMCS in the presence of 10 μCi/μl [γ-32P]GTP (37Titus D.E. Second Ed. Promega Protocols and Applications Guide. Promega Corp., Madison, WI1991Google Scholar). The nucleotide concentrations used in 100-μl reactions were 0.5 mm for ATP, CTP, and UTP but 0.25 mm for GTP. The 32P-end labeled RNA was subjected to partial digestion with S1 and V1 nuclease as described previously (38Zhang Y. Dolph P.J. Schneider R.J. J. Biol. Chem. 1989; 264: 10679-10684Google Scholar, 39Evstafieva A.G. Ugarova T.Y. Chernov B.K. Shatsky I.N. Nucleic Acids Res. 1991; 19: 665-671Google Scholar). RNA markers were synthesized by limited alkaline hydrolysis and RNase T1 nuclease digestion of the labeled RNA. The initial characterization of IRES activity in eIF4G mRNA employed a portion of the 5′-UTR representing 357 nt of the total 368 nt (10Gan W. Rhoads R.E. J. Biol. Chem. 1996; 271: 623-626Google Scholar). However, it was not clear how much of this sequence was required for IRES activity. We therefore prepared a series of plasmids expressing forms of the 5′-UTR that were deleted from either 5′- or 3′-ends. To serve as a guide in choosing segments to be deleted, we examined the distribution of single- and double-stranded regions in the 5′-UTR. Subjecting the 357-nt sequence to a folding algorithm (M-fold, GCG Sequence Analysis Package) produced a series of structures, all having free energies of folding between −105 and −110 kcal/mol. Many of the predicted stems and loops were common among the structures. To test the validity of these various structures, a 5′-end labeled transcript representing nt 1–357 was subjected to partial enzymatic digestion with single- and double-strand-specific nucleases (S1 and V1; see "Experimental Procedures"). In the theoretical structure exhibiting the best agreement with experimental results, 75% of the digestion sites were in the predicted single- and double-stranded regions (data not shown). These results indicate that additional experimental information will be required to arrive at a precise secondary and tertiary structure of the 5′-UTR. Nonetheless, the regions of double- and single-stranded structure were used as a guide for making deletions in the 5′-UTR. Each successive 5′-deletion (Fig. 1 A) corresponded to the removal of an additional region of double-stranded RNA (based on nuclease sensitivity). The truncated DNA segments were inserted between CAT and luciferase coding regions to permit the use of CAT activity as an internal control for mRNA levels (Fig. 1 A). Surprisingly, deletion of the first 57 nt enhanced IRES activity by 80% rather than diminishing it (Fig. 1 B, C/4GΔ1–57/L versus C/4G/L). IRES activity decreased progressively from this value with further deletions, except for C/4GΔ1–256/L, which expressed the same level of luciferase activity as the wild type IRES. There were no consistent differences between IRESs containing upstream AUGs (Fig. 1 A, asterisks) and those that did not, in keeping with the earlier observation that mutation of these AUGs to AAGs had no effect on IRES activity (10Gan W. Rhoads R.E. J. Biol. Chem. 1996; 271: 623-626Google Scholar). An RNA with three-fourths of the 5′-UTR deleted was considerably reduced in IRES activity (C/4GΔ1–276/Lversus C/4G/L) but was still 55-fold higher than the bicistronic mRNA containing no IRES (C/4GΔ1–276/Lversus C/L). Northern blot analysis showed that these bicistronic mRNAs were intact (data not shown), indicating that the reduction in activity was not due to RNA cleavage. Previous studies also showed that monocistronic RNAs containing the luciferase coding region downstream of the eIF4G IRES, or variants of it, were intact (10Gan W. Rhoads R.E. J. Biol. Chem. 1996; 271: 623-626Google Scholar). It was possible that some of the changes in luciferase activities observed in Fig. 1 B were underestimated because the amount of expression of bicistronic mRNA exceeded the cell's capacity to translate it, i.e. the assay was not in the linear range. To test this, different amounts of plasmid DNA were used for transfection of K562 cells (Fig. 1 C). Luciferase activity increased in response to the amount of transfected DNA for the parent (wt) plasmid (C/4G/L), for a mutant exhibiting higher activity than the parent (C/4GΔ1–57/L), for a mutant exhibiting the same activity as the parent (C/4GΔ1–256/L), and for a mutant exhibiting lower activity (C/4GΔ1–230/L). These results indicate that the changes in activity observed in Fig. 1 B are not distorted due to saturation of some limiting component necessary for expression. The 3′-boundary of the IRES was similarly defined by deletion analysis (Fig. 2 A). Removal of 58 nt from the 3′-end completely abolished IRES activity (Fig. 2 B, C/4GΔ300–357/L versus C/4G/L). However, an RNA with only 38 nt deleted, transcribed from plasmid C/4GΔ322–357/L, restored wild-type levels of luciferase. This indicates that the sequence between nt 300 and 321 is critical for IRES activity. More than half of this sequence consists of a 12-nt PPT. The experiments described in Figs. 1 and 2 utilized vectors which produced an RNA containing nt 1–357 of the 5′-UTR, but the complete 5′-UTR is 368 nt (30Yan R. Rychlik W. Etchison D. Rhoads R.E. J. Biol. Chem. 1992; 267: 23226-23231Google Scholar). It was conceivable that the 11 nt from 358 to 368 played a role in IRES activity. In hepatitis C virus, for instance, the first 10 codons are required for internal initiation and hence are considered part of the IRES (40Reynolds J.E. Kaminski A. Kettinen H.J. Grace K. Clarke B.E. Carroll A.R. Rowlands D.J. Jackson R.J. EMBO J. 1995; 14: 6010-6020Google Scholar). To study the IRES activity of the complete 5′-UTR plus the initial coding region, we fused DNA corresponding to the full-length eIF4G 5′-UTR and the first three codons of eIF4G (nt 1–377) upstream of, and in-frame with, the initiating ATG of luciferase (Fig. 3 A, C/4G+/L). This construct generated the same level of luciferase expression as C/4G/L (nt 1–357) within experimental error (data not shown), indicating that the sequence responsible for IRES activity does not extend beyond nt 357. Variations of this extended construct (4G+) were used for subsequent studies. Since deletion of the region containing the PPT abolished IRES activity (Fig. 2), it was of interest to determine whether a PPT per se was needed or merely a spacer between the IRES and the initiation codon. We therefore substituted a polypurine stretch consisting of six A and six G residues for the 12-nt PPT (Fig. 3 A, C/4G+Pu/L). This polypurine tract completely abolished IRES activity (Fig. 3 B, C/4G+Pu/Lversus C/4G+/L), indicating the need for a PPT. To determine whether a specific sequence in the PPT was required, we inverted the PPT sequence (Fig. 3 A, C/4G+yP/L). This reduced IRES activity by 50%, but the modified 5′-UTR was still 50-fold more effective in directing luciferase expression than the negative control (Fig. 3 B, C/4G+yP/L versus C/L). The plasmid expressing this inverted PPT contains a central sequence of CTTTC that is the same as in the original plasmid (Fig. 3 A). To rule out the possibility that this precise sequence was important, we made a different construct in which the original six T residues and six C residues were rearranged into a TC repeat sequence (Fig. 3 A, C/4G+TC/L). Expression from C/4G+TC/L was further decreased but still 30-fold higher than expression from C/L (Fig. 3 B). These results suggest that a PPT is needed, but that the wild-type sequence is not critical for IRES activity. The extended construct C/4G+/L (Fig. 3 A) expresses an RNA containing two in-frame AUG codons immediately upstream of the luciferase coding region, either one of which could, in principle, initiate translation of luciferase. The first of these, at nt 369 of the eIF4G mRNA (30Yan R. Rychlik W. Etchison D. Rhoads R.E. J. Biol. Chem. 1992; 267: 23226-23231Google Scholar), is presumed to be the in vivoinitiation codon for eIF4G, although this has not heretofore been proven. This supposition is based on its location at the beginning of a 4191-nt open readin
Referência(s)