Initiation Factor eIF2-independent Mode of c-Src mRNA Translation Occurs via an Internal Ribosome Entry Site*

Overexpression and activation of the c-Src protein have been linked to the development of a wide variety of cancers. The molecular mechanism(s) of c-Src overexpression in cancer cells is not clear. We report here an internal ribosome entry site (IRES) in the c-Src mRNA that is constituted by both 5′-noncoding and -coding regions. The inhibition of cap-dependent translation by m7GDP in the cell-free translation system or induction of endoplasmic reticulum stress in hepatoma-derived cells resulted in stimulation of the c-Src IRES activities. Sucrose density gradient analyses revealed formation of a stable binary complex between the c-Src IRES and purified HeLa 40 S ribosomal subunit in the absence of initiation factors. We further demonstrate eIF2-independent assembly of 80 S initiation complex on the c-Src IRES. These features of the c-Src IRES appear to be reminiscent of that of hepatitis C virus-like IRESs and translation initiation in prokaryotes. Transfection studies and genetic analysis revealed that the c-Src IRES permitted initiation at the authentic AUG351, which is also used for conventional translation initiation of the c-Src mRNA. Our studies unveiled a novel regulatory mechanism of c-Src synthesis mediated by an IRES element, which exhibits enhanced activity during cellular stress and is likely to cause c-Src overexpression during oncogenesis and metastasis.

assembly of productive initiation complexes on these IRESs is energy-efficient and can ignore the need of several critical translation initiation factors (eIFs 4F/4A/4B/1/1A) that are controlled by a variety of external and internal cellular regulators (15,19). This "40 S-binding signature" has not been reported for the known cellular IRESs.
Translational dysregulation of a whole host of mRNAs has been observed in many diseases, including cancer. This is caused by a breakdown of the translational control mechanism, aberrant levels of translation factors, and/or undesirable mutations in these factors (2,9,20). The level of c-Src protein, a prominent member of the nonreceptor tyrosine kinase family, is known to increase in a variety of tumors (21)(22)(23)(24)(25). However, it is not known whether the enhanced expression is regulated by transcriptional and/or post-transcriptional mechanisms. The c-Src protein promotes cell differentiation, tumor growth, metastasis, and angiogenesis (24 -27). It activates STAT3, which transcriptionally regulates expression of Bcl-X L , c-Myc, and cyclin D1 leading to activation of anti-apoptotic and cell cycle progression pathways (28,29). It has been shown that the activated c-Src-focal adhesion kinase complex promotes cell mobility, cell cycle progression, and cell survival. The c-Src activities are also important for promoting vascular endothelial growth factor-associated tumor angiogenesis and protease-associated metastasis (30).
Post-translational modifications such as phosphorylation and myristoylation are key regulators of the c-Src activities. Although nonmyristoylated c-Src readily moves to the nucleus at G 0 and at the G 1 /S phase, myristoylation at the N terminus is required for its membrane attachment and transforming activities (31,32). The intramolecular interaction between its Src homology 2 domain and phosphorylated Tyr-530 residue (numbered according to GenBank TM accession number NM_198291) at the C terminus induces closed or inactive conformation in the c-Src molecule. Under basal conditions in vivo, 90 -95% of Src is found in this state (33). The dephosphorylation of Tyr-530 by protein-tyrosine phosphatase and autophosphorylation of Tyr-419 by its kinase domain causes induction of an enzymatically active, open conformation (25,27).
The Src gene is composed of 14 exons (34,35). Transcription of this gene in hepatoma cells from two different promoters and alternative splicing results in mature transcripts that differ only in the extreme 5Ј ends but encode the same 60-kDa c-Src protein (Fig. 1A). The c-Src type 1A mRNA contains a 350-nt-long 5Ј-noncoding region (5ЈNCR) or a 5Ј-untranslated region with multiple AUGs located at nt positions 147, 179, and 351 (Fig.  1B). However, only AUG351 is used to initiate translation of the c-Src open reading frame (ORF). The type 1␣ c-Src transcript contains a 451-nt-long 5ЈNCR and differs with type 1A only in the first exon (type 1A or type 1␣) (35). The second and third exons (1B and 1C) are shared in both transcripts. Hepatoma cells have been shown to express both transcripts (35). The regulatory role(s) of these noncoding/untranslated elements during translation of c-Src mRNAs and their role(s) in c-Src overexpression are not known. Here, we report the presence of an IRES element in the c-Src mRNAs that is constituted by both noncoding and coding regions. This IRES possesses many unique attributes not found in the known cellular IRESs. The data presented here show that the c-Src IRES features are reminiscent of that of HCV-like IRESs and/or translation initiation in prokaryotes. Thus, our findings have opened up new avenues for investigation on the translation control of c-Src synthesis and its effects on tumorigenesis.

EXPERIMENTAL PROCEDURES
Plasmid Constructs-Total RNA was isolated from the hepatoma-derived Huh7 cell line using Qiagen RNeasy kit. Nucleotides 1-383 of the type 1A c-Src mRNA were amplified from the total RNA using the high fidelity reverse transcription-PCR kit (Promega) and a pair of primers (sense, 5Ј-CATA-GCAAGCTTGCCGGAGCGGCCAGGCCGCCGTCTGC-3Ј; antisense, 5Ј-GCGCCGCTGTCATGAGGCATCCTTGGGC-TTGCTCTTGTTGCTACC-3Ј, restriction enzymes sites are underlined). The amplified DNA contains wild type full-length 5ЈNCR and a 33-nt coding region that encodes the first 11 amino acids of the c-Src protein. The PCR products were digested with HindIII and BspHI for ligation into a vector backbone. The backbone was created by digestion of a previously described plasmid pT7C1-DC29 -332 (36) with HindIII and NcoI, and the PCR products were ligated at this site after restriction digestion. The resulting plasmid p5ЈSrc-FLuc contains a T7 promoter at the 5Ј end of the c-Src sequence that is followed by ORF of firefly luciferase (FLuc) and ends with an oligo(A) tail. The in vitro transcription of the HpaI-digested plasmid by T7 RNA polymerase produces an RNA containing the entire 5ЈNCR followed by coding sequences representing the N-terminal 11 amino acids (MGSNKSKPKDA) of c-Src fused with the FLuc ORF ending with a poly(A) tail. The plasmid p5ЈSrc⌬C-FLuc was similarly constructed as described for p5ЈSrc-FLuc except that it lacks 33 nt of the c-Src coding sequence. To construct this plasmid, the c-Src PCR-amplified DNA described above was digested with HindIII and NcoI and cloned at the same site in the vector backbone. In both cases, the Kozak sequence context was maintained at the translation site. The sequence of each construct was verified by restriction enzyme digestion and the Big Dye DNA sequencing method (Applied Biosystems).
The plasmid 5ЈPV-FLuc contains wild type, full-length poliovirus (PV, Mahoney strain) 5ЈNCR (nt 1-742) cloned in-frame with FLuc ORF followed by the poly(A) tail. The mutant plasmid 5ЈPV(⌬286 -605)-FLuc was constructed by restriction enzyme digestion of the 5ЈPV-FLuc plasmid with BlpI and BsaBI followed by filling with T4 DNA polymerase and religation with Quick T4 DNA ligase (New England Biolabs). The resulting plasmid contains a deletion of 319 bp (nt 286 -605) in the PV 5ЈNCR.

c-Src IRES-mediated Translation Initiation
In Vitro Transcription-The plasmids p5ЈSrc-FLuc, p5ЈSrc⌬1-FLuc, p5ЈSrc⌬2-FLuc, p5ЈSrc⌬3-FLuc, and p5ЈSrc⌬C-FLuc were linearized with HpaI and transcribed with T7 RNA polymerase to produce luciferase reporter RNAs. The uncapped RNAs were prepared with RiboMax large scale RNA production kit (Promega). The capped RNAs were synthesized in the presence of the ARCA cap analogue using mMessage mMachine ultra kit (Ambion) in accordance with the manufacturer's instructions. The transcribed RNAs were passed through G-25 column and purified by extraction with phenol: chloroform:isoamyl alcohol followed by water-saturated cold ether. Following precipitation and washing with 70% ethanol, the final preparations were dissolved in RNase-free water and checked for integrity of RNAs by formaldehyde-agarose gel electrophoresis. Concentrations of RNA were determined spectrophotometrically. For preparation of the c-Src NCR probe, 5ЈSrc-FLuc was linearized with XbaI and transcribed with T7 RNA polymerase in the presence of [␣-32 P]CTP. The 5ЈPV(⌬286 -605)-FLuc was similarly digested with XbaI and transcribed for preparation of an inactive IRES control probe. The probes were purified using Qiagen RNeasy purification method. The plasmid pRL-HCV1b encodes upstream Renilla luciferase followed by the HCV IRES (nt 1-357 of the HCV genotype 1b) linked to the second reporter FLuc (37). The plasmid was linearized with HindIII and transcribed in the presence of cap analogue using T7 RNA polymerase.
HeLa translation lysates (S10) and lysates containing initiation factors (IFs) were prepared according to the protocol described by Barton et al. (38). The rabbit reticulocyte lysate (RRL) nuclease-treated was purchased from Promega. The total lysates from cultured Huh7 cells were prepared using M-PER kit (Pierce) as instructed.
In Vitro Translation of RNAs-The in vitro transcribed wild type 5ЈSrc-FLuc and its mutant derivatives were translated in HeLa cell-free system. The standard HeLa cell-free translation mixtures contain 20 l of S10, 10 l of initiation factors, 5 l of 10ϫ buffer (155 mM HEPES-KOH, pH 7.4, 600 mM potassium acetate, 10 mM ATP, and 2.5 mM GTP, 300 mM phosphocreatine, 4 mg/ml creatine phosphokinase), 20 units of RNasin, 5-10 g of RNA template in a 40-l final volume. One microliter of [ 35 S]methionine was added for radiolabeling of the newly synthesized proteins. The translation mixtures were incubated for 1-2 h at 30°C, and the FLuc activity was assayed using 2-l aliquots. For detection of protein bands, the samples were subjected to SDS-PAGE followed by autoradiography. For detection of the 5ЈSrc-RFLuc RNA expression, a Dual-Luciferase assay protocol (Promega) was employed, and Renilla and firefly luciferase activities were simultaneously assayed. Varying amounts of m 7 GDP or m 7 GTP were added in the standard HeLa translation mixtures for inhibition of cap-dependent translation. Unmethylated GDP or GTP served as negative control. Translation of the RNA in RRL was carried out as described in the supplier's protocol (Promega).
RNA Stability Assay-Equal amounts of 32 P-labeled wild type or mutant reporter RNAs were incubated in standard HeLa translation reactions, and total RNAs were extracted from each sample using the RNeasy kit (Qiagen). The recovered RNAs were subjected to formaldehyde-agarose gel electrophoresis followed by autoradiography of the dried gel. The bands of 18 S or 28 S rRNA in each lane were measured by ethidium bromide staining before drying the gels. During transfection experiments, 32 P-labeled reporter RNAs (1-2 ϫ 10 6 dpm) were transfected into Huh7 cells using the standard transfection method, and total RNAs were isolated. The radioactive full-length RNAs were detected by autoradiography.
Transfection of RNA into Cells-Huh7 cells were transfected with in vitro transcribed RNAs using Lipofectamine 2000 reagent (Invitrogen) according to the manufacturer's protocol. After 3,5,8,24, and 48 h post-transfection, the cells were harvested and resuspended with lysis buffer (100 mM potassium phosphate pH 6.8, 1 mM dithiothreitol, 0.5% Igepal). The samples were then subjected to two freeze-thaw cycles, and supernatants were assayed for Luc activities. For fluorescence microscopy, the cells were grown on coverslips (Fisher) followed by RNA transfection. The cells were fixed with 4% formaldehyde 48 h post-transfection, permeabilized, and stained with anti-firefly luciferase monoclonal antibody (Bionovus). A fluorescein isothiocyanate-labeled secondary conjugate was used to visualize the FLuc distribution in the transfected cells.
Isolation of 40 S Ribosomal Subunit-HeLa S10 lysate was prepared from HeLa S3 cells grown in a spinner flask as described by Barton et al. (38). The ribosomes were pelleted from S10 lysate by centrifugation in Ti70.1 rotor (Beckman, 45,000 rpm) for 3 h at 4°C, and the pellet was resuspended in buffer A (20 mM Tris-HCl, pH 7.5, 2 mM dithiothreitol, 50 mM KCl, and 4 mM MgCl 2 ) at a concentration of 150 units/ml measured at A 260 as described by Pisarev et al. (39). Puromycin (1 mM) and KCl (0.5 M) were added, stirred in an ice bath for 10 min, followed by incubation for 10 min at 37°C. The mixture was then loaded onto a 10 -30% sucrose density gradient and centrifuged for 16 h at 4°C in a Beckman SW28 rotor (22,000 rpm). The peak fractions containing 40 S ribosomes (as determined by the presence of only 18 S rRNA) were pooled and concentrated in Ultracell-100 K (Millipore). The final preparation was dialyzed in buffer C (20 mM Tris-HCl, pH 7.5, 2 mM dithiothreitol, 100 mM KCl, 2 mM MgCl 2 , 0.25 M sucrose), aliquoted, and stored at Ϫ80°C (39).
Sucrose Density Gradient Analysis-Capped or uncapped 32 P-labeled mRNAs were incubated in standard HeLa translation lysates that were treated with 1 mM GMP-PNP for 5 min in an ice bath. The mixtures were then incubated for 15 min at 30°C, layered onto a 10 -30% sucrose gradient in buffer K (20 mM Tris-HCl, pH 8.0, 100 mM potassium acetate, 5 mM magnesium acetate, and 2 mM dithiothreitol), and centrifuged for 3 h at 45,000 rpm and at 4°C in an SW-51 rotor. Fractions (250 l) were collected from the bottom of the gradient and analyzed by scintillation counter. Total RNAs from peak fractions were iso-c-Src IRES-mediated Translation Initiation FEBRUARY 19, 2010 • VOLUME 285 • NUMBER 8 JOURNAL OF BIOLOGICAL CHEMISTRY 5715 lated using Qiagen RNeasy mini column for analysis of the RNA contents in the fractions.
For 40 S-IRES binary interaction assay, the 32 P-labeled c-Src IRES or a nonspecific RNA probe derived from 5ЈPV(⌬286 -605) as scrambled IRES was mixed with purified HeLa 40 S subunit in buffer K containing 20 units of RNasin in a final volume of 40 l and incubated for 15 min at 30°C. The reaction mixtures were analyzed by sucrose density gradient method as described above. A sample of RNA probe without 40 S was used to specify the position of the free probes during centrifugation.

RESULTS
Characteristics of the Computer-generated c-Src RNA Structures-Our preliminary investigations on c-Src translation (not shown) and several published reports (21,22,26,28) indicate that c-Src level is enhanced in many cell types during stress conditions that impair cap-dependent translation. This observation prompted us to examine cap-independent translation of the c-Src mRNAs. The 5ЈNCRs in the c-Src transcripts have been shown to be relatively longer than those in most cellular mRNAs (35). Sequence analysis revealed multiple pyrimidine-rich motifs and two cryptic AUGs with short ORFs at positions 147 and 179 in the 350nt-long type 1A 5ЈNCR (Fig. 1). Only the AUG located at position 351 is known to serve as the initiator codon in this mRNA. Our nucleotide blast search using the BLASTN 2.2.20ϩ program (40) revealed that the exon 1C and the 11-amino acid N-terminal coding sequences are highly conserved in humans, chimpanzees, and rhesus monkeys (94 -100% identity), whereas mouse sequences in this region are 76 -80% identical to the reference human c-Src mRNA sequences (GenBank TM accession numbers NM_005417.3 and NM_198291).
Using the M-Fold program (41), we examined a number of predicted secondary structures representing three segments (nt 1-353, 1-383, and 1-410) of the type 1A c-Src mRNA. A representative structure (dG ϭ Ϫ135) for nt 1-383, which was found to be similar to the structure obtained for nt 1-410 segment, is shown in Fig. 2. This Y-shaped secondary structure appears to contain three domains designated as domains I-III. We further observed that a large portion of domains I and II were conserved in the structures predicted for all three c-Src segments. In addition, a high degree of conservation in the apical loops contributed by AACAAGA (nt 360 -366), GUGCCA (SL II, nt 289 -294), and UAUUUC (SL III, nt 255-260) motifs was also noticed in the predicted structures. The structures in domain III, however, showed less conservation among various structures generated for the three c-Src segments. A 14-nt pyrimidine (Py)-rich motif (nt 330 -344) was located 6 nt upstream of the initiator AUG, which conforms with Py tracts found in many viral IRESs. These characteristics and the Y-shaped architectural features are considered as important elements of many viral and cellular IRESs (42). The predicted structure for the c-Src nt 1-353 that represents the entire 5ЈNCR and AUG codon lacked a significant portion of domain II structure (structure not shown).
c-Src mRNA Motif Supports Cap-independent Translation of Reporter RNAs-FLuc-based reporter mRNAs were engineered to test if the c-Src 5ЈNCR supports cap-independent translation (Fig. 3A). Because a 33-nt sequence motif downstream of the initiator AUG in the c-Src mRNA forms a conserved stem-loop structure at the translation initiation site (Fig. 2), we included the region with the 5ЈNCR for engineering a parent reporter 5ЈSrc-FLuc RNA. The RNA contains c-Src nt 1-383 (fulllength 5ЈNCR plus 33 nt of the coding region) that is fused in-frame with luciferase ORF and ends with the poly(A) tail (Fig. 3A). In vitro transcribed capped and uncapped RNAs were translated in rabbit reticulocyte cell-free lysate (45,46) in the presence of [ 35 S]methionine, and the synthesized products were visualized by autoradiography. As expected, the capped 5ЈSrc-FLuc RNA was translated to produce active luciferase protein (Fig. 3B, lane 3). Interestingly, the uncapped 5ЈSrc-FLuc RNA was also translated but with higher efficiency (Fig. 3B, lane 2) than its capped counterpart (lane 3). During the assay, we used a reporter RNA [5ЈPV(⌬286 -605)-FLuc] that contains PV 5ЈNCR but lacks a major portion of its IRES element (from nt 286 to 605). Thus, the noncoding region (420 nt) in the mutant PV construct represents nt 1-285 followed by nt 606 -

c-Src IRES-mediated Translation Initiation
746 of the PV 5ЈNCR. This PV region, although known to form stable stem-loop structures, provide stability to the RNA, and bind a number of cellular factors (62), failed to support translation of FLuc in RRL (Fig. 3B, lane 4). However, synthesis of FLuc was successfully achieved when the same RNA had 5Јm7G cap structure (Fig. 3B, lane 5). These results clearly demonstrate that the c-Src nt 1-383 allow cap-independent translation of downstream ORF, and its activity is enhanced when cap function is absent.
The translatability of the mRNA constructs was further tested in HeLa cell-free translation lysates that have been widely used for investigating IRES-mediated translation initiation (38). An uncapped reporter FLuc RNA that contains wild type, full-length PV 5ЈNCR (5ЈPV-FLuc, Fig. 3C, lane 4) and 5ЈSrc-FLuc (lane 2) were efficiently translated, whereas the mutant 5ЈPV(⌬286 -605)-FLuc again failed to support synthesis of luciferase (lane 3). Next, we examined the c-Src 5ЈNCR-promoted translation in the context of a dicistronic mRNA (5ЈSrc-RFLuc). The RNA is similar to the monocistronic 5ЈSrc-FLuc RNA except that it contains RLuc ORF and stop codon upstream of the c-Src sequence (Fig. 4A). The in vitro transcribed capped RNA was transfected into hepatoma Huh7 cells for 3 h, and the lysates were subjected to Dual-Luciferase assay. As shown in Fig. 4B, both ORFs were translated in these cells. The capped dicistronic RL-HCV1b and RL-vector RNAs were used as positive and negative controls, respectively, during the transfection. The RL-HCV-1b is similar to the 5ЈSrc-RFLuc (Fig. 4A) except that the translation of downstream FLuc ORF is controlled by the HCV IRES instead of the c-Src IRES. In RL-vector, the c-Src IRES between upstream RLuc and downstream FLuc ORF is deleted. Both of the RNAs produced results as expected (supplemental Fig. S2). In a parallel experiment, total RNAs isolated from the dicistronic 5ЈSrc-RFLuc RNA-transfected cells were subjected to Northern blot analysis using a 32 P-labeled oligonucleotide probe that detects the 3Ј end of FLuc ORF. The result showed that the dicistronic RNA was intact in the transfected cells (Fig. 4C, lane 3). The migration of isolated RNA was similar to that of in vitro transcribed dicistronic RNA  FEBRUARY 19, 2010 • VOLUME 285 • NUMBER 8 (Fig. 4C, lane 2) and was not cleaved into the monocistronic form (lane 1).

c-Src IRES-mediated Translation Initiation
The 5ЈSrc-FLuc RNA encodes a chimeric firefly luciferase with amino acids 1-11 (MGSNKSKPKDA) of the c-Src protein at its N terminus (supplemental Fig. S1A). This c-Src motif has been shown to play an important role in membrane localization and translocation of the protein into the nucleus (31,47). Transfection of an uncapped 5ЈSrc-FLuc RNA into Huh7 cells resulted in the synthesis of luciferase protein that was primarily localized in the nucleus and perinuclear membranes (supple-mental Fig. S1B). This observation was in sharp contrast to the diffused cytoplasmic localization of luciferase that was encoded by 5ЈHCV-FLuc RNA in which translation of FLuc occurs under the control of HCV IRES. The resulting FLuc lacks the c-Src amino acid 1-11 motif. These results suggest that the luciferase synthesized from 5ЈSrc-FLuc RNA contains the c-Src protein motif, which is possible when translation is initiated at the authentic AUG351 (also see Fig. 5 for translation of 5ЈSrc⌬1-FLuc).
Identification of an IRES Element in the c-Src mRNA-We introduced several deletion mutations in the c-Src motif of the 5ЈSrc-FLuc RNA to determine its putative IRES function. The mutant 5ЈSrc⌬C-FLuc is similar to the wild type 5ЈSrc-FLuc RNA except that it lacks the c-Src coding sequence (nt 354 -383; Similarly, the 5ЈSrc⌬2-FLuc mutant RNA that contains a large deletion (nt 95-348) upstream of initiator AUG also failed to support efficient synthesis of FLuc (Fig. 5B, lane 5). Unlike these mutants, a 5ЈSrc⌬3-FLuc RNA that maintains nt 1-47 and 216 -383 of the c-Src mRNA showed cap-independent translation of FLuc (Fig. 5B, lane 6) and was comparable with that of wild type 5ЈSrc-FLuc RNA. The predicted structure of this mutant c-Src motif (data not shown) by the M-Fold program showed significant similarities in the domains I and II of the wild type structure (Fig. 2).
To determine the stability of the reporter constructs, we translated 32 P-labeled uncapped mutants and wild type RNAs in HeLa cell-free lysates as described above. Total RNAs from each reaction were isolated by the RNeasy column method, and the input probes were visualized by autoradiography. As shown in Fig. 5C (upper panel), the amounts of full-length mutant RNAs recovered (lanes 2-5) were similar or better than that of

c-Src IRES-mediated Translation Initiation
the wild type 5ЈSrc-FLuc (lane 1). The quantity of 18 S rRNA (internal control) in each lane had minor variations (Fig. 5C,  lower panel). This observation suggests that the mutant RNAs were present in the lysates, yet these RNAs were unable to support translation of FLuc due to absence of essential elements in the c-Src sequence motif. The different band intensities observed for the RNA probes may likely be due to minor differences in stability and/or loss during the purification process. A similar observation was also made during transfection of three mutant RNAs (⌬2, ⌬3, and ⌬C) into Huh7 cells. Although fulllength mutant RNA probes were purified from the transfected cells (Fig. 5F), only ⌬3 mutant showed efficient synthesis of reporter FLuc (Fig. 5E). These results further suggest that a functional IRES that is represented by the c-Src motif in ⌬3 mutant (Fig. 2, Domains I and I,) was capable of directing translation by a cap-independent mechanism in cells as well as in the cell-free lysates. Unlike known cellular IRESs, this IRES requires a coding region for its optimal function.
We carried out kinetic analysis of translation promoted by the wild type and mutant c-Src motifs in HeLa translation lysates. The time course experiment presented in Fig. 5D shows that translation of the 5ЈSrc-FLuc RNA exponentially increased  FEBRUARY 19, 2010 • VOLUME 285 • NUMBER 8

JOURNAL OF BIOLOGICAL CHEMISTRY 5719
with time, whereas the mutant 5ЈSrc⌬2-FLuc was translated inefficiently at all time points. During our investigations (Fig. 5, D and E), a minor translation was consistently observed for ⌬2 mutant RNA. It is possible that sequence motifs that form domain I and a translation site in this RNA might have played a role in the residual translation. These motifs are, however, absent in ⌬1 and ⌬C mutants that are completely incompetent for translation initiation.

Assembly of 80 S Initiation Complex on the c-Src IRES in HeLa Cell-free Translation Lysates Is Not Affected by
Inhibition of eIF2-Sucrose density gradient analysis was carried out to study the assembly of 80 S translation initiation complex on the c-Src IRES. The 32 P-labeled uncapped 5ЈSrc-FLuc RNA or capped FLuc was incubated for 15 min in HeLa cell-free translation lysates containing 1 mM GMP-PNP, a nonhydrolyzable GTP analogue, which causes accumulation of 48 S complex and inhibition of 80 S formation on a capped mRNA (18,43). The reaction mixtures were analyzed for ribosome assembly on the mRNAs by a 10 -30% sucrose density gradient centrifugation method. The initiation complexes in the gradient fractions were determined by incorporation of the input RNA probe into the ribosomal complexes (Fig. 6A). A single peak (Fig. 6A, Peak I, solid line with squares) of ribosomal complex containing the input RNA probe, 18 S, and 28 S rRNA was obtained for the 5ЈSrc-FLuc RNA (Fig. 6B, lane 4). The result clearly established assembly of the 80 S complex on the c-Src 5ЈNCR motif, which was not inhibited by 1 mM GMP-PNP (Fig. 6, A and B). In a similar translation reaction, we further reduced GTP and ATP concentrations by omitting the 10ϫ reaction buffer from the translation mixture. This omission caused 10ϫ increase in the GMP-PNP to GTP ratio during translation. Interestingly, the 80 S assembly on the 5ЈSrc-FLuc mRNA probe occurred (Fig. 6A, Peak I, broken line with squares; Fig. 6B, lane 5) similar to the standard reaction conditions described above. On the contrary, when a capped FLuc mRNA probe that lacks the c-Src motif was used in a standard translation reaction supplemented with 1 mM GMP-PNP, only 48 S complex was obtained as expected (Fig. 6A, Peak II, solid line with triangles). This conclusion was based on the observation that peak II lacks 28 S RNA, showed lower sedimentation than peak I, and contains only input RNA probe and 18 S rRNA (Fig. 6B, lane 7). In addition, the 48 S complex assembly was considerably reduced when the cap structure was absent in the FLuc RNA probe (Fig.  6, A, Peak II, broken line with triangles, and B, lane 6). The GMP-PNP is known to inhibit GTPase function that is required for the assembly of the 80 S complex on a capped mRNA. Thus, the observed cap-independent 80 S assembly on the c-Src IRES is most likely to be independent of eIF2 function. In a similar

c-Src IRES-mediated Translation Initiation
experiment, the assembly of ribosomal complex on a mutant poliovirus 5ЈNCR containing FLuc (PV⌬286 -605-FLuc) was compared with the 5ЈSrc-FLuc RNA in the presence of 1 mM GMP-PNP (Fig. 6, C and D). The uncapped 5ЈSrc-FLuc RNA showed assembly of 80 S in a reproducible manner (Fig. 6, C,  Peak I, squares with solid line, and D, lane 3). However, the uncapped PV⌬286 -605-FLuc RNA showed a peak at lower sucrose density (Fig. 6, C, Peak II, triangles with broken line; D, lane 4), and the complexes were spread over a wide range of sucrose density, most likely due to varying compositions of ribonucleoprotein complexes formed with the mutant PV RNA.
The assembly of 80 S on the c-Src IRES was further strengthened by our finding that purified HeLa 40 S ribosomal subunit directly interacts with the c-Src IRES (Fig. 7A). The purified 40 S subunit was mixed with uncapped 32 P-labeled IRES fragment of the 5ЈSrc-FLuc, and bound complex was separated from the free probe by sucrose density gradient centrifugation. Characterization of the peak fractions revealed the presence of 18 S rRNA and the probe in the same peak (Fig.  7A, inset, lane 1), although the free IRES probe showed a peak at lower sucrose density. A nonspecific RNA probe of similar length and nucleotide contents (PV⌬286 -605 5ЈNCR) failed to form the 40 S-RNA binary complex in this assay (Fig. 7B  inset, lane 2). These results together provide evidence that the c-Src mRNA contains an IRES element that directly interacts with the 40 S ribosomal subunit and is capable of assembling the 80 S complex during conditions when eIF-2 function is significantly compromised.

c-Src IRES-mediated Translation Is Enhanced when Cap-dependent Translation Is Inhibited-The eIF4E
protein is a key translation initiation factor that binds the 5Ј cap structure of an mRNA and initiates assembly of the 48 S preinitiation complex (1). It has been shown that m 7 GDP inhibits eIF4E function by occupying its cap-binding site. Therefore, cap-dependent translation is efficiently inhibited by the m 7 GDP cap analogue (44). The cap-dependent translation of a FLuc RNA, which contains the 5Јcap and 3Ј poly(A) tail at the respective ends of the luciferase ORF but lacks an IRES (5ЈCap-FLuc), was inhibited by m 7 GDP in a dose-dependent manner in RRL (Fig. 8A). In contrast, the HCV IRES-controlled translation of a reporter FLuc ORF (5ЈHCV-FLuc) was stimulated until a threshold concentration (10 g) of m 7 GDP was reached. Above this concentration, both cap-dependent as well as HCV IRESdependent translations were inhibited. Interestingly, translation of the 5ЈSrc⌬3-FLuc RNA (genetic organization shown in Fig. 5A) was considerably enhanced in the presence of m 7 GDP as observed for the HCV IRES-mediated translation initiation. Similar observations were also made for the uncapped wild type 5ЈSrc-FLuc RNA (not shown).
Next, we examined translation of a capped dicistronic RNA (5ЈSrc-RFLuc; Fig. 4A) in RRL in the presence of increasing concentrations of m 7 GDP. The wild type c-Src IRES-controlled translation of FLuc was initially enhanced in the presence of m 7 GDP (5-10 g) as observed for its monocistronic counterpart, whereas the cap-dependent translation of the upstream RLuc ORF continued to decline with increasing concentrations of m 7 GDP (Fig. 8B).  FEBRUARY 19, 2010 • VOLUME 285 • NUMBER 8

JOURNAL OF BIOLOGICAL CHEMISTRY 5721
Although the requirement of inhibitor concentration to inhibit overall translation was a little higher than that of the monocistronic RNAs, the stimulation pattern of the c-Src IRES in the presence of m7GDP was similar for both monoand dicistronic templates. During several control experiments (data not shown), we observed that the m 7 GTP cap analogue also causes stimulation of the c-Src IRES in HeLa and RRL cell-free translation systems, whereas the unmethylated nucleotides (GTP or GDP) had no effects within the concentration range used in our studies. These results together with those described above (Figs. 3, 5, and 6) established the presence of a functional IRES in the c-Src mRNA that can be activated when cap function is absent or significantly inhibited and/or eIF-2 activity is inadequate in the translation system.

c-Src IRES-controlled Translation Is Modestly Enhanced during Cellular Stress That Blocks Cap-dependent Translation-
Thapsigargin (TG) causes eIF-2 phosphorylation resulting in global translation inhibition and endoplasmic reticulum stress response due to inhibition of Ca 2ϩ -ATPase activities (48,49). We treated Huh7 cells with 1 M TG for different time points (0.5-6 h) and monitored the status of eIF2␣ by Western blot. We observed a considerable increase in the phosphorylation level of eIF2␣ within 30 min of TG treatment, which remained elevated for 6 h (Fig. 9A, lanes 2-6, upper panel). The total eIF2␣ level, on the other hand, was not affected by this treatment (Fig. 9A, lower panel). The enhanced eIF2␣ phosphorylation may be considered as an indicator for TG-induced cellular stress and reduction in global cap-dependent translation in the treated Huh7 cells. In the subsequent experiments, Huh7 cells were pretreated with 1 M TG for 3 h before transfection with the capped 5ЈSrc-RFLuc or RL-HCV1b RNAs while maintaining 1 M TG. The RL-HCV1b is similar to the 5ЈSrc-RFLuc except that the translation of the FLuc ORF in the RNA is driven by an HCV IRES. The FLuc and RLuc activities were assayed in the cytoplasmic fractions 3 h post-transfection. The translation of reporter luciferases in untreated transfected cells (control) was considered as 100% and compared with that of DMSO-or DMSO-TG-treated cells (Fig. 9B). Both the cap-dependent and IRES-dependent (HCV or c-Src) translation was not affected by DMSO treatment of the cells. In contrast, the cap-dependent translation of RLuc was dramatically reduced for both the RNAs due to DMSO-TG treatment of the cells. In these cells, however, the c-Src or HCV IRES-controlled translation of FLuc was moderately enhanced. We also determined the expression level of natural c-Src protein in these lysates. A modest increase in the total c-Src protein level was observed in DMSO-TG-treated cell lysates as compared with the untreated (control) or DMSO alone-treated lysates (Fig. 9C). The c-Src mRNA level remained unchanged in these cells (Fig. 9D).
Serum deprivation of cells also causes suppression of cap-dependent translation due to competitive inhibition of eIF4E binding to the 5Ј cap structure by poly(A)-specific ribonuclease (50) and phosphorylation of eIF2␣ (4). When Huh7 cells were subjected to serum starvation, the total c-Src protein level was moderately enhanced within 72 h (Fig. 9E, minus) as compared with the control (plus), although the c-Src mRNA level was not affected (Fig. 9F). These results together complement our findings that the c-Src level in Huh7 cells is modestly enhanced or maintained to steady levels under varying cellular stress conditions that are unfavorable for cap-dependent translation. This effect is likely to occur due to increase in the c-Src IRES activities.

DISCUSSION
An overwhelming majority of reports, including polysomeprofiling data, strongly advocate for IRES-dependent translation initiation of a subset of cellular mRNAs during cell division, apoptosis, cellular stress, and viral infections where cap-dependent translation initiation is compromised (reviewed in Ref. 15). Unlike known cellular IRESs, the c-Src IRES demonstrated here exhibits many unique attributes that are analogous to the characteristics of HCV-like IRESs. To identify an FIGURE 8. Stimulation of the c-Src IRES-controlled mRNA translation when eIF4E function is inhibited. A, capped RNAs as follows: 5ЈSrc-FLuc (triangle) and FLuc (5ЈCap-FLuc, circle) or uncapped 5ЈHCV-FLuc (square) RNAs were translated in triplicate in the presence of increasing amounts of m 7 GDP in RRL for 1 h, and one-tenth of each reaction mixture was assayed for FLuc activity. Average FLuc activity of three reactions is shown for each m7GDP concentration. The FLuc activity in samples without m 7 GDP was considered as 100% translation and compared with those containing the cap analogue (inhibitor). Similar translation reactions were carried out twice to confirm the results. B, translation of a capped Dual-Luciferase RNA construct (5ЈSrc-RFLuc) in RRL in the presence of increasing amounts of m 7 GDP as described above. Relative cap-dependent translations of RLuc and c-Src IRES-dependent FLuc synthesis are shown. Each translation mixture was carried out in triplicate. The results were confirmed by three independent experiments.

c-Src IRES-mediated Translation Initiation
IRES function in viral and cellular mRNAs, mono-and dicistronic RNA-expressing plasmids have been extensively used during transfection studies. This approach, however, has been a subject of criticism due to expression via cryptic promoters and faulty transcription and splicing of the reporter constructs (51). To avoid spurious results generated by this method, we have used only in vitro transcribed capped and uncapped reporter RNAs for cell-free translation assays, transfection studies, and sucrose density gradient analyses. The transcription reactions were digested with DNase I prior to purification and checked by agarose gel electrophoresis for the absence of DNA contamination in the final RNA preparations. Furthermore, the reporter RNA transcription is under the control of the T7 promoter, and transcription of the RNA from plasmid DNA contamination is not possible in any of the systems used here. We also demonstrated that the wild type and mutant c-Src motif containing reporter RNAs were intact during various translation assays. These measures permitted us to present reliable data for the identification of c-Src IRES.
We established here that the c-Src IRES-controlled translation can be stimulated similar to the HCV IRES when eIF4E function is blocked (Fig. 8). The initiation factor eIF4E has been shown to be a negative modulator of the IRES-mediated translation, and translation of IRES-containing RNAs is accelerated when eIF4E availability is reduced (63). This is likely attributed to a decrease in eIF4F complex formation that may be accompanied by an increased availability of eIF4G/eIF4A or eIF4A RNA helicase or other initiation factors. Based on these observations, we believe that a direct binding of m7GDP to eIF-4E in our assay may lead to increased availability of translation factors that are required for efficient activities of the HCV or c-Src IRESs.
Our in vitro studies that defined the presence of an IRES in the c-Src mRNA were further corroborated by the results of transfection of mono-and dicistronic reporter RNAs into hepatoma-derived cells and induction of cellular stress in the transfected cells. Uncapped reporter RNAs containing nt 1-383 of c-Src mRNA at their 5Ј ends were efficiently translated in two cell-free translation systems (RRL and HeLa lysates) and in Huh7 cells. Our genetic analysis shows that nt 200 -383 of the c-Src mRNA, which harbors initiator AUG (at nt 351), plays a pivotal role in promoting cap-independent translation. An extensive analysis of the secondary and/or possible higher order structures within this region is, however, needed for accurate understanding of its role in loading productive initiation complex. We found that the c-Src IRES promotes assembly of stable 80 S complexes in the absence of cap structure and in the presence of 1 mM GMP-PNP. Under similar conditions, however, only 48 S complex can be trapped on a capped reporter RNA lacking a 5ЈNCR or contains a scrambled IRES. Furthermore, similar to the HCV-like IRESs, a direct binding of purified HeLa 40 S with the c-Src nt 1-383 was detected in the absence of initiation factors. These evidence together strongly support the existence of a physiologically relevant IRES element at the 5Ј end of c-Src mRNA. The c-Src IRES appears to be functionally similar to the HCV IRES as both IRES elements directly interact with the purified 40 S subunit, require coding region for their functions, promote eIF2-independent assembly of 80 S complex (Figs. 3, 5 and 6) (see Refs. 19,43,52,53 for the HCV IRES function), and are stimulated when eIF4E or eIF2␣ function is impaired (Figs. 8  and 9). Therefore, our studies reported here present several unique attributes of a cellular IRES that have been demonstrated only for HCV-like IRESs. . c-Src IRES-controlled translation is not inhibited during cellular stress. A, phosphorylation of eIF2␣ by TG treatment. The cytoplasmic lysates (40 g) from Huh7 cells that were treated with 1 M TG for 0.5, 1, 2, 4, and 6 h (lanes 2-6) were subjected to Western blot analysis using anti-(phospho-eIF2␣(Ser-51)) antibody (Cell Signaling, upper panel) and anti-eIF2␣ antibodies (Sigma, lower panel). Lane 1, 0-min treatment. B, Huh7 cells in triplicate were treated with DMSO alone or 1 M thapsigargin dissolved in DMSO (DMSO-TG) for 3 h followed by transfection with in vitro transcribed capped 5ЈSrc-RFLuc RNA (solid black bar) or RL-HCV1b (solid gray bar). The upstream RLuc in RL-HCV1b RNA is translated by cap-dependent mechanism, whereas HCV IRES mediates downstream FLuc translation. The cytoplasmic lysates were assayed for FLuc (IRES-dependent translation) and RLuc (cap-dependent translation) activities 3 h post-transfection. The activities of FLuc and RLuc in untreated (control) samples were considered as 100% and compared with the solvent alone (DMSO) or TG-treated cells. C, cytoplasmic lysates from experiments described in B (above) were subjected to Western blot for the total c-Src protein with monoclonal anti-Src antibody (clone 327, Santa Cruz Biotechnology, upper panel) or anti-actin antibodies (lower panel). D, Northern blot analysis of total RNA extracted from Huh7 cells described in B and probed with 32 P-labeled oligonucleotide corresponding to nt 320 -350 of the c-Src 5ЈNCR (upper panel). Lower panel shows 18 S rRNA in the same samples. E, total c-Src level in Huh7 cell lysates cultured for 72 h in serum-deprived (indicated as minus) or 10% serum containing (plus) media. Western blot was carried out with anti-Src antibody as described above. F, Northern blot for probing c-Src mRNA (as described above) in total RNA extracted from Huh7 cells cultured in serum-starved (minus) and serum supplemented (plus) regular media. The 18 S rRNAs in each lane is shown in the lower panel.
The sucrose gradient analyses further provided insights into the mechanism of ribosome assembly on the c-Src IRES. The nonhydrolyzable GTP analogue, GMP-PNP, blocks eIF2dependent initiation pathway at the 48 S complex stage (18,39). This effect was clearly evident for the cap-dependent translation initiation of the FLuc mRNA in HeLa cell-free lysates in which the 48 S complex was trapped by 1 mM GMP-PNP treatment (Fig. 6). Thus, the GMP-PNP concentration used here during translation initiation assembly was sufficient to block 80 S assembly by cap-dependent initiation mechanism in the translation mixture. In sharp contrast, assembly of 80 S complex took place on the c-Src IRES in the presence of GMP-PNP or in a reaction mixture containing the analogue but was also deficient in exogenously added ATP and GTP. Generally, 60 S subunit joins the 48 S complex to form 80 S only after eIF5-induced GTP hydrolysis and dissociation of the eIF2-GDP complex (53). This step is preceded by ATP-dependent scanning by the 48 S complex to locate the AUG codon (1). From the data presented here, c-Src IRES appears to evade both of the critical energy-dependent steps that are needed for the 80 S assembly by the cap-dependent mechanism. Because the c-Src IRES directly binds the 40 S subunit (Fig. 7) and the structural motifs from flanking regions of the initiator AUG are required for efficient function of the c-Src IRES (Fig. 5), it is highly likely that the 48 S complex formed at this element may not require energy-dependent scanning for the initiator AUG. This notion is supported by the genetic analysis of the c-Src IRES. A 19-nt deletion at translation site in 5ЈSrc-⌬1-FLuc RNA resulted in complete impairment of the IRES function despite the presence of upstream AUG147 and AUG179. Recently, the HCV IRES was shown to switch from classical eIF2-dependent initiation to the eIF2-independent pathway under cellular stress that favors inactivation of eIF2 due to phosphorylation of its ␣ subunit. This alternative pathway was further shown to require only eIF3 and eIF5B (an analogue of bacterial IF2) for Met-tRNA i Met delivery at the P site. Based on these observations, it was proposed that the 80 S assembly on the HCV IRES is analogous to bacteria-like mode of translation initiation (53). In this context, the c-Src IRES appears to follow HCV IRES-like mode of translation initiation when the GTPase function of the ternary complex is blocked. This conclusion is further supported by RNA transfection studies in which thapsigargin-led induction of cellular stress in Huh7 cells failed to inhibit the c-Src IRES despite increased Ser-51 phosphorylation of eIF2␣ as compared with the normal (unstressed) cells (Fig. 9).
The studies presented here demonstrate significant resistance of the c-Src IRES activities to the reduced level of ternary complex and eIF-2␣ phosphorylation. In contrast to the eIF2dependent initiation pathway in which the eIF-2 complex delivers Met-tRNAi to 40 S subunits in a GTP-dependent manner, the eIF2A has been shown to deliver the Met-tRNAi to 40 S subunits by AUG-dependent and GTP-independent mechanisms (19,64). In addition, a number of RNA-binding proteins have been shown to stabilize IRES structure and/or promote ribosomal complex assembly (13,14). A comprehensive analy-sis is needed to ascertain whether these factors contribute to the reduced ternary complex dependence of the c-Src IRES.
In the cells, stress and serum deprivation causes inhibition of phosphatidylinositol 3-kinase/AKT/mammalian target of rapamycin pathway-dependent phosphorylation of eIF4E-BP. The unphosphorylated protein forms a tight complex with eIF4E and prevents its binding to eIF4G and the cap structure (8). Similarly, hypophosphorylation of eIF4E that is controlled by Ras-mitogen-activated protein kinase (MAPK) pathway also reduces its cap-binding ability. Both of these events culminate into suppression of global cap-dependent translation. In addition, phosphorylation of eIF2␣ by cellular kinases (e.g. PKR, PERK, HR1, and GCN2) in response to various cellular stress and viral infections leads to reduction in the level of ternary complex (eIF2-GTP-Met-tRNA i Met ) due to inhibition of guanine nucleotide exchange factor activity (54). Our investigations revealed that 80 S assembly on the c-Src IRES occurs when the functions of eIF2 and eIF4E are inhibited. Therefore, it is possible that the c-Src mRNA can easily escape from tight regulation of both of these translation initiation factors, which may ultimately lead to continued c-Src protein synthesis during adverse conditions (e.g. endoplasmic reticulum stress and starvation). Enhanced c-Src level has been shown to correlate with its activated state in hepatocellular carcinoma (55). Activated c-Src is known to induce phosphorylation of 4E-BP1 via phosphatidylinositol 3-kinase/mammalian target of rapamycin and eIF4E via Ras/Raf/extracellular signal-regulated kinase (ERK) pathway, both of which favor cell survival and proliferation (56,57). Thus, the c-Src IRES controlled translation provides an important recovery mechanism from translational blockade during cellular stress.
It has been shown that the cap-dependent translation of c-Src mRNA is regulated by elements located in its long 3ЈNCR through interaction with heterogeneous nuclear ribonucleoprotein K (59). It would be interesting to investigate if heterogeneous nuclear ribonucleoprotein K or microRNAs can affect the c-Src IRES-controlled translation through the 3ЈNCR interactions. Both transcripts of the c-src gene (type 1A and type 1␣, Fig. 1A) contain conserved sequences that constitute most parts of the IRES element. However, the extreme 5Ј ends in these mRNAs are dissimilar in length and nucleotide composition. It is not known if these sequences play any role in regulating the c-Src translation.
c-Src is an important player in signal transduction pathways that control oncogenesis, cell proliferation, and metastasis (24 -27). The emerging strategies for treatment of breast, lung, prostate, skin, and other cancers are focused on the inhibition of c-Src activities (23,58,60) but not the enhanced supply of c-Src in the tumor cells. Many of the small molecules that target c-Src activities also inhibit other protein kinases and/or show high degrees of cytotoxicity (60). Adaptation for growth during cellular stress is a hallmark feature of many cancer cells, and c-Src has been shown to a play very important role during this process (61). Our study presents c-Src IRES as a new therapeutic target for treatment of cancer. Because the c-Src IRES is located downstream of the cap structure in the mRNA, interference with the IRES structure and/or function will likely result in the inhibition of cap-dependent as well as IRES-de-c-Src IRES-mediated Translation Initiation pendent c-Src synthesis. This strategy will prevent unabated c-Src supply in the cancer cells and hence is likely to reduce the chances of cancer cell survival.