Quantitative Assessment of EF-1α·GTP Binding to Aminoacyl-tRNAs, Aminoacyl-viral RNA, and tRNA Shows Close Correspondence to the RNA Binding Properties of EF-Tu*

A ribonuclease protection assay was used to determine the equilibrium dissociation constants (K d ) for the binding of various RNAs by wheat germ EF-1α·GTP. Aminoacylated fully modified tRNAs and unmodified tRNA transcripts of four specificities (valyl, methionyl, alanyl, and phenylalanyl) from higher plants or Escherichia coli were bound with K d values between 0.8 and 10 nm. A valylated 3′-fragment of turnip yellow mosaic virus RNA, which has a pseudoknotted amino acid acceptor stem, was bound with affinity similar to that of Val-tRNAVal. Uncharged tRNA and initiator Met-tRNAMet from wheat germ, RNAs that are normally excluded from the ribosomal A site in vivo, bound weakly. The discrimination against wheat germ initiator Met-tRNAMet was almost entirely due to the 2′-phosphoribosyl modification at nucleotide G64, since removal resulted in tight binding by EF-1α·GTP. A 44-nucleotide RNA representing a kinked acceptor/T arm obtained by in vitroselection to bacterial EF-Tu formed an Ala-RNA·EF-1α·GTP complex with a K d of 29 nm, indicating that much of the binding affinity for aminoacylated tRNA is derived from interaction with the acceptor/T half of the molecule. The pattern of tRNA interaction observed for EF-1α (eEF1A) therefore closely resembles that of bacterial EF-Tu (EF1A).

The translational elongation factors EF-1␣ (eEF1A) 1 and EF-Tu (EF1A) are GTP-binding proteins that serve similar roles in protein synthesis in prokaryotes and eukaryotes, respectively. Their normal role is to deliver aminoacylated tRNAs into the A site of the ribosome. While dispensing this function for tRNAs carrying all the standard 20 amino acids (including methionine) that are elongationally inserted into proteins, elongation factors must discriminate against non-aminoacylated tRNAs and against the methionylated initiator tRNA Met that is loaded into the ribosomal P site by initiation factors.
The basis for these functions is well understood in bacterial systems, for which extensive ligand binding and structural studies have been performed. The reported equilibrium dissociation constants for the interactions of Escherichia coli EF-Tu⅐GTP with aminoacylated elongator tRNAs fall between 0.2 and 7 nM (1-3), varying over a 13.8-fold (4) or a 34-fold (2) range when determined in a single set of experiments. These tight associations forming aminoacyl-tRNA⅐EF-Tu⅐GTP ternary complexes contrast strongly with the weak affinity of E. coli EF-Tu⅐GTP for uncharged tRNA (K d ϭ 2.6 -2.8 M; Ref. 3). The 2250-fold weaker affinity for tRNA Phe lacking the esterified phenylalanyl moiety (3) effectively prevents uncharged tRNA from reaching the ribosomal A site unless the normal aminoacylation status of cellular tRNAs is severely impaired. A considerably smaller binding differential exists in the case of methionylated initiator tRNA Met ; the K d values for binding elongator methionyl-tRNA Met and initiator formylmethionyl-tRNA Met were reported as 1.8 and 136 nM, respectively (3), a 76-fold difference. Exclusion of initiator formylmethionyl-tRNA Met from the A site thus relies only partially on EF-Tu⅐GTP discrimination, but also on A site competition by elongator methionyl-tRNA Met ⅐EF-Tu⅐GTP ternary complex, complex formation with IF2⅐GTP, and selective interaction of initiator tRNA with the ribosomal P site (5).
The recently solved crystal structure of the phenylalanyl-tRNA Phe ⅐EF-Tu⅐GTP complex from Thermus aquaticus (6) has shown that the protein contacts the aminoacyl-tRNA only in the acceptor/T half of the tRNA molecule. Accordingly, aminoacylated tRNA half-molecules and tRNAs lacking the anticodon domain have been reported to bind Thermus thermophilus EF-Tu⅐GTP with affinities similar to those of the parental tRNAs (7,8).
The overall similarity of EF-1␣ function to that of EF-Tu (see, e.g., Refs. 9 -11) has been convincingly established by the interchangeability of EF-Tu and EF-1␣ in binding to, although not in supporting protein synthesis by, bacterial and mammalian ribosomes (12). EF-1␣, like EF-Tu, has been shown to form ternary complexes with GTP and aminoacyl-tRNA (9,10,13). However, in contrast to the rich information available for EF-Tu, quantitative data on the interaction of EF-1␣ with tRNAs are almost non-existent. The only estimate for the stability of aminoacyl-tRNA⅐EF-1␣⅐GTP ternary complex we have found in the literature is one of "about 10 nM" for Phe-tRNA (11), and the extent of binding discrimination against uncharged tRNA is not known. With regard to the exclusion of initiator methionyl-tRNA Met from the A site of eukaryotic ribosomes, it has been shown that a 2Ј-phosphoribosyl modification of the purine at position 64 of the yeast and wheat germ initiator tRNAs is an important antideterminant of interaction with EF-1␣⅐GTP (14,15). After removal of this modification, both tRNAs could serve as elongators in in vitro protein synthesis (15); in the yeast case, the demodified methionyl-tRNA showed an increased ability to enter ternary complex with EF-1␣⅐GTP (14), although no dissociation constants were reported.
In order to improve our understanding of the detailed function of EF-1␣ in eukaryotic protein synthesis, we have studied the binding of wheat germ EF-1␣⅐GTP to various charged and uncharged RNAs. Our results emphasize the similarity in the RNA binding properties of higher eukaryotic EF-1␣⅐GTP to those of its bacterial homologue EF-Tu⅐GTP.

EXPERIMENTAL PROCEDURES
Preparation of RNAs-Mature tRNA Phe from E. coli strain MRE600 was purchased from Boehringer Mannheim, total wheat germ tRNA was purchased from Sigma, and purified yeast initiator tRNA Met was a gift from Drs. C. Florentz and R. Giegé (Institut de Biologie Moléculaire  et Cellulaire, Strasbourg, France).
Mature tRNA Val , and initiator and elongator tRNAs Met were purified from total wheat germ tRNA by hybridization affinity using a procedure based on those of Tsurui et al. (16) and Mörl et al. (17). 3Ј-Biotinylated deoxyoligonucleotides complementary to the 30 nucleotides upstream and including the discriminator base (nucleotide 73) were synthesized by automated chemistry; these oligomers were based on the sequences of higher plant tRNA Val (lupine; Ref. 18 (21) and analysis of ribonuclease T1 digestion products by 20% sequencing polyacrylamide gel electrophoresis; no products apart from those expected from the selected sequence were observed (data not shown).
All other tRNAs were generated in vitro by transcription with T7 RNA polymerase in the presence of 10 mM 5Ј-GMP and 1 mM GTP to produce 5Ј-monophosphate termini (22). RNAs were purified by denaturing polyacrylamide gel electrophoresis (8%), recovered by electroelution, and dialyzed against water. Transcriptional templates were either plasmid DNAs linearized at a BstNI restriction site coincident with the 3Ј-CCA, or DNA amplified in vitro by polymerase chain reaction (23). Plasmid templates were used to make unmodified lupine tRNA Val (synthetic clone pTVAL, which encodes tRNA Val with a CAC anticodon in place of the modified IAC (I ϭ inosine) of the mature tRNA), E. coli tRNA Ala and short derivatives thereof (8), and wheat germ elongator tRNA Met (24).
Preparation of Aminoacylated RNAs-RNAs (30 -50 pmol) were preparatively aminoacylated with appropriate radiolabeled amino acids and aminoacyl-tRNA synthetases. The completed reactions were acidified by adjusting to 75 mM sodium acetete (pH 5.2), deproteinized with phenol/chloroform equilibrated to pH 5.2 with sodium acetate solution, and ethanol-precipitated. After recovery by centrifugation, the valylated RNAs were redissolved in 5 mM sodium acetate (pH 5.2) and stored at Ϫ80°C until use. The plant tRNAs, yeast initiator tRNA Met , and the viral TYSma RNA were aminoacylated with activities present in a partially purified wheat germ extract (26) (8).
EF-1␣⅐GTP Binding Assays-Purified wheat germ EF-1␣ (27) was activated by incubation with 20 M GTP in EF buffer (40 mM HEPES (pH 7.5), 100 mM NH 4 Cl, 10 mM MgCl 2 , 1 mM dithiothreitol) plus 25% (v/v) glycerol at 30°C for 20 min. Various concentrations (0 -250 nM) of EF-1␣⅐GTP were incubated together in multiwell strips (Nunc) with valylated RNAs (0.2-5 nM) in EF buffer containing 12.5% glycerol and 0.5 mg/ml fragmented salmon sperm DNA on ice for 15 min (after Ref. 13) in 20 l reactions. The amounts of ternary complex were estimated with a ribonuclease protection/ trichloroacetic acid filter precipitation assay adapted from Louie et al. (4). Incubations were terminated by addition of 4 l of ice-cold 5 mg/ml ribonuclease A (Sigma) and incubation for 15 s before quenching the ribonuclease activity with the further addition of 4 l of 50 mg/ml tRNA or total torula RNA (Sigma). Total reaction mixtures were then immediately pipetted onto dried 1-cm filter paper squares (3MM, Whatman) impregnated with 20% trichloroacetic acid containing 2 mM of the appropriate amino acid, and immediately plunged into excess 10% trichloroacetic acid to be held on ice for at least 5 min. After several washes with 5% trichloroacetic acid and 95% ethanol, the dried filters were subjected to counting in a liquid scintillation spectrometer.
Equilibrium dissociation constants (K d ) were calculated from binding curves comprising data from 14 concentrations of EF-1␣⅐GTP. Each binding assay was performed in duplicate, and repeated at least twice. The concentration of active EF-1␣⅐GTP present in binding experiments was determined by binding in the presence of excess [ 3 H]Val-tRNA Val transcript. Aminoacyl-tRNA⅐EF-1␣⅐GTP ternary complex formation follows Equation 1.
In terms of total concentrations of aminoacylated tRNA (t) and EF-1␣⅐GTP (e), and defining the concentration of ternary complex as c, this equation becomes K d ϭ ((t Ϫ c)(e Ϫ c))/c, which can be rearranged to c 2 Ϫ c(t ϩ e ϩ K d ) ϩ t⅐e ϭ 0. Solution for c, the quantity assayed in the binding reaction, by means of the general quadratic equation yields c ϭ The plotting program KaleidaGraph (Adelbeck Software) was used to solve the plot of c (filter cpm) versus e, yielding K d , t, and the y intercept representing the level of background counts. Note that, in practice, t (total concentration of charged tRNA competent to enter ternary complex) is an unknown quantity, since not all the aminoacyl-tRNA is capable of forming ternary complexes, presumably due to denaturation. Typically, 80 -95% of aminoacylated tRNAs were bound, but in some instances lower proportions of charged tRNAs were competent for binding (see Tables II and  III). The plotted equation incorporated a specific activity constant converting tc filter cpm to moles. This specific activity constant was determined by assaying 3 H-or 35 S-labeled aminoacylated tRNA immediately after desalting with a Sephadex G50 spin column. Counts (cpm) from direct counting in scintillation fluid at known efficiency were compared with counts obtained after taking an aliquot of the labeled tRNA in 20 l of binding buffer through a mock assay termination procedure (including addition of RNA and spotting onto filter paper). This specific activity measurement thus included adjustment for losses during transfer to the filter paper and from incomplete precipitation onto the paper.
Competition Binding Assays-Assays were performed as above, except that all EF-1␣⅐GTP concentrations were represented in triplicate, and two binding curves for the [ 35 S]methionyl-tRNA Met transcript were performed in parallel in the presence and absence of unlabeled competitor RNA. The same quadratic function used for fitting standard binding curves gave a good fit to the data determined in the presence of competitor RNA (similar chi-square value), and was used as an approximation of the extremely complex function describing the competition. The difference in K d estimated in the presence and absence of competitor ("K d offset"), which represents the concentration of competitor RNA⅐EF-1␣⅐GTP complex when the reporter is half-bound, was then used to calculate the K d for the competitor RNA. Equation 1 applied to the competitor RNA is shown by Equation 1a; when the reporter is half-bound, [free EF-1␣⅐GTP] ϭ K d (reporter), this latter term being determined in the binding assay lacking competitor.
Thus, the above equation becomes Equation 2 .
The competition binding assay was compared with the direct assay by determining the K d for a [ 3 H]valylated viral RNA both ways (data not shown). The direct assay yielded K d ϭ 2.9 Ϯ 0.5 nM, while the competition assay yielded K d values of 1.5 and 2.0 nM in two runs. The similarity of these estimates validates the competition assay.
De-modification of Initiator tRNA Met -Periodate oxidation and ribose removal by ␤-elimination were performed on wheat germ and yeast initiator tRNA Met as described (15), except that the reaction products were not subjected to chromatographic purification. The treatment results in loss of the phosphoribosyl modification of nucleotide 64 as well as removal of the 3Ј nucleotide. Demodification was judged com- 2 The abbreviation used is: TYMV, turnip yellow mosaic virus.
plete, since the 3Ј nucleotides were shown after 3Ј-labeling to have been quantitatively removed. The treated tRNAs were methionylated by a wheat germ methionyl-tRNA synthetase activity in the presence of (CTP,ATP):tRNA nucleotidyltransferase, which is able to regenerate the eliminated 3Ј-A residue.

RESULTS
Equilibrium Dissociation Constants for Aminoacylated tRNAs-A ribonuclease protection assay based on that described for the EF-Tu studies of Louie et al. (4) was developed to measure the dissociation constants for the interactions between EF-1␣⅐GTP and various aminoacylated tRNAs. This assay relies on the protection of the RNA bearing an amino acid labeled with 3 H or 35 S against ribonucleolytic degradation that occurs when the RNA is bound by EF-1␣. To make the technique more convenient, cheaper, and more accessible to RNAs available in limited quantities, the binding reactions were reduced in size to 20 l. This permitted the entire reaction to be spotted onto paper filters at the end of the assay, avoiding cumbersome filtration. After treatment with the high levels of ribonuclease A necessary to digest unprotected phosphodiester bonds within 15 s on ice, rapid quenching of the ribonuclease was needed to prevent further digestion while the RNA was precipitated onto filter paper squares. This was achieved by addition of ice-cold RNA, rapid spotting onto filter paper that had been dried after impregnation with 20% trichloroacetic acid, and immediate addition to a 10% trichloroacetic acid bath held on ice. To permit binding assays at subnanomolar concentrations of aminoacylated tRNA, necessary when using [ 35 S]methionyl-tRNA Met as a reporter in competition binding assays (see below), fragmented salmon sperm DNA (0.5 mg/ml) was included in all binding reactions in order to prevent adsorptive losses of labeled RNA to the plastic walls of the incubation vessel.
Representative binding curves of four of the aminoacylated RNAs studied are shown in Fig. 1, and the dissociation constants determined from at least three replicates are reported in Table I. The tRNAs assayed were of four specificities (valine, methionine, alanine, and phenylalanine), and included tRNAs of higher plant or bacterial origin; the tRNAs were either fully modified mature tRNAs purified from cells, or in vitro transcripts lacking post-transcriptional modifications but with the natural 5Ј-GMP termini. The dissociation constants measured for the aminoacylated tRNAs varied between about 1 and 10 nM. Only small differences in binding affinity were observed between the fully modified and unmodified forms of plant tRNA Val and tRNA Met (both forms were derived from the same gene in the case of tRNA Met ). No significant difference was observed between the binding affinities of the plant (Arabidopsis thaliana) and bacterial (E. coli) tRNA Ala transcripts (5.3 versus 6.5 nM). Both alanyl-tRNA and phenylalanyl-tRNA were bound considerably less tightly than valyl-tRNA and methionyl-tRNA.
The rather narrow range in the dissociation constants of the aminoacylated tRNAs studied in Table I   with tRNAs of widely differing sequences. Despite varying primary sequence, however, tRNAs all share very similar Lshaped tertiary conformations. Different structural motifs (a pseudoknotted amino acid acceptor stem; Refs. 28 -30) are found in some aminoacylatable plant viral RNAs that have been shown to interact with EF-1␣ (29, 31), although dissociation constants for these interactions have never been reported. TYMV RNA has a 3Ј-terminal 82-nucleotide-long tRNA-like structure that can be aminoacylated with valine (26,32). To assess the affinity of EF-1␣ for pseudoknotted RNA, we prepared 264-nucleotide-long TYSma transcripts (25) that included the tRNA-like structure shown in Fig. 2, and assayed the binding affinity of the valylated RNA to wheat germ EF-1␣⅐GTP. The dissociation constant (1.9 nM; Table I) measured for the valylated viral RNA was comparable to that of the unmodified valyl-tRNA Val transcript (2.3 nM), indicating a capacity for EF-1␣⅐GTP to form tight complexes with RNAs of varying acceptor stem architecture. Tight complex formation was dependent on EF-1␣ activation and was not observed for EF-1␣⅐GDP, and valylated viral RNA competed with valylated tRNA Val for binding by EF-1␣⅐GTP (data not shown). These observations confirm that the valylated viral RNA was bound by the same binding surface as aminoacyl-tRNA.
Discrimination between Charged and Uncharged tRNA-The high specific activity labeling achievable for tRNA aminoacylated with [ 35 S]methionine made this an excellent reporter species in a competition variant of the standard assay, permitting a study of the affinity of wheat germ EF-1␣⅐GTP for uncharged wheat germ tRNA. The addition of 10 M uncharged wheat germ tRNA to 0.47 nM [ 35 S]methionyl-tRNA Met transcript resulted in a slightly weakened apparent binding affinity of the reporter methionyl-tRNA Met . The calculated K d of 15.2 Ϯ 6.3 M for the uncharged tRNA (Table I) means that the affinity of EF-1␣⅐GTP for uncharged tRNA is 10 3 -fold to 10 4 -fold weaker than for charged tRNA.
Phosphoribosyl Modification of Initiator tRNA Met as an Antideterminant of Interaction with EF-1␣⅐GTP-Mature initiator tRNA Met and elongator tRNA Met were purified by a hybridization affinity method from total wheat germ tRNA. 3Ј-Labeling with 32 P revealed no detectable cross-contamination of signature oligonucleotides generated by ribonuclease T1 between these two tRNA Met preparations (data not shown). In contrast to the tight binding of elongator [ 35 S]methionyl-tRNA Met (K d ϭ 0.83 nM; Table I), initiator [ 35 S]methionyl-tRNA Met bound weakly to EF-1␣⅐GTP. At accessible concentrations of EF-1␣⅐GTP, the binding curves were incomplete, preventing reliable estimation of the dissociation constant. However, the shape of the curve suggests a dissociation constant Ͼ100 nM (Table II). Similarly weak binding was observed to yeast initiator [ 35 S]methionyl-tRNA Met (data not shown).
Both the wheat germ and yeast initiator tRNA Met were oxidized with periodate in order to remove the 2Ј-ribosyl modification of nucleotide 64 by ␤-elimination (15,33). The oxidized, phosphatase-treated tRNAs were preparatively aminoacylated with [ 35 S]methionine using a methionyl-tRNA synthetase activity from wheat germ that also contains (CTP,ATP):tRNA nucleotidyltransferase. This latter activity replaced the 3Ј-terminal A residue removed by the ␤-elimination procedure, permitting subsequent aminoacylation. Both oxidized initiator methionyl-tRNAs formed high affinity complexes with EF-1␣, and produced binding curves similar to those of elongator methionyl-tRNA Met , with K d values of 2.5 and 12 nM for the wheat germ and yeast tRNAs, respectively (Table II). There was a consistently higher background when using the oxidized tRNA, and a rather low proportion of the oxidized methionyl-tRNAs was capable of forming ternary complex (33% and 34% for two different preparations of oxidized wheat germ tRNA Met ); the latter was probably due to incomplete removal of the phosphoribosyl group or other damage to the molecule resulting from the chemical treatment. Nevertheless, these results clearly show that the phosphoribosyl modification of nucleotide 64 is a powerful antideterminant of EF-1␣⅐GTP binding.
EF-1␣⅐GTP Interacts Primarily with the Acceptor/T Arm of tRNA-Three truncated derivatives of E. coli tRNA Ala that had previously been used in studies with EF-Tu (8) were used to assess the part of tRNA interacting with wheat germ EF-1␣⅐GTP (Fig. 3). The ⌬ Anticodon RNA has the 17-nucleotide anticodon stem/loop replaced with a UUAA spacer. The Ala minihelix RNA lacks all D-and anticodon arm sequences, with the 3Ј-82-nucleotide-long tRNA-like structure that is present in TYSma transcripts. To assist in following the linear sequence through the pseudoknotted amino acid acceptor stem, the TYMV nucleotides are numbered (from the 3Ј-end). Loops L1 and L2, which cross the major and minor grooves of the helix, respectively (30), are indicated. Nucleotide 64 is arrowed on the tRNA Val structure to indicate the position of the 2Ј-phosphoribosyl modification of initiator tRNA Met (see "Results"). the remaining acceptor/T arm built as a continuous 12-base pair helix. RNA 12 is an RNA that emerged from sequential in vitro selections as a tight ligand of T. thermophilus EF-Tu⅐GTP (8), and contains a 9-residue bulge loop interrupting the Ala minihelix RNA. It was thought that the bulge mimicked a known bend between the T and acceptor stems of tRNA (34), thereby improving the stability of ternary complexes formed with charged minihelix-type RNAs.
Alanyl-⌬ Anticodon RNA was bound by wheat germ EF-1␣⅐GTP with a K d only about 3-fold higher than that for the alanyl-tRNA Ala transcript (Table III), indicating that the anticodon domain is not essential for binding. As observed in studies with T. thermophilus EF-Tu⅐GTP, a rather low fraction of the deleted alanyl-RNA formed ternary complexes (Table III), suggesting a tendency for this RNA to assume a conformation not recognized by EF-1␣⅐GTP. Note that the alanylated transcripts of E. coli tRNA Ala were themselves bound to only 56% in our assays, considerably lower than the 80 -95% typical of charged tRNAs. Alanyl-RNA 12 was bound somewhat less tightly, with a K d 4.5 times higher than that of the alanyl-tRNA Ala transcripts, but some 3-fold tighter than alanyl-Ala minihelix (Table III). Incomplete binding curves for this latter RNA indicated poor interaction with EF-1␣⅐GTP. These results demonstrate that, as with EF-Tu, a bent acceptor/T arm is preferred for binding, and that most of the binding affinity for aminoacyl-tRNA is derived from interaction with the acceptor/T portion of the tRNA.
Residues Contacting Aminoacyl-tRNA Are Conserved between EF-Tu and EF-1␣-The close similarity in RNA binding properties between bacterial EF-Tu and wheat germ EF-1␣ prompted a detailed sequence comparison of these proteins to determine whether sequence conservation is especially high in regions of T. aquaticus EF-Tu that are in contact with the aminoacyl-tRNA. The best studied EF-Tu proteins, from E. coli, T. thermophilus, and T. aquaticus, which are known to share the same folded structure (35,36), each have about 33% sequence identity and 55% weighted similarity in primary sequence with wheat EF-1␣. 3 The sequence identity is scattered throughout the protein, permitting ready alignment of all EF-Tu and EF-1␣ sequences in the data base (Fig. 4). Secondary structure prediction by the PHDsec program at the Pre-dictProtein server (37) for T. aquaticus EF-Tu was in good agreement with the crystal structure (6), and the secondary structure elements predicted for human and wheat EF-1␣ could be readily fitted to the EF-Tu structure. Alignment based on predicted structure violated no alignments of universally conserved amino acids (data not shown). Higher eukaryotic EF-1␣s differ by some 7 insertions and 3 deletions from EF-Tu. Mammalian and plant EF-1␣s are 18% and 14% longer, respectively, than E. coli EF-Tu, one of the shortest. Insertions or deletions are almost exclusively placed in the loops and at the N and C termini of the EF-Tu structure, rather than within helix or sheet structural elements (Fig. 4), such that overall structure would be expected to be conserved.
In Fig. 4, highly conserved amino acids have been marked along with the secondary structure elements and ligand contacts taken from the EF-Tu ternary complex crystal structure (6). Amino acids that contact the GTP⅐Mg 2ϩ and aminoacyl-tRNA noticeably coincide with or are close to clusters of conserved residues (Fig. 4). This is clearly the case for most of the GTP⅐Mg 2ϩ contacts (in domain 1) and the contacts with the aminoacyl-and 5Ј-termini of the bound RNA (in domains 1 and 2; Fig. 4). The interactions of T. aquaticus residues Tyr-47, Asp-51, and Ser 164 -Ala 165 -Leu 166 with GTP⅐Mg 2ϩ , and of residues Lys 52 with the phosphates in the 3Ј-CCA of the tRNA (6), are an exception, although these residues are close to a single conserved amino acid. Contacts with the T-stem of the bound tRNA in domain 3 involve less conserved amino acids, partly because domain 3 is considerably less conserved than the other two domains. Nevertheless, the T-stem contacting T. aquaticus residues, Gln 341 and Gly 391 , fall within clusters of conserved residues.
The general coincidence of ligand-contacting amino acids 3 Information was obtained using the PredictProtein server (http://www.embl-heidelberg.de/predictprotein/ppDoPredDef.html).  with conserved residues means that the length heterogeneities and regions of low sequence conservation are separated from the elements crucial for ternary complex formation. The combined sequence and structure alignment in Fig. 4 is consistent with the model that EF-1␣ has a core structure close to that of EF-Tu, bearing superficial elaborations mainly in external loops not in contact with aminoacyl-tRNA. This interpretation is consistent with the closely similar RNA binding properties of EF-Tu and EF-1␣. Supportive experimental evidence comes from cross-linking studies between rabbit EF-1␣⅐GTP and aminoacyl-tRNA (38) that indicate contact sites corresponding to T. aquaticus EF-Tu residues 270 -275, 345-355 and 375, sites that are in contact with the RNA in the ternary complex crystal (6). The rather large sequence variability surrounding the GTP⅐Mg 2ϩ binding site (insertion in the helix near T. aquaticus residue 35 to produce a predicted helix-loop-helix, and variability in the helix between residues 144 and 161; Fig. 4) may be responsible for the different nucleotide binding properties of EF-Tu and EF-1␣. EF-1␣ has similar affinity for GTP and GDP, while EF-Tu has a 100-fold higher affinity for GDP than for GTP (13). DISCUSSION High Affinity Binding of EF-1␣⅐GTP to Aminoacylated tRNAs-The equilibrium dissociation constants of ternary complexes formed by wheat germ EF-1␣⅐GTP with tRNAs of four specificities (valyl, methionyl, alanyl, and phenylalanyl) were between 0.8 and 10 nM (Table I), indicating a high affinity interaction similar to that between E. coli EF-Tu⅐GTP and aminoacyl-tRNAs (K d values of 0.2-7 nM; Refs. 1-3). Phenylalanyl-tRNA Phe from E. coli was bound by EF-1␣⅐GTP with the weakest affinity of the aminoacyl-tRNAs tested. Since there was no significant difference between the binding of plant and E. coli alanyl-tRNA Ala transcripts to EF-1␣⅐GTP (5.3 and 6.5 nM, respectively; Table I), the weaker binding of Phe-tRNA Phe is unlikely to be due to the fact that this tRNA and the EF-1␣ are from different sources. The 12-fold range of K d values we have observed for the aminoacyl-tRNA⅐EF-1␣⅐GTP complexes for a subset of the 20 amino acid specificities falls within the 13.8-fold (4) and 34-fold (2) ranges measured for aminoacyl-tRNA complexes with E. coli EF-Tu⅐GTP. However, the relative affinities of different aminoacyl-tRNAs do not appear to strictly follow those observed for EF-Tu  Table I). The stabilities of ternary complexes formed with higher eukaryotic EF-1␣⅐GTP are thus clearly similar to those formed by bacterial EF-Tu⅐GTP.
Only about a 2-fold difference in the binding affinities to wheat germ EF-1␣⅐GTP were observed between the fully modified and unmodified forms of Val-tRNA Val and Met-tRNA Met (Table I). Similar approximately 2-fold differences were observed between the K d values for EF-Tu⅐GTP ternary complexes containing modified and unmodified aminoacyl-tRNAs (aspartate, Ref. 7; phenylalanine, Ref. 39). Thus, in both the eukaryotic and prokaryotic systems, modified nucleotides, such as the thymine and pseudouracil bases present in the T-loops of virtually all tRNAs, exert little or no influence on ternary complex formation.
Binding Discrimination against Two Types of tRNAs That Are Normally Excluded from the Ribosomal A Site-Since the role of EF-1␣⅐GTP is to deliver aminoacyl-tRNAs to the ribosomal A site, it was of interest to determine the extent to which two A site non-ligands (uncharged tRNA and initiator Met-tRNA Met ) are excluded from the A-site by virtue of weak interaction with EF-1␣⅐GTP. Uncharged wheat germ tRNA interacted very weakly with EF-1␣⅐GTP (Table I). The 10 3 -fold to 10 4 -fold difference in the K d values of ternary complexes containing charged and uncharged tRNAs is similar to the 2250fold difference in K d values for the equivalent ternary complexes formed with E. coli EF-Tu⅐GTP (3). Thus, in both the eukaryotic and prokaryotic systems, the poor ability to form ternary complexes with EF⅐GTP is a major factor in excluding uncharged tRNA from the ribosomal A site.
Discrimination at the level of ternary complex formation with EF-1␣⅐GTP is clearly also a factor in excluding initiator FIG. 4. Sequence alignment between EF-Tu and EF-1␣. The EF-1␣ sequences from wheat, yeast, and humans (accession numbers Q03033, P02994, and P04720, respectively) were aligned with the EF-Tu sequences from T. aquaticus and E. coli (accession numbers Q01698 and P02990, respectively). Amino acids conserved between the two EF-Tu species shown and at least 90% of the 64 EF-1␣ sequences retrieved from the data base are shaded. Also shown are sheet (b) and helix (a) secondary structure elements for T. aquaticus EF-Tu (6), and points of contact with GTP⅐Mg 2ϩ (G; including Mg 2ϩ contacts through water molecules), the CCA-phenylalanyl moiety (C), the 5Ј-terminus (5), and T-stem (T) of the bound aminoacyl-tRNA (data taken from crystal structures in Refs. 6, 35, and 36 Table I) from the A site. Although a better estimate for the ternary complex K d will need to be obtained in the future, it appears that the role of discrimination by EF-1␣⅐GTP in plants is at least as significant as that of EF-Tu⅐GTP discrimination in E. coli (the K d for the EF-Tu⅐GTP ternary complex with initiator formylmethionyl-tRNA Met was reported as 136 nM; Ref. 3), but probably not as dominant as in the case of uncharged tRNA. Removal from the wheat germ initiator tRNA of the phosphoribosyl modification of nucleotide 64 that is a characteristic of plant and yeast initiator tRNAs (33) led to high affinity binding by EF-1␣⅐GTP (K d ϭ 2.5 nM; Table I). Since the ratio of this K d to that for the ternary complex with elongator Met-tRNA Met is only 3, it is evident that the phosphoribosyl modification acts as a powerful antideterminant blocking EF-1␣⅐GTP binding. This modification is positioned at the base of the T stem (see Fig. 2) in a position that is in close proximity to the protein in the EF-Tu⅐GTP ternary complex (6). Our results are in agreement with the experiments of Sprinzl and co-workers (14,15) on the critical role of modified nucleotide 64.
EF-1␣⅐GTP Interacts Primarily with the Acceptor/T Half of Aminoacyl-tRNA-The relatively tight complexes formed between EF-1␣⅐GTP and two alanylated deletion variants derived from tRNA Ala (⌬ Anticodon and 12 RNAs; Table III) show that regions of the tRNA outside the acceptor/T arm play at most a minor role in complex formation. Both alanyl-RNAs were far superior EF-1␣⅐GTP ligands to uncharged tRNA. The tighter ternary complex formation of alanyl-12 over alanyl-Ala minihelix RNA (Table III), also observed with T. thermophilus EF-Tu⅐GTP (8), suggests that a kinked acceptor/T arm as revealed by x-ray crystallographic studies for tRNAs (34) is more favorable for binding than a straight helix. An uninterrupted acceptor/T helix may not readily undergo the conformational change needed to assume the generalized conformation that EF-Tu⅐GTP is suggested to impose on the various aminoacyl-tRNAs on binding (3,40). Interestingly, the high affinity binding to valylated-TYSma RNA (Table I), which has a pseudoknotted acceptor stem (Fig. 2), indicates that minor structural variations in the acceptor/T arm are compatible with tight binding to EF-1␣⅐GTP. Perhaps the observed flexibility between the stacked helices of the TYMV acceptor/T arm (30) helps to make this RNA such an excellent ligand to EF-1␣⅐GTP.
Our findings indicate strong similarities in the way that eukaryotic and prokaryotic EF⅐GTPs interact with aminoacyl-tRNAs to form ternary complex, supporting an overall structure of the EF-1␣ ternary complex similar to that of T. aquaticus EF-Tu (6), rather than the structure proposed by Kinzy et al. (38). The demonstrated interaction of EF-Tu⅐GTP with the acceptor/T half of the tRNA (6) appears to be applicable also to ternary complex formation by EF-1␣⅐GTP, as deduced by the interference role of the phosphoribosyl modification in the T stem of initiator tRNA Met and ternary complex formation by the tRNA half-molecule 12 RNA. The binding discrimination against initiator tRNA and uncharged tRNA, and tight ternary complex formation with various charged tRNAs are also closely similar functions of EF-1␣⅐GTP and EF-Tu⅐GTP. These RNAinteraction properties are strongly conserved between EF-Tu and EF-1␣ even though these proteins have only about 33% sequence identity. Notably, however, most of the sequence variability between EF-Tu and EF-1␣ is away from the amino acids contacting the aminoacyl-tRNA. These contact sites include many highly conserved residues, particularly among those contacting the aminoacyl-and 5Ј-termini, while much of the sequence variability is accommodated in loops (Fig. 4) distant from the RNA (6). This arrangement presents a picture of a protein with a core structure that interacts with aminoacyl-tRNA much like EF-Tu does, but with unique surface features that presumably are involved in the cytoskeleton interaction (41) and various roles in cellular regulation (42) that are characteristic of higher eukaryotic EF-1␣.