Mutational analysis of the functional domains of the large subunit of the isozyme form of wheat initiation factor eIF4F.

The isozyme form of plant eukaryotic initiation factor 4F (eIF(iso)4F) contains two subunits: p28, a cap-binding protein, and p86. To identify the functional domains of p86, truncations of the p86 cDNA were made, and the protein was expressed in Escherichia coli and purified. The deletion mutants were tested for the ability to bind the p28 subunit by two methods. In addition, these deletion mutants were evaluated in vitro by the ability to catalyze eIF4A and RNA-dependent ATP hydrolysis and to support polypeptide synthesis. The loss of the ability to bind p28 occurs within the first 90 amino acids of the N terminus and abrogates the ability of p86 to participate in translation initiation and bind to eIF4A, but does not affect ATP hydrolysis. Up to 299 amino acid residues from the C terminus of p86 must be deleted before an effect is observed on the ATP hydrolysis activity. Thus, the p28 binding and ATP hydrolysis activities appear to lie on two separate domains and are functionally uncoupled. In addition, at least a portion of the eIF4A binding domain appears to be in close proximity to the p28 binding domain and is also uncoupled from the ATP hydrolysis activity.

The initiation of eukaryotic protein synthesis is a complicated process and involves several initiation factors (for recent reviews, see Refs. 1 and 2). The precise mechanism of binding the mRNA and bringing it to the 40 S subunits is unknown, and several models have been proposed (3)(4)(5). One of the initiation factors, eIF4F, 1 contains two subunits, eIF4E, a cap-binding protein and eIF4G, a protein of largely unknown function. Another initiation factor, eIF4A, is frequently found associated with eIF4F depending upon the method of purification (6,7).
The subunits of mammalian eIF4F are believed to be involved in regulation of protein synthesis. The large subunit of mammalian eIF4F, eIF4G, is cleaved by picornaviral proteases, severely affecting the ability of the cell to translate capped mRNAs (8). These proteases cleave mammalian eIF4G into N-terminal and C-terminal fragments (9,10). The N-terminal fragment binds the mammalian cap-binding protein, and the C-terminal fragment binds eIF4A and eIF3 (9). The Cterminal fragment alone has been shown to promote cap-independent translation of mRNAs containing an internal ribosome binding site (10).
Phosphorylation of mammalian eIF4E is believed also to be involved in regulation of initiation. However, there is no clear difference in the ability of phosphorylated eIF4E to either bind m 7 GTP-Sepharose or interact with eIF4G (1). In studies where the phosphorylation state of eIF4E was altered, there were also changes in the phosphorylation state of other initiation factors (11)(12)(13)(14). Consequently, the exact mechanism of this regulation remains to be elucidated. Recently, another protein was identified that binds to mammalian eIF4E. This protein, eIF4E-BP1 or PHAS-I, is phosphorylated in response to insulin in mammalian cells (15,16). There is now considerable evidence that this protein sequesters eIF4E and prevents eIF4E from interacting with eIF4G (17). The phosphorylated form of eIF4E-BP1 is unable to bind eIF4E and releases any bound eIF4E (15). Several kinases in the signal transduction pathways have been implicated in this process (18).
Wheat translation initiation factor eIF(iso)4F, the isozyme form of eIF4F, contains two subunits, p28 and p86 which are antigenically distinct from the two eIF4F subunits isolated from wheat germ (19). The isozyme form of eIF4F appears to be unique to higher plants and has not been identified as yet in other eukaryotes. The protein has been observed in wheat, maize, and cauliflower (20), and cDNA expressed sequence tags for the subunits have been identified for rice and Arabidopsis thaliana. 2 The p28 subunit is a cap-binding protein (19). Analogous to mammalian eIF4G, no specific function(s) has been assigned to p86. The p86 subunit is considerably smaller (86 kDa) than the predicted molecular mass for human eIF4G (154 kDa (21)) or yeast eIF4G (104 kDa (22)). There are seven regions of similarity between p86 and mammalian and/or yeast eIF4G (see Fig. 1A). eIF(iso)4F has the same functional properties as wheat germ eIF4F: 1) it substitutes for eIF4F in an in vitro translation system deficient in eIF4F; 2) it substitutes for eIF4F in supporting the binding of mRNA to 40 S ribosomal subunits; 3) it exhibits RNA-dependent ATP hydrolysis activity in the presence of eIF4A; and 4) it exhibits ATP-dependent RNA unwinding activity in the presence of eIF4A (19,(23)(24)(25).
Plant eIF(iso)4F carries out the same functions as eIF4F, but with a large subunit approximately one-half the size of mammalian eIF4G. The smaller size and the ability to express a functionally active protein in Escherichia coli for p86, makes it ideal for identifying the functional domains of this protein. In the present paper, we initiate mapping of the domains of the p86 subunit to better understand its interaction with the capbinding protein and other initiation factors. Truncation mutations of p86 were made and tested for the ability to bind to p28, to participate in ATP hydrolysis, to support polypeptide synthesis, and to interact with eIF4A.

EXPERIMENTAL PROCEDURES
Materials-Wheat germ high-salt-washed ribosomes, 40 -70% ammonium sulfate fraction, and highly purified fractions of eIF3 (26), eIF4A (26), eIF4C (27), eIF(iso)4F (20), and recombinant p28 (28) were prepared as described previously. Purification of eIF4B is described elsewhere. Preparation of satellite tobacco necrosis virus RNA was as described previously (29). Restriction enzymes, T4 DNA ligase, and oligonucleotide primers were from Life Technologies, Inc. Sequencing reagents were from United States Biochemical Corp. DNA amplification reagents were from P/E Express. SDS-PAGE was performed as described previously (30). Protein concentrations were determined by the method of Bradford (31) with bovine serum albumin as a standard.
Construction of Truncated Forms of p86 -Construction of the plasmid for expression of wild-type p86 (pET3D-p86) was as described previously (28). The plasmid containing the cDNA encoding p86 was used as a template for amplifying portions of the coding region. Oligonucleotide primers introducing an NcoI site at the desired N-terminal end and a BamHI site at the desired C-terminal end were used to amplify the cDNA (see Table I). The amplified DNA products were restricted, purified, and ligated into NcoI/BamHI-cut pET3d (Novagen) as described previously (28). Construction of each truncation mutant was verified by DNA sequencing.
Expression and Purification of Mutant Proteins-p86 wild-type and mutant proteins were expressed and purified as described previously (28) except the clarified cell extract was applied to a 2-ml phosphocellulose column (Whatman P11) equilibrated in buffer B (20 mM HEPES-KOH, pH 7.6, 1 mM dithiothreitol, 0.1 mM EDTA, 10% glycerol) containing 100 mM KCl. Expressed protein was eluted from the column with a 20-ml linear gradient from 100 to 250 mM KCl in buffer B (wild-type, N-52, N-90, N-136, N-186) or from 150 to 400 mM KCl in buffer B (C-462, C-489, C-511). To remove a nonspecific E. coli ATP hydrolysis activity, 0.5 mg of the phosphocellulose column eluate was dialyzed into buffer B-100 and applied to a 1-ml DEAE-cellulose column (Whatman DE52) equilibrated in buffer B-100. The flow-through fractions (500 l) contained the p86 mutants, and the E. coli ATP hydrolysis activity was retained on the column. The flow-through fractions were analyzed by SDS-PAGE and pooled.
Chromatography of Recombinant Complex on m 7 GTP-Sepharose-Each p86 mutant (3000 pmol) was combined with purified p28 subunit (2000 pmol) and applied to a 1-ml m 7 GTP-Sepharose column (Pharmacia Biotech Inc.) equilibrated in buffer B-100 (28). The complex was eluted with 0.1 mM m 7 GTP in buffer B-100. Fractions (500 l) were collected and analyzed by SDS-PAGE.
eIF4F-dependent Polypeptide Synthesis-Polypeptide synthesis was carried out as described previously (20,26) and contained 10 pmol of recombinant p28 and 10 pmol of p86 or mutant protein (DEAE flowthrough). The values reported are the average of three experiments.
ATP Hydrolysis-The standard ATP hydrolysis reaction was carried out as described previously (24) and contained 1 g of poly(U), 5 g of purified wheat germ eIF4A, and 20 pmol of p86 or mutant protein (DEAE flow-through). The ATP hydrolysis in this assay is dependent upon the addition of both eIF4A and poly(U) (24). The values reported are the average of six experiments.
Cloning of p28, p86, and eIF4A into Yeast Two-hybrid Vectors-Matchmaker 2-hybrid system vectors (pGBT-9 and pGAD-424) and control plasmids (pLAM5Ј, pTD1, and pVA3) were from Clontech. Two oligonucleotides that introduced an EcoRI site directly in front of the initiator AUG of p86 and a BamHI site directly 3Ј to the stop codon were used to amplify the coding region of the plasmid containing the p86 cDNA. The amplified DNA product was cloned into EcoRI/BamHI site of pGTB-9 (binding domain) by standard methods. Similar methods were used to place the p28 and eIF4A coding regions into yeast two-hybrid vectors. All the constructs were confirmed by DNA sequencing.
Construction of the pGBT-9/N Vector-The pGBT-9 plasmid was altered to include an NcoI site between the EcoRI and BamHI polylinker sites to allow direct subcloning of the p86 mutant cDNAs and eIF4A cDNA. The resulting vector was termed pGBT-9/N. Construction of the vector was confirmed by DNA sequencing across the polylinker site on both strands. The p86 deletion mutants were excised from the pET3d expression vector with NcoI and BamHI and inserted into pGBT-9/N restricted with NcoI and BamHI by standard procedures.
Two-hybrid Interaction Assay-Sets of two plasmids were co-transformed as indicated (see Tables III and IV) into yeast strain SFY-526 as per the manufacturer's instructions (Clontech). Interaction in Table III was scored by the appearance of blue color using the ␤-galactosidase filter assay as per the manufacturer's instructions (Clontech). Interaction in Table IV was quantitated by the ␤-galactosidase liquid assay as per the manufacturer's instructions (Clontech).

RESULTS AND DISCUSSION
A series of deletion mutants from the N terminus and C terminus of p86 were made to identify the functional domains. The maximum deletions, N-186 and C-462, were designed to include only the region of maximum similarity (approximately amino acids 204 to 432, blocks 2-6 in Fig. 1A) to mammalian and yeast eIF4G (28). Within this region are blocks that range from 31-50% similarity to mammalian and/or yeast eIF4G (see Fig. 1A). The conservation of these blocks suggest that they are involved in the function of the protein: interaction with other   components of the translational machinery or enzymatic activity. There are three sets of blocks (3, 4, and 7) that have similarity to mammalian eIF4G or yeast eIF4G, but not both. The N-terminal end of p86 is much shorter than either mammalian or yeast eIF4G, suggesting that these regions of mammalian and yeast eIF4G may be involved in regulation rather than function. The C terminus of yeast eIF4G is much shorter than mammalian eIF4G or p86 (see Fig. 1A). There is a block of similarity (Fig. 1A, block 7) shared by p86 and mammalian eIF4G that occurs after the yeast eIF4G terminates. The function of this region is unknown at this time; however, this region of p86 is implicated in the binding of microtubules (32).
A schematic of the N-terminal and C-terminal truncation mutants of the p86 subunit of eIF(iso)4F is shown in Fig. 1B. These mutants were expressed in E. coli and purified. An SDS-PAGE of the mutant forms is shown in Fig. 2. There are some minor degradation products in the protein preparations, particularly of the C-terminal truncation mutants. The presence of these products suggests that certain regions in the C terminus are probably more exposed in the truncation mutants and are more susceptible to protease digestion.
The ability of the mutants to form a complex with p28 was measured by retention of the mutant protein on a m 7 GTP-Sepharose column. As shown in Table II, N-terminal deletions past residue 52 (N-90, N-136, N-186) were unable to form a complex with p28. The C-terminal deletions were all able to form a complex with p28.
The yeast two-hybrid system was used to confirm the p28-p86 interactions. A very intense blue color was obtained when wild-type p86 (binding domain vector) and p28 (activation domain vector), as well as the positive controls (pVA3 and pTD1), were co-transformed into yeast cells (see Table III). The negative controls or the plasmids alone (p86, p28, pLAM5Ј, pTD1, or pVA3) gave no color (see Table III). Combinations of p86 with pTD1 or pVA3 or combinations of p28 with pTD1 or pVA3 were also negative. When the p28 was placed into the binding domain vector and p86 into the activation domain vector, an interaction of similar intensity was observed. These results show that the regions of the two molecules that interact are not impaired by the addition of any vector-specific sequences.
A new binding domain vector (pGBT-9/N) was constructed to facilitate the subcloning of the p86 truncation mutations from pET3d into the binding domain vector of the yeast two-hybrid system. As shown in Table III, results identical to the retention on m 7 GTP-Sepharose were obtained for p28 binding by p86 deletions mutants with the two-hybrid system. Interaction of p28 was no longer possible when mutants truncated past residue 52 were used, whereas all the C-terminal mutants were able to interact with p28. These data indicate that the site of interaction of p86 with p28 resides between residues 52 and 90. This result is consistent with the observation that the sequence from amino acids 62 to 73 in p86 (block 1 in Fig. 1A) contains a motif identified as necessary for the binding of the mammalian eIF4E to mammalian eIF4G or eIF4E-BP1 (33).
The purified mutants were assayed for the ability to support polypeptide synthesis in the presence of p28 in an eIF(iso)4Fdependent system from wheat germ (28). As shown in Table II, N-terminal deletions past amino acid 52 lost the ability to support translation in the assay system. The N-52 mutant consistently showed a slightly higher activity in this assay system. The removal of these amino acids may be changing the In binding domain vector, pGBT-9/N. b In activation domain vector, pGAD-424. c Positive and negative controls for two-hybrid systems. conformation of the protein such that certain domains are more exposed and increase the activity of the complex. The mutants that were unable to form a complex with p28 (N-90, N-136, and N-186) were also unable to support translation in vitro. These results suggest that a complex of p86 and p28 is necessary to initiate translation. As shown in Table II, truncations from the C-terminal end of the p86 subunit did not adversely affect the ability of p86 to support translation up to residue 462 (C-462).
The ATP hydrolysis activity of p86 is both RNA-dependent and eIF4A-dependent, but does not require the presence of p28 (data not shown). None of the truncations from the amino terminus affected the ATP hydrolysis activity of p86 (see Table  II). However, the C-489 mutant, while still supporting translation as well as wild-type, showed ATP hydrolysis activity at 58% of wild-type. C-terminal truncations to C-462 almost completely abrogated ATP hydrolysis activity (to 14% of wild-type) and seriously reduced the ability of the p86 mutant to support translation (36% of wild-type). This suggests that at least a portion of the ATP hydrolysis domain on p86 is located in a central region of the protein, around residues 450 -500, and a partial loss of ATP hydrolysis activity is tolerated before protein synthesis activity is affected. Consequently, this region may define the site(s) or portions of the site(s) of interaction of RNA, ATP, and/or eIF4A with p86 during the initiation process. This region is very close to the start of the conserved blocks (Fig. 1A, block 6) between p86 and mammalian and yeast 4G, suggesting that the interactions of eIF4G, RNA, ATP, and/or eIF4A may share a common motif.
The interaction of the truncation mutants with eIF4A was examined in the two-hybrid system. The interaction of the p86 truncation mutants with eIF4A or p28 was quantitated using a liquid assay for ␤-galactosidase as shown in Table IV. Interestingly, differences in the relative binding of the p86 mutants to p28 were observed in the quantitative two-hybrid assay that were not obvious in the filter assay. The C-terminal deletion mutants have an approximately 2-fold higher interaction with p28 and suggest that the C terminus of p86 may hinder binding of p28 at the N terminus. Surprisingly, the N-terminal mutants do not interact with eIF4A, and the C-terminal truncation mutants have a slightly increased level of interaction. These results suggest that the N terminus of p86 interacts with eIF4A in a manner that is not accurately reflected in the ATP hydrolysis assay. Interaction of eIF4A with p28 was not observed in either the qualitative or quantitative two-hybrid assays. These results would indicate that at least part of the site of interaction of eIF4A with p86 is in close proximity to the p28 binding site. If the quantitative two-hybrid data are taken as a reflection of the affinity of the interaction of two proteins, then the interaction of eIF4A with p86 is at least 5-fold less than with p28. Clearly, further investigation is needed to understand the interactions of these three proteins and to quantitate the affinity of interaction.
Considering both the ATP hydrolysis and the p28 binding data together indicates that these functions lie on separate domains on p86 and can be uncoupled. The binding of p28 is not required for ATP hydrolysis activity, and, conversely, ATP hydrolysis activity is not required for p28 binding. Furthermore, ATP hydrolysis activity and the eIF4A binding appear to lie on separate domains as well. It will be interesting to contrast the ATP hydrolysis domain and eIF4A binding domain with the domain(s) for RNA binding and/or RNA helicase activity. Both the p86 and p28 in the presence of eIF4A and ATP are required for RNA helicase activity (34). Experiments are in progress to determine which portions of p86 are involved in RNA helicase activity.
We have demonstrated that the site of interaction of the cap-binding protein (p28) with its large subunit (p86) resides within a region defined by residues 52-90 that contains a motif shown to be required for binding of mammalian eIF4E to mammalian eIF4G (9,33). We have shown that this region also appears to be required for interaction with eIF4A. This is a novel observation that the N terminus of p86 may also be involved in the binding of eIF4A. The C-terminal portion of mammalian eIF4G has been shown to interact with mammalian eIF4A (9). The ability of p86 to bind the p28 subunit is absolutely required for in vitro protein synthesis activity. However, the ability of the p86 subunit to bind p28 or eIF4A appears to be uncoupled from its ability to stimulate RNA-dependent ATP hydrolysis in the presence of eIF4A. We have also demonstrated that the use of the yeast two-hybrid system will be a very powerful tool in elucidating the complex interactions of the subunits of eIF(iso)4F with each other and with other initiation factors.