Identification of a Nucleic Acid Binding Domain in Eukaryotic Initiation Factor eIFiso4G from Wheat*

Higher plants have two complexes that bind the m7G-cap structure of mRNA and mediate interactions between mRNA and ribosomal subunits, designated eIF4F and eIFiso4F. Both complexes contain a small subunit that binds the 5′-cap structure of mRNA, and a large subunit, eIF4G or eIFiso4G, that binds other translation factors and RNA. Sequence-specific proteases were used to cleave native cap-binding complexes into structural domains, which were purified by affinity chromatography. We show here that eIFiso4G contains a central protease-resistant domain that binds specifically to nucleic acids. This domain spans Gln170 to Glu443 and includes four of the six homology blocks shared by eIFiso4G and eIF4G. A slightly shorter overlapping sequence, from Gly202 to Lys445, had no nucleic acid binding activity, indicating that the N-terminal end of the nucleic acid binding site lies within Gln170 to Arg201. The binding of the central domain and native eIFiso4F to RNA homopolymers and double- and single-stranded DNAs was studied. Both molecules had highest affinity for poly(G) and recognized single- and double-stranded sequences.

The group 4 translation initiation factors, eIF4A, 1 B, E, and G, bind and recruit mRNA to 43 S pre-initiation complexes for translation (reviewed in Refs. 1 and 2). Association of eIF4A and eIF4B confers ATPase, RNA helicase, and RNA annealing activities on the intact complex (1)(2)(3) and is thought to be necessary for melting of RNA secondary structure and ribosome scanning to the AUG start codon. eIF4E binds to the 5Ј-cap structure of mRNAs and associates tightly with eIF4G. These two factors together form the core of the mRNA-binding complex, designated eIF4F. Higher plants have two such complexes: eIF4F, which is homologous to the mammalian complex, and eIFiso4F, which is ϳ10-fold more abundant (4,5). eIF4G and eIFiso4G differ ϳ2-fold in molecular mass and show only 25-30% sequence similarity overall, with ϳ50% similarity in six small homologous regions (6). Yet, despite this substantial divergence, these two factors have very similar functions (1). Either factor alone is sufficient for cap-dependent initiation in vitro (7,8). eIFiso4G and eIFiso4E have been cloned and sequenced from wheat and have deduced molecular masses of 86.5 and 23.5 kDa, respectively (6).
The eIF4G and eIFiso4G subunits of the mRNA-binding complexes act as a central scaffold for binding of other translation factors, proteins, co-factors, and RNAs. These include binding sites within mammalian eIF4G for eIF4E (aa 409 -457) (9), eIF3 (aa 480 -886) (10), and two sites for eIF4A (aa 887-1402 and 478 -883) (10,11). Plant eIFiso4G also binds eIFiso4E near its N terminus (aa 52-90) (12) in a small region with sequence homology to the mammalian factor. Binding of eIF4G and eIFiso4G to RNAs, distinct from the cap recognition function of the intact complex, has been demonstrated by several laboratories and appears to play a critical role in the cap-independent initiation of viral RNAs. Sha et al. (13) calculated the equilibrium binding constants of recombinant wheat eIFiso4G for both capped and uncapped mRNAs. They found that the binding constants for uncapped message were comparable to those of eIFiso4E and eIFiso4F for capped mRNA, clearly indicating that the large subunit also binds mRNA efficiently and that it does not discriminate between capped and uncapped messages. These findings are consistent with the role of eIF4G in the cap-independent translation of viral mRNAs, following cleavage and loss of the eIF4E binding domain by viral proteases (10,14). In functional assays for initiation complex formation, Pestova et al. (15) have shown that only eIF4A and the central third of human eIF4G (aa 457-932) are required for binding of encephalomyocarditis virus RNA to 43 S ribosomal complexes, and that this central eIF4G peptide binds, and cross-links to, the viral internal ribosome entry site (IRES). Wheat eIF4F and eIFiso4F were also reported to bind poly(A) mRNA (16), which implies that the cap-binding complex interacts with both the 5Ј and 3Ј termini of intact mRNAs, perhaps ensuring that only intact messages are translated. The binding of eIFiso4G to other proteins, in addition to translation factors, has also been reported. The N-terminal region of eIFiso4G appears to bind poly(A)-binding protein (17), thereby increasing its activity. Additionally, recombinant wheat eIFiso4G binds to, and promotes the bundling of microtubules in vitro, and co-localizes with microtubules in vivo, suggesting a role for eIFiso4F in localizing the translational machinery to the cytoskeleton (18). The specific sites and structural motifs in eIFiso4G or eIF4G that mediate most of these RNA and protein interactions have not yet been identified.
In this study, we have used limited proteolysis and affinity chromatography to isolate functional domains of the mRNAbinding complexes of wheat. Protease-resistant domains were fractionated by affinity chromatography on phosphocellulose (phosphate interactions), oligo(dT)-cellulose (phosphodiester or base interactions), and m 7 GTP-sepharose (cap recognition). Peptides eluted from the different affinity columns were purified by reverse phase HPLC, sequenced to determine their N terminus, sized by mass spectrometry, and compared with published sequences of cloned factors to localize each peptide within the primary amino acid sequence. We report here the identification of a central protease-resistant domain in eIFiso4G, which binds both nucleic acids and phosphocellulose. The specific affinities of both intact eIFiso4F and the central domain of eIFiso4G were determined for RNA homopolymers and ss-and dsDNAs.

EXPERIMENTAL PROCEDURES
Reagents-Sequencing grade trypsin and endoproteinase glu-C were from Boehringer Mannheim. Oligo(dT)-cellulose and m 7 GTP-Sepharose 4B were from Amersham Pharmacia Biotech. Phosphocellulose was from Whatman. Agarose-conjugated poly(A), poly(C), poly(G), poly(U), ssDNA, and dsDNA were from Sigma. Oligonucleotides were from Sigma. All other chemicals were purchased from Sigma, U. S. Biochemicals, Bio-Rad, or Fisher Scientific.
Protein Purification-mRNA cap-binding complexes were purified from wheat germ by a modification of the method of Lax et al. (Ref. 19; note that eIFiso4G is designated eIF-4B in this reference). Wheat germ (150 g) was ground with 10% each fine sand and polyvinylpolypyrrolidone (20), homogenized in four volumes of iced buffer (0.25 M sucrose, 200 mM Tris/Hepes, pH 8.6, 400 mM potassium acetate, 50 mM magnesium acetate, 5 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride, and 0.07 mg/ml soybean trypsin inhibitor), and centrifuged 10 min at 10,000 ϫ g. The supernatant was decanted through 70 m nylon mesh, centrifuged 2 h at 368,000 ϫ g, precipitated with 70% ammonium sulfate, and dialyzed against buffer B120 (buffer "B" contains 20 mM Hepes/KOH, pH 7.6, 1 mM dithiothreitol, 0.1 mM EDTA, 10% glycerin, and the concentration of KCl, in mM, indicated by the suffix). Material from five or six preparations was adsorbed to a 50-ml m 7 GTP-Sepharose 4B column, and mRNA cap-binding complexes were eluted with 100 M m 7 GTP in B120. Approximately 20 mg of protein was recovered and applied to a 30-ml phosphocellulose column. eIF4A did not bind, and eIFiso4F eluted with a linear salt gradient (B80 to B1000) at approximately 150 mM KCl. Fractions containing eIFiso4F were concentrated and desalted in Centricon-10 spin cartridges (Amicon). Preparations of cap-binding complexes at this stage contained both eIFiso4F and eIF4F, and were used to identify peptides from both complexes which bound to affinity resins (Figs. 1-3). In replicates of these experiments and all subsequent binding specificity studies (Figs. 5 and 6 and Tables I and II), eIFiso4F was further purified by HPLC using a DEAE column (Waters AP Mini, 5 ϫ 50 mm) eluted with a linear salt gradient (B80 to B500). At each step, fractions were analyzed for protein content by Bradford assays or spectrophotometry at 280 nm, and by SDS-PAGE (21) on 10% gels stained with Coomassie Brilliant Blue R250 or silver (22).
Proteolytic Digestions-Limited protease digestion under native conditions was according to Geisler and Weber (23). Digestion reactions contained 10 mM Hepes/KOH, pH 7.5, 50 mM KCl, 30% glycerol, 10 mM ␤-mercaptoethanol, cap-binding complexes or purified eIFiso4F (0.6 mg/ml for trypsin or 0.9 -1.5 mg/ml for endoproteinase glu-C), and either 1% trypsin or 2% endoproteinase glu-C (by mass, relative to protein sample). Trypsin cleaves on the carboxyl side of arginine and lysine residues, and endoproteinase glu-C cleaves on the carboxyl side of aspartate and glutamate residues. Incubations were at room temperature for the times indicated. Reactions were stopped by freezing in liquid nitrogen, and the cleavage products analyzed on 14% gels by SDS-PAGE and silver staining.
Affinity Fractionation of Protein Digests-For the results reported in Fig. 2, 300 -500 g of digested cap-binding complexes were chromatographed on 2-ml phosphocellulose, oligo(dT)-cellulose, or m 7 GTP-Sepharose columns, and eluted with a 12-ml linear salt gradient (B100 to B1000). All fractions were analyzed for protein composition by SDS-PAGE on 14% gels with silver staining, and by fluorescence spectroscopy at excitation and emission wavelengths of 280 and 350 nm, respectively. Only selected fractions are shown.
Binding studies with ribonucleotide homopolymer and ss and dsDNA resins were carried out in spin columns using 20 -100 g of protein and 100 l of packed resin in microcentrifuge filter cups (MSI, Westboro, MA; 0.45-m acetate filters). Solutions (300-l aliquots) were spun through the resin at 500 -1,000 rpm for 2-5 min; at least three washes (nine column volumes) were used at each step. Columns were equilibrated in B50 and eluted sequentially with 0.1, 1, or 4 mM NTPs or dNTPs, 200 M oligo(dT) 6 or oligo(dT) 12 , or with increased KCl, as indicated. All fractions were analyzed as above, but only selected fractions are shown. eIFiso4F, endoproteinase glu-C digests of eIFiso4F, and purified E31 were chromatographed on all affinity resins at least twice.
Purification, Sequencing, and Mass Spectrometric Analysis of Peptide Fragments-Proteolytic cleavage products that bound affinity resins were further purified by reverse phase HPLC on a C18 column (Waters, 3.9 ϫ 150 mm) equilibrated in 0.1% trifluoroacetic acid. Elution was with a linear acetonitrile gradient (0 -80%) containing 0.1% trifluoroacetic acid. Absorbance was monitored at 205 nm. Peaks were concentrated in a Centrivap concentrator and analyzed by SDS-PAGE and silver staining. The N-terminal sequence of purified peptides was determined by Edman degradation at the UCR Biotechnology Instrumentation Facility, or at the Stanford Biotechnology Facility. Peptide masses were determined by time of flight mass spectrometry by Curtis Monnig in the Department of Chemistry at UCR, or at the Harvard Biotechnology Facility. The locations of peptides within wheat eIFiso4G and eIFiso4E were determined by comparison with published sequences (Ref. 6; GenBank accession numbers M95747 and M95819, respectively). Other sequences used for comparison were; maize eIFiso4G, 2 thale cress eIFiso4G (AC005170), 3 (29) and refined by hand.

Identification of Protease-resistant Domains in Wheat mRNA
Cap-binding Complexes Which Bind Phosphocellulose, Oligo(dT)-Cellulose, and m 7 GTP-Sepharose-Preparations of wheat cap-binding complexes contain eIFiso4F, a small amount of eIF4F, and a trace amount of eIF4A which is bound to eIF4F ( Fig. 1 left), as determined by immunoblotting (data not shown). These preparations were the starting point for identi- fying peptides from either complex that bound to affinity resins. Purified eIFiso4F (Fig. 1, left) was used in the nucleic acid binding experiments described below.
Cap-binding complexes were cleaved into protease-resistant domains by digestion with trypsin or endoproteinase glu-C under native conditions (Fig. 1). Several large peptides were reproducibly generated that withstood even prolonged digestion. After 16 h of incubation with trypsin, two main bands remained, which migrated on SDS-PAGE with apparent molecular masses of 28 and 26 kDa. Minor bands between 20 and 24 kDa were more abundant in some preparations. The breakdown pattern generated by endoproteinase glu-C was more complex; approximately 10 protease-resistant bands, with apparent molecular masses between 14 and 31 kDa, were generated in a 9-h digest. Digestion times of 12 h for trypsin and 8 h for endoproteinase glu-C were used in the experiments shown below.
In order to identify and purify peptides with specific binding affinities, digests of cap-binding complexes were chromatographed in parallel on phosphocellulose, oligo(dT)-cellulose, and m 7 GTP-Sepharose. When trypsin digests were applied to phosphocellulose only T28, the largest stable fragment, bound and was eluted from the column in ϳ150 mM KCl ( Fig. 2 top  left, lane 4). In contrast, none of the peptides generated by trypsin digestion bound to oligo(dT)-cellulose (Fig. 2 bottom  left). The second major band at ϳ26 kDa was primarily undigested eIFiso4E, which was resistant to trypsin digestion for up to ϳ16 h. It bound to m 7 GTP-Sepharose (data not shown) and had no affinity for either phosphocellulose or oligo(dT)-cellulose (Fig. 2). This binding pattern was the same as for native eIFiso4E (data not shown).
Cap-binding proteins digested with endoproteinase glu-C were chromatographed as above (Fig. 2 right). Major bands with apparent masses of 31, 23, and 22 kDa were retained on phosphocellulose, and eluted in 150 -200 mM KCl (Fig. 2 top  right, lanes 4 and 5). The two smaller bands were also retained on m 7 GTP-Sepharose (data not shown), while the 31-kDa band bound to oligo(dT)-cellulose (Fig. 2, bottom) and eluted at ϳ150 mM KCl. Further fractionation of the 31-kDa band by RP-HPLC revealed two peptides with very similar size and hydrophobicity, running as a barely resolved doublet on SDS-PAGE, and eluting from a C18 column at only a 1% difference in acetonitrile concentration (Fig. 3). Only the larger of these two peptides, corresponding to the second RP-HPLC peak, was present in digests of purified eIFiso4F, and was designated E31. E31 binding to phosphocellulose and oligo(dT)-cellulose was due to its direct interaction with the resins, rather than to a protein:protein interaction with the second peptide, or with one of the other peptides retained on the columns. Purified E31, isolated from digests of purified eIFiso4F by phosphocellulose chromatography and DEAE-HPLC, had the same affinity and elution profile on phosphocellulose and oligo(dT)-cellulose as did E31 in digests (data not shown).
Peptide Assignments-Amino acid sequencing and mass spectrometry were used to identify those peptides that bound to affinity resins. Seven amino acid residues were sequenced from the N terminus of each fragment; each was sequenced twice from separate purifications, and E31 was sequenced following both phosphocellulose and oligo(dT)-cellulose purification. The C-terminal residues were calculated from the total mass and amino acid composition. T28 and E31 were found to be overlapping peptides from the central portion of eIFiso4G with 242 amino acids, from Gly 202 to Glu 443 , in common (Table I). The additional 32 amino acids at the N terminus of E31, from Gln 170 to Arg 201 (Fig. 4A), are required for oligo(dT)-cellulose binding activity. This sequence thus appears to form part of a nucleic acid recognition site.

FIG. 2. Affinity fractionation of peptides from cap-binding complexes.
Trypsin (left) and endoproteinase glu-C (right) digests were chromatographed on either phosphocellulose (top, PC) or oligo(dT)-cellulose (bottom, dT). Elution profiles were monitored by fluorescence. Insets, SDS-PAGE of selected fractions, run on 14% gels and silver-stained. i4F, eIFiso4F standard. T, trypsin digest of cap-binding complexes. E, endoproteinase glu-C digest of cap-binding complexes. Numbered lanes correspond to labeled fractions in the elution profiles. The molecular masses of standard proteins in kDa are indicated to the left of each gel.

FIG. 3. Separation of E31 and a homologous eIF4G domain by RP-HPLC.
An endoproteinase glu-C digest of cap-binding proteins was sequentially chromatographed over phosphocellulose and oligo(dT)-cellulose to yield the 31-kDa band, which was further fractionated by RP-HPLC. Only part of the 60-min RP-HPLC elution profile is shown; no other peaks were detected. Inset, SDS-PAGE of the initial endoproteinase glu-C digest (E), the 31-kDa band (31), and the two RP-HPLC peaks (1 and 2). Fractions were analyzed by SDS-PAGE on a 14% gel and silver-stained. The molecular masses of standard proteins in kDa are indicated to the left.
The second 31-kDa peptide, present in endoproteinase glu-C digests of cap-binding complexes, originated from eIF4G. Its N-terminal sequence was KKYVVGK, which is located 20 amino acids N-terminal to the second region of homology shared by eIFiso4G and mammalian eIF4G (12), beginning at Lys 215 in eIFiso4G (Fig. 4A, Region 2). When the sequences of wheat eIF4G and eIFiso4G were compared in this region, the start site of the eIF4G peptide was seen to lie between those of E31 and T28, and to include only seven of the 32 additional residues present in E31 (Fig. 4A). There is limited sequence conservation between factors and species in this region, but several short conserved sequences and amino acids are present in almost all of the factors sequenced to date (Fig. 4B). Most prominent are two small charge clusters, corresponding to KAD 194 and EKDR 209 in eIFiso4G, which flank conserved Trp, Ser, and Arg residues present at the N terminus of the eIF4G peptide. These elements are also present in yeast eIF4G, but are much more widely spaced (Fig. 4B).
Digestion with endoproteinase glu-C produced two peptides of ϳ22-23 kDa, which were retained on phosphocellulose (Fig.  2, top right, lanes 4 and 5). These bands were identified as eIFiso4E peptides by their affinity for m 7 GTP-Sepharose (data not shown), their co-migration with the two main fragments generated by digestion of purified eIFiso4E (data not shown), their intense fluorescence (eIFiso4E has an unusually high tryptophan content), and by sequencing of the smaller peptide, which was designated E21. The N-terminal sequence of E21 was AKGPHKL and its mass was 20,852 Da, corresponding to Ala 29 -Val 209 of eIFiso4E (6). Truncation of the N-terminal 28 amino acids did not alter the affinity of E21 for the cap structure, as determined by fluorescence quenching analysis (data not shown). This finding agrees with results reported by Cha-van et al. (30), who identified a trypsin-resistant core domain in human eIF4E (aa 47-182), which also retained its affinity for m 7 GTP. Native eIFiso4E does not bind either phosphocellulose or oligo(dT)-cellulose (data not shown), but is retained on both columns when bound to eIFiso4G. Retention of the eIFiso4E peptides on phosphocellulose could thus have resulted from their interaction with E31. This proved not to be the case, however, as purified E21 alone bound to phosphocellulose (data not shown). Apparently the loss of seven Asp and Glu residues from the N terminus leaves several unpaired positive charges, which can interact with phosphate groups.
Purified eIFiso4F and E31 Bind Specifically to Nucleic Acids-Intact eIFiso4F could not be eluted from oligo(dT)-cellulose by free dNTPs, NTPs, or NMPs, even at concentrations as high as 4 mM (data not shown), but were readily eluted by 200 M oligo(dT) 12 (Fig. 5, top). As the concentration of phosphate groups in 4 mM NTP is 5-fold higher than that of the oligo solution, this interaction cannot be due simply to charge. In addition, oligo(dT) 6 failed to elute eIFiso4F (Fig. 5,  bottom), indicating that the minimum length of the target sequence required for eIFiso4F binding was between 7 and 12 nucleotides.
The binding specificities of intact eIFiso4F and E31 were determined for RNA homopolymers and random sequence ssand dsDNAs (Table II and Fig. 6). The binding specificity of purified E31 was the same as for E31 in digests, and in each case where binding occurred, free nucleoside mono or triphosphates failed to elute eIFiso4F or E31 from the columns (Fig. 6 and data not shown). eIFiso4F and E31 bound to poly(G) with the highest affinity, and showed no interaction with poly(C). Unlike eIFiso4F, E31 showed no affinity for poly(U) and weak affinity for poly(A), being retarded on poly(A)-agarose but eluting in the wash fractions (data not shown). T28, which did not bind oligo(dT) (Fig. 2), also did not bind poly(G) (data not shown). These results indicate that the central domain of eIFiso4G mediates binding of the eIFiso4F complex to poly(G), oligo(dT), and phosphocellulose, and that this domain recognizes both single-and double-stranded sequences. The lack of interaction of E31 with poly(A) and poly(U) suggests that these lower affinity binding sites lie outside of the central domain. However, it is possible that these nucleic acids are also bound FIG. 4. Peptide start sites and amino acid sequence comparison of eIF4G isoforms from various species. A, localization of the start sites for the phosphocellulose (T28) and oligo(dT)-cellulose (E31) binding domains of wheat eIFiso4G (wheat i4G), and for the homologous domain from wheat eIF4G (4G pep , wheat 4G). The eIFiso4G sequence is shown from Gln 170 through Leu 218 , with the corresponding eIF4G sequence below (see Footnote 2). Residues obtained from N-terminal sequencing are double underlined. The second block of homology shared by mammalian and plant factors begins at Lys 215 in eIFiso4G (Region 2). B, conserved sequences around the start site of the oligo(dT) binding domain are shown in bold. Cress, thale cress (Arabidopsis thaliana); i4G, eIFiso4G; 4G, eIF4G; 4G2, eIF4G2. Where known, amino acid residues are numbered at the right. A consensus sequence is given, with spacing between the conserved residues calculated as the average number of intervening amino acids from all species except yeast. by the central domain, but that additional sequences located elsewhere in the molecule are required to stabilize their interactions. DISCUSSION We have shown that wheat eIFiso4G contains an endoproteinase glu-C-resistant structural core, which binds nucleic acids with base specificity and poly(G) with highest affinity. This domain is centrally located in eIFiso4G, and spans four of the six small regions of homology it shares with mammalian eIF4G (Fig. 7) (1, 6). A corresponding domain from eIF4G was also purified, and our results suggest that it has many of the same properties; it was resistant to endoproteinase glu-C digestion, bound oligo(dT)-cellulose and phosphocellulose, migrated closely with E31 on SDS gels, and had very similar hydrophobicity. The N-terminal end of the nucleic acid recog-nition site in eIFiso4G lies within Gln 170 -Arg 201 , because the shorter tryptic peptide (T28) lacks these 32 residues and fails to bind nucleic acids. Only the later third of this sequence is conserved among eIF4G factors, and only the last seven amino acids, corresponding to Val 195 -Arg 201 of eIFiso4G, are also present in the eIF4G peptide (Fig. 4). This suggests that the N-terminal end of the nucleic acid binding site lies within these seven residues, which form part of an additional short block of homology centered at the N terminus of the tryptic peptide (Figs. 4B and 7). This sequence, from Lys 192 to Arg 209 , is ϳ50% similar in eIF4G isoforms from all species, a comparable level of homology to the other six conserved regions (1). Further studies of purified eIF4G are required to define structural and functional properties of its central domain.
The central third of eIF4G is known to have several important functions. First, Pestova et al. (15) have shown that eIF4A and the central third of eIF4G are alone sufficient for capindependent initiation of encephalomyocarditis virus RNA, and that the eIF4G fragment cross-links to the viral IRES. The wheat eIF4G and eIFiso4G core domains identified here would be included in this larger 476-amino acid fragment, and could be involved in IRES recognition or binding.
Second, it has been known for some time that chromatography of cap-binding proteins on phosphocellulose depletes the complex of eIF4A. The extent to which this occurs varies between laboratories. We have found that, for the plant factors, eIF4F is substantially depleted of eIF4A, while eIFiso4F is completely stripped of eIF4A, both on phosphocellulose (Fig. 1) and oligo(dT)-cellulose (data not shown). Why the binding of eIFiso4G to eIF4A and phosphocellulose should be mutually exclusive was not clear, given the lack of overlap in their recognition sites (Fig. 7); the eIF4A binding site on wheat eIFiso4G was mapped to the N terminus of the molecule by Metz and Browning (12), using eIFiso4G deletion constructs expressed in a yeast two-hybrid system. In contrast, two eIF4A binding domains have now been identified in mammalian eIF4G, one in the C-terminal third of the molecule (10), and a second in the central third (11). Only the eIF4A binding site in the central third of eIF4G is conserved in eIFiso4G, and mutations in this sequence of the mammalian factor abolished eIF4A binding activity (11). This site corresponds to the second FIG. 5. Competitive elution of eIFiso4F from oligo(dT)-cellulose by nucleic acids. eIFiso4F was applied to oligo(dT)-cellulose spin columns, which were successively washed with B50, either 200 M oligo(dT) 12 (top) or oligo(dT) 6 (bottom), and with B200. i4F, eIFiso4F; i4G, eIFiso4G; i4E, eIFiso4E; FT, flow-through fraction from protein application.  (9) is indicated by a white box within the schematic, and regions which bind eIF3 (10) and eIF4A (10,11) are labeled. The locations of T28 and E31 within wheat i4G are represented by black lines above the factor schematic, and their binding specificities for phosphocellulose (PC) and nucleic acids (NA) are indicated. block of homology shared by eIF4G and eIFiso4G (Fig. 4, Region 2; Fig. 7). As RNA recognition motifs are typically 50 -100 amino acids long (31), the eIF4A and nucleic acid binding sites may well overlap, with the loss of eIF4A during chromatography resulting from direct competition. This has interesting implications for the mode of action of the mRNA cap-binding complexes, which perhaps deliver eIF4A to the message, and then bind mRNA directly through the same site to stabilize the preinitiation complex.
Third, Lamphear et al. (10) have shown that a ϳ407-amino acid peptide from the central third of eIF4G binds to eIF3 (Fig.  7). Further proteolysis, resulting in the removal of ϳ50 amino acids from the C terminus of the peptide, abolished this activity. Thus, at least part of the eIF3 recognition sequence in the mammalian factor lies carboxyl to and outside of the core domain defined here, while the nucleic acid binding site in both plant factors begins at the N-terminal boundary of this structural core. Based on these observations, it seems unlikely that these two sites overlap.
Finally, one can speculate on why a translation factor involved in the binding and recruitment of mRNA to ribosomes should also bind ss-and dsDNA. Such dual specificity in RNAbinding proteins is not unusual, and may have no biological significance. However, several examples of RNA-binding proteins that recognize specific DNA sequences have now come to light (32)(33)(34). The cap-binding translation factors may also have multiple functions. eIF4E has been found localized to the nucleus in both mammalian (35) and yeast cells (36), and it appears to regulate the expression of cyclin D1 at both the transcriptional and post-transcriptional levels (37). eIF4E has also been shown to cooperate with two oncogenes, v-myc and E1A, in the transformation of primary fibroblast cultures (38). Additional functions for eIFiso4G have also been reported. eIFiso4G was shown to bind, bundle and co-localize with microtubules in maize cells (18), and to promote their growth by end-to-end annealing in vitro (24), suggesting possible roles for eIFiso4F in the localization of translational components to the cytoskeleton (18) or in mRNA transport (1). These relatively ancient RNA-binding proteins may therefore have other functions in both the nucleus and cytoplasm, beyond formation of the 43 S preinitiation complex in translation.