Eukaryotic initiation factors 4A (eIF4A) and 4G (eIF4G) mutually interact in a 1:1 ratio in vivo.

mRNA translation in eukaryotic cells involves a set of proteins termed translation initiation factors (eIFs), several of which are involved in the binding of ribosomes to mRNA. These include eIF4G, a modular scaffolding protein, and eIF4A, an RNA helicase, of which two closely related forms are known in mammals, eIF4A(I) and eIF4A(II). In mammals, eIF4G possesses two independent sites for binding eIF4A, whereas in other eukaryotes (e.g. yeast) only one site appears to be present, thus raising the issue of the stoichiometry of eIF4G.eIF4A complexes in different eukaryotes. We show that in human embryonic kidney cells eIF4G is associated with eIF4A(I) or eIF4A(II) but not with both simultaneously, suggesting a stoichiometry of 1:1 rather than 1:2. To confirm this, eIF4A(I) or eIF4A(II) was expressed in a tagged form in these cells, and complexes with eIF4G were again isolated. Complexes containing tagged eIF4A(I) or eIF4A(II) contained no endogenous eIF4A, supporting the notion that eIF4G binds only one molecule of eIF4A. Each binding site in eIF4G can bind either eIF4A(I) or eIF4A(II). The data imply that the second binding site in mammalian eIF4A does not bind an additional eIF4A molecule and that initiation factor complexes in different eukaryotes contain one eIF4A per eIF4G.

Translation of mRNA in mammalian cells involves a set of proteins termed "translation factors." A number of the proteins are involved in the initiation phase of translation, the main control point in this process, and are accordingly termed "initiation factors" (eIFs). 1 Several of the eIFs play a role in the recruitment of mRNAs to ribosomes and in the events leading to the recognition of the initiation codon. One such protein is eIF4A, an ATP-dependent RNA helicase of the "DEAD-box" family (1,2). It is thought to act to remove regions of secondary structure within the 5Ј-untranslated region of the mRNA to facilitate ribosome binding and scanning (1,3). As revealed by the use of dominant negative mutants (4), eIF4A appears to be essential for cap-dependent and also for cap-independent initiation directed by picornavirus internal ribosome entry sites. eIF4A interacts with a scaffold protein, eIF4G, to form complexes that also contain the cap-binding protein eIF4E, which binds the cap structure (m 7 GpppN . . . ) at the 5Ј-end of the mRNA (for reviews see Refs. 1 and 5)). These complexes are termed "eIF4F." eIF4A shows much higher helicase activity as part of such complexes than as free eIF4A (6).
Three related proteins have been designated as distinct forms of eIF4A. eIF4A I and eIF4A II are closely related and are both known to be cytoplasmic and to interact with eIF4G (7). Some evidence for a preferential association of eIF4G with eIF4A II as compared with eIF4A I was presented (7). More recently, a further protein termed eIF4A III has been identified and reported to inhibit translation (8). However, this protein is more distantly related and appears to be primarily nuclear (9). Its involvement in translation initiation has therefore been questioned (8,9).
Recently, two independent binding sites were identified within mammalian eIF4G for eIF4A (10). One lies in the central region of eIF4G whereas the other lies toward its C terminus (10 -12). In contrast, both species of yeast eIF4G lack the C-terminal domain that interacts with eIF4A and thus appear to contain only one site for interaction with this factor within the central region of the yeast eIF4G polypeptide (13,14). Mammalian eIF4A can interact with the remaining site in the central region of eIF4G (15) but cannot functionally replace the yeast eIF4A gene (16). Recent data from Morino et al. (12) suggest that the C-terminal region of mammalian eIF4G may function as a modulatory domain.
The existence of two eIF4A binding sites in eIF4G suggested that mammalian eIF4F may contain two copies of eIF4A, and a recent report from Korneeva et al. (17) using surface plasmon resonance to explore eIF4G/eIF4A interactions in vitro offered support to this idea. If true in vivo, this could reflect a fundamental difference in the composition and function of eIF4F complexes in mammals compared to eukaryotes, given that yeast eIF4G only possesses one site for binding eIF4A (13,14). We therefore considered it important to study the interactions between mammalian eIF4G and eIF4A in vivo, and our data indicate that in fact eIF4G can only bind one molecule of eIF4A at a time under in vivo conditions. Thus, eIF4F complexes in yeast and mammals do not appear to differ in their eIF4A content, despite the presence in mammalian eIF4G of the second eIF4A binding site.

MATERIALS AND METHODS
Chemicals and Biochemicals-General laboratory chemicals were from Sigma-Aldrich (Gillingham, United Kingdom) or BDH Merck (Dorset, UK). Materials for cell culture were from Life Technologies, Inc. m 7 GTP-Sepharose was from Amersham Pharmacia Biotech. Oligonucleotides were synthesized from MWG Biotech (Milton Keynes, UK). Restriction enzymes were purchased from either Promega (Southampton, UK) or New England Biolabs (Hitchin, UK).
Production of Antisera-Polyclonal antibodies were generated by immunizing sheep (Diagnostics Scotland) with the appropriate peptide immunogen coupled (via the C-terminal cysteinyl residue) to keyhole limpet hemocyanin. Antibodies were purified over columns consisting of the relevant peptide coupled to Affi-Gel (Bio-Rad). Peptides used were as follows, the C-terminal Cys being used for coupling: ASQD-SRSRDNGPDAC (residues 3-16 of human eIF4A I ); GGSADYNREHG-GPEC (residues 3-16 of human eIF4A II ) and NRENYIHRIGRGGRF-GRKC (sequence common to eIF4A I (residues 352-369) and eIF4A II (residues 354 -371)). The anti-eIF4G I antisera were raised in a similar manner using the following peptides as immunogens: for the middle domain of eIF4G I , LKEELEEARDIARRC (corresponding to residues 841-854), and for the C-terminal domain of eIF4G I , RARTPATKRSF-SKEVEC (residues 1137-1152).
Vectors-pETHM-eIF4A I and pETHM-eIF4A II are vectors encoding His-and Myc-tagged eIF4A I and eIF4A II , respectively. pETHM-eIF4A I was made by ligating an NheI-KpnI-digested PCR fragment encoding the N terminus of eIF4A I and a KpnI-SalI fragment from pGem3Z-4A into the NheI-SalI sites of pET28b (Novagen). The PCR primers used to generate the N terminus of eIF4A I were 5Ј-ATGGCTAGCGAACAAAA-ACTCATCTCAGAAGAGGATCTGGGATCCATGTCTGCGAGTCAG GATTC-3Ј (forward) and 5Ј-GGCCAGGGGTACCCACGATGAT-3Ј (reverse). The upstream primer provides a BamHI site (underlined) between the Myc-tag and the start codon of eIF4A to facilitate further cloning. pETHM-eIF4A II was made by ligating a BamHI-SpeI-digested PCR fragment of N-terminal eIF4A II and a SpeI-SacI fragment from p4AIIms (a kind gift from P. Nielsen (18)) into the BamHI-SacI sites of pETHM. pETHM is a modified version of the expression vector pET28, containing a Myc-tag instead of the T7-tag found in pET28. It was made by replacing the SphI-BamHI region of pET28 with the SphI-BamHI fragment from pETHM-eIF4A I . The primers used for amplification of the N terminus of eIF4A II were 5Ј-ACGGATTCATGTCTGGTGGCTC-CGCG-3Ј (forward) and 5Ј-TTGAGCCAGTTCTCTGGTGGG-3Ј (reverse). For expression in human embryonic kidney (HEK) 293 cells, the vector pcDNA3.1(ϩ) was used. pETHM-eIF4A I was cut with XbaI for insertion into the NheI site of pcDNA3.1(ϩ), and pETHM-eIF4A II was cut with XbaI/NotI for insertion into the corresponding sites of pcDNA3.1(ϩ). Similarly, expression plasmids for fragments of eIF4G I were created by cutting pET-eIF4G I -M (kindly provided by Dr. C. U. T. Hellen, SUNY, Brooklyn, NY) with XbaI/XhoI (insertion into pcDNA3.1(ϩ) between the NheI-XhoI sites) and by cutting pET-eIF4G I -C (also from Dr. Hellen) with XbaI/HindIII for insertion into the same sites of pcDNA3.1(ϩ).
Cell Culture, Transfection, and Extraction-HEK 293 cells were maintained and transfected (20 -30 g of DNA/10-cm dish) as described earlier (19). The extractions were performed using our standard buffer (19), except that for Ni-NTA-agarose pulldown experiments the EDTA concentration was decreased from 1 to 0.2 mM.
Other Procedures-For immunoprecipitations, 3 l of antisera for eIF4A I or eIF4A II were mixed with 10 l (packed volume) of protein G-agarose in extraction buffer for 30 min at 4°C. The beads were then washed with extraction buffer and mixed with cell lysates (about 500 g of protein). Gel electrophoresis and Western blotting were performed as described earlier (20) using gels containing 12.5% (w/v) acrylamide and Immobilon-P membrane (Millipore).

RESULTS
The main aim of this study was to investigate the composition of eIF4F complexes in vivo in mammalian cells. To determine whether HEK 293 cells express both eIF4A I and eIF4A II , we made use of antisera that specifically recognize these proteins. The antisera were raised against peptides corresponding to residues within the N termini of eIF4A I and eIF4A II , which are completely different between the two polypeptides (see "Materials and Methods"). To verify that the antisera were indeed specific for each protein and to compare their potencies, we studied their abilities to recognize recombinant eIF4A I and eIF4A II in Western blots. The anti-eIF4A I antibody recognized eIF4A I but showed no reaction with eIF4A II , whereas the converse was true for the anti-eIF4A II antibody (Fig. 1A). Both proteins were recognized by a third antibody raised against a peptide sequence common to the C termini of both proteins. The antisera were calibrated using differing amounts of recombinant eIF4A I and eIF4A II (Fig. 1B). The anti-eIF4A C antibody should detect each form equally, and this was borne out by the data in the upper part of Fig. 1B, where a signal of similar strength was seen for the same amounts of the two proteins. In the lower part of Fig. 1B it can be seen that similar signals were seen for 150 ng of eIF4A I and 75 ng of eIF4A II (and so on for the three amounts of protein tested) indicating that the anti-eIF4A II antibody is roughly twice as sensitive as the antibody to eIF4A I .
Western blot analysis using the specific antibodies revealed that both eIF4A I and eIF4A II are present in HEK 293 cells (Fig.  1C). In the cell extracts, the strengths of the signals for eIF4A I and eIF4A II were similar suggesting (given the data in Fig. 1B) that the level of eIF4A I is probably slightly (about 2-fold) higher than that of eIF4A II . Cells were treated with insulin for 20 min prior to extraction to enhance the formation of eIF4F complexes (21). Samples of cell extract were also subjected to chromatography on m 7 GTP-Sepharose. This resin binds eIF4E, and thus eIF4A that is part of eIF4F complexes is also retained on the resin through the interaction of eIF4A with eIF4G. Western blotting with the above antibodies revealed that both FIG. 1. eIF4F complexes derived from HEK293 cells contain eIF4A I or eIF4A II . A, samples of recombinant eIF4A I and eIF4A II (as indicated; 1 g of each) were subjected to SDS-PAGE and Western blotting using the antisera raised against peptides from eIF4A I or eIF4A II or peptides common to both, as shown. B, different amounts of recombinant eIF4A I and eIF4A II were subjected to SDS-PAGE and immunoblotting using the indicated antibodies. C, HEK 293 cells were treated with insulin (100 nM, 20 min) and then extracts were prepared. Samples of extract were applied to m 7 GTP-Sepharose beads, and pulldown experiments were performed as described under "Materials and Methods." A sample of the bound material was retained for analysis, and the m 7 GTP-Sepharose beads were then washed twice with extraction buffer containing 200 M m 7 GDP to elute the bound proteins. Samples of eluate were incubated with anti-eIF4A I or anti-eIF4A II antibodies, and the immunoprecipitates were analyzed by SDS-PAGE/ Western blotting. Lane 1, cell lysate; lane 2, material bound to m 7 GTP-Sepharose; lanes 3 and 4, proteins eluted from m 7 GTP-Sepharose (200 M m 7 GDP, two washes which were then combined) and then immunoprecipitated with anti-eIF4A I (lanes 3) or anti-eIF4A II (lanes 4). samples were analyzed by SDS-PAGE followed by Western blotting using anti-eIF4G IC , anti-eIF4A I , or anti-eIF4A II (as indicated). eIF4A I and eIF4A II were found in the material retained on this resin, showing that both eIF4A I and eIF4A II become incorporated into eIF4F. The signals for eIF4A I and eIF4A II were again similar, suggesting that a similar proportion of each is incorporated into eIF4F complexes (Fig. 1C).
Because eIF4G possesses two independent binding sites for eIF4A (10), it was possible that eIF4F complexes could contain a mixture of these species of eIF4A. To address this possibility, material retained on m 7 GTP-Sepharose was eluted (with m 7 GDP) and then subjected to immunoprecipitation with anti-eIF4A I or anti-eIF4A II . The analysis clearly revealed that the resulting immunoprecipitates contained only the isoform of eIF4A against which the antibody was raised and that the other species of eIF4A was absent (Fig. 1C). Immunoprecipitates contained full-length eIF4G, thus ruling out the possibility that the eIF4G had been degraded to produce fragments that contained only one eIF4A binding site (Fig. 1C, top section). This finding indicates that individual eIF4F complexes do not contain a mixture of eIF4A I and eIF4A II .
Two possible explanations can be advanced to explain this observation: (i) eIF4F complexes each contain only one molecule of eIF4A or (ii) each complex may contain two molecules of eIF4A but these must both be either eIF4A I or eIF4A II . To distinguish between these possibilities, HEK 293 cells were transfected with vectors encoding eIF4A I or eIF4A II , each possessing Myc and hexahistidine (His 6 ) tags. Transfection experiments using a vector encoding the green fluorescent protein revealed that the transfection efficiencies were routinely about 95% (data not shown). As shown in Fig. 2, these tagged forms of eIF4A are expressed efficiently in HEK 293 cells, as assessed using the antiserum that recognizes both forms of eIF4A and at levels rather above those of the endogenous factors. eIF4F complexes were isolated from cells transfected with eIF4A I -Myc/His 6 or eIF4A II -Myc/His 6 . Western blotting showed that this material contained both endogenous eIF4A and the recombinant tagged polypeptide, which migrates more slowly because of the tags. The eluted material was subjected to chromatography on Ni-NTA agarose, and the bound material was analyzed by immunoblotting using the anti-eIF4A C antibody. Only the recombinant tagged eIF4A was detected, and none of the endogenous eIF4A was seen. This finding strongly suggests that the eIF4F complexes contain only one molecule of eIF4A. If complexes had contained two molecules of eIF4A, the endogenous protein would also have been seen in the Ni-NTA agarose-bound fraction.
The above data support the idea that eIF4G just binds a single eIF4A molecule, either eIF4A I or eIF4A II . They therefore imply that each of the two binding sites for eIF4A on eIF4G can bind either to eIF4A I or eIF4A II and that a single eIF4A molecule might bind simultaneously to both sites. It has, however, not previously been shown that eIF4A II binds both sites in eIF4G. To test this, HEK 293 cells were transfected with plasmids encoding the middle ("eIF4G M ") and C-terminal ("eIF4G C ") fragments of eIF4G. These polypeptides include His 6 and Myc tags and were efficiently expressed in HEK 293 cells (Fig. 3A). Samples of extract from these cells were subjected to chromatography on Ni-NTA-agarose, and the bound material was analyzed by SDS-PAGE/immunoblotting. The tagged eIF4G fragments were clearly retained on the resin (Fig. 3A), whereas full-length, untagged eIF4G was not (data not shown). Analysis of the Ni-NTA-agarose-bound material showed that eIF4A was retained on this resin when either of the tagged eIF4G fragments was expressed but not in the absence of these fragments. eIF4A does not therefore bind nonspecifically to the resin (Fig. 3B). To determine whether both forms of eIF4A were bound by the eIF4G fragments, we made use of the isoform-specific anti-eIF4A antisera. Western analysis revealed that endogenous eIF4A I and eIF4A II were each retained on the resin when samples from cells transfected with either the middle or C-terminal fragment of eIF4G were analyzed, implying that each protein can bind to both sites in eIF4G (Fig. 3C). This experiment also demonstrates that the interactions between eIF4A and eIF4G are relatively stable, thus ruling out the possibility that our observation that only the eIF4A molecule is found in each eIF4F complex is an artifact because of dissociation of eIF4A from one of the binding sites in eIF4G.
If eIF4A can bind simultaneously to both sites in eIF4G, it might be able to act as a bridge between two different eIF4G molecules effectively bringing about the dimerization of eIF4G. To test this, we transfected HEK 293 cells with plasmids encoding the His 6 -tagged eIF4G M or eIF4G C polypeptides. These fragments of eIF4G do not contain the binding site for eIF4E that is present in native eIF4G and should therefore only be  2, 4, and 6) or eIF4A II (lanes 3, 5, and 7). After 48 h, extracts were prepared and subjected to m 7 GTP-Sepharose chromatography, followed by elution of the bound materials (as described under "Materials and Methods") and a further purification of this material on Ni-NTA-agarose. Samples were analyzed by SDS-PAGE and Western blotting using the anti-eIF4A C antibody that detects both eIF4A I and eIF4A II . Lanes 1-3 show extracts from control cells or cells transfected with eIF4A I or eIF4A II , respectively. Lanes 4 and 5 show m 7 GTP-Sepharose-bound material from cells transfected with eIF4A I or eIF4A II , respectively, whereas lanes 6 and 7 show the material derived from those samples after elution from the resin and chromatography on Ni-NTA agarose. Western blotting employed antisera against eIF4A (eIF4A C antibody that recognizes eIF4A I and eIF4A II ; positions of endogenous eIF4A and recombinant eIF4A(His/ Myc-tagged) are shown) and against eIF4G (position shown). isolated on the cap resin if bridging to native eIF4G occurred. m 7 GTP-Sepharose pull-down assays were therefore performed using extracts of transfected cells, and blots were carried out to examine whether eIF4G M or eIF4G C could interact indirectly, e.g. via eIF4A and eIF4G/4E, with the cap resin. Western blot analysis revealed that both fragments of eIF4G were expressed at similar levels (Fig. 4A). As expected, the endogenous fulllength eIF4G was retained on m 7 GTP-Sepharose (Fig. 4A) by virtue of its binding site for eIF4E. In contrast, no signal was detected for either eIF4G M or eIF4G C in these experiments (Fig. 4A), implying that there is no bridging, e.g. via eIF4A, from these fragments to complexes containing endogenous eIF4G. The converse experiment was also performed, i.e. pulling out the His 6 -tagged eIF4G fragments on Ni-NTA agarose and analyzing the bound material for full-length eIF4G. Both fragments were bound by the nickel resin, but full-length eIF4G was not observed (Fig. 4A) again suggesting that no bridging occurs, although both eIF4G domains can bind to eIF4A.
As a further test of both "bridging" and the stoichiometry of eIF4G/eIF4A interaction, cells were transfected with plasmids encoding the tagged forms of eIF4A I or eIF4A II . These proteins were expressed at similar levels to the endogenous eIF4A proteins as judged by Western blotting using the anti-eIF4Ac antibody (Fig. 4B). Cell extracts were applied to Ni-NTA-agarose, and the bound material was analyzed by SDS-PAGE/ Western blotting using anti-eIF4Ac. The tagged eIF4A molecules were clearly retained on the resin, but no endogenous eIF4A was observed, ruling out dimerization of eIF4F complexes containing multiple copies of eIF4A and again demonstrating that eIF4F complexes contain only one molecule of eIF4A (had they contained two, then endogenous eIF4A would also have been observed here).
These data suggest that no such bridging between eIF4G molecules (mediated, e.g. by eIF4A) occurs. This could reflect either (i) a strong "preference" for eIF4A to interact in cis with the two binding sites on a single eIF4G molecule rather than in trans with sites on two different eIF4G molecules or (ii) that only a single eIF4A binding site can actually be occupied at one time on a given eIF4G molecule. Our experiments do not allow us to distinguish between these two possibilities. DISCUSSION The present study provides strong evidence that the mammalian eIF4F complex contains only one molecule of eIF4A, despite the finding that mammalian eIF4G actually possesses two independent sites for interaction with eIF4A (10). Two possible explanations for the 1:1 stoichiometry are (i) that eIF4A cannot bind simultaneously to the two sites, so that it is associated with one or the other, giving rise to the 1:1 stoichiometry indicated by our results or (ii) that eIF4A does bind to both sites on eIF4G at the same time, through different binding sites in the eIF4A protein, so that it is "sandwiched" between these sites as originally proposed by Imataka and Sonenberg (10). Earlier data (10,12) indicate that the C-terminal binding site enhances cap-dependent translation and the binding of ribosomes to mRNA in vitro. However, these studies and those of de Gregorio et al. (22) (who showed that the central "core" of eIF4G could support translation, albeit rather inefficiently) indicate that this site is not essential for mRNA translation. In contrast, binding of eIF4A to the central region is both essential and sufficient for translation (10). Based on these considerations, Morino et al. (12) have suggested that the C-terminal domain of eIF4G plays a modulatory role.
Our data lend support to the ideas initially put forward by Sonenberg and coworkers (10,12) that eIF4G binds only one molecule of eIF4A. Thus, the eIF4F complex of mammals does not appear to differ in overall composition from the likely situation in yeast, where both isoforms of eIF4G each contain only one binding site for eIF4A (23). The residues within the C terminus of eIF4G identified as being required for eIF4A binding (Phe-Val-Arg (10)) are not conserved in the orthologs from wheat (24) or Drosophila melanogaster (25). As noted before (17), the existence of two binding sites in mammalian eIF4G may result in more stable binding of eIF4A than for the forms of eIF4G that only have one such site. This could explain why eIF4A does not co-purify with eIF4G from yeast (26) or Drosophila (27), even though an interaction between eIF4G and eIF4A can be demonstrated in vitro for these species (8,14,15). Mutation of just one of the sites in mammalian eIF4G was enough to reduce by about 10-fold the association of eIF4A with eIF4G in toe-printing experiments (12) although there is ample data showing that eIF4A can interact stably with fragments of eIF4G containing only one binding site (see, for example, Refs. 10, 12, and 17 and this study).
Recent studies from Korneeva et al. (17) used surface plasmon resonance techniques, with purified proteins, to determine the binding affinities of the middle and C-terminal domains of eIF4G for eIF4A and the association and dissociation constants for these interactions. It was found that the middle domain of eIF4G binds eIF4A with a 20-fold higher affinity than the C-terminal fragment. Each domain of eIF4G could bind one molecule of eIF4A, and they reported that the region of eIF4G containing residues 642-1560 (with both binding sites) could bind two molecules of eIF4A. This last conclusion is in disagreement with our own data showing a 1:1 interaction of eIF4A with endogenous full-length eIF4G in vivo. It could be argued that the our approach involving immunoprecipitation may miss weak protein/protein interactions, which dissociate during the work-up of the samples, and that this could account for the FIG. 4. There is no bridging between eIF4G and fragments of eIF4G containing binding sites for eIF4A. A, HEK 293 cells were transfected with plasmids encoding the middle (4G M ) and C-terminal (4G C ) domains of eIF4G, each containing a His 6 tag, or with empty vector. Samples of cell extract were applied to m 7 GTP-Sepharose or Ni-NTA-agarose to purify eIF4F complexes or the tagged eIF4G fragments, respectively. These samples and aliquots of cell lysate were analyzed in parallel by SDS-PAGE followed by Western blotting using the anti-eIF4G C or anti-His 6 antisera as indicated. B, HEK 293 cells were transfected with plasmids encoding eIF4A I or eIF4A II or with empty vector (Ϫ). Expression of the recombinant proteins was verified by SDS-PAGE/Western blotting with the anti-eIF4Ac antibody that detects both eIF4A I and eIF4A II . Samples of extract were applied to Ni-NTA-agarose to purify the tagged eIF4A molecules, and the bound material was subjected to SDS-PAGE/Western blotting using the same antibody. The positions of the endogenous and recombinant (tagged) proteins are indicated. difference between our studies. However, it should be noted that in our studies (see Fig. 3) both domains of eIF4G M and eIF4G C did appear to interact with similar efficiencies with eIF4A I and eIF4A II . Thus, we do not appear to be "losing" eIF4A that is bound to one of the sites during the isolation of complexes from the cell extracts. This would be consistent with the finding of Korneeva et al. (17) that although the K d values for the two interactions do markedly differ, this is due to differences in the association rate constants rather than in the dissociation constants. The lack of bridging between two fragments of eIF4G by eIF4A observed by Korneeva et al. (17) in solution is also entirely consistent with our own findings. The implication of our study that eIF4G only binds one eIF4A molecule at a time is also in agreement with the data of Li et al. (8) who showed that, when endogenous eIF4G is bound to eIF4A III (which binds only to the middle domain of eIF4G), eIF4A I cannot bind to this eIF4G molecule. If eIF4G could simultaneously bind to two eIF4A molecules, both species of eIF4A should be found.
Our data show further that both eIF4A I and eIF4A II can each bind to both domains of mammalian eIF4G independently. Early data (7) suggested that eIF4A II might be preferentially incorporated into eIF4F complexes; our data reveal no gross difference between the binding of eIF4A I and eIF4A II to the two domains of eIF4G. The possibility that eIF4A is sandwiched between the two binding sites in eIF4G requires that eIF4A possesses two different sites for interaction with eIF4G rather than binding through a single site to both regions of eIF4G. To date, the two binding sites in eIF4G have been partially defined. The FVR motif in the C-terminal domain has been noted above, and the key residues in the central region of eIF4G that are required for the binding of eIF4A are hydrophobic ones (10). An important goal of further studies will be to define the binding sites within eIF4A for the two domains of eIF4G. This may be aided by the recent determinations of the three-dimensional structure of yeast eIF4A (28 -30).