X-ray Structure of Translation Initiation Factor eIF2 (cid:1) IMPLICATIONS FOR tRNA AND eIF2 (cid:1) BINDING*

The x-ray structure of the (cid:1) -subunit of the heterotrimeric translation initiation factor eIF2 has been determined to 2.4-Å resolution. eIF2 is a GTPase that delivers the initiator Met-tRNA to the P site on the small ribosomal subunit during a rate-limiting initiation step in translation. The structure of eIF2 (cid:1) closely resembles that of EF1A (cid:1) GTP, consisting of an N-terminal G domain followed by two (cid:2) -barrels arranged in a closed configuration with domain II packed against the G domain in the vicinity of the Switch regions. The G domain of eIF2 (cid:1) has an unusual zinc ribbon motif, not previously found in other GTPases. Structure-based site-directed mutagenesis was used to identify two adjacent features on the surface of eIF2 (cid:1) that bind the (cid:3) -subunit and Met-tRNA iMet , respectively. These structural, biochemi- cal, and genetic results provide new insights into eIF2 ternary complex assembly. Translation initiation involves assembly of a large protein-RNA complex at the initiation codon of a mRNA in three main steps: binding of initiator Met-tRNA and mRNA with associated factors to the small ribosomal subunit, pairing of the anticodon glutathione S -transferase (GST) fusion protein was expressed in Esch- erichia coli strain BL21(DE3). Cells were grown to an OD 600 of (cid:3) 0.8 in minimal medium containing 0.2 mg/ml ampicillin at 37 °C, and the expression was induced by 0.5 m M isopropyl- o -thiogalactoside. After 6 h of induction at 30 °C, cells were harvested and resuspended in 20 m M TrisCl, pH 7.5, 300 m M NaCl, 1 m M aprotinin, and disrupted using the French press. After centrifugation, the protein was loaded on a GST affinity column. Following cleavage of the GST tag with Precision protease, eIF2 (cid:3) was further purified with heparin and gel filtration chromatography. Free enzyme crystals (native and Se-Met) were grown at room temperature via hanging drop vapor diffusion against 100 m M HEPES pH 7.5, 10%(w/v) PEG 6000, 5%(v/v) methyl pentanediol, using a protein concentration of 15 mg/ml. The crystals grow in the monoclinic space group P2 1 with one protein per asymmetric unit (unit cell: a (cid:4) 52.8 Å, b (cid:4) 52.4 Å, c (cid:4) 74.1 Å, (cid:2) (cid:4) 92.5°, diffraction limit (cid:4) 2.4-Å resolution). Crystal cryoprotection was achieved by adding glycerol to a final concentration of 20% (v/v). YCplac33 eIF2 (cid:3) constructed the Altered Sites mutagenesis mutagenic oligonucleotides: EcoRI-PstI digestion, J212. was generated by subcloning a (cid:3) 2.4-kb BamHI fragment from plasmid p1097 (22) to the vector pRS423. The high copy number HIS3 plasmid encoding IMT4 (pC1683) was generated by transferring a (cid:3) 0.8-kb SalI-BamHI fragment from the plasmid p1776 (20) to the plasmid pRS423. Standard methods were used for culturing yeast strains (23). Yeast growth analyses were performed by streaking purified transformants on S.D. minimal medium supplemented with essential nutrients or on YPD rich medium. Growth rates were qualitatively determined by comparing the size of isolated single colonies in the streaks after growth at 30 °C for 3 days.

Translation initiation involves assembly of a large protein-RNA complex at the initiation codon of a mRNA in three main steps: binding of initiator Met-tRNA and mRNA with associated factors to the small ribosomal subunit, pairing of the anticodon of Met-tRNA i Met with the AUG start codon, and joining of the two ribosomal subunits to form the 80 S initiation complex. In eukaryotes, this process requires at least twelve distinct translation initiation factors (eIFs) 1 and hydrolysis of two molecules of GTP (reviewed in Ref. 1). Translation initiation factor eIF2 is a heterotrimeric GTPase that delivers the Met-tRNA i Met to the small ribosomal subunit as part of a ternary complex with GTP. Pairing between the anticodon of the Met-tRNA i Met and the AUG start codon triggers hydrolysis of GTP by eIF2 and eIF2⅐GDP is released, leaving Met-tRNA i Met bound in the P site. eIF2⅐GDP cannot bind Met-tRNA i Met and is converted to eIF2⅐GTP by the heteropentameric exchange factor eIF2B (reviewed in Ref. 2). eIF2 is a stable complex of three subunits (␣, ␤, and ␥), each essential for viability in yeast. Since their discovery, highly conserved eIF2 subunits have been identified in all eukaryotes (reviewed in Ref. 2) and also archaebacteria (3). Biochemical and genetic analyses have demonstrated that the three subunits have distinct activities during translation initiation. The ␥-subunit binds GTP and Met-tRNA i Met (reviewed in Ref. 4). GTP hydrolysis by the eIF2 ternary complex bound to the 40 S ribosomal subunit is stimulated by eIF5, which interacts with eIF2␤ (5). The ␤-subunit also interacts with Met-tRNA i Met and has been reported to bind mRNA (6). Given the pivotal role eIF2 plays in eukaryotic translation initiation, it is not surprising that it represents an important target for regulation. In response to various environmental stressors (viral infection, amino acid starvation, heme deficiency, ER stress etc), the ␣-subunit is phosphorylated on Ser 51 , abolishing translation initiation by preventing eIF2B catalyzed GDP-GTP exchange.
The ␥-subunit of eIF2 belongs to the superfamily of GTPbinding proteins and is most closely related to the translation elongation factor eEF1A and its eubacterial counterpart EF1A (previously EF-Tu), which form ternary complexes with GTP and aminoacylated elongator tRNAs. eIF2␥ sequences are conserved from archaebacteria to mammals (pairwise amino acid sequence identities, 43-59%; Fig. 1), suggesting that they share a common three-dimensional structure (7). The highest sequence conservation occurs within the G domain (pairwise identities, 46 -69%). Mutations of residues implicated in GTP binding are lethal in the yeast Saccharomyces cerevisiae (8), confirming the functional importance of the G domain and, by inference, GTP hydrolysis in translation initiation.
Here we present the x-ray structure of Methanococcus jannaschii eIF2␥ at 2.4-Å resolution. Apo eIF2␥ consists of three domains structurally homologous to those in EF1A and arranged in a closed configuration, similar to that seen in EF1A⅐GTP and not EF1A⅐GDP. A similar domain arrangement was observed in the structure of eIF2␥ from Pyrococcus abyssi (9). Our structure of an archaeal eIF2␥ from a different organism and in a different space group shows that the closed configuration is not due to crystal packing effects and confirms that domain coupling differs between eIF2␥ and EF1A, despite high sequence identity. Interestingly, comparison of our structure with the published one reveals that domains II and III are rotated by 14°with respect to the G domain revealing a hinge point around the Switch 2 region of the G domain, around which domains II and III move as a rigid body. In addition, using structure-based mutational analyses combined with genetic and biochemical experiments we mapped on eIF2␥ conserved and adjacent binding sites for the 3Ј-end of Met-tRNA i Met and eIF2␣, providing new insights into the mechanism of ternary complex assembly.

EXPERIMENTAL PROCEDURES
Protein Preparation and Crystallization-A pUC18 derivative (AMJBZ12) containing M. jannaschii genomic DNA encoding eIF2␥ was obtained from American Type Culture Collection (ATCC), Inc. (Manassas, VA). DNA corresponding to residues 35-437 of M. jannaschii eIF2␥ was amplified by PCR using primers designed to introduce a 5Ј-BamHI site and a 3Ј-EcoRI site. Following digestion the DNA was inserted into the pGEX-6P-1 (Amersham Biosciences) expression vector generating the plasmid pC1399. The N-terminal 35 residues of eIF2␥ are not conserved and the corresponding region in yeast was shown to be dispensable for function (10), indicating that the critical functional domains of eIF2␥ are present in our construct. The resulting N-terminal FIG. 1. Structure-based sequence alignment of selected eIF2␥ homologues from Archaea and eukaryotes and EF1A. Sequence numbering corresponds to M. jannaschii eIF2␥ (GenBank TM accession code: gi 1591895), with GTP binding motifs (G1, G2, G3, G4, and G5) enclosed in red. Secondary structural elements are denoted as follows: ␣or 3 10 -helices, cylinders; ␤-strands, arrows; random coil, lines. Gray circles represent portions of the polypeptide chain that were not well resolved in the electron density maps. The four cysteines involved in zinc ion binding are indicated with blue stars. BLOSUM62 sequence similarity is colored coded using a gradient from white (Ͻ40% identity) to dark green (100% identity). The sequence of Thermus aquaticus EF1A is also color coded according to its similarity with M. jannaschii eIF2␥. glutathione S-transferase (GST) fusion protein was expressed in Escherichia coli strain BL21(DE3). Cells were grown to an OD 600 of ϳ0.8 in minimal medium containing 0.2 mg/ml ampicillin at 37°C, and the expression was induced by 0.5 mM isopropyl-o-thiogalactoside. After 6 h of induction at 30°C, cells were harvested and resuspended in 20 mM TrisCl, pH 7.5, 300 mM NaCl, 1 mM aprotinin, and disrupted using the French press. After centrifugation, the protein was loaded on a GST affinity column. Following cleavage of the GST tag with Precision protease, eIF2␥ was further purified with heparin and gel filtration chromatography. Free enzyme crystals (native and Se-Met) were grown at room temperature via hanging drop vapor diffusion against 100 mM HEPES pH 7.5, 10%(w/v) PEG 6000, 5%(v/v) methyl pentanediol, using a protein concentration of 15 mg/ml. The crystals grow in the monoclinic space group P2 1 with one protein per asymmetric unit (unit cell: a ϭ 52.8 Å, b ϭ 52.4 Å, c ϭ 74.1 Å, ␤ ϭ 92.5°, diffraction limit ϭ 2.4-Å resolution). Crystal cryoprotection was achieved by adding glycerol to a final concentration of 20% (v/v).
Data Collection, Structure Determination, and Refinement-Diffraction data were measured at the SGX CAT Beamline of Argonne National Laboratory. Se-Met single wavelength anomalous dispersion (SAD) data (11,12) were collected at an x-ray wavelength corresponding to the white line of the selenium K-absorption edge. Data were processed using DENZO/SCALEPACK (13). Because of radiation induced decay, diffraction data from two different crystals were merged in SCALEPACK. All seven possible selenium sites were found using SnB (14), followed by anomalous difference Fourier syntheses with preliminary SAD phases. Definitive experimental phases were calculated at 2.35-Å resolution with SHARP (15), giving a final figure of merit of 0.58 for accentric reflections (Table I). After density modification, 80% of the polypeptide chain could be built into the electron density map using O (16). Refinement of this partial model using CNS (17) and calculation of difference Fourier syntheses allowed completion of protein structure building. The current refinement model of eIF2␥ consists of residues 35-437, one zinc ion, and 182 water molecules. Two regions of the polypeptide chain (residues 63-71 and 206 -213) were not well resolved in the electron density map, and are presumed disordered. The working and free R factors at 2.4-Å resolution are 21.2 and 26.4%, respectively. PROCHECK (18) revealed 2 unfavorable (,) combinations, and main chain and side chain structural parameters consistently better than average (overall G-factor ϭ 0.2).
eIF2␣ Binding Analysis-Derivatives of pGEX-6P-expressing mutant forms of M. jannaschii eIF2␥ were constructed by mutating the appropriate codons in the plasmid pC1399. The M. jannaschii eIF2␥-R242D and -G319I mutants were generated using the Altered Sites mutagenesis kit (Promega, Inc.) according to the vendor's protocols and the following mutagenic primers: R242D, 5Ј-ATGTATGTTGCAGATA-GCTTTGATATCAACAAACCAGGA; G319I, 5Ј-AGAAAAGCTCATCCC-GGGGGTTTGATTATTGTTGGGACAACA. The eIF2␥-L256D mutant was generated by PCR using the primers 5Ј-CCCCCCAAGCTTTGAT-ATCAACAAACCAGGAAACTGAGATTAAGGATGACAAAGGAGGGG and 5Ј-CCAATCTCAGCACATATTGGAAGCTTTAATTTTATATCCGC. The PCR product was digested with HindIII and used to replace the corresponding HindIII fragment in pC1399. The eIF2␥-D325A mutant was generated by PCR using a 5Ј-primer containing a BamHI site immediately upstream of the codon for residue 35 and the 3Ј primer: 5Ј-CCCCCCATGCATCTGATTTTGTTAAGTATGGGGCTAATGTTGT-CCCAACCCCAATCAAAC. The PCR product was digested with BamHI and NsiI and used to replace the corresponding fragment in pC1399. All mutant and wild-type constructs were confirmed by DNA sequencing. GST-M. jannaschii eIF2␥ (35-437) fusions (wild-type and mutants) and GST alone were expressed in Escherichia coli, purified by glutathione-Sepharose chromatography as previously described, and dialyzed against binding buffer (25 mM TrisCl, pH 7.5, 300 mM NaCl, 10 mM MgCl 2 , 5 mM dithiothreitol, 10%(v/v) glycerol). A pUC18 derivative (AMJIN75) encoding M. jannaschii eIF2␣ was obtained from ATCC, Inc. DNA encoding full-length M. jannaschii eIF2␣ was amplified by PCR using primers to insert a 5Ј-NdeI and a 3Ј-XhoI restriction site. The resulting product was subcloned into the pET28a vector (Novagen) and expressed untagged in E. coli using the protocol described for eIF2␥. The protein was purified on a heparin Sepharose column followed by dialysis against binding buffer. For binding studies, GST and GST-eIF2␥ (wild type and mutants) were each immobilized on 50 l of glutathione-Sepharose resin (Amersham Biosciences), and unbound protein removed by washing. 14 g of eIF2␣ were added to 50 l of GST-resin or GST-eIF2␥-resin mixture. Binding reactions were diluted to 100 l with binding buffer and incubated at 4°C for 1 h. After washing twice with 500 l of binding buffer for 20 min, the resin was harvested by centrifugation, and bound proteins were eluted with elution buffer (50 mM TrisCl, pH 8.0, 300 mM NaCl, 10 mM MgCl 2 , 5 mM dithiothreitol, 20 mM reduced glutathione) and detected by SDS-PAGE electrophoresis.
Standard methods were used for culturing yeast strains (23). Yeast growth analyses were performed by streaking purified transformants on S.D. minimal medium supplemented with essential nutrients or on YPD rich medium. Growth rates were qualitatively determined by comparing the size of isolated single colonies in the streaks after growth at 30°C for 3 days.

RESULTS AND DISCUSSION
Crystallization and Structure Determination-M. jannaschii eIF2␥ yielded high quality crystals containing one molecule/ asymmetric unit (see "Experimental Procedures"). The structure of eIF2␥ was determined via SAD (11, 12)(see "Experimental Procedures" and Table I for a complete description of the crystallographic structure determination and subsequent refinement.) Structural Overview- Fig. 2A shows the structure of eIF2␥, which consists of three domains (G, II, and III). Structurebased sequence alignments of various eukaryotic and archaeal homologs ( Fig. 1) demonstrate that conserved residues map to the hydrophobic cores of each of the three domains, whereas insertions and deletions map to random coil portions of the structure. The remarkable level of sequence identity and the pattern of amino acid differences across phylogeny allow us to conclude that all known eIF2␥s share the three-dimensional structure illustrated in Fig. 2A.
eIF2␥ also contains a zinc ribbon (residues 84 -107) that protrudes from the main body of the G domain, forming an overhang between the G domain and domain II. The zinc ribbon motif is formed by two ␤-hairpin turns, each contributing a pair of cysteines ( Fig. 2B; Cys 87 and Cys 90 ; Cys 99 and Cys 102 ). The zinc ion appears to stabilize the relative positions of ␤-strands ␤3, ␤4, and ␤5, which do not make extensive hydrophobic contacts with one another. Mutation of invariant Cys 155 to Ala in S. cerevisiae (corresponding to Cys 87 in M. jannaschii) causes a severe slow growth phenotype (10), underscoring the importance of zinc ion binding. Interestingly, in mammals three of the four cysteines are mutated to Thr and Leu. Loss of zinc binding during evolution has also been observed among transcription factor zinc-containing motifs (25). In these cases, the integrity of the motif is preserved by main chain hydrogen bonds and van der Waals interactions among the hydrophobic amino acids replacing the cysteines, which may be the case for mammalian eIF2␥s.
Hydrophobic interactions and hydrogen bonds contribute to the positioning of the zinc ribbon with respect to the main body of the G domain (Fig. 2C). Conserved Tyr 93 is sandwiched between the aliphatic side chains of conserved Lys 86 and Lys 219 , with its backbone amide hydrogen bonded to the side chain of conserved Asp 215 . Mutation of Tyr 161 in S. cerevisiae eIF2␥ to Ala (corresponding to Tyr 93 in M. jannaschii eIF2␥) causes a slow growth phenotype (data not shown), underscoring the importance of preserving the integrity of the interface between the zinc ribbon motif and the body of the G domain.
In all GTPases for which structures of the GDP and GTPbound enzymes are available, significant conformational changes are restricted to two polypeptide chain segments denoted Switch 1 and 2 (Fig. 1). Switch 1 is involved in interactions with effectors (in the case of eIF2, these could be Met-tRNA i Met , the ribosome and/or eIF5). Switch 1 lies at the interface between the G domain and domain II and interacts with residues from ␤-strand ␤17 stabilizing the relative orientation of the two domains. Part of Switch 1 (residues 63-71) is disordered in our structure. Switch 2 lies at the heart of the structure where it makes contacts with domains II and III.  Fig. 3, A and B illustrate the structures of EF1A⅐GTP (29) and EF1A⅐GDP (30). Residues Gly 44 -Gly 59 of eIF2␥, forming the P loop and part of ␣-helix ␣1, are structurally similar to the corresponding regions of EF1A (pairwise ␣-carbon r.m.s.d. ϭ 0.6 -0.8 Å; Fig. 3D). In both cases, structural superpositions of these regions result in close overlap of the remaining nucleotide binding segments, with the exception of the switch regions. Fig. 3C also demonstrates that the spatial arrangement of the three domains in apo eIF2␥ is similar to that seen in EF1A⅐GTP and not EF1A⅐GDP. In EF1A⅐GDP domain II is disengaged from the G domain, while in EF1A⅐GTP domain II is closely packed against the Switch 1 region of the G domain (Fig. 3, A and B).

Similar Domain Arrangements in Apo eIF2␥ and EF1A⅐ GTP-
The orientation of the Switch 2 helix ␣2 of apo eIF2␥ is identical to that seen in EF1A⅐GDP (Fig. 3D). When EF1A binds GTP (Fig. 3, A and B) the C terminus of ␣2 moves 7.5 Å and rotates 45°, causing domain II to rotate 90°and pack against the G domain. Thus domain coupling differs between eIF2␥ and EF1A, despite high sequence conservation (22% identity; 42% similarity).
The closed domain configuration of apo eIF2␥ was initially seen in the structure of Pyrococcus abyssi (P. abyssi) eIF2␥ (9) where it could have resulted from lattice packing effects (9). Our structure of another archaeal eIF2␥ in a different crystal form confirms that apo eIF2␥ adopts the closed domain configuration. We presume, therefore, that the conformational changes undergone by eIF2␥ on nucleotide binding and hydrolysis differ from those seen for EF1A (Fig. 3, A-D). Domains II and III of the M. jannaschii and P. abyssi eIF2␥ can be superimposed with an r.m.s.d. of 1.2 Å (Fig. 3E). Following this superposition the G domain of M. jannaschii eIF2␥ is rotated by 14°relative to its counterpart in P. abyssi, revealing a hinge region near the Switch 2 helix ␣2 around which domains II and III move as a rigid body.
Differences in the orientation of the G domain with respect to domains II and III have also been observed among EF1A⅐GDP structures (31), indicating that this conformational plasticity is a common feature for these two GTPases. Interestingly, the structure of a mutant form of P. abyssi eIF2␥ showed no domain rearrangements upon GTP binding (9), while its close homolog EF1A shows large domain movements (29). This finding is surprising since it fails to explain the nucleotide dependence of the tRNA binding activity of this factor. One explanation put forth in the report describing the P. abyssi eIF2␥ structures is that the other subunits of the complex are necessary for this activity (9). However, we have found that the mutation introduced in P. abyssi eIF2␥ to obtain the structure of the GTP bound complex is lethal in S. cerevisiae (data not shown).

Molecular Modeling and Functional Identification of the Met-tRNA i
Met Binding Site on eIF2␥ -Our structure of M. jannaschii eIF2␥ provides a rational basis for directed studies of ternary complex formation. Fig. 4, B and D illustrate the solvent-accessible surface of eIF2␥ color coded for amino acid conservation. The green portion of the molecular surface corresponds to surface residues that are invariant among the sequences of known eIF2␥s (Fig. 1). The ventral surface of eIF2␥ (Fig. 4B) shows a remarkably high level of conservation and is almost certainly responsible for interactions with conserved components of the translation machinery.
The x-ray structure of EF1A in complex with GTP and Phe-tRNA Phe (26) revealed that domain II of EF1A contributes to recognition of the 3Ј-end of the tRNA acceptor helix. Given the marked similarity between domains II of EF1A and eIF2␥ (Fig.  3C) it is reasonable to suggest that eIF2␥ interaction with the tRNA 3Ј-end resembles that observed in the EF1A ternary complex. Superposition of domains II of eIF2␥ and EF1A in the ternary complex places the 3Ј-end of the tRNA in a binding pocket between the G domain and domain II of eIF2␥, and results in few predicted steric clashes, the most notable being between Phe of the Phe-tRNA Phe of the EF1A ternary complex and Leu 317 of eIF2␥. The predicted tRNA binding surface of eIF2␥ (Fig. 5A) corresponds to a positively charged/hydrophobic surface (Fig. 4A) that could support phosphate and/or base recognition. This model enables us to have a clearer understanding of the phenotypes of previously identified mutants and identify new residues important for tRNA recognition.
Deletion of the chromosomal GCD11 gene encoding eIF2␥ in yeast is lethal; however, cell viability can be maintained by a plasmid-borne GCD11 allele. As described under "Experimental Procedures," we generated the gcd11⌬ strain J212 in which FIG. 4. Surface properties of M. jannaschii eIF2␥. GRASP (35) representations of the solvent-accessible surface of eIF2␥ calculated using a water probe radius of 1.4 Å. Surface electrostatic potential is colored red and blue, denoting ϽϪ13 and Ͼϩ12k B T, respectively, where k B is the Boltzmann constant and T is the temperature. The calculations were performed with an ionic strength of 0 mM NaCl and dielectric constants of 80 and 2 for solvent and protein, respectively (36). Ventral (A) and dorsal (C) surfaces of eIF2␥, color-coded for electrostatic potential. Ventral (B) and dorsal (D) surfaces of eIF2␥, color-coded for sequence similarity as in Fig. 1. the essential eIF2␥ function is provided from a GCD11 gene on a plasmid also containing the URA3 gene. To test the impact of eIF2␥ mutations on yeast cell growth, we engineered desired mutations into a GCD11 allele on a LEU2 plasmid and introduced these mutant alleles into the strain J212. Following plasmid-shuffling, as described under "Experimental Procedures," the GCD11 mutant allele on the LEU2 plasmid is the sole source of eIF2␥. Mutations that impair eIF2␥ function will be identified as causing slow growth or lethal phenotypes following plasmid shuffling.
In our model of Phe-tRNA Phe bound to M. jannaschii eIF2␥ (Fig. 5, A and B), the terminal adenosine and Phe of Phe-tRNA Phe project into a pocket on the surface of domain II, with the amino acid stacking against invariant Tyr 79 from Switch 1 (this residue is Tyr 142 in yeast eIF2␥, and His-in E. coli EF1A). Consistent with this model, Erickson and Hannig (8) previously showed that substitution of His for Tyr 142 in yeast eIF2␥ caused a slow growth phenotype that could be suppressed by overexpression of tRNA i Met , and that purified yeast eIF2 containing the mutant eIF2␥ -Y142H subunit showed decreased binding to Met-tRNA i . Invariant Gly 319 in M. jannaschii eIF2␥, corresponding to Gly 397 in yeast eIF2␥, lies at the bottom of this putative tRNA binding pocket (Fig. 5B). Mutation to a larger residue would be predicted to result in a steric clash with the ribose of A76 of the tRNA. Fortuitously, the yeast eIF2␥ Gly 397 to Ala mutation was isolated previously as a mutant that derepresses GCN4 mRNA translation (gcd11-507 mutation) (32). The eIF2␥-G397A mutation causes a modest slow growth phenotype (32), and here we show that this slow growth phenotype can be fully suppressed by overexpression of Met-  Fig. 4B. tRNA (tRNA Phe from EF1A ternary complex) is shown with a ball and stick representation (carbon, white; oxygen, red; nitrogen, blue; and phosphorous, yellow). B, the surface proposed to be involved in recognition of the 3Ј-end of Met-tRNA i Met (colored as in Fig. 5A). Mutated residues that impair tRNA binding are labeled in magenta. Other conserved residues on this surface are labeled in white. C, yeast expressing the individual eIF2␥ mutants in the presence or absence of overexpressed eIF2␣ or Met-tRNA i Met , as indicated, were streaked for single colonies on minimal medium and incubated at 30°C. Relative growth rates were scored by comparing colony sizes after 3-6 days with 5ϩ indicating wild-type growth (large colonies at 3 days) and 1ϩ indicating only small colonies after 6 days; ND, not determined. The column labeled eIF2␣ binding summarizes the results of pull-down assays, shown in Fig. 6A, examining the ability of M. jannaschii GST-eIF2␥ fusion proteins to bind eIF2␣. D, suppression of mutant eIF2␥ growth phenotypes in yeast by overexpression of Met-tRNA i Met or eIF2␣. Upper plate, the gcd11-507 (eIF2␥ -G397A) mutant yeast strain J208, or its isogenic wild-type parent J210, was transformed with the high copy number plasmid YEp24 (vector), or YEp24 carrying the SUI2 gene encoding yeast eIF2␣ (H.C. eIF2␣) or the IMT4 gene encoding tRNA i Met (H.C. Met-tRNA i Met ). The indicated strains were streaked on YPD rich medium and incubated at 30°C for 3 days. Middle and lower plates, derivatives of the gcd11⌬ strain J212 expressing eIF2␥, eIF2␥ -D403A, or eIF2␥ -V402A,D403A,L406A, as indicated, from a low copy number LEU2 plasmid were transformed with the high copy number HIS3 plasmid pRS423 (vector), or pRS423 carrying the SUI2 gene (H.C. eIF2␣) or IMT4 gene (H.C. Met-tRNA i Met ). The indicated transformants were streaked on S.D. minimal medium and incubated at 30°C for 3 days. tRNA i , but not eIF2␣ (Fig. 5, C and D, upper panel), consistent with the idea that the G397A mutation decreases Met-tRNA i Met binding affinity. Furthermore, we found that mutation of Gly 397 in eIF2␥ to the bulkier Ile is lethal in yeast (Fig. 5C). Finally, based on the modeled tRNA binding pocket (Fig. 5B), we propose that the conserved Arg 242 of M. jannaschii eIF2␥ (corresponding to S. cerevisiae Arg 319 ) contributes to tRNA binding through phosphate neutralization. Consistent with this hypothesis, we found that mutation of Arg 319 in yeast eIF2␥ to the negatively charged Asp is also lethal (Fig. 5C) Asp 325 ) to Ala resulted in a pronounced slow growth phenotype that was suppressed by overexpression of eIF2␣, but not Met-tRNA i , consistent with defects in eIF2␣ binding (Fig. 5, C and  D, middle panel).
To directly assess the impact of eIF2␥ mutations on eIF2␣ binding, we expressed in E. coli and purified wild-type and mutant forms of M. jannaschii eIF2␥ fused to GST, and then tested the fusion proteins for the ability to pull-down bacterially expressed and purified M. jannaschii eIF2␣. Consistent with the yeast mutants results, M. jannaschii eIF2␥-D325A and -L256D mutants exhibited dramatic reductions in eIF2␣ binding ( Fig. 6A; lanes 8, 12 versus 6, 10). To further define the eIF2␣ binding site on eIF2␥, we generated the M. jannaschii eIF2␥-N247A/K248A/G250A triple mutant (see Fig. 6B). This triple mutation, which altered residues between the mapped eIF2␣ and Met-tRNA i Met binding sites, only slightly impaired eIF2␣ binding (data not shown), indicating that the eIF2␣ binding site is localized to the side of eIF2␥.
The locations of mutations that affected eIF2␣ binding are illustrated in Fig. 6B (labeled in magenta). Despite the close proximity of the tRNA and eIF2␣ binding sites on eIF2␥, mutation of residues in the putative tRNA binding pocket of M. jannaschii eIF2␥ (Gly319 3 Ile and Arg242 3 Asp) did not affect binding to eIF2␣ (Fig 6B; lanes 14 and 16 versus 10). However, it is possible that eIF2␣ enhances Met-tRNA i Met binding, as yeast eIF2 complexes lacking eIF2␣ show lowered Met-tRNA i Met binding (33), and the lethal phenotype associated with loss of eIF2␣ in yeast is partially suppressed by overproduction of Met-tRNA i Met , eIF2␤, and eIF2␥ (34). Further supporting the idea that eIF2␣ contributes to Met-tRNA i Met binding by eIF2␥, we found that overexpression of either eIF2␣ or Met-tRNA i Met partially suppressed the slow growth phenotype of a yeast strain expressing the eIF2␥-V402A/D403A/L406A triple mutant (Fig. 5, C and D).  6. eIF2␣ binding. A, GST pulldown assays to assess binding of M. jannaschii eIF2␥ to M. jannaschii eIF2␣. The GST and GST-eIF2␥ fusion proteins (wild-type and mutants) expressed in E. coli were purified on glutathione-Sepharose beads and incubated with eIF2␣. Following binding and washing, supernatant fractions (S) and pellet fractions (P) were resolved by SDS-4 to 20% PAGE and visualized by Coomassie staining. The input lane (I) contains 3.3 g of eIF2␣. B, the surface required for eIF2␣ binding (colored as in Fig. 5A). Mutated residues that impair eIF2␣ binding are labeled in magenta. Other conserved residues on this surface are labeled in white.
binding sites are in close proximity on the eIF2␥ surface, and they provide additional suggestive evidence that eIF2␣ contributes to Met-tRNA i Met binding to eIF2. In conclusion, we have determined the crystal structure of an archaeal eIF2␥ that serves as the foundation of the eIF2 ternary complex on which, the ␣and ␤-subunits and Met-tRNA i Met assemble. A combination of in vivo and in vitro studies has allowed us to locate conserved, adjacent binding surfaces for eIF2␣ and tRNA. Our work provides a starting point for further systematic biochemical, genetic and structural studies, aimed at understanding interactions between the eIF2 subunits, tRNA, mRNA, and the small (40 S) ribosomal subunit.