The Position Dependence of Translational Regulation via RNA-RNA and RNA-Protein Interactions in the 5′-Untranslated Region of Eukaryotic mRNA Is a Function of the Thermodynamic Competence of 40 S Ribosomes in Translational Initiation*

Cap proximity is a requirement to enable secondary structures and RNA-binding proteins to repress translational initiation via the 5′-untranslated region (5′-UTR) of mammalian mRNAs. We show that in Saccharomyces cerevisiae, unlike mammalian cells, the in vitro translational repressive effect of the mammalian iron regulatory protein 1 (IRP1) is independent of the site of its target in the 5′-UTR, the iron-responsive element (IRE). In vitro studies demonstrate that the binding affinity of IRP1 is also unaffected by the position of the IRE. Using IRE loop mutants, we observe an almost complete loss of IRP1-dependent repression in yeast concomitant with a 150-fold reduction in binding affinity for the IRE target. This mirrors the natural quantitative range of iron-induced adjustment of IRE/IRP1 affinity in mammalian cells. By enhancing the stability of the IRE stem-loop, we also show that its intrinsic folding energy acts together with the binding energy of IRP1 to give an additive capacity to restrict translational initiation. An IRE·IRP1 complex in a cap-distal position in yeast blocks scanning 40 S ribosomes on the 5′-UTR. It follows that the position effect of mammalian site-specific translational repression is dictated by the competence of the mammalian preinitiation complex to destabilize inhibitory structures at different steps of the initiation process.


Cap proximity is a requirement to enable secondary structures and RNA-binding proteins to repress translational initiation via the 5-untranslated region (5-UTR) of mammalian mRNAs. We show that in Saccharomyces cerevisiae, unlike mammalian cells, the in vitro translational repressive effect of the mammalian iron regulatory protein 1 (IRP1) is independent of the site of its target in the 5-UTR, the iron-responsive element (IRE).
In vitro studies demonstrate that the binding affinity of IRP1 is also unaffected by the position of the IRE. Using IRE loop mutants, we observe an almost complete loss of IRP1-dependent repression in yeast concomitant with a 150-fold reduction in binding affinity for the IRE target. This mirrors the natural quantitative range of iron-induced adjustment of IRE/IRP1 affinity in mammalian cells. By enhancing the stability of the IRE stem-loop, we also show that its intrinsic folding energy acts together with the binding energy of IRP1 to give an additive capacity to restrict translational initiation. An IRE⅐IRP1 complex in a cap-distal position in yeast blocks scanning 40 S ribosomes on the 5-UTR. It follows that the position effect of mammalian site-specific translational repression is dictated by the competence of the mammalian preinitiation complex to destabilize inhibitory structures at different steps of the initiation process.
RNA-RNA and RNA-protein interactions are dominant features of posttranscriptional gene expression. In eukaryotic translational initiation, ribosomes and a large number of initiation factors follow a complex series of steps leading to recognition of the first codon of a reading frame on the mRNA (1-3). The primary pathway of eukaryotic translational initiation for the majority of cellular mRNAs is currently described by a working model which envisages that 40 S ribosomal subunits progressively scan the 5Ј-UTR 1 from the 5Ј end in search of start codons (4). Global regulation of translational initiation can be achieved via modifications of the initiation factors (5). Control of initiation on individual mRNAs, on the other hand, is related to specific properties of the structure of mRNA and in particular the 5Ј-untranslated region (5Ј-UTR). The most frequently observed form of control mediated by the 5Ј-UTR is attributed to intrinsic structural properties that impose restrictions on the movement of the ribosomal preinitiation complex along the mRNA (6,7). For example, sequence elements within the 5Ј-UTR that are internally complementary can form stemloop structures that most likely have to be completely disrupted if an initiation complex is to scan through them. The ability of a given secondary structure in the 5Ј-UTR to act as a barrier to translational initiation is a function of this structure's free energy of formation; increased stability generally leads to stronger inhibition (7)(8)(9)(10)(11). However, the sensitivity of the host translational apparatus to this type of inhibition depends on the organism. Thus, mammalian ribosomes are not detectably impeded by structures that are strongly inhibitory to ribosomes of the yeast Saccharomyces cerevisiae (12). Indeed, a stem-loop structure with an estimated stability of approximately Ϫ18 kcal/mol Ϫ1 inhibits translation in S. cerevisiae by approximately 90% (8 -11), whereas an equivalent degree of inhibition in animal cells can only be achieved by a stem-loop with a stability exceeding Ϫ50 kcal/mol Ϫ1 provided this is not capproximal (6).
The impact of intramolecular folding within the 5Ј-UTR on translational initiation can be a function of the position of the resulting structures relative to other key elements of the mRNA. In higher eukaryotic cells, a given stem-loop structure is more inhibitory when it is cap-proximal than when it is positioned nearer the start codon (13). A proposed explanation for this is that a structure close to the 5Ј end of the mRNA interferes with the initial steps of ribosome mRNA binding, whereas a structure further away from the 5Ј end interferes with the scanning process (13). This implies that either the free energy changes of the early ribosome-mRNA interactions are smaller than the thermodynamic driving force intrinsic to the scanning process, or ribosomes can "skip" structural barriers more readily in a cap-distal position. However, neither of these principles seems to apply to S. cerevisiae, in which the position of a stem-loop structure within the 5Ј-UTR is of little significance in terms of the degree of translational inhibition observed (10,11). This discrepancy must reflect inherent differences in the pathways of translational initiation in the respective higher and lower eukaryotic systems.
The above considerations are relevant to the function of trans-acting regulators of translation that interact with the 5Ј-UTR. In particular, RNA-binding proteins are instrumental in the regulation of translational initiation in both prokaryotes (14) and eukaryotes (15,16). In contrast to the constitutive limitation of translational initiation that can be imposed by intramolecular RNA-RNA interactions, the inducible synthesis or activation of an mRNA-binding repressor protein allows negative regulation within a defined period. The best characterized eukaryotic example of this type of regulation is based on the binding of the iron regulatory proteins (IRP1 and IRP2) to the iron-responsive element (IRE) in the 5Ј-UTRs of the mRNAs encoding ferritin and erythroid 5-aminolevulinic acid synthase in vertebrate cells (17). Moreover, both vertebrate and insect IRPs bind to an IRE in the 5Ј-UTR of the succinate dehydrogenase subunit b mRNA of Drosophila melanogaster (18). Although most studies of the vertebrate system have focused on the role of IRP1, it was shown that both IRPs are expressed in all tissues and bind consensus IREs present in eukaryotic transcripts with equal affinity (19). The affinity of IRP1 for IREs in these mRNAs is high (K d ϭ 10 Ϫ10 -10 Ϫ11 M) at low iron concentrations and sufficient to block translation but is reduced 50-to 100-fold in the presence of iron levels in excess of the requirements of the cell. The expression of the IRP1 gene was also found to be sufficient for strong translational repression of an mRNA bearing an IRE-containing 5Ј-UTR in S. cerevisiae thus demonstrating that regulation requires no mammalian components other than IRP1 and IRE to function (20).
The influence of the binding of two other RNA-binding proteins to the 5Ј-UTR has also been investigated in yeast (21). By inserting the appropriate recognition elements in the 5Ј-UTR of a reporter gene in S. cerevisiae, it was possible to demonstrate translational repression induced by the synthesis of either the human spliceosomal protein U1A or the bacteriophage MS2 coat protein in vivo. This confirmed that a translational regulatory system can in principle be created using the binding energy of any RNA-binding protein for its target.
As with most other RNA-RNA and RNA-protein interactions involved in eukaryotic gene expression, the mechanistic basis of translational control is poorly understood. In this paper, we use the IRP-IRE interaction as a tool in experiments designed to address the principles governing the access of eukaryotic 40 S ribosomal subunits to the first initiation codon in an mRNA. Analogously to a stem-loop structure, the function of a protein acting as a repressor of initiation depends on its ability to regulate, directly or indirectly, this access. Translational regulation by IRP1 in higher cells is mediated by binding to an IRE that is proximal to the 5Ј end of the mRNA (reviewed in Ref. 17). Indeed, increasing the distance of the IRE beyond more than 50 nucleotides from the cap greatly reduces the repressive effect of IRP1 binding (22). Moreover, in vitro experiments have indicated that IRP1 binding to an IRE in a suitably cap-proximal position prevents the formation of stable 43 S complex-mRNA interactions at the 5Ј end (23). This seems to indicate that IRP1 can only function as a repressor if it acts directly on the initial 43 S-mRNA interaction, which raises the question of whether the impact of the binding of a trans-acting factor such as IRP1 is dictated by specific binding properties of the protein. Alternatively, the ability of 40 S ribosomal subunits to overcome structural resistance (which is a form of "thermodynamic competence") at different steps of initiation has been suggested to play a key role (15). It is therefore evident that resolution of this issue will provide insight into the mechanism of translational regulation via the 5Ј-UTR. In the current study we have used the specific behavior of the S. cerevisiae translational apparatus to show how the significance of position and repressor binding affinity can only be understood in terms of the thermodynamic principles governing the interactions between the host translational apparatus and the mRNA.
Plasmid Constructs-The IRE-containing sequences (Table I) were inserted into the AflII site of the YCp22FL1 leader sequence ( Fig. 1B; Ref. 19). The spacer sequence was cloned into the BamHI site of the YCp22FL1 plasmid carrying IRE-wt (Fig. 1B).
The in vitro transcription vector used was pHST7 (derived from pHST0 (25)). The BamHI and EcoRI fragments from the YCp22FL1 constructs containing the leader sequence and a part of the luciferase gene (in the luciferase gene the EcoRI site is located at 599 nucleotides from the AUG start codon) were inserted into the BglII and EcoRI sites of the pHST7 vector immediately downstream of the T7 polymerase promoter.
DNA/RNA Preparation and Luciferase Assay-DNA cloning and sequencing were performed using standard methods (26). Oligodeoxyribonucleotides were synthesized using an Applied Biosystems DNA synthesizer. Total yeast RNA was isolated using the hot phenol method (27) from 20 ml of yeast culture. The Northern blots and luciferase assays were performed as described previously (20).
Polysomal Analysis in Vivo-Yeast cell extracts were prepared from cultures in 50 ml of YEP-Gal medium (A 600 ϭ 0.8 -1.0) and loaded on 12-ml 15-38% linear sucrose gradients (28). The gradients were centrifuged at 60,000 ϫ g for 2.5 h at 4°C using a Beckman SW40 rotor. RNA was extracted from 600-l fractions and analyzed by Northern blot analysis as described previously (28).
In Vitro Transcription, Translation, and Sucrose Gradient Analysis-DNA templates linearized with NarI (located 33 nucleotides downstream of the LUC start codon) were transcribed in vitro using T7 RNA polymerase (Promega) according to the manufacturer's specifications in the presence of the cap analogue m 7 GpppG and [␣-32 P]CTP (Amersham Corp.). The RNA transcripts were separated from free nucleotides by electrophoresis on a 6% acrylamide (acrylamide:bisacrylamide of 38:2),

FIG. 1. Expression vectors and IRE sequences used in this
work. The YCpSUP-IRP1 plasmid contains the human IRP1 gene under the control of the inducible GAL-PGK1 fusion promoter (A). The YCp22FL1 plasmid with the constitutive TEF1 promoter carries the firefly luciferase reporter gene (LUC) preceded by an IRE containing 5Ј-UTR (B). Introduction of a 50-nucleotide spacer generated a derivative in which the IRE is cap-distal (C). A range of mutations was introduced into the loop of the minimal IRE (D). The wild-type IRE was also extended and stabilized by adding G:C base pairs (E and F).
8 M urea gel. The full-length bands were cut out after visualization by autoradiography and eluted overnight in 400 l of 300 mM NaOAc, pH 5.2, 100 l of phenol, shaking at 37°C. Eluates were extracted with phenol-chloroform and precipitated by adding 3 volumes of 100% ethanol at Ϫ70°C. Extracts from yeast strains W303 and W303 bearing the YCpSUP-IRP1 plasmid grown on YEP medium with 2% galactose were prepared as described (29). 5 ng of labeled RNA were incubated with 12 l of each extract at 20°C under translation conditions (29) in the presence of 700 M methionine, 0.25 mM spermidine, and 400 g ml Ϫ1 yeast tRNA (30) for 5 min and immediately loaded onto the 5-30% linear sucrose gradients. Gradient analysis was performed as described previously (31).
Competition Experiments, Gel Retardation Assays, and Quantification-The relative binding affinities of mutated IRE RNAs were determined by their abilities to compete with IRE-wt for IRP1 binding. The sizes of the in vitro synthesized RNAs were equivalent to the lengths of the leaders indicated in Table I. In vitro transcription vectors linearized with NdeI (the NdeI site contains the ATG start codon of the luciferase gene) were used as templates for the synthesis of the full-length leader RNAs bearing the IREs. The IRE-wt-bearing RNA was labeled with [␣-32 P]CTP (Amersham Corp.) to a specific activity of 10,000 cpm/ng. The competitor RNAs were labeled with [5,6-3 H]UTP (DuPont) to a specific activity of approximately 500 cpm/ng. Synthesized RNAs were visualized by autoradiography in the case of 32 P-labeled RNAs and by ethidium bromide staining in the case of 3 H-labeled RNAs and purified as described above. After purification, RNAs were resuspended in diethyl pyrocarbonate-treated water, and the RNA concentrations were measured by scintillation counting. Gel-retardation assay mixtures were prepared in 20-l reaction volumes. 1 ng of 32 P-labeled RNA and serially diluted 3 H-labeled competitor RNAs were mixed in a final volume of 10 l, heated to 65°C for 5 min, and chilled in ice. Yeast extracts were prepared from cells bearing the YCpSUP-IRP1 plasmid that had been grown on YEP-Gal medium for 8 h. 0.5-1.75 g of extract in 10 l of binding buffer (100 mM potassium phosphate, pH 7, 10 mM MgCl 2 , 50 mM KCl, 2 mM dithiothreitol) were pretreated for 10 min with 2% 2-mercaptoethanol and 20 units of RNasin ribonuclease inhibitor (Promega) (32) and then incubated with RNA probe at 25°C for 30 min. 2-Mercaptoethanol improved the sharpness of the shifted bands but did not influence the measured binding affinities (20). After adding heparin to a final concentration of 5 mg/ml, the samples were incubated for an additional 10 min, mixed with 3 l of loading buffer (100% glycerol, 0.05% bromphenol blue), and loaded on 6% nondenaturing polyacrylamide gels. Electrophoresis was performed on 1 ϫ TBE for 3 h at 60 V. The gels were dried, and the amount of RNA⅐protein complex was determined on a Molecular Dynamics PhosphorImager using the Im-ageQuant software, version 3.3. The data were analyzed as described previously (33,34).

RESULTS AND DISCUSSION
Defining the Limits of Binding Affinity Required for Translational Repression by an RNA-binding Protein in Yeast-The 5Ј-UTR has been shown to act as the site of translational control mediated by secondary structures or protein binding sites. To investigate the relationship between RNA protein binding and the regulation of translation we used the IRE/IRP1 system, which has previously been found to function in both mammalian cells and yeast (17,20). Since other work has also shown that the binding of either the human U1A spliceosomal protein or the bacteriophage MS2 coat protein to their respective RNA target elements provides an analogous mechanism of translational regulation in yeast (21), the IRE/IRP1 system can be expected to provide representative data on the principles of action of a range of RNA-binding proteins in eukaryotic translation. At the outset of this work, specific changes in the loop sequence were known to affect the affinity of IRP1 for in vitro synthesized IRE stem-loop structures (33). In the first phase of the present work, we wanted to determine how translational regulation via the IRE-IRP1 interaction in yeast responded to changes in the binding affinity of the repressor/target pair. We therefore made use of a number of the IRE loop mutants already described in vitro by Jaffrey et al. (33). Since our objective was to examine the role of IRE/IRP1 binding affinity in translational regulation in vivo, we sought to obtain binding affinity data that would take into account the potential influence of the RNA environment of the IRE in the 5Ј-UTR used in FIG. 2. Assessment of relative IRE/IRP1 binding affinities via competitive gel band-shift assays. Representative gel retardation assay data (A and B) and the resulting competition curves for IRE-wt and IRE-mut2, respectively (C and D). The band-shift assay was performed using 32 P-labeled IRE-wt leader RNA and the indicated amounts of 3 H-labeled competitor 5Ј-UTR RNAs. The amounts of bound 32 P-labeled probe were quantitated using a PhosphorImager. The positions of the IRE⅐IRP1 complex and the unbound probe are indicated by arrowheads. The competition curves can be used to calculate IC 50 values (Table II) and relative dissociation constants (34). Under the conditions used here, the latter can be deduced directly from the plotted data. our expression studies. We therefore synthesized the complete leaders containing the respective mutant versions of IRE as 3 H-labeled transcripts in vitro and examined their individual abilities to compete with the equivalent 32 P-labeled leader containing the wild-type IRE in a band-shift assay (Figs. 1 and 2, Tables I and II). Competition curves were constructed using the relative band-shift intensities obtained at different ratios of mutant IRE leader RNA to wild-type IRE leader RNA (see Fig.  2 for example). The concentration of each competitor RNA required in the standard assay to attenuate wild-type IRE binding by 50% (IC 50 ) was assessed graphically. By using an excess of IRE ligands (labeled and unlabeled) over IRP in the extracts, we achieved conditions in which the relative IC 50 values corresponded directly to relative dissociation constants (34).
The 3 H-labeled IRE-wt RNA as self-competitor yielded an IC 50 value of 1, while the substitution or deletion of single nucleotides in the IRE loop (Fig. 1D) decreased the relative affinity of IRP1 by a factor ranging from 3 to 158. This fully covers the range of decrease in IRP binding affinity observed in mammalian cells under conditions of nonlimiting iron. We conclude that the context for the IRE provided by the leader sequence was of secondary importance with regard to the determination of IRP1 binding affinity since the relative values we obtained are similar to those of Jaffrey et al. (33) who used only the IRE sequences. This would be less likely to apply to leader sequences that are capable of forming secondary structures that interfere with IRE folding or the interaction with IRP1 (see also Ref. 35).
We also investigated the influence of changes in two additional properties of the IRE on the binding to IRP1: its distance to the 5Ј end of the mRNA and the G/C content and stability of its stem. Both properties are relevant to the mode of action of the IRE/IRP1 system. The introduction of a spacer between the cap and the IRE (IRE-wt ϩ 50sp; Fig. 1D) did not change the binding affinity of IRP1 (Table II). On the other hand, increasing the stability of the stem by introducing G:C base pairs (IRE-s1 and IRE-s2; Fig. 1, E and F) did influence IRP1 binding. These additional G:C pairs are likely to change the overall structure of the IRE stem significantly (see Ref. 36) also possibly reducing the ability of the IRE to accommodate itself to an "induced-fit" type of binding to IRP. These findings are also consistent with previous work emphasizing the importance of the overall structure of the IRE in determining its affinity for IRPs (37,38).

IRE/IRP Affinity as a Determinant of Translational Inhibition and mRNA Stability in Yeast Cells-
The effect of the interaction between IRP1 and the various IRE-containing leaders could be tested in yeast using the two-plasmid system established in previous work (20) (Fig. 1). One of the centromeric plasmids carries the repressor gene under the control of an inducible promoter (Fig. 1A), and the second encodes the firefly luciferase gene with the IRE (wild-type or mutated) sequence in the 5Ј-UTR of its mRNA (Fig. 1B). The luciferase activity of the construct lacking an IRE (FL) after induction of IRP1 gene expression in galactose medium was normalized to 100%. The binding of IRP1 to the respective IRE mutants resulted in different degrees of partial inhibition (Fig. 3). The strength of inhibition correlated with the relative binding affinity (Table II).
In further experiments, we investigated to what extent additional stabilization of the IRE stem could substitute for the binding of IRP1 in terms of translational inhibition and whether the free energies of RNA folding and of IRP1 binding, respectively, function additively. Increasing the stability of IRE folding to a predicted Ϫ15 kcal/mol Ϫ1 (IREs1) inhibited translation by 95%, whereas the binding of IRP for this version of IRE gave a greater combined total level of inhibition of 98%. The even more stable IREs2 imposed a constitutive level of inhibition of 99.3% so that the additional influence of IRP1 gene induction was barely detectable (Fig. 3B). These data show that the binding of IRP1 to these extended IRE structures does indeed provide an additive inhibitory effect. Interestingly, however, the affinity of IRP1 was reduced by a factor of 10 and 13, respectively, relative to the wild-type IRE. In any analysis of the effect of 5Ј-UTR function on the translation of mRNA, it is essential to determine whether there are concomitant changes in mRNA stability. This may be of significance to any regulatory mechanism involved and is certainly relevant to the accurate assessment of the translational effects under study. Previous work has shown that translational inhibition can modulate mRNA stability; where this occurs the type of response depends on the nature of the gene. LUC mRNA is stabilized upon inhibition of its translation (20,39). Therefore we analyzed the steady-state abundance of each of the constructs. Whereas the mRNA levels of the respective constructs in glucose medium were all the same, the induction of IRP1 synthesis (in galactose) resulted in degrees of stabilization that were a positive function of the binding affinities for the respec- The samples loaded were from cells in which IRP1 synthesis had been induced by growth in galactose medium. The relative abundance data were used to correct the luciferase activities presented in Fig. 3. Northern blot hybridization was performed using radioactive probes for LUC and PGK1 mRNAs, the latter taken as an internal control.
line of inhibition was greatly increased relative to that observed with the normal IRE structure used in the other constructs, and the IRP1 affinity for these types of IRE mutant was reduced (Table II). The introduction of an additional spacer sequence (IRE-wtϩ50sp) did not change the inhibitory effect of IRP1 binding on translation (A and B). The original leader sequence FL lacking the IRE was taken here as a reference point (normalized to 100%). This leader contains no IRE sequence and therefore supports a higher absolute level of initiation than those leaders bearing an IRE. All values presented are the results of averaging the data from at least three independent measurements and are corrected for mRNA abundance (B and C, compare Fig. 4).  Tables I and II and Fig. 1 tive IREs (Fig. 4). These values were used to correct the raw luciferase activity data to obtain reliable estimates of relative translation rates (Fig. 3).
Having determined the steady-state levels of translation and mRNA abundance of the constructs, we proceeded to examine the interactions between ribosomes and mRNA. We initiated this part of the work with sucrose gradient analysis of in vivo polysome distributions. The first set of experiments revealed that the LUC mRNA was shifted into the monosomal region upon binding of IRP1 to the wild-type IRE (Fig. 5A) as would be expected where translation is inhibited at the initiation step. This accumulation of monosomes was not observed using an mRNA lacking an IRE (Fig. 5B) or under conditions in which IRP1 synthesis is not induced (data not shown). The extent of the shift to the monosomal fractions was strongly reduced with IRE loop mutants that had attenuated IRP1 binding affinity (Fig. 5C). Therefore we conclude that the degree of exclusion of 40 S ribosomal subunits from polysomal fractions is related to the affinity of IRP1 for its binding site in the 5Ј-UTR.
The Properties of the Host Translational Apparatus Dictate the Position Dependence of Translational Repression via an RNA-binding Protein-It is known that for efficient regulation of translation by the IRE⅐IRP1 complex in higher eukaryotes, the IRE must be located within 40 nucleotides of the 5Ј end (22). If it is located at a position more distal from the cap, the regulation in vivo and in vitro is very much weakened. At least in vitro a similar position effect was observed in the case of translational repression by the U1A protein and bacteriophage MS2 coat protein (40). In contrast to these data, our results indicate that there is no position requirement for IRE function in yeast. Increasing the length of the spacer between the 5Ј end of the mRNA and the IRE-wt (IRE-wtϩ50sp) did not reduce the degree of translational inhibition in vivo in comparison with that of the control IRE-wt construct (Fig. 3). Polysomal gradient analysis revealed that the LUC mRNA with the extended leader was primarily found associated with monosomes and FIG. 6. Sucrose gradient profiles of translation in vitro performed in cell extracts prepared from S. cerevisiae. The 32 P-labeled shortened mRNA transcripts of IRE-wt (B), IRE-wtϩ50sp (C), and FL (D) were added to a wild-type (W303) extract (q) and to an extract from a strain containing overexpressed IRP1 (E). The ribosome-RNA interactions were analyzed on 5-30% linear sucrose gradients. Fractions collected from the gradients were counted in a liquid scintillation counter, and the counts were plotted against the fraction numbers. The A 264 absorption profile (A) indicates the positions of 40 S, 60 S, and 80 S ribosomal particles. The presence of IRP1 in a cellfree extract prevented formation of a stable complex between the 43 S preinitiation complex and an mRNA with a capproximal IRE. However, the extension of the 5Ј-UTR upstream of the IRE allowed stable ribosome binding, which manifested in the appearance of a peak in the monosome region (compare B and C).
show the fractional distribution of optical absorbance in a sucrose gradient experiment above a Northern blot prepared using the numbered fractions. These data indicate how the binding of ribosomal subunits to the respective mRNAs is affected by the IRP1-IRE interaction. In these blots, cohybridization with a PGK1-specific probe reveals the polysomal distribution of chromosomally encoded PGK1 mRNA. The data shown are those obtained with the following LUC constructs: IRE-wt (A), the control construct FL, which lacks an IRE (B), IRE-mut5 (C), and IRE-wtϩ50sp (D). To confirm that the observed luciferase mRNA is the full-length nondegraded species containing the spacer region, samples were hybridized with a spacer probe (E and F). The lanes in panel F marked IREwtϩ50sp and IREwt are controls demonstrating that the spacer probe (E) hybridized only with mRNA including the 5Ј spacer region. disomes (see Fig. 5D). To be certain that the full-length extended mRNAs of the construct IRE-wtϩ50sp were being excluded from polysomes, we repeated the Northern analysis of the gradient fractions using a probe specific for the additional spacer region (Fig. 5, E and F).
The gel retardation assay revealed that the binding of IRP1 was unaffected by the position of the IRE; the IC 50 values for IRE-wt and IRE-wtϩ50sp were identical (Table II). This means that IRP1 binding is not position-dependent and also that the binding characteristics of this repressor are not the source of the position effect seen in higher eukaryotic cells. In contrast, in yeast the scanning 40 S ribosomal subunit and associated initiation factors apparently do not possess the necessary energy to dissociate the wild-type IRE⅐IRP1 complex. A prediction arising from this is that a cap-distal IRE bound by IRP1 should be able to cause the accumulation of ribosomal subunits on the spacer. We tested this prediction by performing sucrose gradient analysis on in vitro synthesized mRNA in cell-free extracts derived from S. cerevisiae. This procedure allowed us to achieve a higher resolution analysis of the effects on ribosome-mRNA interactions of the various leaders. Extracts were prepared from W303 or from W303 containing the IRP1 expression plasmid (Fig. 6). The presence of the wild-type IRE in the 5Ј-UTR greatly reduced the population of 80 S ribosomes associated with the mRNA (Fig. 6B). This effect was not observed using the control mRNA bearing either no IRE (Fig. 6D) or the mutant form, IRE-mut5 (data not shown). In contrast, the addition of the 50 nucleotide spacer 5Ј of the IRE allowed accumulation of 40 S subunits on the leader (Fig. 6C). Under the conditions of these in vitro experiments, the 40 S subunits manifested themselves as a single additional peak thus indicating that each mRNA molecule had only one "queuing" 40 S subunit. Given that the spacer region would be expected to be capable of accommodating at least two 40 S subunits, this may be attributable to the reduced availability of active 40 S subunits in the cell-free extracts.
Thermodynamic Principles Dictate the Lack of a Position Effect in the Yeast 5Ј-UTR-In conclusion, our data indicate how the molecular basis of translational repression by an RNAbinding protein in the eukaryotic cell is governed by thermodynamic principles (Fig. 7). The position effect of the IRE/IRP1 system is not an intrinsic property of these components but rather derives from the nature of the host translational machinery. The restriction of translation in cap-proximal and cap-distal positions in S. cerevisiae corresponds to what are apparently thermodynamically equivalent processes with there being no effective distinction between initial cap-related mRNA binding of the 40 S ribosomal subunit and scanning in terms of the free energy required to achieve a given level of inhibition. In this respect, the mode of action of the IRE-IRP1 interaction is mechanistically equivalent to that of a stem-loop structure placed either proximally or distally to the 5Ј end.
As we have seen, the rate or frequency with which the preinitiation complex moves through the site of an IRE⅐IRP1 complex in the 5Ј leader depends on the binding energy of the repressor interaction (Figs. 2 and 3). In yeast, the binding affinity of IRP1 for the wild-type IRE is sufficient to block progression toward the initiation codon irrespective of whether the IRE is cap-proximal or cap-distal (Fig. 7, A and B). In the present work we have shown that this applies to a cap-to-IRE distance of 59 nucleotides, a distance that would nullify the repression effect in vertebrate cells (22). While demonstrating that yeast and vertebrate cells differ markedly in this respect, this does not rule out the possibility that extending the distance further might reduce repression in the yeast system. It should be noted that the IRE used in this work is a minimal FIG. 7. Model of inhibition of translation initiation via IRP1 binding to an IRE in the 5-UTR of an mRNA in yeast. The binding of IRP1 to a cap-proximal IRE structure prevents the association of higher eukaryotic 43 S translation preinitiation complexes at least in vitro (23) (A). This is also likely to apply to yeast (see "Results and Discussion"). IRP1 bound to an IRE that is located in a position distal from the cap can be easily displaced by the scanning 40 S ribosomal subunit (possibly with the help of the RNA helicase activities of eIF4A/ eIF4B) in the higher eukaryotic system (C), but in yeast the IRE⅐IRP1 complex still blocks its movement toward the initiation codon (B). The difference in position dependence between higher and lower eukaryotes is a manifestation of the differing abilities of the respective preinitiation complexes to overcome physical barriers at the mRNA-association and scanning steps of the initiation process. The yeast-scanning 40 S ribosomal subunit is only weakly thermodynamically competent to overcome the IRE⅐IRP1 complex, irrespective of its position. In contrast, the scanning mammalian 40 S ribosomal subunit can dissociate a cap-distal IRE⅐IRP1 complex and unwind the IRE stem-loop (C), whereas the free energy of formation of the initial complex with the mRNA is insufficient to drive this event. This is analogous to the position dependence of inhibition by a stable stem-loop structure (7). The mutational analysis of IRE/IRP1 binding affinity in yeast shows how the free energy required for disruption of this complex can be progressively reduced thus allowing scanning to proceed (B3C). This lowering of the binding energy is not required by the mammalian scanning complex, which already possesses full thermodynamic competence to force dissociation of IRP1 from wild-type IRE. Unlike the yeast preinitiation complex, the scanning of the mammalian ribosome is driven by a significantly greater thermodynamic force. The binding affinity of IRP1 to IRE is not affected when this element is placed in a more distal position (panel B and Table II), and therefore the binding characteristics of IRP1 itself do not contribute to the position effect seen in mammalian cells. In conclusion, the thermodynamic competence of the host translational apparatus dictates the conditions under which an inhibitory element can participate in translational regulation. As a consequence, position dependence is not a fixed parameter for any given type of control element. form of the wild-type structure and may not be as tightly bound by IRP1 as the full-size natural IRE (compare with Refs. 20 and 22). The wild-type IRE used here, however, restricts translational initiation by at least 92%. Mutations in the IRE element that reduce the binding affinity of IRP1 allow increased rates (frequencies) of ribosomal scanning through the site of the IRE (Fig. 7C). Figs. 2 and 3 therefore provide us with an indirect measure of the relationship between the thermodynamic stability of the blocking element (IRE/IRP1) and the rate at which ribosomes can proceed through a regulatory site on the 5Ј-UTR in yeast cells (Fig. 3C). This relationship is shifted considerably toward higher free energy values in higher eukaryotic cells. In the latter, the translational apparatus is capable of destabilizing a much more stable inhibitory structure/complex, albeit only when this is placed in a distal position relative to the cap structure. The spectrum of IRP1 binding affinities provided by the mutant IREs studied in this work reflects a variation in relative binding affinity similar to that induced in IRP1 by changes in iron concentration in mammalian cells. On the other hand, only a 10-fold reduction in IRP/IRE binding affinity sufficed to abolish iron-dependent regulation in rat fibroblasts (38).
The experiments with the extended and stabilized IRE mutants (IRE-s1 and IRE-s2) have demonstrated that the effect of intramolecular RNA-RNA interactions and of intermolecular RNA-protein interactions (at the same site) are effectively additive. The introduction of the additional G:C base pairs into the IRE stem stabilizes a structure that, in itself, is otherwise relatively inconspicuous in terms of the scanning ribosome. G:C base pairs are particularly effective at blocking yeast translational initiation (10). At the same time they alter the overall structure of the IRE, reducing its affinity to IRP1 (Table II). Despite this, the binding of IRP1 gives an additional repressive effect on translational initiation (Fig. 3B). The additional IRP1induced repression is approximately equivalent to the inhibitory ratio expected from an IRE-IRP1 interaction whose affinity has been reduced 10-fold relative to that of the wild-type IRE (compare with Table II, IRE-s1 in Fig. 3B, and IRE-mut2 in Fig. 3A). These data are also consistent with the model presented in Fig. 7. They indicate that the intramolecular RNA-RNA interactions and intermolecular mRNA-protein interactions act in thermodynamically and mechanistically equivalent ways with respect to the scanning ribosome. Progress through the inhibitory site requires unwinding of the RNA-RNA structure and/or disruption of the RNA⅐protein complex. In both cases, the ability of the scanning ribosome to achieve this is dictated by the free energy required to disrupt the element or complex at the given site.
The large discrepancy between higher and lower eukaryotic cells in terms of the translatability of mRNAs with structured leaders must reflect mechanistic differences in the mode of action of the respective translational apparatuses. For example, since scanning is apparently an ATP-driven process, mammalian ribosomes may be coupled to the turnover of a larger number of ATP molecules per given length of mRNA scanned than their lower eukaryotic counterparts. On the other hand, in higher eukaryotic cells, the thermodynamic driving force coupled to the initial interaction of the preinitiation complex with the 5Ј region of the mRNA is effectively weaker than that coupled to scanning. In contrast, in yeast, no distinction is apparent. An alternative way of looking at this is to assume that the association of the preinitiation complex with the 5Ј end of the mRNA is thermodynamically equivalent in higher and lower eukaryotic cells, and that the difference is in the scanning process. The greater thermodynamic driving force that is associated with higher eukaryotic scanning could feasibly be derived from the closer association of eIF4A/eIF4B with eIF4F and/or the preinitiation complex (2), which might allow more free energy from ATP hydrolysis to be brought to bear on the scanning process. The experiments described in this paper have now provided a framework for examining the mechanistic basis for these thermodynamic effects in more detail.