eIF4G functionally differs from eIFiso4G in promoting internal initiation, cap-independent translation, and translation of structured mRNAs.

Eukaryotic initiation factor (eIF) 4G plays an important role in assembling the initiation complex required for ribosome binding to an mRNA. Plants, animals, and yeast each express two eIF4G homologs, which share only 30, 46, and 53% identity, respectively. We have examined the functional differences between plant eIF4G proteins, referred to as eIF4G and eIFiso4G, when present as subunits of eIF4F and eIFiso4F, respectively. The degree to which a 5'-cap stimulated translation was inversely correlated with the concentration of eIF4F or eIFiso4F and required the poly(A)-binding protein for optimal function. Although eIF4F and eIFiso4F directed translation of unstructured mRNAs, eIF4F supported translation of an mRNA containing 5'-proximal secondary structure substantially better than did eIFiso4F. Moreover, eIF4F stimulated translation from uncapped monocistronic or dicistronic mRNAs to a greater extent than did eIFiso4F. These data suggest that at least some functions of plant eIFiso4F and eIF4F have diverged in that eIFiso4F promotes translation preferentially from unstructured mRNAs, whereas eIF4F can promote translation also from mRNAs that contain a structured 5'-leader and that are uncapped or contain multiple cistrons. This ability may also enable eIF4F to promote translation from standard mRNAs under cellular conditions in which cap-dependent translation is inhibited.

Protein synthesis requires the participation of numerous eukaryotic initiation factors (eIFs) 1 that assist the binding of 40 S ribosomal subunits to an mRNA and the assembly of the 80 S ribosome at the correct initiation codon. The 5Ј-cap structure (m 7 GpppN, where N represents any nucleotide) serves as the binding site for the cap-binding protein eIF4E, the small subunit of eIF4F. eIF4G, the large subunit of eIF4F, interacts with several proteins in addition to eIF4E, including eIF4A (which is required to remove secondary structure within the 5Ј-leader sequence that would otherwise inhibit scanning of the 40 S ribosomal subunit), eIF3 (which promotes 40 S ribosomal subunit binding to the mRNA), and the poly(A)-binding protein (PABP; which stabilizes eIF4F binding to the 5Ј-cap) (1)(2)(3)(4)(5). The N-terminal domain of eIF4G is responsible for binding eIF4E and PABP; the middle domain binds eIF3 and eIF4A; and in mammalian eIF4G, the C-terminal domain binds a second molecule of eIF4A as well as Mnk1, a MAPK-activated protein kinase responsible for phosphorylating eIF4E (6 -8). Consequently, eIF4G functions as a scaffold protein that recruits many of the factors involved in stimulating 40 S ribosomal subunit binding to an mRNA.
Two related but highly distinct eIF4G proteins were first identified in plants (9). The two plant eIF4G proteins, referred to as eIF4G and eIFiso4G, differ in size (165 and 86 kDa, respectively). Two forms of eIF4G are also observed in yeast and mammals (10,11), but do not differ substantially in molecular mass and are more conserved. Mammalian eIF4GI and eIF4GII are 46% identical (11), and yeast eIF4G1 and eIF4G2 are 53% identical (10), in contrast to plant eIF4G and eIFiso4G, which are only 30% identical. 2 Mammalian eIF4GII functionally complements eIF4GI to a significant extent (11), and yeast can tolerate the deletion of either gene encoding eIF4G, although at least one gene is required for viability (10). Although these studies suggest that both eIF4G proteins in eukaryotic species are largely functionally similar, differences also have been reported. For example, deletion of the gene encoding yeast eIF4G1 leads to a synthetic lethal interaction with cdc33-1, an eIF4E temperature-sensitive mutant, whereas deletion of the gene encoding eIF4G2 does not (12). Moreover, yeast eIF4G2 supports translation of uncapped polyadenylated mRNA to a greater extent than does eIF4G1 (12). Whether this is due to functional differences or to the in vivo levels of each eIF4G is unknown. Cleavage of mammalian eIF4G occurs following infection with poliovirus or human rhinovirus, resulting in the inhibition of protein synthesis (13,14). For both viruses, cleavage of eIF4GI occurs prior to that of eIF4GII, and the cleavage of the latter correlates with the loss of protein synthesis following viral infection (13)(14)(15). In contrast, the cleavage of eIF4GI and eIF4GII that occurs during apoptosis is temporally similar (16). Although wheat eIF4F and eIFiso4F support translation in vitro and exhibit RNA-dependent ATP hydrolysis activity and ATP-dependent RNA unwinding activity (17)(18)(19)(20), the affinity of eIF4F for hypermethylated cap structures is lower than that of eIFiso4F (21), and the ATPase activity of eIF4F is greater than that of eIFiso4F when mRNA is used to stimulate the RNA-dependent activity (20). Moreover, binding studies with oligonucleotides suggest that eIF4F binding is sensitive to the presence of secondary structure and that eIFiso4F exhibits a binding preference for linear structures (22).
Whether the two types of eIF4G present in eukaryotic cells exhibit specialization in determining which mRNAs are translated or whether they differ in the efficiency in which they support translation has not been investigated for any species. In this study, the functional differences of plant eIF4G and eIFiso4G were investigated during the translation of capped or uncapped mRNAs, mRNAs containing a structured 5Ј-leader, or dicistronic mRNAs. The addition of eIF4F or eIFiso4F (in which eIF4G and eIFiso4G are present as subunits, respectively) to lysates depleted of eIF4F and eIFiso4F supported the translation of an unstructured mRNA; however, only eIF4F significantly supported translation from an mRNA with a structured 5Ј-leader. eIF4F increased the translation of an uncapped mRNA and stimulated the translation from the second cistron of a dicistronic mRNA to a greater extent than did eIFiso4F, suggesting that eIF4F has evolved to promote translation from nonstandard mRNAs, i.e. those that lack a cap, contain a structured 5Ј-leader, or contain multiple cistrons, or has evolved to promote translation from standard mRNAs under cellular conditions in which cap-dependent translation is inhibited, whereas eIFiso4F may be largely limited to facilitating translation from standard mRNAs. Additionally, the concentration of eIF4F, eIFiso4F, and PABP determined the extent to which the cap stimulated translation: PABP was required for the cap to stimulate translation efficiently; however, increased levels of PABP, eIF4F, or eIFiso4F substantially reduced the competitive advantage that a cap conferred to an mRNA. These observations suggest that eIF4F and eIFiso4F have undergone functional specialization that allows them to discriminate between mRNAs. Moreover, these observations suggest that developmental changes in the cellular concentration of eIF4F, eIFiso4F, or PABP may influence the extent of cap-dependent translation.

MATERIALS AND METHODS
Plasmid Constructs and in Vitro RNA Synthesis-The T7-based monocistronic and dicistronic luciferase constructs have been described previously (23). DNA concentration was quantitated spectrophotometrically following linearization and brought to 0.5 mg/ml. In vitro transcription was carried out as described previously (24) using 40 mM Tris-HCl (pH 7.5), 6 mM MgCl 2 , 100 g/ml bovine serum albumin, 0.5 mM ATP, 0.5 mM CTP, 0.5 mM UTP, 0.5 mM GTP, 10 mM dithiothreitol, 0.3 units/l RNasin (Promega), and 0.5 units/l T7 RNA polymerase. The constructs used terminated in a A 50 tail. Capped RNAs were synthesized using 3 g of template in the same reaction mixture as described above, except that GTP was used at 160 M, and 1 mM m 7 GpppG was included. Under these conditions, Ͼ95% of the mRNA is capped. The free energy of secondary structures used in this study was calculated at a temperature of 37°C using MFOLD of GCG Software Package Version 10, which is based on the Zuker algorithm for determining multiple optimal and suboptimal secondary structures (25) with the folding parameters as described (26).
Proteins from control and depleted wheat germ lysates were resolved using standard SDS-polyacrylamide gel electrophoresis, and the proteins were transferred to 0.22 M nitrocellulose membrane by electroblotting. Following transfer, the nitrocellulose membranes were blocked in 5% milk and 0.01% thimerosal in TPBS (0.1% Tween 20, 13.7 mM NaCl, 0.27 mM KCl, 1 mM Na 2 HPO 4 , and 0.14 mM KH 2 PO 4 ), followed by incubation with primary antibodies diluted typically 1:1000 to 1:2000 in TPBS with 1% milk for 1.5 h. The blots were then washed twice with TPBS and incubated with horseradish peroxidase-conjugated goat anti-rabbit antibodies (Southern Biotechnology Associates, Inc.) diluted to 1:10,000 for 1 h. The blots were washed twice with TPBS, and the signal was detected typically between 1 and 15 min using chemiluminescence (Amersham Pharmacia Biotech).
In Vitro Translation Assays-200 l of wheat germ extract (Promega) was added to 300 l of m 7 GTP-Sepharose (Amersham Pharmacia Biotech) or 100 l of poly(A)-agarose (Sigma) and incubated with rotation at 4°C for 30 min. The lysate was collected by centrifugation (800 ϫ g for 1 min) through a spin column (Promega) and used immediately. The extent of depletion of eIF4G, eIF4E, eIFiso4G, eIFiso4E, eIF4A, eIF4B, eIF3, eEF2, PABP, or Hsp101 was determined by Western analysis following resolution of the extract by SDS-polyacrylamide gel electrophoresis. mRNA constructs were translated using complete or depleted wheat germ lysate as described by the manufacturer, except that all amino acids were unlabeled. The lysates were supplemented with recombinant initiation factors or factors purified from wheat germ extract as indicated. In wheat germ lysate, eIF4A is present in a Ͼ30-fold molar excess relative to eIF4G (31). Consequently, a similar ratio was used when lysates were supplemented with eIF4A. The ratio of eIF4B to eIF4G has not been measured; and therefore, the ratio used for supplementation was determined empirically. The reactions were incubated for 3 h, and 2-l aliquots were assayed in a MonoLight 2010 luminometer for luciferase activity. Each mRNA construct was translated in triplicate, and the mean Ϯ S.D. for each construct is reported.

RESULTS
eIF4F Directs Translation from an mRNA with a Structured 5Ј-Leader to a Greater Extent than Does eIFiso4F-An alignment of eIF4G from wheat, human, and Saccharomyces cerevisiae revealed that plant eIF4G and eIFiso4G are most conserved with eIF4G of other eukaryotes in the region responsible for interaction with eIF4A and eIF3 ( Fig. 1) (1). A second conserved region is the eIF4E-binding domain (1,32). The Mnk1-binding domain and a second eIF4A-binding site have been mapped to the C-terminal region of human eIF4G (6,8), but are not present in yeast and plant eIF4F proteins. However, plant eIF4G and eIFiso4G do contain a domain near their C terminus that shares limited conservation with the human FIG. 1. Graphical representation of the protein sequence alignments for eIF4G proteins from eukaryotes. Protein alignments were carried out using MACAW Version 1 (61). The degree of similarity between eIF4G proteins from wheat (Triticum aestivum (Ta)), human (Homo sapiens (Hs)), and S. cerevisiae (Sc) is indicated by the size and darkness of the shading. The scale at top is in amino acids. Protein interaction domains identified in mammalian eIF4G for human PABP (hPABP), eIF4E, eIF4A, eIF3, and Mnk1 are indicated below the alignment. The interaction domain for PABP is not conserved among eIF4G proteins from different species; consequently, the domain that is involved in interaction with yeast PABP (yPABP) and the putative domain identified in wheat eIFiso4G (wPABP) are indicated below the mammalian domains. eIF4G proteins and is absent from the yeast orthologs. Plant eIFiso4G differs most from eIF4G in that it lacks an ϳ700amino acid-long N-terminal region present in eIF4G. In this respect, plant eIF4G is more similar to human eIF4G than is eIFiso4G. The yeast eIF4G proteins also contain an N-terminal region, although it is shorter than that present in plant eIF4G or in either mammalian eIF4G protein. Although the domain responsible for interaction with PABP is not conserved among eIF4G proteins, it is located within this N-terminal region of human and yeast eIF4G proteins. The PABP interaction domain within plant eIF4G has not been identified precisely, but the putative site in eIFiso4G has been mapped to its N-terminal region (33).
To examine the function of eIF4G or eIFiso4G in vitro, it was necessary to generate an eIF4G-and eIFiso4G-dependent lysate. This was accomplished by depleting wheat germ lysate of eIF4F (composed of eIF4G and eIF4E) and eIFiso4F (composed of eIFiso4G and eIFiso4E) through their binding to m 7 GTP-Sepharose. Western analysis confirmed that the level of eIF4E and eIFiso4E was reduced by 90 -95%, as was that of eIF4G and eIFiso4G (Fig. 2). To examine whether the depleted lysate was eIF4F-or eIFiso4F-dependent, capped luc-A 50 mRNA was translated in a fractionated lysate supplemented with increasing amounts of purified eIF4F or eIFiso4F. The extent to which the reporter mRNA was translated was determined by measuring luciferase activity. A reduction in the level of eIF4F and eIFiso4F reduced translation by Ͼ95% (compare translation in complete and fractionated lysates) (Fig. 3), as would be expected following a reduction in those initiation factors that are normally required for efficient translation. Residual translational activity of the fractionated lysate may be the result of the low level of either eIF4G and eIFiso4G remaining in the lysate (Fig. 2) or a result of PABP (see below). Supplementation with 16 nM eIF4F increased translation from the reporter mRNA nearly 10-fold in the fractionated lysate, but did not affect translation of the same mRNA in the unfractionated lysate (Fig. 3). Translation in the fractionated lysate was also dependent on eIFiso4F, whereas supplementation of the unfractionated lysate with eIFiso4F did not affect translation (Fig. 3). A comparison of their relative effects on translation reveals that eIF4F was more stimulatory than was eIFiso4F. Native eIF4F and recombinant eIF4F are equally active in supporting translation in vitro, as are native eIFiso4F and recombinant eIFiso4F (30), 3 suggesting that there is no significant difference in the fraction of each purified factor that is active. The stimulation afforded by eIFiso4F was more nonlinear at the highest concentration used for this factor than that observed for eIF4F, suggesting that even higher concentrations of eIFiso4F would not yield a level of translation comparable to that observed for eIF4F. The nonlinearity of the activity of these factors has been reported previously (34). The greater nonlinearity of eIF4F at lower concentrations may indicate its higher affinity for RNA compared with eIFiso4F, a possibility that is supported by the observation that the eIFiso4F doseresponse curve is slightly sigmoidal ( Fig. 3; see Fig. 5C). These data suggest that the endogenous level of eIF4F and eIFiso4F in the unfractionated lysate is necessary for maximum translational activity and that their removal from the lysate renders the fractionated lysate dependent upon the addition of either eIF4F or eIFiso4F.
To examine whether eIF4F and eIFiso4F differ in the extent to which they support translation of an mRNA with a structured 5Ј-leader, it was first necessary to determine the translational characteristics of mRNAs with or without secondary structure in the eIF4F/eIFiso4F-dependent lysate. A stable stem-loop structure containing a 24-base pair stem of ⌬G ϭ Ϫ42.9 kcal/mol was introduced 4 nucleotides downstream of the 5Ј terminus of the luc reporter mRNA (referred to as SL 24luc-A 50 ). In addition, deletions were made within the 24-base pair stem-loop to generate less stable structures with 19-, 13-, and 7-base pair stems of ⌬G ϭ Ϫ31.8, Ϫ21.3, and Ϫ4.5 kcal/ mol, respectively (referred to as SL 19 -luc-A 50 , SL 13 -luc-A 50 , and SL 7 -luc-A 50 , respectively). The presence of these structures was shown previously to inhibit translation in unfractionated lysate as a function of their stability (35). To determine the effect of the stem-loop on translation, each construct was synthesized in vitro as a capped mRNA containing a A 50 tail and translated at three different concentrations in a lysate reduced in eIF4F and eIFiso4F. When the eIF4F/eIFiso4F-dependent lysate was programmed with a 9.6 ng/l concentration of each construct, the presence of a 24-, 19-, 13-, or 7-base pair stemloop (i.e. SL 24 -luc-A 50 , SL 19 -luc-A 50 , SL 13 -luc-A 50 , and SL 7 -luc-A 50 , respectively) inhibited translation to 15.7, 12.3, 16.3, and 61.5% of that of the control mRNA with an unstructured leader (Fig. 4). As expected, decreasing the RNA concentration reduced the amount of luciferase produced (compare the absolute levels of expression in Fig. 4 and note the difference in scale on the x axis). However, the presence of the same stem-loop structures was progressively less inhibitory as the concentration of the input mRNA decreased (see the relative values to the right of each histogram in Fig. 4, which are relative to the expression from the control luc-A 50 mRNA), and translation from the SL 7 -luc-A 50 and SL 13 -luc-A 50 mRNAs was actually higher than 3 K. S. Browning, unpublished data. Each lysate was supplemented with the indicated amounts of eIF4F or eIFiso4F, and the amount of luciferase synthesized was measured in a luminometer. eIF4A was included at a ratio of 1:70 (for eIF4F assays) or 1:30 (for eIFiso4F assays). Luciferase expression is indicated as light units from 2 l of each reaction. eIF4G and eIFiso4G Functionally Differ that from the control mRNA at the lowest RNA concentration tested. Consequently, secondary structure of lower stability (e.g. SL 7 -luc-A 50 and SL 13 -luc-A 50 mRNAs) was inhibitory only at a high RNA concentration, whereas more stable secondary structure (e.g. SL 19 -luc-A 50 and SL 24 -luc-A 50 mRNAs) was inhibitory at all RNA concentrations tested, albeit less inhibitory at low RNA concentrations. These data suggest that the translational machinery, such as eIF4A, which functions as an RNA helicase and as a subunit of eIF4F (or eIFiso4F), is required during translation in direct proportion to the degree of secondary structure present in the 5Ј-leader of an mRNA (36), is sufficient to unwind moderately stable secondary structure when an mRNA is present at a low concentration. However, higher concentrations of mRNA may act to titrate the RNA helicase activity such that the structure cannot be removed from all of the input RNA, resulting in the inhibition of translation by secondary structure of even moderate stability.
To examine the effect of eIF4F or eIFiso4F on the translation of a structured mRNA, the SL 13 -luc-A 50 mRNA construct was selected as an mRNA with moderately stable secondary structure. This mRNA also exhibited the greatest potential range in expression as exemplified by the degree to which it was translated at high or low RNA concentrations (Fig. 4). Supplementation of the eIF4F/eIFiso4F-dependent lysate programmed with eIF4F increased translation from SL 13 -luc-A 50 mRNA up to 14.7-fold (Fig. 5B), whereas supplementation with eIFiso4F did not increase translation from the same mRNA (Fig. 5D). In contrast, translation from the control (i.e. unstructured) mRNA increased when the lysate was supplemented with either eIF4F (Fig. 5A) or eIFiso4F (Fig. 5C). For the structured mRNA (Fig.  5B), eIF4F was less stimulatory at high concentrations than at lower concentrations. This effect is also observed for unstructured mRNAs if eIF4F is added to the lysate to a level higher than was used for these studies (data not shown). An excess of eIF4F (i.e. a level of eIF4F that is in excess of the binding capacity of the input mRNA) may compete with the bound eIF4F for other factors needed for translation. The secondary structure present in the construct of Fig. 5B may reduce the amount of eIF4F that can bind the mRNA, which would result in a condition of excess eIF4F at a lower concentration than normally observed for an unstructured mRNA. Once in excess, free eIF4F may compete with the bound eIF4F for other necessary factors, resulting in an apparent lower level of stimulation. As eIF4F is considered to be present in limiting amounts in vivo in eukaryotes, such in vitro conditions of excess eIF4F are unlikely to be present in vivo. These data suggest that eIF4F and eIFiso4F differ considerably in their ability to promote translation from an mRNA containing a structured 5Ј-leader.
The introduction of each stem-loop increases the length of the 5Ј-leader relative to the construct lacking the secondary structure, which could have an effect on translation efficiency. For example, introduction of SL 24 , SL 19 , SL 13 , or SL 7 results in a leader length of 64, 53, 41, or 29 nucleotides, respectively. To determine whether the length of the 5Ј-leader influences the degree to which eIF4F or eIFiso4F increases translation, capped luc-A 50 mRNAs containing a 17-, 72-, or 144-nucleotide unstructured 5Ј-leader were translated in the eIF4F/eIFiso4Fdependent lysate and supplemented with eIF4F or eIFiso4F. Expression was affected only to a small extent as a function of the 5Ј-leader length, with a 2-fold increase in translation as a function of the length of the leader when increased from 17 to 72 nucleotides (Table I). The addition of eIF4F increased translation 4.4 -6-fold, whereas an ϳ2-fold increase was observed for eIFiso4F. The length of the leader did not substantially alter the stimulatory effect of either factor, suggesting that neither eIF4F nor eIFiso4F discriminates between mRNAs based on their 5Ј-leader length.
eIF4F Promotes Cap-independent Translation to a Greater Extent than Does eIFiso4F-mRNAs that initiate protein synthesis using an internal ribosome entry site, such as those of the picornaviral family, often have long structured leader sequences. The observation that secondary structure proximal to the 5Ј-cap inhibits eIFiso4F function to a greater extent compared with eIF4F suggests that eIF4F may be able to promote cap-independent translation to a greater extent compared with eIFiso4F. To examine this possibility, capped and uncapped luc-A 50 mRNAs were translated in the eIF4F/eIFiso4F-dependent lysate supplemented with increasing amounts of either factor. The addition of eIF4F at a concentration of just 2 nM increased translation from uncapped mRNA by 4.8-fold and from capped mRNA by 6.8-fold (Table II). Higher concentrations of eIF4F increased translation up to 11-fold. In contrast, eIFiso4F at a concentration of 5 nM increased translation from uncapped mRNA by only 1.4-fold, and no increase was observed for capped mRNA. Higher concentrations of eIFiso4F resulted in up to a 3.7-fold stimulation of translation from uncapped mRNA and 2.8-fold from capped mRNA. These data suggest that both eIF4F and eIFiso4F can promote cap-independent translation; however, eIF4F facilitates cap-independent translation to a greater extent than does eIFiso4F, consistent with its greater ability to promote translation from mRNAs that contain secondary structure proximal to the 5Ј-cap. Interestingly, despite the reduction in the level of eIF4F and eIFiso4F, there was little alteration in the extent to which translation was stimulated by the presence of a 5Ј-cap (Table II).
Wheat germ lysate is highly message-dependent because of a low concentration of endogenous transcript and a high level of unengaged translational machinery. As a consequence, those features that increase the competitiveness of an mRNA, such a 5Ј-cap, would not be expected to confer a translational advantage under the noncompetitive conditions that prevail in nor- eIF4G and eIFiso4G Functionally Differ mal lysate. Accordingly, competitive translation might be achieved by increasing the concentration of mRNA to a level that exceeds the translational capacity of the lysate or by removing the excess capacity of those factors most important in determining competitive translation. To test the former prediction, complete and eIF4F/eIFiso4F-dependent lysates were programmed with increasing amounts of capped and uncapped luc-A 50 mRNAs, and the cap dependence was determined by the ratio of protein synthesis from capped to uncapped mRNAs (Table III). The presence of a 5Ј-cap did not confer a translational advantage when the mRNA was present at a low concentration in either the complete or eIF4F/eIFiso4F-dependent lysate, results consistent with those in Table II. Note that the lower translation from capped mRNA relative to uncapped mRNA reflects a lower yield of in vitro transcribed mRNA when made in the capped versus uncapped form. As mRNA concen-tration was increased in the lysate, the presence of the 5Ј-cap became increasingly important in stimulating translation. These results suggest that even following a reduction in the level of eIF4F and eIFiso4F, translation is not competitive at low RNA concentration. Increasing the concentration of the RNA increased the competitive advantage that a cap conferred to an mRNA, although the degree to which translation in the lysate was cap-dependent remained considerably lower than that typically observed in vivo. This suggests that despite the reduction in eIF4F and eIFiso4F, translation remains largely less competitive than that observed in vivo and indicates another factor, such as PABP, that may be important in influencing cap function remains in excess in the fractionated lysate.
Because PABP interacts with the eIF4G and eIFiso4G subunits of eIF4F and eIFiso4F, respectively (33, 37); recruits eIF4G to an mRNA in the absence of a cap or functional eIF4E FIG . 5. eIF4F (but not eIFiso4F) supports the translation of an mRNA with 5-proximal secondary structure. The eIF4F/eIFiso4Fdependent lysate was programmed with 7.5 ng/l luc-A 50 (A and C) or SL 13 -luc-A 50 (B and D) mRNA and supplemented with the indicated amounts of eIF4F (A and B) or eIFiso4F (C and D). eIF4A was included at a 1:70 ratio of eIF4A to eIF4F or at a 1:30 ratio of eIF4A to eIFiso4F. Each mRNA construct was translated in triplicate, and the mean Ϯ S.D. for each construct is reported. The -fold stimulation provided by eIF4F or eIFiso4F is indicated above each histogram. Luc, luciferase.

TABLE I eIF4F and eIFiso4F stimulate translation independently of 5Ј-leader length
Capped luc-A 50 mRNAs containing a 17-, 72-, or 144-nucleotide 5Ј-leader were translated in the eIF4F/eIFiso4F-dependent lysate. The lysate was supplemented with the indicated amounts of eIF4F or eIFiso4F, and the amount of luciferase synthesized was measured in a luminometer. eIF4A was included at a ratio of 1:70 (for eIF4F assays) or 1:30 (for eIFiso4F assays). Luciferase expression is indicated as light units (LU) from 2 l of each reaction. -Fold increase is relative to translation in depleted lysate without the addition of any factor. nt, nucleotides.

eIF4G and eIFiso4G Functionally Differ
(the cap-binding subunit of eIF4F) (12,38); and promotes translational initiation both in cis and in trans (39), we examined whether the level of PABP influences the extent to which the lysate is cap-dependent. The lysate was depleted of PABP following incubation with poly(A)-agarose, which was confirmed by Western analysis (Fig. 6). Because eIF4G, eIFiso4G, and eIF4B also bind poly(A) RNA, albeit with a considerably lower affinity than does PABP, and PABP is known to physically interact with eIF4G, eIFiso4G, and eIF4B, which in turn interact with eIF4A and eIF3, these initiation factors were expected to be partially depleted from the lysate following incubation with poly(A)-agarose. Their partial depletion, which was not as great as that observed for PABP, was confirmed by Western analysis (Fig. 6). Of the initiation factors, eIF4G was reduced to the greatest extent. Depletion of eEF2 was less than that observed for initiation factors, and no reduction was observed for the heat shock protein Hsp101 (Fig. 6). The PABP-depleted lysate was then programmed with capped or uncapped mRNAs to determine the degree to which translation was cap-dependent. The presence of the cap increased translation 10-fold (Table IV), a higher degree of cap dependence than that observed in the eIF4F/eIFiso4F-depend-ent lysate at this mRNA concentration (Table III). Supplementation of the PABP-depleted lysate (which was also reduced in the level of eIF4F and eIFiso4F) (Fig. 6) with increasing amounts of eIF4F or eIFiso4F reduced the cap dependence of translation. These data indicate that the depletion of PABP in combination with the partial reduction in eIF4F and eIFiso4F increases cap-dependent translation to a greater extent than does a reduction in eIF4F and eIFiso4F alone. Interestingly, the basal level of translational activity in the PABP-depleted TABLE II eIF4F promotes translation of capped and uncapped mRNAs to a greater extent than does eIFiso4F Capped or uncapped luc-A 50 mRNAs were translated at 3.4 ng/l in the eIF4F/eIFiso4F-dependent lysate. The lysate was supplemented with the indicated amounts of eIF4F or eIFiso4F, and the amount of luciferase synthesized was measured in a luminometer. eIF4A was included at a ratio of 1:70 (for eIF4F assays) or 1:30 (for eIFiso4F assays). Luciferase expression is indicated as light units (LU) from 2 l of each reaction.   3-6 for eIF3, eEF2, and Hsp101, as these membranes represent those used for the analysis of eIF4G, eIFiso4G, and PABP that had been stripped and reprobed. eIF4G and eIFiso4G Functionally Differ lysate was substantially lower (Table IV) than that observed in the eIF4F/eIFiso4F-dependent and complete lysates (Fig. 3), suggesting that in addition to eIF4F and eIFiso4F, PABP contributes to the overall translational activity of the lysate.
Both PABP and eIF4E function to recruit eIF4G to an mRNA, which is a necessary prerequisite for the recruitment of the 43 S complex. Although PABP may recruit eIF4G to an mRNA whether a 5Ј-cap is present or not, the effect of PABP on eIF4F recruitment would be expected to be greater for a capped mRNA, in which eIF4E (which binds to the cap) could contribute more to the recruitment of eIF4G, than for an uncapped mRNA, to which eIF4E would not be able to bind. Consistent with the prediction of a preferential recruitment of eIF4F by PABP to a capped mRNA, the presence of PABP at an appropriate concentration would be expected to increase the extent to which translation is cap-dependent. To test these predictions, capped and uncapped mRNAs were translated in the PABP-depleted lysate in the presence of increasing amounts of PABP (Fig. 7). In the lysate depleted of PABP and reduced in the level of eIF4F and eIFiso4F (Fig. 6), the presence of a 5Ј-cap stimulated translation by 4.2-fold (Fig. 7). Supplementation with low concentrations of PABP increased the translation of both the capped and uncapped mRNAs, but did so preferentially for the mRNA with a 5Ј-cap, resulting in an increase in the stimulation afforded by a cap of up to 11.2-fold. As the concentration of PABP was increased further, the competitive advantage conferred by a cap was increasingly lost, resulting in a decline in translation specifically from the capped mRNA and, consequently, a reduction in cap function. The data presented in Table IV and Fig. 7 suggest that the concentration of PABP contributes to the extent to which a cap stimulates translation from an mRNA: the presence of a moderate amount of PABP preferentially stimulates the translation of capped mRNA, presumably through the recruitment of eIF4G as a result of the combined efforts of PABP and eIF4E. However, the competitive advantage provided by a 5Ј-cap is lost with an increase in the concentration of eIF4F, eIFiso4F, or PABP.
eIF4F Promotes Internal Initiation to a Greater Extent than Does eIFiso4F-The results presented in Fig. 5 and Table II showing that eIF4F promotes translation of mRNA with secondary structure proximal to the 5Ј-cap or uncapped mRNA to a greater extent than does eIFiso4F suggested that eIF4F might also promote internal initiation to a greater extent compared with eIFiso4F. To test this possibility, a dicistronic mRNA construct containing the uidA gene, which encodes ␤-glucuronidase (GUS), as the 5Ј-proximal cistron and luc as the second cistron was translated as an uncapped mRNA in the eIF4F/eIFiso4F-dependent lysate in the presence of increasing amounts of eIF4F or eIFiso4F. Because the level of eIF4A and eIF4B was reduced in the eIF4F/eIFiso4F-dependent lysate, the supplementation of the lysate with eIF4F or eIFiso4F was performed in the presence or absence of additional eIF4A and eIF4B (Fig. 8). Translation from the second cistron was then determined by measuring the resulting level of luciferase synthesized. The addition of 16 nM eIF4F (together with eIF4A and eIF4B) increased translation from the second cistron up to 63-fold, whereas the addition of 40 mM eIFiso4F (together with eIF4A and eIF4B) increased translation only 9-fold (Fig. 8). When eIF4F or eIFiso4F was added in the absence of additional eIF4A and eIF4B, eIF4F increased translation from the second cistron up to 16-fold (at 4 nM), whereas the addition of eIFiso4F increased translation only 3-fold (at 7.5 nM) (Fig. 8). Interestingly, the maximum increase in translation from the second cistron occurred at 4 nM eIF4F or 7.5 nM eIFiso4F, and higher levels of either factor resulted in lower levels of internal initiation. This is in contrast to the data obtained when eIF4A and eIF4B were present: translation from the second cistron continued to increase as a function of the increase in eIF4F or eIFiso4F (up to 16 and 40 nM, respectively), suggesting that eIF4A and/or eIF4B increased the activity of eIF4F and eIFiso4F over a greater concentration range perhaps because the endogenous level of eIF4A and eIF4B had become limiting.

TABLE IV
Cap dependence is inversely proportional to the concentration of eIF4F or eIFiso4F PABP-dependent lysate was programmed with 1.5 ng/l capped or uncapped luc-A 50 mRNAs. The lysate was supplemented with the indicated amounts of eIF4F or eIFiso4F, and the amount of luciferase synthesized was measured in a luminometer. Luciferase expression is indicated as light units (LU) from 2 l of each reaction. eIF4A was included at a ratio of 1:70 (for eIF4F assays) or 1:30 (for eIFiso4F assays). eIF4B was included at a ratio of 1:4 (for eIF4F assays) or 1:1.6 (for eIFiso4F assays). Stimulation by a cap was determined from the ratio of expression from capped to uncapped mRNA.  7. PABP is required for optimal cap-dependent translation. The PABP-dependent lysate was programmed with (0.6 ng/l) capped or uncapped luc-A 50 mRNAs. The lysate was supplemented with the indicated amounts of PABP, and the amount of luciferase synthesized was measured in a luminometer. Luciferase expression is indicated as light units from 2 l of each reaction.

eIF4G and eIFiso4G Functionally Differ
These data suggest that although eIF4F and eIFiso4F can stimulate internal initiation, they require eIF4A and/or eIF4B for their maximum stimulatory effect.
To determine whether eIF4A or eIF4B was responsible for increasing the ability of eIF4F or eIFiso4F to promote translation from the second cistron, the dicistronic mRNA was translated in the eIF4F/eIFiso4F-dependent lysate with factors added either separately or in combination (Table V). The addition of eIF4A or eIF4B alone increased translation from the second cistron by 2.6-and 2.4-fold, respectively, whereas the addition of eIF4F alone increased second cistron translation 19-fold, and eIFiso4F had a 2.1-fold effect. The combination of eIF4A and eIF4F or of eIF4A and eIFiso4F had a synergistic effect on increasing second cistron translation, as did the combination of eIF4B and eIF4F or of eIF4B and eIFiso4F, although not as great as when eIF4A was employed. An even greater stimulation of translation was observed when a combination of eIF4A and eIF4B was included with eIF4F or eIFiso4F.
To determine whether it was the eIF4G subunit of eIF4F that was responsible for increasing translation from the second cistron, the dicistronic mRNA was translated in the eIF4F/ eIFiso4F-dependent lysate supplemented with increasing amounts of recombinant eIF4G or eIFiso4G. The addition of eIF4G increased translation from the second cistron up to 20-fold, whereas the addition of eIFiso4G increased translation ϳ2-fold (Table VI). eIF4A was less effective in stimulating recombinant eIF4G function than it was for native eIF4F, suggesting that the native form may be more competent for interaction with this partner protein. Similar observations were made previously, in which native eIFiso4F interacted with PABP at a lower concentration than did recombinant eIFiso4G or recombinant eIFiso4F (33). The addition of either eIF4E or eIFiso4E did not increase translation from the second cistron and even decreased translation slightly at high concentrations (Fig. 9). The addition of PABP had only a small stimulatory effect (Fig. 9), suggesting that, in contrast to its ability to stimulate translation from a 5Ј-proximal cistron of a capped mRNA, it does not substantially promote internal initiation under the conditions used. Together, these results suggest that it is the eIF4G (or eIFiso4G) subunit that is responsible for promoting internal initiation and that eIF4G promotes internal initiation to a greater extent than does eIFiso4G. Moreover, the relative effect of recombinant eIF4G compared with eIFiso4G (Table VI) was very similar to that obtained for native eIF4F and eIFiso4F (Table V), supporting the conclusion that the observed differences are inherent properties of these proteins and not a result of differences in the active fraction of each factor preparation. DISCUSSION The presence of two highly divergent eIF4G proteins in plants has suggested the possibility of their functional specialization. The results presented in this study indicate that the greatest difference between these two factors was observed for capped mRNAs containing a structured 5Ј-leader: eIF4F supported translation from such an mRNA to a greater extent than did eIFiso4F. The presence of secondary structure within a leader blocks efficient scanning of the 40 S ribosomal subunit (40,41). eIF4F exhibits higher activity than does eIFiso4F when mRNA is used to stimulate the RNA-dependent ATPase activity in the presence of eIF4A (20). This helicase activity is required for the ATP-dependent unwinding of mRNA secondary structure (17)(18)(19)(20). The higher ATP hydrolysis activity of eIF4F is consistent with its greater ability to promote translation from structured mRNA. The present observations and the FIG. 8. eIF4F promotes internal initiation to a greater extent than does eIFiso4F. Uncapped GUS-SL-Con 134 -luc-A 50 mRNA (containing 134 nucleotides of control sequence in the intercistronic region) was translated in the eIF4F/eIFiso4F-dependent lysate. To be comparable to conditions under which the stem-loop in the construct would have maximum inhibitory effect on upstream ribosome scanning (as determined in Fig. 4), the dicistronic mRNA was translated at 25 ng/l because the construct is twice as long as the corresponding monocistronic mRNA. Left panel, the lysate was supplemented with the indicated amounts of eIF4F or eIFiso4F in which eIF4A and eIF4B were included at ratios of 1:70 and 1:4, respectively, for eIF4F and 1:30 and 1:1.6, respectively, for eIFiso4F. Right panel, the lysate was supplemented with the indicated amounts of eIF4F or eIFiso4F in the absence of eIF4A and eIF4B. The degree to which the second cistron of the dicistronic mRNA was translated in each assay was determined by the amount of luciferase synthesized. Luciferase expression is indicated as light units from 2 l of each reaction. V eIF4A and eIF4B assist eIF4F in promoting internal initiation Uncapped GUS-SL-Con 134 -luc-A 50 mRNA was translated at 25 ng/l in the eIF4F/eIFiso4F-dependent lysate. The lysate was supplemented with the indicated amounts of initiation factors and the degree to which the second cistron of the dicistronic mRNA was translated was determined by the amount of luciferase synthesized. Luciferase expression is indicated as light units (LU) from 2 l of each reaction.

eIF4G and eIFiso4G Functionally Differ
previous studies suggest that eIF4F facilitates translation from structured mRNAs as a result of its higher hydrolysis activity that is used to remove the secondary structure or by binding downstream of the structure. In contrast, the function of eIFiso4F during translation and in RNA binding studies suggests that it exhibits a preference for mRNAs with an unstructured 5Ј-leader. Moreover, eIF4F was observed to stimulate translation from uncapped mRNAs or from the 5Ј-distal cistron of a dicistronic mRNA to a greater extent than was eIFiso4F. When compared on an equal molar basis, eIF4F was 20 -30-fold more active than eIFiso4F in promoting translation from the 5Ј-distal cistron of a dicistronic mRNA or from an uncapped mRNA containing a structured 5Ј-leader. In wheat embryos, eIFiso4F is present at an ϳ9-fold greater level compared with eIF4F (31). This higher level may be necessary to compensate for its lower activity or for the translation of specific classes of mRNAs. Although eIF4F (or eIFiso4F) exhibits low levels of RNA-dependent ATP hydrolysis activity and ATP-dependent RNA helicase activity, eIF4A has been shown to significantly stimulate these activities (17)(18)(19)(20). In good agreement with these previous studies, we observed that the basal activity of eIF4F or eIFiso4F was substantially enhanced by the addition of eIF4A. Our observation that eIF4B also improved the basal activity of eIF4F or eIFiso4F is also consistent with previous studies reporting that eIF4B functions to increase the ATP hydrolysis and ATP-dependent RNA helicase activities of eIF4A, eIF4F, and eIFiso4F (42)(43)(44)(45)(46)(47)(48). However, eIF4A and eIF4B increased eIF4F and eIFiso4F activity equally, which suggests that the functional divergence of eIF4G and eIFiso4G does not affect the extent to which they depend on eIF4A or eIF4B. Multiple phosphorylated isoforms of eIF4B and two isoforms of eIF4A have been observed in plants (49). Whether a specific isoform may exhibit a preference for eIF4F or eIFiso4F remains to be determined.
The degree to which a 5Ј-cap stimulated translation from an mRNA was partially dependent on the presence of PABP. The addition of a moderate level of PABP to a PABP-depleted lysate preferentially increased translation from capped mRNAs, thereby increasing the cap dependence of the lysate. PABP promotes the recruitment of 40 S ribosomal subunits to an mRNA through its physical contact with eIF4G and increases eIF4F binding to the 5Ј-cap (4, 7, 33, 50 -53), suggesting that the stimulatory effect on cap function observed in Fig. 7 resulted from an increase in eIF4F recruitment to a capped mRNA. Although a moderate level of PABP stimulated cap function, higher concentrations (equivalent to those present in unfractionated wheat germ lysate) failed to stimulate cap-dependent translation. Similarly, increasing the concentration of eIF4F or eIFiso4F resulted in a decrease in cap dependence. Consequently, the high endogenous level of unengaged eIF4F, eIFiso4F, and PABP in wheat germ lysate appears to be responsible for the low level of cap dependence typically observed in the unfractionated lysate. These observations suggest that, in addition to eIF4F and eIFiso4F, the concentration of PABP also influences the extent of cap-dependent translation in a cell.
Under what circumstances might the functional differences observed between eIF4F and eIFiso4F be important in plants? mRNAs that contain secondary structure or one or more small open reading frames in the 5Ј-leader are just two strategies that have evolved to limit the amount of protein synthesized from an mRNA (reviewed in Refs. 54 and 55). Stable secondary structure within a leader can impede the scanning of the 40 S ribosomal subunit in its search for the initiation codon and thereby inhibit translation (40,41,56,57), whereas upstream open reading frames require the involvement of re-initiation or internal initiation mechanisms if the downstream cistron is to be translated (58). Secondary structure within the 5Ј-leader of an mRNA may serve to regulate translation in response to alterations in the amount or activity of the translational machinery. As eIFiso4F did not stimulate translation substantially from an mRNA containing a structured leader, eIF4F (or eIF4G) and its cellular concentration may be largely responsible for determining the extent to which structured mRNAs are translated. In plants, the level of eIF4G (but not eIF4E, eIFiso4G, or eIFiso4E) increases substantially during late seed development (59), a developmental stage in which most soluble proteins, including eIFiso4G, undergo proteolysis as the seed prepares to enter a metabolically quiescent stage as part of its maturation program. The developmentally late increase in eIF4G concentration may be required to facilitate the translation of mRNAs under the conditions that prevail during late seed development. The level of eIF4G (but not eIF4E) also increases following a heat shock as a function of the severity of the stress (59). Heat stress results in a loss of cap function and a preferential increase in the translation of uncapped mRNAs to a level equal that of capped mRNAs (60). These in vivo observations are similar to the loss-of-cap function and a preferential increase in the translation of uncapped mRNAs following an increase in eIF4F concentration (Table IV).
In conclusion, our results suggest that eIF4G (as part of eIF4F), in addition to supporting translation from normal capped mRNAs, may function to facilitate translation from nonstandard mRNAs, i.e. those that contain secondary structure proximal to the 5Ј-cap, those that lack a cap, and those that contain more than one cistron. In contrast, the function of eIFiso4G (as part of eIFiso4F) may be largely limited to supporting translation from standard mRNAs, i.e. capped mRNAs with an unstructured 5Ј-leader. FIG. 9. eIF4E, eIFiso4E, or PABP do not stimulate internal initiation. Uncapped GUS-SL-Con 134 -luc-A 50 mRNA was translated at 25 ng/l in the eIF4F/eIFiso4F-dependent lysate (top and middle panels) or in the PABP-depleted lysate (bottom panel). The lysate was supplemented with the indicated amounts of each protein, and the degree to which the second cistron of the dicistronic mRNA was translated was determined by the amount of luciferase synthesized. Luciferase expression is indicated as light units from 2 l of each reaction. eIF4G and eIFiso4G Functionally Differ