Introduction

The recruitment of the small ribosomal subunit as part of the 43S translation pre-initiation complex normally represents the rate-limiting step in mRNA translation (Hershey, 1991; Sachs et al., 1997). The 5' cap structure and the 3' poly(A) tail, with their respective binding proteins eIF4E and PABP/Pab1p, have been shown to play critical roles in translation initiation (Gallie, 1991; Iizuka et al., 1994; Tarun and Sachs, 1995; Preiss and Hentze, 1998). The function of the protein with which both eIF4E and PABP/Pab1p interact, eIF4G, in ribosome recruitment is less clearly defined (Hentze, 1997). eIF4G is a subunit of the cap binding complex eIF4F which includes the cap recognition factor eIF4E and the RNA-dependent ATPase eIF4A (Figure 1A, upper scheme). Stimulated by eIF4B, eIF4A is thought to unwind secondary structure in the 5' UTR of the mRNA (Rozen et al., 1990). eIF4G has a modular structure (see Figure 5A). It interacts with eIF4E (Mader et al., 1995) and PABP/Pab1p (Tarun and Sachs, 1996; Imataka et al., 1998) through its N-terminal third, thus circularizing a capped and polyadenylated mRNA (Wells et al., 1998). The central region of mammalian eIF4G bears a putative RNA recognition motif (RRM) (Goyer et al., 1993; De Gregorio et al., 1998) and binding sites for eIF4A (Imataka and Sonenberg, 1997) as well as eIF3, a multiprotein complex directly associated with small ribosomal subunits (Benne and Hershey, 1976; Lamphear et al., 1995). The C-terminal third of eIF4G harbours a binding site for the eIF4E kinase Mnk1 (Pyronnet et al., 1999; Waskiewicz et al., 1999), and a second binding site for eIF4A (Lamphear et al., 1995; Imataka and Sonenberg, 1997). The multitude of interactions involving eIF4G suggests a central function for this factor in translation.

Not surprisingly, eIF4G and its binding partners are frequent targets for translational regulation. For example, eIF4G is cleaved by the proteases L or 2A during picornaviral infection to inhibit cellular protein synthesis and to favour the translation of viral RNAs (Jackson et al., 1995; Gradi et al., 1998). Furthermore, the eIF4E binding proteins, which competitively inhibit the binding of eIF4E to eIF4G, interfere with cap-dependent translation in vivo and in vitro (Haghighat et al., 1995; Altmann et al., 1997). Recently, it was also shown that the eIF4G–eIF3 interaction is controlled by the IRE/IRP system that regulates the translation of ferritin and other IRE-containing mRNAs: IRP-1 binding to an IRE located in the proximity of the cap structure blocks the productive assembly of translation initiation complexes beyond the step of eIF4F binding (Muckenthaler et al., 1998).

The wealth of biochemical data and in vitro functional assays clearly suggest an important role for eIF4G in translation. However, little was known about the exact role of eIF4G in translation in the context of a living cell, where mRNAs compete for limiting translation initiation factors. To address the role of eIF4G in vivo, we adopted a tethered function approach in transfected HeLa cells: the N-terminal eIF4E- and PABP binding region of human eIF4GI was replaced by a specific RNA binding protein (RBP) and directed to a cognate binding site (B.S.) in the intercistronic space of a bicistronic reporter mRNA (Figure 1A, lower scheme). By assaying the translation of the second cistron, which is normally translated very poorly, this approach permits assessment of eIF4G function independent of eIF4E- and PABP binding and in the context of wild-type cells.

Here we show that a region of human eIF4GI (amino acids 642–1560) fused to the RNA binding protein IRP-1 (IRP–4G) suffices to direct translation of the downstream cistron of a bicistronic mRNA bearing an iron-responsive element (IRE) in the intercistronic space. The position and number of IREs modulate this effect. C-terminal deletions of eIF4GI demonstrate that the central region of eIF4GI (amino acids 642–1091) represents the critical core as further deletion into the eIF4A/eIF3 binding region on eIF4GI renders the fusion protein (amino acids 642–877) functionally inactive. Activation of translation of the downstream cistron is preserved when translation of the upstream cistron is inhibited by means of a stem–loop structure placed in the 5'UTR. The activity of the 4G/IRE–IRP module in driving translation in transfected HeLa cells ranges within an order of magnitude of the potency of natural viral internal ribosome entry sites (IRES).

Results

IRP–4G mediates productive in vivo translation of the downstream cistron of a bicistronic mRNA in a binding-site-specific manner

To assay the effect of tethering eIF4G domains to a reporter mRNA in vivo, we constructed two classes of plasmid vectors: the first, termed effector plasmids, express human iron regulatory protein-1 alone (pSGIRP), or fused to parts of human eIF4GI (depicted in Figure 5A); the second, termed reporter plasmids, produces bicistronic mRNAs coding for the reporter proteins luciferase (LUC, upstream) and chloramphenicol acetyltransferase (CAT, downstream). Binding sites for IRP-1 (iron-responsive elements, IREs) or for the bacteriophage MS2 coat protein as a negative control (MSC) were introduced into the intercistronic space (Figure 1A, intercistronic sequences are specified in Table I). These plasmids were co-transfected into HeLa cells together with a third plasmid expressing -galactosidase (-Gal), which served as a transfection efficiency control.

CAT and LUC activities were corrected for transfection efficiencies and the data expressed in relation to specificity controls in two ways: first, in relation to a transfection with the 'empty' effector plasmid pSG5 (termed 'relative CAT or LUC expression'; relative CAT expression: bar graphs, relative LUC expression: numbers below the histogram in Figure 1B); second, as a ratio between the expression of IRP-1 alone versus the IRP–4G fusion protein (CAT: white numbers inside the histograms, LUC: bold numbers below). Both methods of evaluation yielded consistent results. As expected, cotransfection with IRP-1 alone does not elicit CAT activity (white bars). By contrast, IRP–4G activates CAT expression 3- to 5-fold when an IRE mediates its binding to the intercistronic space. Importantly, it fails to do so when the IRE is replaced by the MSC site (Figure 1B) which binds neither IRP-1 nor IRP–4G. A further activation in CAT expression is observed when the bicistronic mRNA contains three IREs instead of one (Figure 1B, see Table I and scheme in Figure 4). The effect of the IRP–4G fusion protein is specific for the downstream cistron, because any changes in LUC expression are only minor (1.2- to 1.6-fold) and independent of the binding sites in the intercistronic region (Figure 1B). Inhibition of LUC translation by the effector plasmids was never observed (data not shown).

Metabolic labelling followed by immunoprecipitation with antisera directed against IRP-1 (IRP, Figure 2A) shows that the expression levels of IRP-1 and IRP–4G in cells transfected with the corresponding effector plasmids are similar (lanes 4 and 5). Gel mobility shift analysis (Figure 2B) demonstrates that overexpressed IRP-1 and IRP–4G both bind to an IRE probe (lanes 1–10) but not to an MSC probe (lanes 11–20). The lower mobility of complexes formed with IRP–4G (lanes 8–10) compared with IRP-1 (lanes 5–7) indicates that the IRP–4G fusion protein binds to the IRE probe. This was confirmed using specific antisera (Figure 2C). Incubation with IRP disrupts the complexes formed with extracts containing either overexpressed IRP-1 (lane 7) or IRP–4G (lane 4), while antisera directed against human eIF4GI (4G) specifically 'supershift' the complex with overexpressed IRP–4G but not IRP alone (compare lanes 3 and 6). We also assessed the steady-state levels and integrity of the reporter mRNAs in the cotransfection experiments by Northern analysis (Figure 2D). Overexpression of IRP-1 (lane 3) or IRP–4G (lane 4) does not affect the level of the IRE reporter mRNAs. The same was found for the other reporter mRNAs used in this study (data not shown). A formal possibility was that monocistronic (uncapped) cleavage products could contribute to CAT protein expression. Such shorter products of >1 kb were undetectable even on long autoradiographic exposures (Figure 2D), or when the analysis was extended to polysome-associated 3IRE mRNA (data not shown).

The data presented in Figures 1 and 2 collectively demonstrate that the IRP–4G fusion protein specifically activates the translation of the downstream CAT cistron in a binding-site-dependent manner. To exclude that our choice of the bulky IRP-1 (98 kDa) as an RNA binding domain could affect the binding of physiological interactors to the eIF4G moiety of the fusion protein and to establish the generality of these findings, we replaced IRP-1 by the short arginine-rich N-terminal domain (amino acids 1–22) of the bacteriophage transcriptional antiterminator protein N [N-(1–22)] (Tan and Frankel, 1995) (Figure 3A). The intercistronic IRE was replaced by the cognate hairpin, called boxB (Figure 3A). This complete exchange of components constituting the tether for eIF4G yielded site-specific activation of CAT translation of similar or even slightly higher potency (Figure 3B). Thus the effect of appropriate eIF4G fusion proteins on the downstream cistron translation reflects authentic functions of eIF4G. Furthermore, this experiment demonstrates the suitability of the N-(1–22)/ boxB couple for tethering approaches in heterologous systems.

IRP–4G-mediated translational activation as a function of the position and number of binding sites in the intercistronic region

We next examined whether the exact position of the IRE to which the IRP–4G fusion protein binds affects the function of this protein. Increasing the distance of the IRE from the upstream cistron by 66 nucleotides (66–IRE) from 70 to 136 nucleotides yields a modest, reproducible improvement in CAT activation (Figure 4A). Reducing the distance between the IRE and the translation initiation codon of the downstream cistron from 46 to 13 nucleotides (IRE33) has a negative effect. We did not assess whether translation initiation in IRE33 was repositioned to AUG codons downstream of the authentic CAT start codon. The putative products of such initiation (N-terminally truncated or out of frame) are not expected to be detected by the CAT-ELISA. Even if the optimal position of the binding site remains to be determined, these data show: (i) that specific activation of translation is observed from differently positioned sites, and (ii) that the positioning and context of the IRP–4G fusion protein within the intercistronic space exert quantitative effects on the translation of the downstream cistron.

In the light of these data and the finding that the 3IRE reporter plasmid yields more potent CAT activation (Figure 1B), we determined whether the latter is caused by the multiplicity of binding sites or an improved positioning of one of them. Derivatives of the 3IRE construct were generated (Figure 4B), which contain in each IRE position either a wild-type IRE or an IRE with a single nucleotide deletion () that reduces its affinity for IRP-1 (Goossen et al., 1990). Mutations of the two upstream (2+1) or the two downstream (1+2) IREs result in markedly weaker activation of CAT expression compared with the wild-type 3IRE construct. The CAT activities of the 2+1 and 1+2 constructs approximate those of the single-IRE constructs 66-IRE and IRE (compare Figure 4A and B), suggesting that increasing the distance between functional IRE and either the LUC (66-IRE/2+1 versus IRE) or the CAT (1+2 versus IRE) cistron only leads to marginal changes of CAT activation. Hence, the presence of multiple binding sites in 3IRE appears to explain better the more potent CAT activation seen with this construct compared with any of the intercistronic regions harbouring a single IRE in various different positions. As expected, mutation of all three IREs in construct 3 profoundly decreases, but does not completely abolish CAT activation. The small but reproducible effect mediated through the 3 sites most likely reflects the residual IRP-1 binding to the mutated IREs (Goossen et al., 1990).

The central third of eIF4GI functions as a 'ribosome recruitment core' in vivo

To map the functional core of eIF4G that directs downstream translation, C-terminal deletions (Figure 5A) were assayed. IRP–4G1091 lacks the C-terminal of the two eIF4A binding sites but retains the central eIF4A binding site, the putative RRM and the binding region for eIF3. The fusion protein comprising IRP-1 and the central third of eIF4GI still activates CAT translation, albeit with lower (60%) efficiency (Figure 5B). A further deletion of eIF4GI sequences (IRP–4G877), which preserves the putative RRM, results in a complete loss of activity (Figure 5B). This excludes the possibility that the activation of translation by the core region of eIF4G is merely due to the presence of the RRM motif, a conclusion that is also supported by the stringent binding site specificity for the translation of the downstream cistron (Figures 1 and 4). As further controls, we confirmed that both IRP–4G1091 and IRP–4G877 fusion proteins bind an IRE probe (Figure 6A) but not the unrelated MSC control (data not shown). The level of complex formation roughly mirrors the steady-state levels of the overexpressed proteins as determined by IRP-1 immunoblot (Figure 6B). This suggests that RNA binding by the IRP-1 moiety of the fusion is unaffected by adding or removing various eIF4GI sequences (Figures 2, 6A and B).

The immunoblot shown in Figure 5C demonstrates that the IRP–4G and IRP–4G1091 fusion proteins are expressed at comparable levels (lanes 3 and 4) when equal amounts of effector plasmids were transfected, whereas IRP–4G877 is less well expressed (lane 5). This prompted us to titrate the amount of effector plasmids used for transfection and to examine translational activation (Figure 6D) as a function of effector protein levels, measured by immunoblotting (Figure 6C). IRP–4G877 failed to stimulate CAT translation at any DNA concentration tested (Figure 6D), even when IRP–4G877 fusion protein levels approached and exceeded those of IRP–4G (Figure 6C, compare transfection of 2 pmol of pSGIRP-4G877 with 0.5 pmol of pSGIRP-4G). Serendipitously, this experiment also revealed that transfection of 0.5–1.0 pmol pSGIRP-4G plasmid yields a further substantial increase in downstream cistron translation over the previously used 2.0 pmol (Figure 6D). Therefore, 1.0 pmol of pSGIRP-4G were used in the experiments shown in Figures 8 and 9.

The data shown in Figures 5 and 6 demonstrate that the central third of eIF4GI is sufficient to direct translation in vivo with an additional contribution by the C-terminus. Further deletion into this core region, as in IRP–4G877, yields a functionally inactive protein. To correlate these findings with the binding of the relevant interaction partners, we performed co-immunoprecipitation assays with immobilized IRP-1 from lysates of cells transfected with all different variants of the effector plasmid (Figure 7). The material recovered from the affinity matrix was subjected to SDS–PAGE and immunoblotting. Probing with IRP-1 confirms that the antisera precipitate the endogenous IRP-1 protein as well as all overexpressed proteins by means of their IRP-1 moiety (Figure 7, upper panel). Reprobing with antibodies directed against human eIF4A (eIF4A) (lower panel) demonstrates the strong and specific interaction of this factor with the long IRP–4G fusion protein. Deletion of the C-terminal eIF4A binding site in IRP–4G1091 strongly reduced the co-immunoprecipitation of eIF4A, while IRP–4G877 allows no detectable co-precipitation of eIF4A (lane 5). Using the polyclonal anti-eIF3 antisera available to us we failed to obtain interpretable results. Thus, the co-immunoprecipitation analysis revealed correlation between the occurrence and magnitude of translational activation by the IRP–4G fusion proteins and their ability and strength of eIF4A binding.

IRP–4G-mediated translational activation in comparison with IRES-mediated and cap-mediated translation

Internal ribosome entry sites that are present within some viral and cellular mRNAs are the only known RNA elements that can drive the translation of a downstream cistron of bicistronic mRNAs (Jang et al.,1988; Pelletier and Sonenberg, 1988). To compare the relative strength of IRP–4G-mediated activation with the potency of a natural IRES, the full-length IRES from hepatitis C virus (Reynolds et al., 1995) including the sequence downstream from the translation start codon (see Materials and methods), was cloned between the LUC and CAT cistrons of our reporter plasmid (HCV, Figure 8A). We chose this IRES, because it is highly active in transfection experiments using several different cell lines including HeLa cells, where it displays similar potency to the other intensively studied IRESes from polio and encephalomyocarditis virus (Borman et al., 1997b). The IRP–4G/3IRE module is 10–15% as efficient as the HCV construct in directing CAT expression under the same conditions (Figure 8B), while luciferase activity was similar in both cases (after correction for transfection efficiency, data not shown).

The cap-dependent translation of the upstream LUC cistron serves as another point of reference for IRP–4G-mediated translation of the downstream CAT cistron. Thus, we determined the absolute quantities of LUC and CAT proteins in cell extracts from the experiment shown in Figure 6C and D using appropriate LUC and CAT protein calibration curves (see Table II and Materials and methods). From these measurements, we estimate that IRP–4G-mediated CAT translation is 5% as efficient as the cap-dependent translation of luciferase (Table II, cotransfection of pSG3IRE with 0.5 or 1 pmol of pSGIRP-4G).

We conclude from these quantitative comparisons that the translation of the downstream cistron driven by the IRP–4G is significant in comparison with a potent IRES and to cap-mediated translation in living cells.

IRP–4G-mediated translational activation is preserved even when translation of the 5' cistron is inhibited

IRES recruit ribosomes internally, independently of the translation of upstream cistrons (Chen and Sarnow, 1995). We finally wanted to address the mechanism of IRP–4G-directed ribosome recruitment. To block the translation of the upstream LUC cistron, a stable stem–loop structure (G = -243 kJ/mol; Kozak, 1986) was inserted into the 5' UTR of 3IRE (Figure 9, right panel). While this insertion represses LUC expression very efficiently (96–97%, Figure 9, right panel), it affects the translational activation of CAT by IRP–4G only by about one third (Figure 9, left panel). This result suggests that, while some of the ribosomes may be re-recruited from the upstream LUC reading frame, ribosomes translating the CAT coding region can be recruited from the free 43S complex pool.

Discussion

Biochemistry, yeast genetics and cell-free translation assays have helped to formulate models of the translation initiation pathway which suggest an important role of eIF4G in the recruitment of the small ribosomal subunit (e.g. Lamphear et al., 1995; Pestova et al., 1996; Tarun and Sachs, 1996; Tarun et al., 1997; Imataka et al., 1998). Information on the proteins that interact directly with eIF4G and their respective roles in translation has allowed the proposal that eIF4G serves as a scaffold or adapter protein between the cap structure and the poly(A) tail of mRNAs on the one hand and the 40S ribosomal subunit on the other (Hentze, 1997; Morley et al., 1997; Sachs et al., 1997). The inhibitory effect of the 4E binding proteins on translation by interference with the eIF4E–eIF4G interaction as well as the negative effect of picornaviral proteolytic cleavage of eIF4G on the translation of host cell mRNAs have underscored the central role of eIF4G in translation initiation (Haghighat et al., 1995; Altmann et al., 1997; Gradi et al., 1998). Moreover, C-terminal fragments of human eIF4G have been shown to be able to stimulate the translation of uncapped mRNAs in a cell-free system by an unknown mechanism (Ohlmann et al., 1996; Borman et al., 1997a; De Gregorio et al., 1998) and to suffice to mediate 40S subunit binding to the IRES of encephalomyocarditis virus (EMCV) (Pestova et al., 1996). Similarly, internal initiation on bicistronic mRNAs lacking conventional IRES elements was stimulated by addition of purified eIF4F in vitro (Anthony and Merrick, 1991). All these activities possibly involve the known general RNA binding capacity of eIF4G, which may also serve to increase binding of eIF4E to the capped end of mRNAs (Haghighat and Sonenberg, 1997) thus contributing to efficient initiation of translation in a competitive cellular environment. In cell-free systems where mRNAs are limiting, the RNA binding capacity of eIF4G may function independently, while the EMCV-IRES may have evolved to engage this functionality of eIF4G with particular efficiency.

eIF4G, like several other initiation factors, appears to be required to mediate ribosome recruitment via the cap structure and the poly(A) tail both in vivo and in vitro (Tarun et al., 1997; Gradi et al., 1998). However, the information available did not permit the prediction of whether productive translation would ensue as a consequence of eIF4G binding to an mRNA, particularly in living cells. We asked whether eIF4G binding would suffice to recruit all necessary 'downstream' translation factors not only to elicit the binding of a functional 43S translation pre-initiation complex to an mRNA, but also to allow the positioning of this complex at a downstream translation initiation codon, and to support the subsequent joining of the 60S subunit for productive translation. To address this question, we devised the bicistronic mRNA and tethered function approach depicted in Figure 1A. This strategy permits examination of the function of eIF4G (and subdomains thereof) in the context of a wild-type cell and without perturbing general translation.

Insights into translation initiation

Our results identify the central region of eIF4GI as an active 'ribosome recruitment core' which requires no more than a means to bind upstream of an open reading frame to recruit all additional factors necessary for at least basal translation in vivo (Figures 1, 3 and 5). Reconstitution experiments with pure translation initiation factors demonstrated that this core region of eIF4GI can bind directly to the EMCV-IRES (Pestova et al., 1996; Kolupaeva et al., 1998) and promote 40S ribosomal subunit binding in vitro (Pestova et al., 1996). However, it was not determined whether this pre-initation complex represented a functional initiation intermediate. Moreover, mutations in the EMCV-IRES have been described which disrupt IRES function (Witherell et al., 1995) without affecting eIF4G binding in vitro (Kolupaeva et al., 1998), indicating that eIF4G binding is not sufficient for translation from the EMCV IRES.

The ribosome recruitment core region of eIF4GI defined in this study retains a putative RRM (Goyer et al., 1993; De Gregorio et al., 1998) and interaction sites for eIF4A (Imataka and Sonenberg, 1997) and eIF3 (Lamphear et al., 1995). Both eIF4A and eIF3 have been assigned central roles in current models of translation initiation (see introduction) and likely represent essential components of a minimal productive initiation complex for cellular mRNAs. In support of this, we show here that the ability of the IRP–4G fusion proteins to bind eIF4A correlates with translational activation. Technical limitations precluded us from ascertaining whether the same is true for eIF3; however, the inactive IRP–4G877 lacks part of a region previously found to be required for both eIF3 and eIF4A binding in vivo (Imataka and Sonenberg, 1997), allowing the prediction that IRP–4G877 does not bind eIF3 either. Refined mutational studies will be required to address possible individual contributions of eIF3- and eIF4A binding to the translational activity of the eIF4G core region. The C-terminal third of eIF4G harbouring the second eIF4A binding site is able to stimulate the function of the core region, and might thus function as an enhancer or regulatory domain.

IRP–4G-mediated translation of the downstream cistron is preserved even when translation via the 5' end is inhibited (Figure 9). This suggests that ribosomes can be recruited from the free 43S pool by the IRE/IRP–4G module, which acts as an 'IRES by design'. How potent is this module in mediating translation in living cells? We have addressed this issue directly in comparison with cap-mediated (Table II) and IRES-mediated (Figure 8) translation on the one hand, and 'background' termination–reinitiation on the other (Table II). As expected, the background translation of the downstream cistron is low in the absence of a functional IRE/IRP–4G module (0.7% of the translation of the upstream cistron; Table II). By contrast, IRE/IRP–4G-mediated translation is 10–15% as efficient as the HCV-IRES, and still 5% as efficient as cap-dependent translation. We consider these values to be highly significant. However, it is also important to ask which aspects of our experimental conditions currently limit the expression of the CAT cistron to these values. Future experiments with the IRE/IRP–4G module will attempt to define the optimal spatial positioning of the 'ribosome recruitment core' in relation to the translation initiation site and to elucidate whether the 'natural' non-covalent binding of eIF4E and PABP to eIF4G is superior to the covalent attachment within the IRP–4G fusion protein. Such experiments may further improve the activity of eIF4G fusion protein-driven translation of the downstream cistron and, importantly, shed light on the role of 'scanning'. It will also be interesting to introduce binding sites for eIF4E and PABP into the IRE/IRP–4G module to assay for potential conformational effects of eIF4E binding to eIF4G or a contribution of the poly(A) tail to the translation of the downstream cistron. Following refinement, we envisage practical utility of our findings in the design of expression vectors that allow control of gene expression at the translational level. Moreover, the sufficiency of eIF4G binding for productive translation of a downstream cistron can be utilized to examine RNA–protein and protein–protein interactions.

Common themes in translation and transcription

The role of the 'ribosome recruitment core' in translation is reminiscent of the function of transcription activation domains, which—when fused to DNA binding proteins—can drive transcription from promoters bearing (multiple) suitable binding sites. This highlights an analogy between these elementary steps in gene expression (Sachs and Buratowski, 1997).

In translation, as well as transcription, cells exploit several strategies to form productive initiation complexes on the nucleic acid templates. The cap structure and the poly(A) tail of a typical cellular mRNA could be viewed as fulfilling a similar function to the IRE in our experiments, while eIF4E and PABP would physiologically play the role adopted by the IRP portion of the IRP–4G fusion protein: to cooperate in the recruitment of eIF4G to the mRNA (Imataka et al., 1998; Preiss and Hentze, 1998; Wells et al., 1998). During translation of rotavirus mRNAs, which are capped but not polyadenylated, the function of PABP appears to be replaced by the viral protein NSP3, which binds to a specific site in the 3' UTR and interacts with eIF4G (Piron et al., 1998). Thus, the various strategies that have evolved to recruit eIF4G to the mRNA could all serve similar purposes.

In summary, eIF4G has long been known to be involved in translation initiation. The most significant advance of the work reported here is the demonstration—in living cells—that eIF4G binding suffices to initiate all downstream events for productive mRNA translation. This finding itself as well as the technical approach used here should pave the way for a better understanding of the translation initiation pathway in mammalian cells, complementing in vitro and genetic approaches.

Materials and methods

Plasmid constructs

Effector plasmids: human eIF4GI cDNA sequences were excised from p220Bam (De Gregorio et al., 1998) with BamHI–HindIII and inserted in-frame behind the IRP-1 coding region, into the MscI–HindIII sites of pSGIRP (Kollmus et al., 1996) to create plasmid pSGIRP-4G. A BamHI–EcoRI fragment from pGEX4G935 (De Gregorio et al., 1998) or a BamHI–NsiI fragment from p220Bam were inserted into the MscI site of pSGIRP to create plasmids pSGIRP-4G1091 and pSGIRP-4G877, respectively. A synthetic DNA fragment coding for the -N-(1–22) peptide (Tan and Frankel, 1995) was cloned in the SacI–BamHI sites of p220Bam to generate p-4G. A SacI–XhoI fragment from p-4G was inserted into the SacI–XhoI sites of pSG5 to generate pSG-4G. Reporter plasmids: p4IREGH, p3+1IREGH and p2+2IREGH (Goossen et al., 1990) were amplified by PCR using the primers: 35: GACGGATCCAAAAAATAC and 33: GGTCTAGAACTCTAGCGTCCAAGCAC; the amplification products were digested with BamHI–XbaI and cloned into the BamHI–XbaI sites of pIRE.CAT (Preiss et al., 1998) creating p3CAT, p2+1CAT and p1+2CAT. SacI–HindIII fragments from pIRE.CAT, pIREscp.CAT (Preiss et al., 1998), pIRE100 (Paraskeva et al., 1999), pMSC-CAT (Stripecke and Hentze, 1992), p3CAT, p2+1CAT p1+2CAT or a BamHI–HindIII fragment from p4IRE-CAT were ligated into the XhoI sites of pSGluc.trs3 (Pantopoulos and Hentze, 1995) to generate the reporter plasmids: pSGIRE, pSGIRE33, pSG66-IRE, pSGMSC, pSG3, pSG2+1, pSG1+2 and pSG3IRE (see Table I for sequences of all intercistronic regions). A synthetic DNA fragment corresponding to the boxB site was inserted into the BamHI–XbaI sites of pIRE.CAT to generate pB-CAT. A SacI–HindIII fragment from pB-CAT was cloned into the XhoI sites of pSGluc.trs3 to generate pSGboxB (see Table I for the sequence of the intercistronic region). pSGG243 was created by inserting a synthetic fragment of dsDNA into the AvrII site of pSG3IRE (Figure 9). The HCV-IRES element (nucleotides 40–373, including the 5'UTR plus 29 nucleotides of the coding sequence) was amplified by PCR and cloned into the BamHI–XbaI sites of pIRE.CAT creating pHCV-CAT. A KpnI–HindIII fragment of pHCV-CAT was inserted into the XhoI sites of pSGluc.trs3 to generate pSGHCV (see Figure 8A). All plasmid manipulations were verified by sequencing.

Cell transfection and protein analysis

HeLa cells were transiently transfected by the calcium phosphate method using 2 pmol of effector plasmids (unless indicated otherwise), 2 pmol of bicistronic vector and 1 pmol of pCMV expressing -Gal. CAT assays were performed using a CAT-ELISA kit (Boehringer Mannheim) following the manufacturer's instructions. Luciferase and -Gal expression were measured by enzymic activity. CAT and LUC calibration curves were obtained using CAT (Boehringer Mannheim) and luciferase (Promega) standards. Cell lysis for mobility shift analysis was carried out as previously described (Pantopoulos and Hentze, 1995). Binding reactions were performed in the presence of 4 mg/ml heparin at room temperature and samples were resolved as described (Pantopoulos and Hentze, 1995).

Proteins were analysed by metabolic labelling followed by immunoprecipitation as described, using polyclonal antisera against recombinant human IRP-1 (Pantopoulos and Hentze, 1995).

For co-immunoprecipitation analysis, cells were lysed by incubation in 800 l of TMGK buffer (20 mM Tris pH 8, 20 mM MgCl₂, 110 mM KCl, 1% Triton X-100, 1 mM PMSF) for 30 min at 4°C. Cellular debris was pelleted by centrifugation at top speed in a microfuge. Supernatants were incubated 1 h at 4°C with antisera against human IRP-1 immobilized on protein A–Sepharose beads with 20 mM dimethylpimelimidate as described (Harlow and Lane, 1988). After incubation the beads were washed three times with TMGK buffer and processed for SDS–PAGE as described (Pantopoulos and Hentze, 1995).

For Western blotting proteins were transferred to a reinforced cellulose nitrate membrane. After blocking for 1 h in TTBS (20 mM Tris pH 7.5, 150 mM NaCl, 0.1% Tween 20) containing 5% dried milk, membranes were incubated with the appropriate dilution of IRP (1/500) or eIF4A (1/50) in TTBS for 1 h at room temperature. Then they were washed four times in TTBS and incubated with a 1/5000 dilution of anti-rabbit or anti-mouse IgG coupled to horseradish peroxidase (Amersham) and subjected to ECL detection (Amersham). Membranes were stripped of the antibodies by incubation in 2% SDS, 68 mM Tris pH 6.7, 100 mM 2-mercaptoethanol for 30 min at 55°C.

Acknowledgements

We thank Martina Muckenthaler, Antje Ostareck-Lederer and Katharina Weber for discussions and experimental help. We thank H.Trachsel for eIF4A. This work was supported by the European Commission Biotechnology Programme (BIO4-CT95-0045).

References

Altmann M, Schmitz N, Berset C and Trachsel H (1997) A novel inhibitor of cap-dependent translation initiation in yeast: p20 competes with eIF4G for binding to eIF4E. EMBO J, 16, 1114–1121. | Article | PubMed | ISI | ChemPort |

Anthony DA and Merrick WC (1991) Eukaryotic initiation factor (eIF)-4F. Implications for a role in internal initiation of translation. J Biol Chem, 266, 10218–10226. | PubMed | ChemPort |

Benne R and Hershey JWB (1976) Purification and characterization of initiation factor IF-E3 from rabbit reticulocytes. Proc Natl Acad Sci USA, 73, 3005–3009. | PubMed | ChemPort |

Borman AM, Kirchweger R, Ziegler E, Rhoads RE, Skern T and Kean KM (1997a) eIF4G and its proteolytic cleavage products: effect on initiation of protein synthesis from capped, uncapped and IRES-containing mRNAs. RNA, 3, 186–196. | PubMed | ChemPort |

Borman AM, Le Mercier P, Girard M and Kean KM (1997b) Comparison of picornaviral IRES-driven internal initiation of translation in cultured cells of different origins. Nucleic Acids Res, 25, 925–932. | PubMed | ChemPort |

Chen C and Sarnow P (1995) Initiation of protein synthesis by the eukaryotic translational apparatus on circular RNAs. Science, 268, 415–517. | PubMed | ISI | ChemPort |

De Gregorio E, Preiss T and Hentze MW (1998) Translational activation of uncapped mRNAs by the central part of human eIF4G is 5' end-dependent. RNA, 4, 828–836. | Article | PubMed | ChemPort |

Gallie D (1991) The cap and poly(A) tail function synergistically to regulate mRNA translational efficiency. Genes Dev, 5, 2108–2116. | PubMed | ISI | ChemPort |

Goossen B, Caughman SW, Harford JB, Klausner RD and Hentze MW (1990) Translational repression by a complex between the iron-responsive element of ferritin mRNA and its specific cytoplasmic binding protein is position-dependent in vivo. EMBO J, 9, 4127–4133. | PubMed | ChemPort |

Goyer C, Altmann M, Lee HS, Blanc A, Deshmukh M, Woolford JJL, Trachsel H and Sonenberg N (1993) TIF4631 and TIF4632: two yeast genes encoding the high-molecular-weight subunits of the cap-binding protein complex (eukaryotic initiation factor 4F) contain an RNA recognition motif-like sequence and carry out an essential function. Mol Cell Biol, 13, 4860–4874. | PubMed | ISI | ChemPort |

Gradi A, Svitkin YV, Imataka H and Sonenberg N (1998) Proteolysis of human eukaryotic initiation factor 4GII, but not eIF4GI, coincides with the shutoff of host protein synthesis after poliovirus infection. Proc Natl Acad Sci USA, 95, 11089–11094. | Article | PubMed | ChemPort |

Haghighat A, Mader S, Pause A and Sonenberg N (1995) Repression of cap-dependent translation by 4E-binding protein 1: competition with p220 for binding to eukaryotic initiation factor 4E. EMBO J, 14, 5701–5709. | PubMed | ISI | ChemPort |

Haghighat A and Sonenberg N (1997) eIF4G dramatically enhances the binding of eIF4E to the mRNA 5'-cap structure. J Biol Chem, 272, 21677–21680. | Article | PubMed | ISI | ChemPort |

Harlow E and Lane D (1988) Antibodies: A Laboratory Manual. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY.

Hentze MW (1997) eIF4G: a multipurpose ribosome adapter? Science, 275, 500–501. | Article | PubMed | ISI | ChemPort |

Hershey JWB (1991) Translational control in mammalian cells. Annu Rev Biochem, 60, 717–755. | Article | PubMed | ISI | ChemPort |

Iizuka N, Najita L, Franzusoff A and Sarnow P (1994) Cap-dependent and cap-independent translation by internal initiation of mRNAs in cell extracts prepared from saccharomyces cerevisiae. Mol Cell Biol, 14, 7322–7330. | PubMed | ISI | ChemPort |

Imataka H and Sonenberg N (1997) Human eukaryotic translation initiation factor 4G (eIF4G) possesses two independent binding sites for eIF4A. Mol Cell Biol, 17, 6940–6947. | PubMed | ISI | ChemPort |

Imataka H, Gradi A and Sonenberg N (1998) A newly identified N-terminal amino acid sequence of human eIF4G binds poly(A)-binding protein and functions in poly(A)-dependent translation. EMBO J, 17, 7480–7489. | Article | PubMed | ChemPort |

Jackson RJ, Hunt SL, Reynolds JE and Kaminski A (1995) Cap-dependent and cap-independent translation: operational distinctions and mechanistic interpretations. Curr Top Microbiol Immunol, 203, 1–29. | PubMed | ChemPort |

Jang SK, Kräusslich H-G, Nicklin MJH, Duke GM, Palmenberg AC and Wimmer E (1988) A segment of the 5' nontranslated region of encephalomyocarditis virus RNA directs internal entry of ribosomes during in vitro translation. J Virol, 62, 2636–2643. | PubMed | ISI | ChemPort |

Kollmus H, Hentze MW and Hauser H (1996) Regulated ribosomal frameshifting by an RNA-protein interaction. RNA, 2, 316–323. | PubMed | ChemPort |

Kolupaeva VG, Pestova TV, Hellen CU and Shatsky IN (1998) Translation eukaryotic initiation factor 4G recognizes a specific structural element within the internal ribosome entry site of encephalomyocarditis virus RNA. J Biol Chem, 273, 18599–18604. | Article | PubMed | ChemPort |

Kozak M (1986) Influences of mRNA secondary structure on initiation by eukaryotic ribosomes. Proc Natl Acad Sci USA, 83, 2850–2854. | PubMed | ChemPort |

Lamphear BJ, Yan R, Yang F, Waters D, Liebig HD, Klump H, Kuechler E, Skern T and Rhoads RE (1993) Mapping the cleavage site in protein synthesis initiation factor eIF-4 of the 2A proteases from human coxsackievirus and rhinovirus. J Biol Chem, 268, 19200–19203. | PubMed | ISI | ChemPort |

Lamphear BJ, Kirchweger R, Skern T and Rhoads RE (1995) Mapping of functional domains in eukaryotic protein synthesis initiation factor 4G (eIF4G) with picornaviral proteases. J Biol Chem, 270, 21975–21983. | Article | PubMed | ISI | ChemPort |

Mader S, Lee H, Pause A and Sonenberg N (1995) The translation initiation factor eIF-4E binds to a common motif shared by the translation factor eIF-4 and the translational repressors 4E-binding proteins. Mol Cell Biol, 15, 4990–4997. | PubMed | ISI | ChemPort |

Morley SJ, Curtis PS and Pain VM (1997) eIF4G: translation's mystery factor begins to yield its secrets. RNA, 10, 1085–1104.

Muckenthaler M, Gray NK and Hentze MW (1998) IRP-1 binding to ferritin mRNA prevents the recruitment of the small ribosomal subunit by the cap-binding complex eIF4F. Mol Cell, 2, 383–388. | PubMed | ISI | ChemPort |

Ohlmann T, Rau M, Pain VM and Morley SJ (1996) The C-terminal domain of eukaryotic protein synthesis initiation factor (eIF) 4G is sufficient to support cap-independent translation in the absence of eIF4E. EMBO J, 15, 1371–1382. | PubMed | ISI | ChemPort |

Pantopoulos K and Hentze MW (1995) Nitric oxide signaling to iron-regulatory protein (IRP): Direct control of ferritin mRNA translation and transferrin receptor mRNA stability in transfected fibroblasts. Proc Natl Acad Sci USA, 92, 1267–1271. | PubMed | ChemPort |

Paraskeva E, Gray NK, Schläger B, Wehr K and Hentze MW (1999) Ribosomal pausing and scanning arrest as mechanisms of translational regulation from cap-distal iron-responsive elements. Mol Cell Biol, 19, 807–816. | PubMed | ChemPort |

Pelletier J and Sonenberg N (1988) Internal initiation of translation of eukaryotic mRNA directed by a sequence derived from poliovirus RNA. Nature, 334, 320–325. | Article | PubMed | ISI | ChemPort |

Pestova TV, Shatsky IN and Hellen CUT (1996) Functional dissection of eukaryotic initiation factor 4F: the 4A subunit and the central domain of 4G subunit are sufficient to mediate internal entry of 43S preinitiation complexes. Mol Cell Biol, 16, 6870–6878. | PubMed | ISI | ChemPort |

Piron M, Vende P, Cohen J and Poncet D (1998) Rotavirus RNA-binding protein NSP3 interacts with eIF4G and evicts the poly(A) binding protein from eIF4F. EMBO J, 17, 5811–5821. | Article | PubMed | ChemPort |

Preiss T and Hentze MW (1998) Dual function of the messenger RNA cap structure in poly(A)-tail-promoted translation in yeast. Nature, 392, 516–520. | Article | PubMed | ISI | ChemPort |

Preiss T, Muckenthaler M and Hentze MW (1998) Poly(A)-tail-promoted translation in yeast: implications for translational control. RNA, 4, 1321–1331. | Article | PubMed | ChemPort |

Pyronnet S, Imataka H, Gingras A-C, Fukunaga R, Hunter T and Sonenberg N (1999) Human eukaryotic translation initiation factor 4G (eIF4G) recruits Mnk1 to phosphorylate eIF4E. EMBO J, 18, 270–279. | Article | PubMed | ChemPort |

Reynolds JE, Kaminski A, Kettinen HJ, Grace K, Clarke BE, Carroll AR, Rowlands DJ and Jackson RJ (1995) Unique features of internal initiation of hepatitis C virus RNA translation. EMBO J, 14, 6010–6020. | PubMed | ISI | ChemPort |

Rozen F, Edery I, Meerovitch K, Dever TE, Merrick WC and Sonenberg N (1990) Bidirectional RNA helicase activity of eucaryotic translation initiation factors 4A and 4F. Mol Cell Biol, 10, 1134–1144. | PubMed | ChemPort |

Sachs AB and Buratowski S (1997) Common themes in translational and transcriptional regulation. Trends Biochem Sci, 22, 189–192. | Article | PubMed | ChemPort |

Sachs AB, Sarnow P and Hentze MW (1997) Starting at the beginning, middle, and end: translation initiation in eukaryotes. Cell, 89, 831–838. | PubMed | ISI | ChemPort |

Stripecke R and Hentze MW (1992) Bacteriophage and spliceosomal proteins function as position-dependent cis/trans repressors of mRNA translation in vitro. Nucleic Acids Res, 20, 5555–5564. | PubMed | ChemPort |

Tan R and Frankel AD (1995) Structural variety of arginine-rich RNA-binding peptides. Proc Natl Acad Sci USA, 92, 5282–5286. | PubMed | ChemPort |

Tarun SZ,Jr and Sachs AB (1995) A common function for mRNA 5' and 3' ends in translation initiation in yeast. Genes Dev, 9, 2997–3007. | PubMed | ISI | ChemPort |

Tarun SZ,Jr and Sachs AB (1996) Association of the yeast poly(A) tail binding protein with translation initiation factor eIF-4G. EMBO J, 15, 7168–7177. | PubMed | ISI | ChemPort |

Tarun SZ,Jr, Wells SE, Deardorff JA and Sachs AB (1997) Translation initiation factor eIF4G mediates in vitro poly(A) tail-dependent translation. Proc Natl Acad Sci USA, 94, 9046–9051. | Article | PubMed | ChemPort |

Waskiewicz AJ, Johnson JC, Penn B, Mahalingam M, Kimball SR and Cooper JA (1999) Phosphorylation of the cap-binding protein eukaryotic translation initiation factor 4E by protein kinase Mnk1 in vivo. Mol Cell Biol, 19, 1871–1880. | PubMed | ISI | ChemPort |

Wells SE, Hillner PE, Vale RD and Sachs AB (1998) Circularization of mRNA by eukaryotic translation initiation factors. Mol Cell, 2, 135–140. | PubMed | ISI | ChemPort |

Witherell GW, Schultz-Witherell CS and Wimmer E (1995) Cis-acting elements of the encephalomyocarditis virus internal ribosome entry site. Virology, 214, 660–663. | Article | PubMed | ChemPort |