J Biol Chem, Vol. 274, Issue 38, 26720-26726, September 17, 1999

Interaction of Translation Initiation Factor eIF4G with eIF4A in the Yeast Saccharomyces cerevisiae^*

Diana Dominguez^§, Michael Altmann, Jörg Benz^¶, Ulrich Baumann^¶, and Hans Trachsel

From the  Institute for Biochemistry and Molecular Biology, University of Bern, Bühlstrasse 28 and the ^¶ Institute for Chemistry and Biochemistry, University of Bern, Freiestrasse 3, 3012 Bern, Switzerland

	ABSTRACT

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

Eukaryotic initiation factor (eIF) 4A is an essential protein that, in conjunction with eIF4B, catalyzes the ATP-dependent melting of RNA secondary structure in the 5'-untranslated region of mRNA during translation initiation. In higher eukaryotes, eIF4A is assumed to be recruited to the mRNA through its interaction with eIF4G. However, the failure to detect this interaction in yeast brought into question the generality of this model. The work presented here demonstrates that yeast eIF4G interacts with eIF4A both in vivo and in vitro. The eIF4A-binding site was mapped to amino acids 542-883 of yeast eIF4G1. Expression in yeast cells of the eIF4G1 domain that binds eIF4A results in cell growth inhibition, and addition of this domain to an eIF4A-dependent in vitro system inhibits translation in a dose-dependent manner. Both in vitro translation and cell growth can be specifically restored by increasing the eIF4A concentration. These data demonstrate that yeast eIF4A and eIF4G interact and suggest that this interaction is required for translation and cell growth.

	INTRODUCTION

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

The initiation phase of translation consists of a series of events leading to the recruitment of translation-competent 80 S ribosomes to the initiation codon of an mRNA. The process is well conserved among eukaryotes and requires the participation of several initiation factors as well as specific features at the 5'- and 3'-ends of the message (reviewed in Refs. 1 and 2). The first step is the recognition of the m⁷G cap structure, present at the 5'-end of all cellular mRNAs, by eukaryotic initiation factor (eIF)¹ 4E (the cap-binding protein), which, together with eIF4G, forms the core of the eIF4F complex. In mammalian cells, eIF4F contains a third subunit, initiation factor eIF4A. The poly(A) tail at the 3'-end of most cellular mRNAs acts, like the cap structure, as a translational enhancer. The complex formed by the poly(A) tail and the associated poly(A) tail-binding protein stimulates 40 S subunit recruitment to the mRNA and, in combination with the cap complex, synergistically activates translation initiation (3-6). RNA secondary structures in the 5'-untranslated region of the mRNA that might impede ribosome binding are then melted by the ATP-dependent RNA helicase eIF4A in conjunction with eIF4B, an RNA-binding protein that stimulates eIF4A activity (7, 8). The eIF4F complex exhibits higher helicase activity than eIF4A alone (9), and eIF4A has been proposed to cycle in and out of the eIF4F complex during translation initiation (10). Once the mRNA is unfolded, the 43 S preinitiation complex binds probably through interaction of eIF4G and eIF4B with ribosome-bound eIF3 (11, 12), and the preinitiation complex scans the 5'-untranslated region for the translation initiation codon.

A key player in translation initiation that coordinates all these steps by acting as a molecular adapter is initiation factor eIF4G (13). Two related eIF4G proteins (eIF4G1 and eIF4G2) encoded by two different genes exist in plants (reviewed in Ref. 14), mammals (15), and the yeast Saccharomyces cerevisiae (16). eIF4G helps to recruit 43 S preinitiation complexes by acting as a bridge between eIF4E bound to the cap, poly(A) tail-binding protein bound to the poly(A) tail, and eIF3 bound to the 40 S ribosomal subunit. Additionally, mammalian eIF4G contains two binding sites for eIF4A, one in the middle (amino acids 478-883) and one in the C-terminal part (amino acids 1045-1404) of the protein (17), that recruit the helicase to the eIF4F complex. Whereas the amino acid sequence of the middle part is conserved among human, yeast, and plant eIF4G, the C-terminal part seems to be unique to the mammalian factor. Yeast eIF4G possess no corresponding region, and plant eIF(iso)4G possess part of the domain with low homology (12%). It is not clear yet whether the two binding sites on mammalian eIF4G bind a single eIF4A molecule or whether they independently interact with two molecules of eIF4A.

Depending on the isolation procedure, plant eIF4A, like its mammalian counterpart, can be recovered in a complex with eIF4G. However, yeast eIF4A could thus far not be detected in eIF4F complexes purified by m⁷GDP affinity chromatography (18). In yeast, eIF4A is an essential protein encoded by two genes, TIF1 and TIF2 (19). It is required for translation of most, if not, all mRNAs (20-22), but its role in translation initiation is not entirely clear. Even mRNAs with AUG start codons positioned close to the cap and with little RNA secondary structure in their 5'-untranslated regions are still dependent on eIF4A for translation (22). The requirement of eIF4A for translation initiation and our inability to recover it as part of the eIF4F complex in yeast prompted us to directly address the question of whether yeast eIF4A and eIF4G interact and whether this interaction is physiologically relevant for translation initiation.

Using purified recombinant proteins, we demonstrate direct binding of eIF4A to eIF4G1 and map the interaction domain to a portion of the eIF4G1 protein between amino acids 542 and 883. We show that this fragment of eIF4G1 functions as a dominant-negative inhibitor of translation in an eIF4A-dependent in vitro system and of cell growth in vivo, most probably because it competes with eIF4G for eIF4A binding. Both translation and cell growth can be rescued by increasing the eIF4A concentration, indicating that, in yeast, the eIF4G-eIF4A interaction is essential for translation and cell growth.

	EXPERIMENTAL PROCEDURES

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

Plasmid Construction-- eIF4G1 constructs were derived from pUC8.4. This plasmid carries the complete TIF4631 open reading frame introduced into pUC8 as an EcoRI fragment from plasmid pCG8.4 (16). Six histidine residues were introduced at the C terminus by polymerase chain reaction. For bacterial expression of glutathione S-transferase (GST) fusion proteins, vectors pGEX-1T, pGEX-2T, pGEX-3X, and pGEX-6P-1 (Amersham Pharmacia Biotech) were used. To obtain eIF4G1-(1-952), the EcoRI fragment of pUC8.4 was subcloned into pGEX-2T. eIF4G1-(542-883) and eIF4G1-(492-539) were subcloned as BglII fragments into the BamHI site of pGEX-1T, and eIF4G1-(441-952) was subcloned as an EcoRV-EcoRI fragment into the SmaI-EcoRI sites of pGEX-3X. The BglII fragment of pUC8.4 encoding eIF4G1-(542-883) was also subcloned into the BamHI site of pGEX-6P-1. eIF4G1-(441-539), eIF4G1-(592-953), and eIF4G1-(592-862) were obtained by polymerase chain reaction using pUC8.4 as template. The construction of eIF4G1-(160-492) and eIF4G1-(883-952) has been described previously (23). Plasmid pYEX-GST-eIF4G1-(542-883) was made by introducing the 1023-base pair BglII fragment of pUC8.4 into the BamHI site of pYEX-4T (AMRAD Biotech). Plasmid pGEX-TIF1 was made by polymerase chain reaction on plasmid pGAL-TIF1 (24), and the fragment was then introduced into pGEX-6P-1. The same fragment was subcloned into vector pEGKG (25), resulting in plasmid pEGKG.eIF4A. p301-HIS3-eIF4A was made by introducing the BamHI-SacI fragment of pGAL-TIF1 into the yeast centromeric vector p301-HIS3 (26).

Yeast Strains, Media, and Genetic Manipulations-- Transformation of yeast cells was done by the lithium acetate method (27). Yeast culture media (YPD, YPGal, SD, and SGal) were made according to standard recipes as described (28). All strains used in this study (see Table I) were, unless otherwise stated, derivatives of the wild-type strain CWO4.

Protein Expression and Purification-- Escherichia coli strain BL21 was used for expression of proteins. Precultures were grown overnight and diluted 1:50 in 1 liter of LB medium. After 2 h of growth, synthesis of the fusion proteins was induced with a 1 mM final concentration of isopropyl--D-thiogalactopyranoside (MBI Fermentas). Cells were collected after 4 h, resuspended in ice-cold Tris-buffered saline (20 mM Tris-HCl (pH 7.4) and 150 mM NaCl), and lysed in a French press at 1000 p.s.i. The lysates were clarified by centrifugation at 8000 × g for 20 min at 4 °C. For the experiments in Figs. 1 and 2, batch purification of GST-eIF4G1 and GST-eIF4G1 fragments was performed by incubating the extracts with 20 µl of glutathione-Sepharose beads (Amersham Pharmacia Biotech) for 3 h at 4 °C, followed by three washes with 1 ml of buffer A (30 mM HEPES (pH 7.4), 100 mM potassium acetate, 2 mM magnesium acetate, and 2 mM dithiothreitol).

For the preparation of yeast recombinant eIF4A, eIF4G1-(542-883), eIF4G1-(592-862), and eIF4G1-(592-953), 300 µl of GSH-Sepharose with the bound GST fusion proteins were incubated in 300 µl of Tris-buffered saline containing 1 mM EDTA, 1 mM dithiothreitol, and 20-40 units of PreScission protease overnight at 5 °C. The resin was spun down at 1000 × g for 1 min, and the supernatant was collected and kept frozen in aliquots at 20 °C.

In Vitro Protein Binding Assays-- About 5 µg of the GST fusion proteins immobilized on 20 µl of GSH-Sepharose resin were resuspended in 300 µl of buffer A, and 30 µl of the suspension were taken for analysis by SDS-PAGE and Coomassie Blue staining. The rest was incubated in buffer A with either 80 µg of ribosomal salt wash (RSW; prepared from strain ABYS as described (29)) (see Fig. 1) or 0.5 µg of purified eIF4A (see Fig. 2) in a final volume of 500 µl for 1.5 h at 4 °C. RNase A (25 µg/ml) was included in the reaction where indicated. Following four washing steps with 1 ml of buffer A each (tubes were changed before the last wash), bound proteins were eluted in sample buffer and resolved by SDS-PAGE.

Surface Plasmon Resonance Assays-- All SPR assays were performed in an IAsys (Affinity Sensors, Thermobio Inc., Cambridge, United Kingdom). Purified eIF4A (60 ng) was immobilized on a 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide/N-hydroxysuccinimide-activated carboxylate cuvette according to the manufacturer's instructions. eIF4G1-(542-883), eIF4G1-(592-862), and eIF4G1-(592-953) were then injected over the chip. Each cycle consisted of a 200-µl injection of the indicated amount of purified proteins in buffer A. The chip was regenerated after each cycle by washing three times with 200 µl of 100 mM HCl (pH 1.5) and five times with 200 µl of buffer A. The response of the eIF4A-coated chip with 200 µl of buffer A was fixed as a base line. Data were analyzed using IAsys FAST-fit software. To calculate the K_d of eIF4G1-(542-883) binding to eIF4A, three concentrations of eIF4G1-(542-883) were used (50, 125, and 250 nM).

In Vivo Protein Binding Assays-- Strains CWO4 and DDY6 were grown overnight in 2 liters of YPGal at 32 °C. Exponentially growing cells (A_600 nm ~ 0.5-1) were pelleted, and the pellet was resuspended in 1 volume of buffer A containing 1 mM phenylmethylsulfonyl fluoride and lysed with glass beads by vortexing eight times for 30 s each. Following clarification of the lysate by centrifugation at 26,000 × g for 30 min, 40 µl of GSH-Sepharose beads were added, and incubation was continued for 2.5 h at 4 °C. After three washes with a total of 12 ml of buffer A (tubes were changed before the last wash), the GST-eIF4A fusion protein and the associated proteins were eluted in sample buffer and resolved by SDS-PAGE.

Strains DDY9 and DDY10 were grown on minimal medium with glucose as carbon source. Cells from a 50-ml culture (A_600 nm ~ 0.7) were collected and lysed in 500 µl of buffer A. The clarified lysates were incubated with 20 µl of GSH-Sepharose beads for 2.5 h at 4 °C and washed as described above. GST and GST-eIF4G1-(542-883) together with the associated proteins were eluted in sample buffer and resolved by SDS-PAGE.

Western Blotting-- Western blotting was performed as described (30). Primary polyclonal antibodies (used in a 1:1000 dilution) were rat anti-eIF4A (24), rabbit anti-eIF4E (31), rat anti-GST-eIF4G1-(542-883) (30), rat anti-eIF4B (32), and rat anti-p20 (23).

eIF4A-dependent Cell-free Translation System-- Preparation of the eIF4A-dependent translation system and translation reactions were as described (24). Recombinant eIF4A and eIF4G1-(542-883) proteins were preincubated at room temperature for 5 min in Tris-buffered saline.

	RESULTS

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

GST-eIF4G1 Interacts with eIF4A from Ribosomal Salt Wash-- To study the putative interaction between yeast eIF4G and eIF4A, the gene encoding eIF4G1 (TIF4631) was inserted into the pGEX-2T vector and expressed in bacteria as an N-terminal GST fusion protein. eIF4G1 immobilized on GSH-Sepharose resin was incubated with an RSW fraction enriched in initiation factors. After extensive washing, bound proteins were eluted in sample buffer and resolved by SDS-PAGE, and the presence of eIF4A was determined by immunoblotting (Fig. 1). The eIF4A polypeptide from RSW was specifically retained when eIF4G1 was bound to the beads (Fig. 1C, lane 2), but not with control GST (lane 9), indicating that the two proteins do interact directly or indirectly.

View larger version (28K): [in this window] [in a new window]   Fig. 1.   GST-eIF4G1 interacts with eIF4A. A, schematic representation of yeast eIF4G1 with the binding sites for eIF4E (23, 35, 42) and poly(A) tail-binding protein (PABP) (35), the RNA recognition motif (RRM) with the RNP1 and RNP2 elements (white boxes) (16), and the two arginine/serine-rich regions (RS). Binding to eIF4A and eIF4E as shown in C is summarized on the right. B, ~5 µg of the full-length GST fusion proteins or GST alone were affinity-purified on GSH-Sepharose beads. One-tenth of the sample was resolved by SDS-PAGE and visualized by Coomassie Blue staining. Full-length products were assigned to the slowest migrating bands on immunoblots (not shown) and are indicated by dots. Molecular mass markers (in kilodaltons) are shown on the left. C, the rest of each of the resin-bound proteins shown in B (~4.5 µg) were incubated with 80 µg of RSW in the presence of 25 µg/ml RNase A as described under "Experimental Procedures." In lane 1, 5% of the input RSW fraction was loaded. Similar results were obtained in three independent experiments.

To map more precisely the region within eIF4G1 responsible for the interaction, several deletion mutants were created and expressed in E. coli as N-terminal GST fusions (Fig. 1, A and B). Similar amounts of protein were purified by GSH-Sepharose affinity column and assayed for eIF4A binding. Most of the GST fusion proteins contained smaller peptides, which are probably C-terminal degradation products, as has been previously observed for both mammalian (33) and yeast (34, 35) eIF4G preparations. The concentration of the full-length products was then estimated from a Coomassie Blue-stained gel, and approximately equal amounts of fusion proteins were used for the binding experiments. The fragment eIF4G1-(160-492) migrated abnormally slowly on SDS gels (Figs. 1B and 2A, lanes 2). This domain is probably responsible for the abnormal migration of eIF4G in SDS-PAGE (16) since the other eIF4G1 fragments displayed normal electrophoretic mobilities. The region of the eIF4G1 polypeptide from amino acids 542 to 883 (Fig. 1B, lane 4) and a larger fragment containing this part (eIF4G1-(441-952); lane 3) were able to retain eIF4A in the resin (Fig. 1C, lanes 4 and 5), but not other parts of the protein.

To test the specificity of the binding experiment, eIF4E binding to the eIF4G1 fragments was used as a control. The binding site for eIF4E on eIF4G1 is known to reside between amino acids 441 and 490 (23). As expected, eIF4E specifically bound to full-length eIF4G1 (Fig. 1C, lane 2) and to those fragments that still harbored the eIF4E-binding site (lanes 3, 4, and 6); very low levels of eIF4E were also recovered after incubation with eIF4G1-(542-883) (lane 5), but they were close to the background levels detected with GST alone (lane 9). The slower migrating band in lane 7 corresponds to a degradation product of eIF4G1-(492-539) (see lane 6 of Fig. 1B), which probably nonspecifically bound the anti-eIF4E antibodies because of the large amount of protein present. Binding of eIF4A to eIF4G1-(542-883) seems to be specific since Prt1, a subunit of eIF3 (29, 36), did not bind to it. On the other hand, Prt1 interacted with full-length eIF4G1 and some of the other eIF4G1 fragments (data not shown). In the experiments shown in Fig. 1C, the binding reaction was carried out in the presence of 25 µg/ml RNase A to exclude interactions mediated by RNA. We conclude from these data that yeast eIF4G1 is capable of interacting with eIF4A either directly or through other factor(s) present in the RSW fraction.

Interaction between GST-eIF4G1 and Purified Recombinant eIF4A-- To exclude the possibility that binding of eIF4A to eIF4G1 is mediated by other factor(s) present in RSW, a GST-eIF4A fusion protein was expressed in bacteria, and the GST moiety was removed with PreScission protease. Binding of purified eIF4A to the eIF4G1 fragments and to GST was then assayed (Fig. 2). Recombinant eIF4A was capable of binding to full-length eIF4G1 (Fig. 2B, lane 2) and to fragments eIF4G1-(441-952) (lane 4) and eIF4G1-(542-883) (lane 5), but not to GST alone (lane 9). In this particular experiment, the amount of eIF4G1-(441-952) immobilized on the resin was at least five times lower than that of other proteins (Fig. 2A, lane 3), explaining the lower recovery of eIF4A. In addition, eIF4G1-(441-539) (Fig. 2B, lane 6), encompassing the eIF4E-binding site and an arginine/serine-rich region, did interact weakly with eIF4A. This interaction was not apparent when RSW was used as the source for eIF4A (Figs. 1C and 2B, compare lanes 6). The significance of this interaction was not further evaluated.

View larger version (71K): [in this window] [in a new window]   Fig. 2.   Interaction between yeast recombinant eIF4A and GST-eIF4G1. A, ~5 µg of the full-length GST fusion proteins or GST alone were purified on GSH-Sepharose beads. One-tenth of the sample was fractionated by SDS-PAGE and visualized by Coomassie Blue staining. The dots indicate the positions of the full-length polypeptides. Size markers (in kilodaltons) are shown on the left. B, the rest of the sample was incubated with 0.5 µg of recombinant eIF4A and treated as described for Fig. 1B. In lane 1 (input), 0.5 µg of eIF4A were loaded. Similar results were obtained in at least three independent experiments.

Although more eIF4A bound to eIF4G1-(542-883) than to the full-length protein, estimation of the relative affinities is complicated by the presence of the numerous eIF4G1 degradation products, some of which might still interact with eIF4A. The same results were obtained when the binding reaction was performed in the presence of 25 µg/ml RNase A (data not shown). Together, these results show that eIF4A binds directly to a domain of eIF4G1 comprising amino acids 542-883.

High Affinity Binding between eIF4A and eIF4G1-(542-883)-- To verify that eIF4G1-(542-883) harbors an eIF4A-binding site, the interaction between the two proteins was analyzed by other techniques. Incubation of equimolar amounts of eIF4A and eIF4G1-(542-883) showed that they form a stable complex that can be purified by gel filtration chromatography. The size of the complex is consistent with a 1:1 ratio between the two proteins (data not shown).

The interaction of eIF4A with some eIF4G1 fragments was then also measured by SPR analysis (Fig. 3). eIF4G1-(542-883) interacted strongly with eIF4A. No binding of eIF4G1-(592-862) was detected, whereas eIF4G1-(592-953) exhibited a greatly reduced binding to eIF4A. Both eIF4G1-(592-862) and eIF4G1-(592-953) showed no binding to eIF4A in in vitro binding assays described in Fig. 2 (data not shown).

View larger version (22K): [in this window] [in a new window]   Fig. 3.   SPR analysis of eIF4A binding to eIF4G1 fragments. A, eIF4A was covalently bound to a carboxylate cuvette, and binding of 2 µg of eIF4G1-(542-883), 2 µg of eIF4G1-(592-953), and 4 µg of eIF4G1-(592-862) was analyzed. Only the on-rates were measured in this experiment. The dimension of the response is in arc seconds. eIF4G1-(542-883) and eIF4G1-(592-862) were added at 3 min, and eIF4G1-(592-953) at 4.5 min. B, shown are the results from Coomassie Blue staining of 0.6 µg of eIF4G1-(542-883) (lane 1), eIF4G1-(592-862) (lane 2), and eIF4G1-(592-953) (lane 3).

In a separate experiment, three different concentrations of eIF4G1-(542-883) were used. The on- and off-rates were measured, and the K_d was estimated to be ~3 × 10⁸ M (data not shown).

These data demonstrate that eIF4G1-(542-883) binds to eIF4A with high affinity to form a stable complex. In addition, they indicate that the region between amino acids 542 and 592 of eIF4G1 contains an important determinant for eIF4A binding.

eIF4A and eIF4G Interact in Vivo-- The absence of eIF4A in eIF4F complexes isolated from yeast cells (18) could be used to argue that the interaction observed in vitro is an artifact. To determine if the eIF4G1-eIF4A interaction occurs in vivo, the yeast strain DDY6 was used (Table I). This strain is deleted for TIF1 and TIF2 (both genes encoding eIF4A) and expresses an eIF4A-thermosensitive protein constitutively from a plasmid (eIF4A^ts (37)) and a GST-eIF4A fusion protein upon galactose induction from another plasmid. The fusion protein is active as shown by in vivo complementation of the thermosensitive phenotype of strain SS13-3A/pSSC120 (Table I) and by its stimulation of translation in an eIF4A-dependent in vitro system (data not shown).

The yeast strain DDY6 and the isogenic wild-type strain CWO4 were grown at 32 °C on rich medium containing galactose as carbon source. Synthesis of GST-eIF4A is induced in DDY6 cells, and under these conditions, only the fusion protein, but not eIF4A^ts, is functional (DDY6 cells do not grow on glucose at 32 °C). Lysates prepared from these cells were incubated with GSH-Sepharose beads, and after extensive washing, GST-eIF4A and the associated proteins were fractionated by SDS-PAGE. Copurification of eIF4G or control proteins was determined by Western blotting (Fig. 4). eIF4G was retained on the resin only in the presence of GST-eIF4A (lane 6). No eIF4G could be recovered in the bound fraction of extracts from wild-type CWO4 cells, in which no GST-eIF4A is expressed (lane 3). Similar results were obtained when the DDY6 strain grown on glucose at 25 °C (no GST-eIF4A is synthesized) was used as a negative control (data not shown).

View larger version (41K): [in this window] [in a new window]   Fig. 4.   eIF4G-eIF4A interaction in vivo. The yeast strain DDY6 and the isogenic wild-type strain CWO4 were grown on YPGal at 32 °C. Equal amounts of lysates from these cells were incubated with GSH-Sepharose beads for 2.5 h at 4 °C, and proteins were eluted in sample buffer and resolved by SDS-PAGE. The same blot was then probed with anti-GST-eIF4G1-(542-883), anti-eIF4B, and anti-p20 antibodies. The former recognizes both eIF4G1 and eIF4G2 (and the GST moiety of GST-eIF4A). Therefore, the protein in this figure is called eIF4G since the eIF4A-interacting factor(s) may consist of both eIF4G1 and eIF4G2. T, 0.5% of the total input; U, 0.5% of the unbound fraction; B, 100% of the bound material. A representative of at least three independent experiments is shown.

Only a small fraction of eIF4G could be isolated in a complex with eIF4A. The association is, however, specific as illustrated by the fact that neither p20 (a negative regulator of eIF4E activity (23)) nor eIF4B, both involved in translation initiation, was retained on the beads (Fig. 4, lane 6).

The amount of eIF4G·GST-eIF4A complex detected in these experiments is probably an underestimation of the real amount present in the cells. First, DDY6 cells encode, in addition to GST-eIF4A, a thermosensitive copy of eIF4A. Despite being inactive at 32 °C, this protein might still interact with eIF4G in competition with GST-eIF4A. Second, for reasons that remain unknown, only part of GST-eIF4A bound to the beads (Fig. 4, compare lanes 4 and 5 and lane 6). Therefore, it is not possible from these experiments to estimate the real levels of the eIF4G·eIF4A complex in yeast cells. However, they demonstrate that yeast eIF4G and eIF4A do interact in vivo.

eIF4G1-(542-883) Inhibits Translation in a Cell-free System-- To evaluate the biological significance of the eIF4G-eIF4A interaction, we intended to disrupt it and to analyze the effects on translation. We reasoned that the eIF4G1-(542-883) protein, if added to a translation mixture, would compete with endogenous eIF4G for eIF4A binding. The consequent disruption of the eIF4G-eIF4A interaction, if significant for the translation process, would result in translation inhibition. Such an inhibition should be specifically rescued by increasing the concentration of eIF4A. To test this hypothesis, an eIF4A-dependent cell-free system was used (24). This system lacks endogenous eIF4A activity; hence, the final level of eIF4A in the extract can be adjusted by addition of purified recombinant protein. As shown in Fig. 5B, the incorporation of [³⁵S]methionine in the absence of exogenous eIF4A (bar 1) was similar to background levels when no RNA was added (bar C). [³⁵S]Methionine incorporation increased as the eIF4A concentration increased, and saturation was reached with 1 µg of eIF4A (bar 4); higher concentrations were slightly inhibitory for translation (bar 5). We calculated that 1 µg of eIF4A in a 15-µl reaction mixture corresponds to two to three copies of eIF4A/ribosome. Four copies of eIF4A/ribosome were shown previously to be required in this system for maximal stimulation of translation (24). A similar eIF4A/ribosome ratio has been estimated for mammalian cells (38). To test the effect of eIF4G1-(542-883) in this system, increasing amounts of the protein were preincubated with 0.25 µg of eIF4A before adding them to the translation mixture. As predicted, eIF4G1-(542-883) inhibited translation in a dose-dependent manner; 0.2 (bar 6), 0.5 (bar 7), and 1 (bar 8) µg of eIF4G1-(542-883) inhibited translation by 20, 50, and 70%, respectively. If eIF4G1-(542-883) acts by sequestering eIF4A, then translation should be rescued by further addition of eIF4A. A dose-dependent recovery of translation by eIF4A can be demonstrated: at concentrations of eIF4A below saturation (0.25-0.5 µg in a 15-µl reaction mixture), the effect of adding 1 µg of eIF4G1-(542-883) was a ~70% reduction in translation (compare bar 2 with bar 8 and bar 3 with bar 9). At saturating eIF4A concentrations, 1 µg of eIF4G1-(542-883) decreased translation by only ~25% (bars 4 and 10), whereas the same amount of the eIF4G1 fragment had even a positive effect on translation when eIF4A was in excess (bars 5 and 11), probably by titrating the excess of eIF4A that per se inhibited translation slightly (bars 4 and 5). These data indicate that eIF4A and eIF4G1-(542-883) interact in this in vitro system and that the eIF4G-eIF4A interaction is important for efficient translation.

View larger version (19K): [in this window] [in a new window]   Fig. 5.   eIF4G1-(542-883) inhibits translation in an eIF4A-dependent cell-free system. A, shown are the results from Coomassie Blue staining of 1 µg of purified eIF4A and eIF4G1-(542-883) proteins used in the in vitro translation system. B, total yeast mRNA was translated in the absence or presence of yeast recombinant eIF4A and eIF4G1-(542-883) proteins as indicated at the top (numbers indicate micrograms of each protein added to a 15-µl reaction mixture). eIF4A and GST-eIF4G1-(542-883) were preincubated for 5 min in Tris-buffered saline at room temperature. Methionine incorporation is expressed relative to the value obtained with 0.25 µg of eIF4A, which was set as 100% (shaded bar, bar 2). Bar C is the negative control without mRNA. Similar results were obtained in four independent experiments.

Expression of eIF4G1-(542-883) Inhibits Cell Growth-- To exclude the possibility that the negative effect of eIF4G1-(542-883) on translation was an artifact of the in vitro system, we tested its effect in vivo. Because eIF4A is an abundant protein, we chose a strain deleted for TIF2 (SS10-3F), assuming lower levels of eIF4A in these cells. Transformation of SS10-3F with pYEX-GST-eIF4G1-(542-883) (GST-eIF4G1-(542-883) under the control of a copper-inducible promoter) or with the empty vector resulted in strains DDY10 and DDY9, respectively (Table I). When these cells were grown on minimal medium in the absence of copper, leaky expression of GST and GST-eIF4G1-(542-883) from the CUP1 promoter was detected (Fig. 6B, lanes 1, 2, 4, and 6). The presence of GST-eIF4G1-(542-883) in cells led to a reduction of growth rate (duplication time for DDY10 cells was 6 h versus 3 h for the control DDY9 cells) (Fig. 6A). When 0.5 mM copper sulfate was added to the medium, DDY9 cells continued to grow, although at a slower rate (4.6-h duplication time), whereas a drastic inhibition of growth was observed for DDY10 cells (Fig. 6A), correlating with slightly higher levels of the eIF4G1 fragment (Fig. 6B). No further growth of DDY10 cells was detected after 20 h of copper treatment (data not shown). To determine whether GST-eIF4G1-(542-883) interacted in vivo with eIF4A, DDY9 and DDY10 cells were grown on SD medium in the absence of copper. Cell lysates were incubated with GSH-Sepharose beads, and bound proteins were resolved by SDS-PAGE (Fig. 6C). eIF4A was recovered when GST-eIF4G1-(542-883) (lane 6), but not GST (lane 3), was bound to the beads. Since for unknown reasons the recovery of GST-eIF4G1-(542-883) was inefficient (compare lanes 5 and 6), the proportion of complex formation in vivo is difficult to estimate. However, these data show that GST-eIF4G1-(542-883) binds to eIF4A in vivo and suggest that this interaction is responsible for the inhibitory effect of GST-eIF4G1-(542-883) on cell growth.

View larger version (20K): [in this window] [in a new window]   Fig. 6.   In vivo expression of GST-eIF4G1-(542-883) results in inhibition of cell growth. A-C, DDY9 and DDY10 cells were grown on minimal SD medium in the presence or absence of 0.5 mM copper sulfate. A, growth was monitored by measuring the absorbance at 600 nm. The duplication times are shown on the right. B, shown are the results from Western blotting with anti-GST-eIF4G1-(542-883) antibody (also recognizes GST). Extracts were prepared by boiling cells (A600 nm = 0.2) taken at the time points (a-d) indicated in A in SDS sample buffer. C, DDY9 and DDY10 cells were grown in 50 ml of SD medium until the cultures reached A600 nm ~ 0.7. Lysates prepared from these cells were incubated with GSH-Sepharose beads for 2.5 h at 4 °C. T, 4% of the input material; U, 4% of the unbound fraction; B, 100% of the bound material. D, DDY15 and DDY16 strains were grown on SD and SGal plates in the presence or absence of Cu2+ as indicated. Cells were allowed to grow for 6 days at 30 °C.

We then tried to determine if growth inhibition by GST-eIF4G1-(542-883) expression could be rescued by increasing the amount of eIF4A in the cell. Since eIF4G1-(542-883) was highly inhibitory for cell growth, the rescue experiment was performed with the CWO4 wild-type strain (both TIF1 and TIF2 genes present). CWO4 cells were cotransformed with plasmid p301-HIS3-eIF4A encoding an additional copy of eIF4A under the control of a galactose-inducible promoter and either pYEX-GST-eIF4G1-(542-883) or the empty vector (strains DDY16 and DDY15, respectively) (Table I). DDY15 and DDY16 cells were grown on SD and SGal plates in the presence or absence of 0.5 mM copper sulfate (Fig. 6D). The presence of copper drastically inhibited growth of DDY16 cells on glucose, with almost no effect on the control DDY15 cells. However, when the cells were grown on galactose to induce the expression of eIF4A from the p301-HIS3-eIF4A plasmid, DDY16 cells did grow despite the presence of copper in the medium. This effect is most likely the result of expressing more eIF4A in DDY16 cells, and not the consequence of growing them on galactose, since a strain that carries an empty vector instead of p301-HIS3-eIF4A did not grow on SGal plates in the presence of copper (data not shown). These results confirm the data obtained in vitro, and together suggest that the eIF4G1-eIF4A interaction is required for translation and cell growth.

	DISCUSSION

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

Interaction between eIF4G and eIF4A has recently been demonstrated for mammalian (11, 17) and plant (39) factors. No such interaction could until now be demonstrated in yeast, suggesting that the model of eIF4G-mediated recruitment of the helicase to the mRNA might not be conserved among eukaryotes. However, the data presented here demonstrate that yeast eIF4G1 does interact with eIF4A both in vivo and in vitro. The region of eIF4G1 from amino acids 542 to 883 is shown to harbor the eIF4A-binding site. This region is conserved at the amino acid level with mammalian eIF4G and plant eIF(iso)4G (17), and the corresponding fragment of mammalian eIF4G (amino acids 478-883) carries one of the two eIF4A-binding sites. The second binding site is located at the C terminus of the molecule and is unique to the mammalian factor (17). Different results were reported for plant eIF(iso)4G, which was shown to bind eIF4A through the N-terminal 90 amino acids (39). However, deletion of part of the middle domain of eIF(iso)4G, which corresponds to the eIF4A-binding sites in mammalian and yeast eIF4G, abrogated the ability of eIF(iso)4G to catalyze eIF4A- and RNA-dependent ATP hydrolysis, suggesting that this region might also be involved, at least partially, in binding eIF4A.

eIF4G1-(542-883) contains a consensus RNA recognition motif (16), raising the possibility that RNA mediates the interaction with eIF4A. We consider this possibility unlikely because (a) eIF4G-eIF4A binding still took place in the presence of RNase A, and (b) we were not able to detect any RNA-binding activity of the eIF4G1-(542-883) protein in filter binding experiments. Under the same conditions, other eIF4G1 fragments were able to bind RNA,² yet they did not interact with eIF4A in our assays.

In an attempt to narrow down the eIF4A-binding site, the region from amino acids 542 to 592 was found to carry an important determinant for eIF4A binding. Deletion of this region (proteins eIF4G1-(592-862) and eIF4G1-(592-953)) (Fig. 3) drastically affected the capacity of the proteins to interact with eIF4A. Unfortunately, we were unable to express the GST-eIF4G1-(542-592) protein in E. coli in order to test its binding to eIF4A directly. However, experiments are under way to map more precisely the determinants of eIF4G1 that are responsible for eIF4A binding.

Our experiments do not allow us to determine the amount of eIF4G·eIF4A complex in the cells due to inefficient binding of GST-eIF4A to the GSH-Sepharose beads and the presence of a non-tagged version of eIF4A. However, we assume that the levels of the eIF4G·eIF4A complex are low. This would explain the failure to detect eIF4A in eIF4F complexes purified by m⁷GDP affinity chromatography (18) and is consistent with the low recovery of the eIF4G·eIF4A complex in our in vivo experiments. The SPR analysis, however, revealed that the interaction in vitro between eIF4A and eIF4G1-(542-883) is strong (K_d ~ 3 × 10⁸ M), i.e. similar to the interaction between yeast eIF4E and the eIF4E-binding domain of eIF4G (K_d = 10⁸ to 10⁹ M) (40). Although we cannot exclude that full-length eIF4G1 has an overall lower affinity for eIF4A than the eIF4G1-(542-883) fragment, we favor the idea that the eIF4G-eIF4A interaction is strong in vivo, but only transient, probably modulated through post-translational modification(s) and/or interaction with other factor(s). In fact, there is evidence for conformational changes in mammalian (31) and yeast (41) eIF4G induced by binding of eIF4E. Interaction of eIF4G with eIF4E or other factor(s) might expose the eIF4A-binding site of eIF4G, creating a transient high affinity for eIF4A. Transient interaction between eIF4A and eIF4G may be important for translation initiation. This is supported by the finding that mammalian eIF4A cycles in and out of the eIF4F complex and that a mutant of mammalian eIF4A probably defective in recycling acts as a dominant inhibitor of translation initiation (10), suggesting that the dissociation of the eIF4G·eIF4A complex is a requisite for translation.

To analyze the biological significance of the eIF4G-eIF4A interaction, we intended to disrupt it by overexpressing the eIF4G1-(542-883) fragment. Indeed, when the eIF4A-binding domain of eIF4G1 was expressed in vivo, it inhibited growth in a dose-dependent manner. Whereas low levels of eIF4G1-(542-883) resulting from leaky expression doubled the duplication time of yeast cells, induction of expression caused an almost complete inhibition of growth. Similarly, eIF4G1-(542-883) had a negative effect on in vitro translation, supporting the idea that its negative effect on cell growth operates at the translational level. We believe that inhibition of translation by eIF4G1-(542-883) is due to complex formation with eIF4A since it could be specifically overcome by increasing the eIF4A concentration in vitro and in vivo.

Yeast cells encode a second copy of eIF4G (eIF4G2 encoded by TIF4632) that is 53% identical to eIF4G1. The two proteins are functional homologues since single gene disruptions, but not the double disruption, are viable (16). The dominant-negative effect of eIF4G1-(542-883) indicates that this protein also interferes with the eIF4G2-eIF4A binding. Therefore, it seems likely that eIF4G2 is also capable of binding eIF4A through the equivalent domain.

The central domain of mammalian eIF4G from amino acids 480 to 886, in addition to binding eIF4A, also binds eIF3 (11). Although this region is conserved at the amino acid level among mammalian, yeast, and plant eIF4G (17), no eIF4G-eIF3 interaction has thus far been described in yeast and plants. The fact that eIF4A alone could rescue the inhibition caused by eIF4G1-(542-883) indicates that eIF4G1-(542-883) sequesters eIF4A, but not other factors required for translation. This fragment of yeast eIF4G1 may contribute to binding eIF3 and/or other translation factors, but may itself not be sufficient for binding. Alternatively, it is possible that another initiation factor binds to this domain of eIF4G1, but that this factor is in large excess or has a higher affinity for full-length eIF4G1 and therefore cannot be competed out by eIF4G1-(542-883). In conclusion, our data demonstrate the capability of yeast eIF4G1 to interact with eIF4A and suggest that this interaction is important for translation and cell growth.

	ACKNOWLEDGEMENTS

We thank Barbara Fischli and Elisabeth Kislig for excellent technical assistance, Patrick Linder for the pGAL-TIF1 plasmid and yeast strains SS13-3A/pSSC120 and SS10-3F, Michael Hall for vector pEGKG, Alan Sachs for the rabbit anti-eIF4E antibodies, and Michael Horn and Monique Vogel for help with SPR.

	FOOTNOTES

^* This work was supported by Grant 31-45528.95 from the Swiss National Science Foundation.The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

^§ To whom correspondence should be addressed. Tel.: 4131-631-4112; Fax: 4131-631-3737; E-mail: diana-ines.dominguez@mci.unibe.ch.

² C. Berset, personal communication.

	ABBREVIATIONS

The abbreviations used are: eIF, eukaryotic initiation factor; GST, glutathione S-transferase; PAGE, polyacrylamide gel electrophoresis; RSW, ribosomal salt wash; SPR, surface plasmon resonance.

	REFERENCES

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES