|Home | About | Journals | Submit | Contact Us | Français|
IGF1R is a proto-oncogene with potent mitogenic and antiapoptotic activities, and its expression must be tightly regulated to maintain normal cellular and tissue homeostasis. We previously demonstrated that translation of the human IGF1R mRNA is controlled by an internal ribosome entry site (IRES), and delimited the core functional IRES to a 90 nucleotide segment of the 5’-untranslated region positioned immediately upstream of the initiation codon. Here we have analyzed the sequence elements that contribute to the function of the core IRES. The Stem2/Loop2 sequence of the IRES exhibits near-perfect Watson-Crick complementarity to the G961 loop (helix 23b) of the 18S rRNA, which is positioned within the E-site on the platform of the 40S ribosomal subunit. Mutations that disrupt this complementarity have a negative impact on regulatory protein binding and dramatically decrease IRES activity, suggesting that the IGF1R IRES may recruit the 40S ribosome by a eukaryotic equivalent of the Shine-Dalgarno (mRNA-rRNA base-pairing) interaction. The homopolymeric Loop3 sequence of the IRES modulates accessibility and limits the rate of translation initiation mediated through the IRES. Two functionally distinct allelic forms of the Loop3 poly(U)-tract are prevalent in the human population, and it is conceivable that germ-line or somatic variations in this sequence could predispose individuals to development of malignancy, or provide a selectable growth advantage for tumor cells.
The type 1 insulin-like growth factor receptor (IGF1R) is a proto-oncogene with potent mitogenic, anti-apoptotic, and transforming capabilities [Gooch et al., 1999; Kim et al., 2007; Kurmasheva and Houghton, 2006; LeRoith and Roberts, 2003; Pandini et al., 1999; Yanochko and Eckhart, 2006]. Precise control of IGF1R expression is critical for maintenance of normal cellular and tissue homeostasis and avoidance of malignant transformation, and this requires specific regulation at both the transcriptional and translational levels [Cooke and Casella, 1994]. The IGF1R mRNA contains an extraordinarily long 5’-untranslated region (1,040 nucleotides) [Cooke et al., 1991], projected to adopt a highly internally base-paired structure (dG>-500kCal/mole) which represents a substantial impediment to ribosomal scanning. We discovered an internal ribosome entry site (IRES) within the human IGF1R 5’-UTR which provides an alternative mechanism for IGF1R translation initiation, allowing the 40S ribosome to bypass the obstacles presented by the highly structured 5’-UTR [Meng et al., 2005]. The IGF1R IRES was authenticated by its sensitivity to deletion of the promoter from a bicistronic construct, its resistance in a monocistronic context to co-expression of a viral 2A protease, and its ability to function in vitro under conditions that block cap-dependent translation initiation [Meng et al., 2008]. The core functional IRES was delimited to the 3’-terminal 90 nucleotides of the 5’-UTR, positioned immediately upstream of the initiation codon. Two sequence-specific RNA-binding proteins, HuR and hnRNP C, which compete for interaction with the IRES and differentially regulate its activity, were identified and characterized, and evidence for pathological dysregulation of the IGF1R IRES in human breast tumor cells relative to non-transformed breast epithelial cells through changes in activities of RNA-binding IRES-regulatory proteins was presented [Meng et al., 2008; Meng et al., 2005].
Here we have performed a detailed examination of the core functional IRES by site-directed mutagenesis within the context of the full-length 5’-UTR. We report two sequence elements which appear to be very important for translation initiation mediated by the IGF1R IRES. The Stem2/Loop2 sequence appears to be a focal point for operation and regulation of the IRES. It serves as the recognition site for a group of IRES-regulatory proteins, and may directly facilitate ribosome recruitment via direct Shine-Dalgarno-like base-pairing interaction with the 18S rRNA. The Loop3 sequence, which exists in two distinct allelic size variants in the human population, appears to govern the maximal rate of translation initiation through the IGF1R IRES by limiting accessibility of the core functional IRES to positive ITAFs.
The human CAGE tag database was accessed at http://gerg01.gsc.riken.jp/cage/hg17 [Carninci et al., 2006]. The CAGE Basic Viewer for Homo sapiens was used to search chromosome 15 sorted on gene symbol for IGF1R, and all of the CAGE tags associated with the IGF1R locus were tabulated. The positions of the tags relative to the ATG initiation codon were calculated, and the number of tags mapping to each site was plotted.
The full-length IGF1R 5’-UTR (1,040 nucleotides) and isolated IRES RNA (90 nucleotides) were synthesized in vitro using T7 RNA polymerase (RiboMax, Promega) and 3’-end-labeled using 32P-pCp and T4 RNA ligase. The RNA was then subjected limited digestion with RNase T1 (cleaves after single stranded G residues) or RNase V1 (cleaves within double stranded RNA). Alkaline hydrolysis and RNase T1 digestion performed following heat denaturation of the RNA were used to generate appropriate reference landmarks.
The parent bicistronic construct containing the full-length IGF1R 5’-UTR cloned between the coding sequences for Renilla and firefly luciferases (pDual IGF1R (1–1038)) has been described [Meng et al., 2005]. The first series of site-directed mutations was designed using mfold and introduced into the parent construct using the QuikChange kit (Stratagene). Inserts for the second series of site-directed mutations were designed using mfold and generated by PCR amplification from T47D (biallelic) genomic DNA, capturing the natural short (U13) and long (U21) variations of the Loop3 sequence, and utilizing mutant primers to introduce compensatory alterations into Stem2.
Base substitutions are as follows: mutLoop2: replace residues 1022–23 (UU) with GC; mutStem2: replace residues 1017–8 (GG) with CC; mutStem2_comp: replace residues 1017–8 (GG) with CC and residues 1027–28 (CC) with GG; minLoop3: reduce 988–1008 (U)21 tract to (U)6 (four residues remaining in Loop3, 2 residues in Stem3); IRESmutStem2a: replace residues 67–69 (GGA) with CCU and replace residues 76–78 (UCC) with AGG; IRESmutStem2b: replace residues 66–68 (GGG) with UAU and replace residues 77–79 (CCC) with AUA; IRESdimLoop3: capture natural (U)13 allele from T47D (biallelic) genomic DNA using wild-type primers.
The T47D human breast carcinoma cell line was obtained from ATCC and grown in RPMI 1640 medium with 10% fetal calf serum, 10 µg / ml insulin, and penicillin/ streptomycin, and maintained in a humidified environment at 37°C with 5% CO2. The T98G human glioblastoma, Saos-2 human osteosarcoma, and MCF-10A non-transformed human mammary epithelial cell lines were propagated according to ATCC recommendations.
T47D cells were seeded at 50% confluence (~100,000 cells per well) into 24 well plates and allowed 36 hours for attachment and recovery prior to transfection using Lipofectamine 2000 (0.67 µl/well) and bicistronic reporter plasmid (100 ng/ well). Forty-eight hours following transfection, cell were harvested, lysates prepared, and firefly and Renilla luciferase activities measured using the Dual Luciferase assay kit (Promega). All assays were performed in triplicate or quadruplicate and repeated at least 2–4 times. IRES activity was calculated as the ratio of firefly to Renilla luciferase activities, and in some figures, presented relative to f/R of the full-length IGF1R 5’-UTR (set = 100%), and in other figures, related to the f/R ratio obtained with the control bicistronic construct pDual (with no IGF1R insert and no IRES, set = 1). To measure the effects of HuR or hnRNP C on activity of the IRES mutants, expression constructs for each of these RNA-binding proteins were cotransfected along with the bicistronic reporter constructs (as previously described) [Meng et al., 2008; Meng et al., 2005].
The three-dimensional structure for the T. thermophilus 30S ribosomal subunit with mRNA (PDB ID: 1JGO) [Yusupova et al., 2001] was accessed from the Structure database of the NCBI. The image was rotated to achieve optimal simultaneous visualization of the G691 loop of the rRNA, Shine-Dalgarno region of mRNA, and anti-Shine-Dalgarno region of rRNA (roughly the intersubunit face view). Segments of interest were highlighted using the interactive viewing program Cn3D.
Human genomic DNA was obtained from commercial sources or prepared from cultured cell lines. The core functional IRES sequence was amplified using forward primer Apa-Seq: 5’-TCGGAGTATTGTTTCCTTCGCCCTTG-3’ (nt 892 – 917) and reverse primer f: 5’-CGACGGTGGCAACTCGGGTTTCGC-3’ (nt 1201 – 1224); yielding a 333 bp product surrounding the core functional IRES (nt 951–1040). We tested several different PCR conditions in attempt to eliminate PCR error (most commonly +1, +2, or −1 U residues) through the long homopolymeric Loop3 sequence, but none was completely effective. We found that AmpliTaq combined with a buffer containing (NH4)2SO4 in place of KCl and 10% glycerol as cosolvent enhanced the specificity and fidelity of this product more so than other polymerase / buffer combinations. PCR products were submitted for direct sequence analysis and also cloned into pSTblue1 and individual clones sequenced.
The northwestern assay was performed essentially as described [Meng et al., 2008] except that a 10% gel was used and the magnesium concentration during incubation with probe was reduced to 2 mM. Briefly, cytoplasmic, nuclear, or whole cell extracts were prepared from T47D or MCF-10A cells, separated by SDS/PAGE, transferred to nitrocellulose, and then renatured sequentially in Tris-buffered saline followed by a detergent-free physiological buffer. Radiolabeled RNA probes (sense orientation) representing the wild-type or mutant full-length IGF1R 5’-UTR / IRES were synthesized using linearized monocistronic constructs as template and T7 RNA polymerase, with internal incorporation of α32P-UTP. The probes were incubated with the blot in renaturation buffer for 2 hours at room temperature, with gentle agitation, then washed three times (the first wash containing 1 mg/ml heparin) prior to autoradiographic exposure.
In their original characterization of the human IGF1R mRNA, Cooke et al determined by primer extension and RNase protection methods that the transcription start site was 1,038 base-pairs upstream of the coding sequence [Cooke et al., 1991]. Data generated more recently by high-throughput genome-scale methodologies such as CAGE tag analysis allow us to re-examine this question (Fig. 1A). Although there is evidence for a minor population of IGF1R mRNA molecules that begin ~700 or ~900 nucleotides upstream of the initiation codon, and an additional small cluster of IGF1R transcription start sites ~50,000 base pairs further upstream (not shown), this new data confirms that the predominantly utilized transcription start site imparts an ~1,043 nucleotide long 5’-untranslated region to the IGF1R mRNA.
The long, G-C-rich 5’-untranslated region of the IGF1R mRNA is projected to adopt a secondary structure with extensive internal base-pairing (Fig. 1B) and high thermodynamic stability (dG=−515.2 kCal/mol). The core functional IRES, which had been localized to the 3’-terminal 90 nucleotides (951–1040) of the 5’-UTR [Meng et al., 2008], is projected to adopt a relatively simple structure composed of one central multi-loop and two terminal stem-loops (see bracket in Fig. 1B). Although the mfold algorithm suggested multiple base-pairing alternatives with similar thermodynamic stabilities available to the 5’-UTR as a whole, the core functional IRES itself appears to represent an independently-folding domain, highly favored within the context of the full-length 5’-UTR. In fact, within the 20 highest scoring structures generated by mfold 3.2 for the 1,040 nucleotide 5’-UTR, 19 exhibited precisely the same 3-Loop folding pattern for the core functional IRES sequence. We were concerned however that the projected structures for the IRES could be influenced by its position so near the end of the sequence entered into the program. Therefore, we repeated the mfold analysis after adding the natural exon 1 and 2 sequences to the 5’-UTR (Fig. 1C). Although the structure projected for the 5’-UTR was dramatically altered by the presence of exons 1 and 2, the functional core IRES, now positioned near the middle of the sequence being analyzed, still preferentially adopted the same 3-Loop structure that had been highly favored in the original analysis.
A magnified image of the 3-Loop structure anticipated for the core functional IRES is presented as it might exist in the context of the full-length 5’-UTR (Fig. 1D), or as an isolated 90 nucleotide fragment (Fig. 1E) which displays even higher IRES activity than the full-length 5’-UTR [Meng et al., 2008]. We have designated the asymmetrical central multi-loop as Loop1, the six nucleotide hairpin loop appended to a four-base pair helix as Loop2, and the large poly(U) sequence appended on a 6-base-pair stem as Loop3. RNase T1 / V1 sensitivity data provide experimental support for the natural existence of this structure both in the intact 5'-UTR (Fig. 1F) as well as the isolated IRES (Fig. 1G). A symmetrical pattern of susceptibility to RNase V1 (which cleaves within double stranded RNA) is observed for residues expected to lie within Stem 2. G-residues within Stem2 (3 out of 4 base pairs are G.C) were partially or completely protected from RNase T1 digestion (which cleaves after single stranded G residues) even in the absence of added magnesium, while G-residues within Stem 3 (of which only 2 out of 6 base pairs are G.C) were more dependent on magnesium for base-pairing and protection from RNase T1 digestion.
We had previously noted that progressive 5’-deletions of the IGF1R 5’-UTR sequence were associated with progressive loss and then return of IRES activity [Meng et al., 2008]. We suspected that the loss of IRES activity observed with the intermediate deletions could be attributed to altered base-pairing patterns negatively impacting the core functional IRES. Indeed, we observe a strong correlation between measured IRES activity and the potential of the RNA to adopt the 3-Loop structure (Table 1).
To further dissect the IRES, we introduced site-directed mutations within the core IGF1R IRES sequence, to identify individual sequence elements important for IRES operation and regulation. The initial series of mutations were introduced and tested within the context of the full-length 1,040 nucleotide 5’-UTR (Fig. 2A, B). We focused our attention on two features: Stem2/Loop2 and the homopolymeric Loop3.
The Stem 2/Loop2 sequence was of particular interest because the hexaloop (AUUUCA) is very similar to a sequence element (UUUCC) thought to contribute to the function of several viral IRESs [Pilipenko et al., 1992; Scheper et al., 1994], and also very similar to the consensus (AUUUA) for the A-U-rich element binding protein HuR [Antic and Keene, 1997], which we had already determined binds to the IRES and potently regulates its activity [Meng et al., 2005]. We found that mutation of two adjacent nucleotides within either Loop2 or Stem2 resulted in ~50% loss of IRES activity. Five additional mutants altering from one to four residues within Loop2 were also associated with ~25–50% decrease in IRES activity (data not shown). A compensatory mutation in Stem2, modifying 4 out of 1,040 nucleotides and with no anticipated change in secondary structure, was associated with >80% loss of IRES activity, suggesting that the primary sequence of this region may be critical for internal ribosome entry.
To determine what effect the large poly(U)-tract comprising Loop3 might exert on IRES function, the (U19) sequence was minimized to a tetraloop. Based on mfold analysis of the mutated sequence, no additional changes to the core IRES structure were anticipated (Fig. 2A). (In fact, the identical 3-Loop structure was projected for each of the top 20 scoring structures generated for both the Stem2 compensatory mutation and the Loop 3 minimization mutation.) Minimization of the Loop3 poly(U)-tract was associated with a dramatic (~2.5X) increase in IRES activity (Fig. 2B), suggesting that, when intact, this long homopolymeric sequence serves as a negative regulator of IGF1R IRES-mediated translation initiation.
To further establish the importance of Stem2/Loop2 and Loop3 to IRES function, we tested the mutant constructs in a series of representative human tumor cell lines in which the IGF1R IRES is active. Very similar results were obtained in T47D (breast carcinoma in which the IGF1R IRES was originally characterized, very high IRES activity), T98G (glioblastoma, moderately high IRES activity), and Saos-2 (osteosarcoma, modest but readily detectable IRES activity) (Fig. 2C). Note that the compensatory Stem2 mutation was severely debilitating to IRES activity, while the Loop3 minimization mutant exhibited a marked increase in IRES activity, in all three cell lines.
To confirm these results, a second series of mutations was designed within the context of the 90 nucleotide core functional IRES. A compensatory substitution within Stem2 retaining the natural G-C-rich sequence composition resulted in near 50% decrease in IRES activity, while a similar mutation which replaced the three G-C base pairs with A-U base pairs exhibited >80% loss of IRES activity (Fig. 2D, E). This result suggests that, in addition to the changes in primary sequence, weakening of this stem may also have severe consequences for IRES function. To further assess the effect of the size of the homopolymeric Loop3 sequence on IRES activity, a more modest diminution of the polyU-tract (from U21 to U13, representing the naturally-occurring allelic variant of this sequence, derived by PCR amplification from T47D genomic DNA, see below) was tested. The decrease in size of Loop3 was again associated with a modest but significant (~20%) increase in IRES activity.
Together, these findings suggest that Stem2/Loop2 may be the key sequence element facilitating internal ribosomal entry, while Loop 3 may serve to modulate the activity of the IRES.
We extended our assays to further assess the effects of the Stem2 and Loop3 mutations on function of the IGF1R IRES over time (Fig. 3). The full-length 1,040 nucleotide IGF1R 5’-UTR and the 90 nucleotide isolated core IRES exhibit nearly identical kinetic profiles, with the isolated IRES exhibiting ~25% higher activity. By the 48 and 72 hour time points, the activity of the IGF1R IRES exceeds that of the well-characterized encephalomyocarditis virus (EMCV) IRES by a considerable margin, although the viral IRES is activated much more rapidly in the cells (peaking at the 4 to 8 hour time points). The Loop3 minimization mutant has lost this intrinsic dampening effect, allowing the IGF1R IRES to accelerate at a level that approaches that of the viral IRES, and ultimately displaying a higher activity than any of the other IRES constructs. The near complete loss of IRES function of the stem 2 compensatory mutant is obvious.
These results further support the conclusion that Stem2/Loop2 contains the sequence most critical for operation of the IRES (recruitment of the 40S ribosomal subunit), while the poly(U) tract of Loop3 may serve to limit the maximal rate of translation initiation mediated through the IRES.
The fundamental purpose of the IRES is to recruit the 40S ribosomal subunit. A direct mRNA-rRNA base-pairing interaction between the IGF1R IRES and the 18S rRNA could conceivably facilitate ribosome recruitment. Upon careful examination, we noted that there exists a near-perfect Watson-Crick complementarity between the Stem2/Loop2 sequence of the IGF1R IRES and the G961 loop / helix 23b of the human 18S rRNA (Fig. 4A, B). The 959–964 region of the human 18S rRNA is known to be accessible to chemical probes, and therefore potentially available for direct base-pairing interactions with the mRNA [Demeshkina et al., 2003; Demeshkina et al., 2000]. Our mutational analyses indicate that alterations to the Stem2Loop2 sequence which disrupt this complementarity are associated with a substantial loss of IRES activity. Furthermore, the G961 loop is projected to occupy an ideal position in three-dimensional space, on the platform of the 40S ribosomal subunit, from which to interact with the mRNA and facilitate translation initiation (Fig. 4C). Based on the relative positions of the mRNA and rRNA deduced from the crystal structure of the T. thermophilus 30S ribosomal subunit [Yusupova et al., 2001], it appears that the G693 loop of the prokaryotic 16S rRNA (homologous to the eukaryotic G961 loop of the 18S rRNA) lies in the immediate vicinity of both the anti-Shine-Dalgarno sequence (which is not present in eukaryotes) as well as the Shine-Dalgarno segment of the proximal 5’-untranslated region of the mRNA at the E-site. Together these findings suggest that the IGF1R IRES may recruit the 40S ribosome by a eukaryotic equivalent of the Shine-Dalgarno interaction, with the G961 loop of the 18S rRNA substituting for the anti-Shine-Dalgarno segment and base-pairing directly with the Stem2/Loop2 sequence of the IRES. The parallels between the mRNA-rRNA base-pairing interaction potentially involved in the operation of the IGF1R IRES in humans and the classical Shine-Dalgarno interaction of prokaryotes are summarized in Table 2.
Examination of multiple GenBank entries containing the human IGF1R 5’-untranslated sequence reveals evidence of natural variation in the size of the Loop3 poly(U)-tract, ranging from 14 to 24 contiguous U residues, with a bimodal distribution centered around U16 and U24 (Fig. 5A). To explore this issue further, we PCR-amplified the core IGF1R IRES sequence from several different sources of human genomic DNA (Fig. 5B). In most cases a doublet (two distinct bands) was observed on agarose gel electrophoresis of the PCR product. Direct sequencing of these PCR products generated results that became challenging to interpret upon reaching the Loop3 sequence from either direction, however, we were able to definitively establish that the results were a composite of two distinct alleles (most commonly U21 and U13) differing by 8 in the number of residues in the Loop3 poly(U)-tract, with additional minor variants generated by polymerase error (most commonly +1, +2, or −1 U residues). However, for several of the human tumor cell lines tested, the PCR-amplified core IGF1R IRES yielded only 1 band on agarose gel, apparently representing only the smaller of the two alleles. To confirm these findings, PCR products generated from two representative cell lines: T47D (biallelic) and MDA-MB-231 (monoallelic) were cloned and multiple individual clones sequenced. The results were indicative of the existence of two distinct alleles, clustered around U13 and U21, in T47D cells, but only 1 allele, clustered around U13, in MDA-MB-231 (Fig. 5C).
We have previously characterized the RNA-binding protein HuR as a potent repressor of the IGF1R IRES [Meng et al., 2005], while hnRNP C appears to activate the IRES [Meng et al., 2008]. We were interested to determine what consequences the representative Stem2 compensatory (decrease in function) and Loop3 minimization (gain in function) IRES mutations would have on the IRES-regulatory capabilities of these RNA-binding proteins. Experiments were performed in which the various IRES reporter constructs were cotransfected with expression vectors for HuR or hnRNP C, as had been done originally to establish the IRES-regulatory activities of these RNA-binding proteins (Fig. 6). We found that each mutant IRES construct remained sensitive to the positive or negative modulatory effects of each of the RNA-binding proteins, with one exception: the severely debilitated Stem2 compensatory mutant IRES could not be activated by hnRNP C.
We have developed a high resolution northwestern protocol to analyze sequence-specific interactions between putative IRES trans-acting factors (ITAFs) and the IGF1R 5’-UTR. The northwestern is a functional assay of RNA-binding activity performed under highly stringent conditions. Utilizing the wild-type sequence as probe, a reproducible pattern of bands representing a series of proteins capable of interacting specifically with the IGF1R 5’-UTR is observed (Fig. 7). To test whether the Stem2 or Loop3 IRES mutations might alter the pattern of protein binding to the IGF1R 5’-UTR, the northwestern assay was repeated using the mutant RNA sequences as probes. A marked change in affinity of multiple individual proteins for binding the mutant 5’-UTR RNA sequences was observed. The compensatory mutation to Stem2, which dramatically decreases IRES activity, was associated with a near complete loss of binding of a number of the putative ITAFs (particularly Bands 2, 3, 4, 4b, 5, 8a, and 8d) detected by the northwestern assay. A distinct subset of bands (7, 9, and 10) remained relatively unchanged or even increased in intensity, confirming that the effect of the Stem2 mutation on protein binding to the IGF1R 5’-UTR was selective. Interestingly, the minimization of Loop3, which dramatically increases IRES activity, was associated with an equally dramatic increase in affinity for the same ITAFs which had been negatively impacted by the Stem2 mutation. Importantly, we have definitively characterized bands 2, 4, and 5 as “internal” ITAFs, binding within the 90 nucleotide core functional IRES sequence, thus their altered affinity for the mutant IRES probes can be rationalized on this basis. Band 8d, which we have determined to be hnRNP C (also an internal ITAF) exhibits the same trend as the other internal ITAFs (decreased affinity with the Stem 2 mutation, increased affinity with minimization of Loop 3). In contrast, we have determined that bands 7, 9, and 10 are “external” ITAFs, binding primarily outside of the core IRES sequence, and indeed these bands show very little variation in intensity accompanying mutation of the core IRES sequence.
It has been widely held that the distance from the beginning of the mRNA to the initiation codon is usually <200 nucleotides, that only a relatively small proportion of eukaryotic mRNAs have 5’-UTRs of substantially greater length, and that these long 5’-UTRs tend to be associated with proto-oncogenes and other factors integrally involved in the control of cell proliferation and survival. A longer 5’-untranslated region provides a larger space for RNA structural features and protein binding sites, and therefore greater opportunities for specific regulation of gene expression at the translational level. Indeed, we have begun to discover many intricate regulatory mechanisms provided by these complex 5’-UTRs which allow the translational efficiency of individual mRNAs to be selectively modulated [Coleman and Miskimins, 2009; Conte et al., 2009; Fernandez et al., 2005; Galban et al., 2008; Jo et al., 2008; Spriggs et al., 2008]. However, emerging results of the ENCODE project [Birney et al., 2007], which involves an unbiased high-throughput mapping of transcriptional activity, have shown that individual annotated genes are associated with an average of 5 distinct transcription start sites, many of which are positioned thousands of base pairs further upstream from the coding sequences [Denoeud et al., 2007], suggesting that complex 5’-untranslated sequences such as that associated with the IGF1R mRNA are likely to be much more common than previously recognized. Such an expansion of the transcriptional landscape potentially provides an even greater opportunity for gene-specific translational regulation. However, if indeed these additional mRNA species with massively extended 5’-untranslated sequences are to function as translatable mRNAs, it is also reasonable to anticipate that IRESs may also be much more commonly utilized than previously suspected [Baird et al., 2006]. Consequently, it is very important that we determine how IRESs operate and how they are regulated in order to understand how these newly discovered transcripts can be effectively translated.
Translation initiation in bacteria inherently requires “internal ribosomal entry” due to the polycistronic nature of the prokaryotic mRNAs [Shine and Dalgarno, 1974; Steitz and Jakes, 1975]. Our findings suggest that the IGF1R IRES may recruit the 40S ribosome by a eukaryotic equivalent of the prokaryotic Shine-Dalgarno interaction. The first point of contact between the IGF1R IRES and the 18S rRNA would likely involve Loop2 of the IRES and the G961 loop of the 18S rRNA. The Stem2 structure may serve an important purpose in presenting the Loop2 sequence in optimal geometrical orientation for the initial “kissing” interaction with the unpaired residues of the G961 loop. Subsequent melting of the Stem2 structure would be required to expose the full Watson-Crick mRNA-rRNA base-pairing potential, a dynamic process likely regulated by sequence-specific RNA-binding proteins. We have observed two different complementary mutations to the Stem2 sequence which result in >80% loss of IRES activity. This data supports the concept that Stem2/Loop2 may be the key determinant of IGF1R IRES function, and that a direct base-pairing interaction with the G961 loop (h23b) of the 18S rRNA at the E-site of the 40S ribosomal subunit may provide the fundamental basis for ribosome recruitment by the IRES.
Compensatory mutation to Stem 2 also markedly decreases affinity for a number of individual bands detected by northwestern analysis, indicating that the Stem 2/Loop2 microdomain serves as a critical recognition sequence for as many as seven distinct RNA-binding proteins (including hnRNP C) which interact specifically with the core functional IRES sequence and may directly influence the efficiency of ribosome recruitment to the IRES. Taken together, these findings suggest that Stem2/Loop2 may be the focal point for ribosome recruitment and its regulation by internal ITAFs.
Minimization of the size of the Loop 3 poly(U)-tract actually increases affinity for several putative IRES-regulatory proteins. Loop3 minimization also dramatically increases IRES activity and alters the kinetics of IRES activation, inducing a much more rapid acceleration. These results suggests that the physiological function of the intact Loop3 sequence may be to limit accessibility of the critical Stem2/Loop2 sequence to internal ITAFs and perhaps also the 40S ribosomal subunit itself, ultimately governing the maximal rate of translation initiation mediated through the IRES. Both HuR and hnRNP C are known to bind preferentially to U-rich sequences [Coleman and Miskimins, 2009; Gorlach et al., 1994; Lopez de Silanes et al., 2004]. However, the fact that the Loop3 minimization mutant remains responsive to both hnRNP C and HuR indicates that the long homopolymeric (U)-tract of Loop3 is not likely the primary binding site for these IRES-regulatory proteins. Rather, these results suggest that hnRNP C and HuR may be directly involved in facilitating or blocking ribosome recruitment respectively, probably through interactions with Stem2/Loop2.
The proposed base-pairing interaction would position the proximal 5’-untranslated region of the IGF1R mRNA at the E-site on the platform of the 40S ribosome. This location on the small ribosomal subunit has been proposed to serve as a dedicated site for a variety of translation-regulatory phenomena involving 5’-untranslated sequences of mRNA [Marzi et al., 2007]. The mRNA exit site, so named because of its role during translation elongation, is bounded by helix 23, helix 26, rpS7 (eukaryotic S5), and rpS11 (eukaryotic S14) [Kaminishi et al., 2007], and actually serves as a docking and entry site during translation initiation. Viral IRESs have also been shown to contact the 40S ribosome in this region [Kieft, 2008].
Shine-Dalgarno-like interactions have been suspected of contributing to the function of several viral IRESs [Le et al., 1992; Pilipenko et al., 1992; Scheper et al., 1994; Yang et al., 2003], and may indeed be utilized by other cellular IRESs as well [Dresios et al., 2006]. In addition, two regulatory mechanisms distinct from internal ribosomal entry have been described which appear to depend on mRNA-rRNA base-pairing to recruit or reposition the ribosome. Ribosomal shunting has been demonstrated for the adenovirus late mRNA, with evidence for Watson-Crick base-pairing of 5’-untranslated sequence to the 3’-terminal hairpin of the human 18S rRNA, located on solvent accessible surface of 40S ribosomal subunit [Yueh and Schneider, 2000]. A similar mechanism has been proposed for the heat shock protein 70, an endogenous cellular mRNA. A process of translation re-initiation is utilized by the feline calicivirus to facilitate expression of the VP2 open reading frame from a naturally bicistronic viral RNA in eukaryotic cells [Luttermann and Meyers, 2009]. This mechanism requires an unpaired sequence element located between the two open reading frames which has the potential for base-pairing interaction (9 contiguous nucleotides of perfect Watson-Crick complementarity) with the loop of helix 26 of the 18S rRNA. Like helix 23 (which we propose may be involved in interaction with the IGF1R IRES), helix 26 of the 18S rRNA is also positioned near the E-site of the 40S ribosomal subunit. For each of these mechanisms (shunting, re-initiation, IRES), the complementarity to rRNA is not the only factor contributing to ribosome recruitment. Secondary structures formed by neighboring sequences within the mRNA appear to be important for optimal presentation of the complementary sequence to ribosomal RNA, and protein-protein interactions likely facilitate the interaction between the mRNA and the ribosome.
However, based on our BLAST analyses, it appears that complementarity to the G961 loop of the 18S rRNA may be unique to the IGF1R IRES. If our interpretation is correct, for such an interaction to be reserved solely for IGF1R would reflect just how biologically important the translational regulation of this critical growth control gene may be to the phenotypic integrity of the cell. Indeed the significance of IGF1R within the hierarchy of genes controlling cell proliferation and survival is not without support, as it appears that many other growth factor / receptor systems such as EGFR, VEGF, and the estrogen receptor, and other growth control genes such as cyclin D1, may function through or be controlled by IGF1R [Coppola et al., 1994; Jones et al., 2008; Stewart et al., 1990]. IGF1R is considered by cellular physiologists to be the single most powerful potentiator of survival signaling in the mammalian cell [Peruzzi et al., 1999], and IGF1R appears to be essentially required for malignant transformation [Sell et al., 1994]. (This is not to suggest that other cellular IRESs may not utilize Shine-Dalgarno-like base-pairing with the 18S rRNA, but rather that interaction with the seemingly ideally positioned G961 loop may be reserved for IGF1R.)
IGF1R levels are low in normal differentiated adult cells [Werner et al., 1989], whereas IGF1R overexpression is implicated in the pathogenesis of a large proportion of human cancers [Kolb et al., 2008; LeRoith and Roberts, 2003; Zhang and Yee, 2000]. Our lab has begun to accumulate substantial evidence that pathological dysregulation of the IGF1R IRES could be responsible for IGF1R overexpression in a proportion of human breast tumors [Meng et al., 2008]. There is clear evidence from multiple GenBank entries and our own focused sequence analyses that the Loop3 U-tract sequence is naturally polymorphic in the human population, with the existence of two distinct alleles, differing by eight in the number of U residues. While each of six different sources of normal human genomic DNA we examined appear to contain both the long and short alleles, three of the seven tumor cell lines screened appear to contain only the small Loop3 allele. As little as 50% increase in IGF1R protein levels may allow cells to proliferate in an anchorage-independent manner (a surrogate assay for tumorigenicity) [Rubini et al., 1997]. Considering the functional relationship we have observed between the size of the Loop3 polyU-tract and IRES activity, it is conceivable that allelic variation of this sequence, whether present in the germ-line or somatically acquired, could be associated with a predisposition to malignant transformation or provide a selectable proliferative/survival advantage for the tumor cells. Studies are currently underway to further explore this question.
Grant sponsor: National Institutes of Health / National Cancer Institute
Grant number: R01 CA108886.