Specific Interaction of Eukaryotic Translation Initiation Factor 5 (eIF5) with the β-Subunit of eIF2*

Eukaryotic translation initiation factor 5 (eIF5) interacts with the 40 S initiation complex (40 S·mRNA· eIF3·Met-tRNAf·eIF2·GTP) and mediates hydrolysis of the bound GTP. To characterize the molecular interactions involved in eIF5 function, we have used 32P-labeled recombinant rat eIF5 as a probe in filter overlay assay to identify eIF5-interacting proteins in crude initiation factor preparations. We observed that eIF5 specifically interacted with the β subunit of initiation factor eIF2. No other initiation factors including the γ subunit of eIF2 tested positive in this assay. Furthermore, both yeast and mammalian eIF5 bind to the β subunit of either mammalian or yeast eIF2. Binding analysis with human eIF2β deletion mutants expressed inEscherichia coli identified a 22-amino acid domain, between amino acids 68 and 89, as the primary eIF5-binding region of eIF2β. These results along with our earlier observations that (a) eIF5 neither binds nor hydrolyzes free GTP or GTP bound as Met-tRNAf·eIF2·GTP ternary complex, and (b) eIF5 forms a specific complex with eIF2 suggests that the specific interaction between eIF5 and the β subunit of eIF2 may be critical for the hydrolysis of GTP during translation initiation.

Eukaryotic translation initiation factor 5 (eIF5) interacts with the 40 S initiation complex (40 S⅐mRNA⅐ eIF3⅐Met-tRNA f ⅐eIF2⅐GTP) and mediates hydrolysis of the bound GTP. To characterize the molecular interactions involved in eIF5 function, we have used 32 P-labeled recombinant rat eIF5 as a probe in filter overlay assay to identify eIF5-interacting proteins in crude initiation factor preparations. We observed that eIF5 specifically interacted with the ␤ subunit of initiation factor eIF2. No other initiation factors including the ␥ subunit of eIF2 tested positive in this assay. Furthermore, both yeast and mammalian eIF5 bind to the ␤ subunit of either mammalian or yeast eIF2. Binding analysis with human eIF2␤ deletion mutants expressed in Escherichia coli identified a 22-amino acid domain, between amino acids 68 and 89, as the primary eIF5binding region of eIF2␤. These results along with our earlier observations that (a) eIF5 neither binds nor hydrolyzes free GTP or GTP bound as Met-tRNA f ⅐eIF2⅐GTP ternary complex, and (b) eIF5 forms a specific complex with eIF2 suggests that the specific interaction between eIF5 and the ␤ subunit of eIF2 may be critical for the hydrolysis of GTP during translation initiation.
Initiation of translation in eukaryotic cells occurs by a sequence of partial reactions requiring the participation of a large number of specific proteins called eukaryotic (translation) initiation factors (eIFs). 1 An obligatory intermediate step in this overall initiation reaction is the binding of the initiator methionyl-tRNA (Met-tRNA f ) as Met-tRNA f ⅐eIF2⅐GTP ternary complex to a 40 S ribosomal subunit, followed by positioning of the 40 S preinitiation complex (40 S⅐Met-tRNA f ⅐eIF2⅐GTP) at the initiation AUG codon of the mRNA to form the 40 S initiation complex (40 S⅐mRNA⅐Met-tRNA f ⅐eIF2⅐GTP). A 60 S subunit then joins the 40 S initiation complex to form the 80 S initiation complex (80 S⅐mRNA⅐Met-tRNA f ) that is active in peptidyl transfer (for a review, see Refs. [1][2][3]. The subunit joining reaction specifically requires the participation of eIF5, an initiation factor that we have purified and characterized from mammalian cells (4 -6) and the yeast Saccharomyces cerevisiae (7). Detailed characterization of the eIF5-catalyzed reaction has shown that eIF5 first interacts with the 40 S initiation complex in the absence of 60 S ribosomal subunits to promote the hydrolysis of ribosome-bound GTP (8). Hydrolysis of GTP causes the release of eIF2 and guanine nucleotide (as an eIF2⅐GDP complex) from the 40 S subunit, an event that is essential for the subsequent joining of the 60 S ribosomal subunit to the 40 S complex (40 S⅐mRNA⅐Met-tRNA f ) to form the 80 S initiation complex that is active in subsequent peptidyl transfer reaction (8 -11).
The mammalian cDNA (rat and human) and the S. cerevisiae gene encoding eIF5 of calculated M r ϭ 48,926 and 45,346, respectively, have been cloned and expressed in Escherichia coli (12)(13)(14)(15)). An interesting feature of the derived amino acid sequences of mammalian (rat and human) and yeast eIF5 proteins is the presence of sequence motifs that have weak homology to characteristic domains present in proteins of the GTPase superfamily (16) which includes the ␣ subunit of Gproteins, H-Ras and Rab 3A, yeast proteins, CDC42 and SEC4, as well as translation initiation and elongation factors, IF2 and EF1A (formerly called EFTu), EF2 (formerly called EFG), and EF1␣. However, unlike these proteins, eIF5 neither binds nor hydrolyzes free GTP or GTP bound to eIF2 as a Met-tRNA f ⅐eIF2⅐GTP ternary complex (8). eIF5 mediates hydrolysis of GTP only when the nucleotide is bound to eIF2 on the 40 S initiation complex containing bound Met-tRNA f ⅐eIF2⅐GTP ternary complex (8). The key question therefore is whether interaction of eIF5 with one or more components of the 40 S initiation complex is required for GTP hydrolysis during translation initiation.
In the work presented in this paper, we have used 32 Plabeled mammalian eIF5 as a probe in filter overlay ("far Western") assays to demonstrate that eIF5 specifically and directly interacts with the ␤ subunit of eIF2. Similar interaction between yeast eIF5 and yeast eIF2 ␤ was also observed. By systematic deletion analysis, the region of the ␤ subunit of eIF2 involved in binding eIF5 was characterized. The implications of this specific interaction between eIF5 and the ␤ subunit of eIF2 in the eIF5-mediated hydrolysis of GTP during translation initiation are discussed.

EXPERIMENTAL PROCEDURES
Purified Proteins and Immunological Methods-Purified eIF2 and eIF5 from rabbit reticulocyte lysates as well as recombinant rat eIF5 were isolated as described (6,13,17). Yeast eIF5 and yeast eIF2 were purified from S. cerevisiae strain, BJ926 as described by Chakravarti et al. (7). IgG antibodies specific for each of the subunits of rabbit eIF2 were isolated from specific rabbit antisera raised against purified rabbit eIF2 (17) by affinity purification using each of the subunits of eIF2 blotted onto aminophenyl thioether paper (Schleicher and Schuell) as an antigen as described (18). Antibodies specific for the ␤ subunit of yeast eIF2 were a kind gift of Dr. Thomas Donahue of Indiana University. Immunoblot analysis was carried out as described (18). Anti-GST antibodies made in mouse were a kind gift from Charles Weaver of our institution.
Construction of Plasmids-The human cDNA corresponding to the ␥ * This research was supported by Grant GM15399 from the National Institutes of Health and by Cancer Core Support Grant P30 CA1330 from the National Cancer Institute. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. This paper is dedicated to the loving memory of Professor Bimal K. Bacchawat of India. He was an outstanding scientist and an extraordinary human being who made enormous contributions to the development of biochemical sciences in India. 1 The abbreviations used are: eIF, eukaryotic translation initiation subunit of eIF2 (eIF2␥-cDNA) was cloned by immunoscreening a HeLa cell cDNA expression library in phage ZAPII (Stratagene) using affinity purified polyclonal anti-eIF2␥ antibodies as probe. The 1.511-kilobase pair cDNA insert present in the recombinant phage was isolated by in vivo excision as a subclone in the plasmid pBlueScript-SK(ϩ) (19). This cDNA insert contained nucleotide sequences that corresponded to the complete open reading frame of the ␥ subunit of eIF2 (nucleotide ϩ 1 to ϩ1417) as well as a 12-nucleotide long 5Ј-untranslated region and part of the 3Ј-untranslated region (19). The amino acid sequence predicted from the nucleotide sequence of the open reading frame with a calculated M r of 51,167 matched completely with that reported by Gaspar et al. (20). Using primers bearing the N-terminal and C-terminal ends of eIF2␥-coding sequences that were flanked by NdeI and EcoRI sites, the eIF2␥-open reading frame was PCR-amplified, and the PCR product was digested with NdeI and EcoRI and cloned into the same sites of pRSET-B vector (Invitrogen) to generate the construct pRSET-eIF2␥ (designated pRSET-2␥). For expression of the ␤ subunit of human eIF2, the open reading frame of human eIF2␤ cDNA (21) was synthesized by reverse transcription-polymerase chain reaction of HeLa poly(A) ϩ RNA using oligonucleotide primers corresponding to the N-terminal and C-terminal ends of the eIF2␤-cDNA (21). The N-terminal primer had BamHI-NdeI sites at the 5Ј end, and the C-terminal primer had an EcoRI overhang. The 1002-base pair long PCR product was sequenced to ensure error-free DNA synthesis and cloned individually into the (a) NdeI/EcoRI sites of pET-5a plasmid (Novagen) and (b) BamHI/EcoRI sites of pGEX-2T (Pharmacia Biotech Inc.) to generate the expression plasmids, pET-5a-eIF2␤ (designated pET-2␤) and pGEX-2T-eIF2␤ (designated pGEX-2␤), respectively. Deletion mutants of eIF2␤ were generated by one-stage PCR amplification of eIF2␤ cDNA sequences using pET-2␤ as the template. The appropriate sense strand oligonucleotide primers used for this reaction had BamHI-NdeI overhangs, thereby introducing an in-frame methionine codon at the N-terminal end, and the antisense oligonucleotide primer corresponding to the C-terminal end of eIF2␤ introduced a translation stop codon followed by an EcoRI restriction site. A BamHI/EcoRI restriction fragment of each PCRamplified eIF2␤ deletion mutant was inserted at the same restriction sites of the vector pGEX-2T. The resulting constructs expressed deleted eIF2␤ mutants as GST fusion proteins. The mutant ␤ construct, ␤ (K2 3 A2) in which the polylysine stretch 79 KKKKKKTKK 87 , designated K2, present in wild-type ␤ was substituted with a polyalanine stretch 79 AAAAAATAA 87 , was generated by a three-fragment ligation as follows. The strategy was to generate a NotI site (G2CGGCCGC) at the position of the polylysine stretch, flanked by GC-rich sequences encoding the amino acid residue alanine whose codons are GCC or GCG. Fragment 1 consisted of a PCR product corresponding to the N-terminal region (amino acid residues 1-80) of eIF2␤ in which a NotI site was introduced at the 3Ј end. The sense strand primer had an NdeI site, whereas the antisense primer used was 5Ј-GGCGGCCGCGGCGGCG-GCTTGATTAAAGAAG-3Ј (the underlined sequence represents the NotI site). The PCR product was digested with NdeI and NotI and purified by agarose gel electrophoresis. Fragment 2 consisted of a PCR product corresponding to the remainder of eIF2␤ (amino acid residues 81-333) in which a NotI site was introduced at the 5Ј end of the PCR fragment. The N-terminal primer used for the PCR was 5Ј-GGCGGC-CGCCACTGCCGCCATATTTGATATT-3Ј, and the C-terminal primer had a downstream EcoRI site. This PCR product was digested with NotI and EcoRI and purified by agarose gel electrophoresis. Purified fragments 1 and 2 were then ligated into pET-5a that had been cut with NdeI and EcoRI. Similar three-fragment ligation strategy was used to construct the mutant ␤ fragments ␤ (K1 3 A1) and ␤ (K3 3 A3) in which the polylysine stretches K1 ( 14 KKKKKKKK 21 ) and K3 ( 124 KKKKKK 129 ) present in wild-type ␤ were substituted with polyalanine stretches A1 ( 14 AAAAAAAA 21 ) and A3 ( 124 AAAAAA 129 ), respectively. For ␤ (K1 3 A1), the N-terminal fragment was generated by annealing complementary strands of oligonucleotides corresponding to the N-terminal 17 amino acids of eIF2␤, and the NotI site was generated at the position of K1, whereas for ␤ (K3 3 A3) the N-terminal fragment of 127 amino acids was PCR-amplified and the NotI site was generated at the position of K3. For construction of mutant ␤ (K2 3 A2 and K3 3 A3), where both the polylysine stretches K2 and K3 in eIF2␤ were substituted with polyalanine stretches A2 and A3, the construct pET-5a-eIF2␤ (K3 3 A3) was digested with AflII and EcoRI. The enzyme AflII cuts the recombinant vector at a unique restriction site between the two polylysine stretches K2 and K3 in eIF2␤, whereas EcoRI cuts it at a site immediately following the C terminus of eIF2␤. The AflII-EcoRI fragment was purified by gel electrophoresis and ligated into the gel-purified vector fragment obtained by digesting pET-5a-eIF2␤ (K2 3 A2) with AflII and EcoRI. All constructs were se-quenced using U.S. Biochemical Corp. kit to ensure error-free DNA synthesis.

Expression and Purification of Recombinant Wild-type or Mutant eIF2␤
Proteins-The pET5a-eIF2␤ series of plasmids containing either the wild-type or mutant eIF2␤ coding sequence and pRSET-eIF2␥ plasmid containing the coding sequence of eIF2␥ were transformed individually into E. coli BL21 (DE3) cells, whereas the pGEX-2T-eIF2␤ series of plasmids in which the coding sequence of wild-type or mutant eIF2␤ was fused at its N terminus to GST were transformed into E. coli DH5␣ cells (Life Technologies, Inc.). In each case, a single ampicillin-resistant colony was grown to mid-logarithmic phase in 3 ml of 2YT medium (22), induced with 1 mM IPTG, and grown for an additional 2-3 h. The cells were centrifuged and bacterial pellets suspended directly in 1 ϫ Laemmli buffer and boiled for 4 min at 100°C, and the proteins in the boiled extract were resolved by SDS-polyacrylamide (15%) gel electrophoresis. The GST or GST-eIF2␤ fusion protein was isolated by inducing E. coli DH5␣ cells harboring either the pGEX-2T or pGEX-2T-eIF2␤ plasmids as described above. The harvested cells were suspended (at approximately A 600 ϭ 1.0) in buffer A (10 mM potassium phosphate, pH 7.0, 150 mM NaCl, 5 mM 2-mercaptoethanol, and protease inhibitor mixture containing 0.5 mM phenylmethanesulfonyl fluoride, pepstatin A (0.7 g/ml), leupeptin (0.5 g/ml)) and sonicated 10 times for 20 s each with cooling on ice between bursts. Following addition of Triton X to 1% final concentration, the cell lysate was clarified by centrifugation at 15,000 ϫ g for 10 min. The lysate containing either GST or GST fusion protein was incubated with pre-equilibrated glutathione (GSH)-Sepharose 4B beads (Pharmacia) in buffer A containing 0.1% Nonidet P-40 for 1 h at 4°C and then washed three times with 1 ml of the same buffer. The amount of protein present in the washed beads containing either GST or GST-eIF2␤ fusion protein was measured by Bio-Rad method. The beads were stored in small aliquots at Ϫ70°C until use.
Preparation and Isolation of 32 P-Labeled eIF5-Recombinant rat eIF5 (20 g) was phosphorylated with [␥-32 P]ATP (5,000 cpm/pmol) and casein kinase II in a reaction mixture similar to that described previously for phosphorylation of rabbit reticulocyte eIF5 (8). 32 P-Labeled eIF5 was isolated by subjecting the reaction mixture to phosphocellulose chromatography as described (8). Similar procedures were used to prepare 32 P-labeled yeast eIF5. It should be noted that eIF5 phosphorylated in vitro by casein kinase II retains full in vitro activity (18).
Filter Overlay Method (Far Western Analysis) for Detection of eIF5 Binding Proteins-eIF5 binding proteins were identified by an adaptation of the procedure previously used by Bregman et al. (23) as follows. Protein samples were subjected to electrophoretic separation in 0.1% SDS, 15% polyacrylamide gels and then electrotransferred to a PVDF membrane. The membrane blot was first rinsed with 20 mM Tris-HCl, pH 7.5, and 150 mM NaCl (TBS) and then placed in TBS containing 5% non-fat dry milk (Blotto) and shaken gently for about 1 h. The blot was then incubated with 32 P-labeled eIF5 (ϳ10 6 cpm of radioactivity in 20 ml of fresh Blotto) for about 12 h at 4°C, after which it was washed for 15 min with 20 ml of Blotto containing 0.05% Tween 20. This washing step was repeated once more. Finally, the blot was washed twice with 20 ml of TBS, air-dried, and subjected to autoradiography using Kodak BIOMAX-MR/X-OMAT films at Ϫ70°C to visualize the binding of 32 Plabeled eIF5 to proteins.
Binding of 32 P-Labeled eIF5 to Glutathione-Sepharose Beads Containing Bound GST-eIF2␤ Fusion Protein-A typical binding reaction mixture consisted of 200 l of buffer A, 20 l of a 30% suspension of GSH beads containing about 4 g of bound protein, and 4 l of purified 32 P-labeled eIF5 (50 ng of protein containing about 50,000 cpm of radioactivity). The reaction mixture was gently mixed in a rotator at 4°C for 1 h and then centrifuged to separate the unbound proteins from the bound proteins. The supernatant was removed and discarded. The beads were then washed three times with 1 ml of buffer A containing 0.1% Nonidet P-40 for 20 min and then resuspended in 20 l of 1% SDS-gel loading buffer. Samples were incubated in a boiling water bath for 3 min and separated on 0.1% SDS-15% polyacrylamide gels. The dried gels were subjected to autoradiography.

Identification of eIF5-interacting Proteins in Crude Initiation
Factor Fractions-To identify eIF5-interacting protein(s), partially purified initiation factor preparations obtained from rabbit reticulocyte lysates as well as a crude cell-free HeLa cell extract were subjected to far Western blot analysis using 32 Plabeled eIF5 as probe (as explained under "Experimental Procedures"). Fig. 1, panel A, shows that in partially purified initiation factor preparations obtained from rabbit reticulocyte lysate (lanes a and b), as well as in HeLa cell extract (lane c), a single major polypeptide band that migrated with an apparent M r of about 54,000 specifically interacted with 32 P-labeled eIF5. When similar far Western blot analysis was carried out with purified initiation factors, 32 P-labeled eIF5 probe detected the 54-kDa polypeptide only in purified eIF2 and in eIF2⅐eIF2B preparations (Fig. 1, panel B). Other initiation factors, e.g. eIF3, eIF1, eIF1A, eIF4F, eIF4A, and eIF6, did not test positive in this interaction assay (only the data for the multi-subunit protein eIF3 is shown in Fig. 1, panel B).
To identify the subunit of eIF2 that specifically interacted with eIF5, we carried out immunoblot analysis of purified eIF2 and eIF2⅐eIF2B preparations using total anti-eIF2 antibodies as probes (Fig. 1, panel C). Comparison of the relative mobilities of the three immunoreactive polypeptides of eIF2 (Fig. 1, panel C) with that of the eIF5-interacting polypeptide in purified eIF2 and eIF2⅐eIF2B preparations (panel B) suggested that the eIF5-interacting 54-kDa polypeptide may be the ␤ subunit of eIF2. It should be noted that in this gel system, while the ␥ subunit of eIF2 of calculated M r ϭ 51,800 (19,20) migrated true to its molecular weight, the ␤ subunit of eIF2 which has a calculated M r ϭ 38,400 (21) migrated with an apparent M r of about 54,000 (17).
Confirmation that the eIF5-interacting protein was indeed the ␤ subunit of eIF2 was derived from the following observations. First, the open reading frame of human eIF2 ␤-cDNA (21) was subcloned into the E. coli expression plasmid, pET-5a under the control of T7 RNA polymerase promoter. When E. coli BL21 (DE3) cells, harboring the pET5a-eIF2␤ (pET-2␤) expression plasmid induced with IPTG and proteins in induced cell lysates, were subjected to Western blot analysis using anti-eIF2␤ antibodies as probes, the synthesis of eIF2␤ was easily observed (Fig. 2, panel A, lane c). The same polypeptide interacted with 32 P-labeled eIF5 when these induced cell lysates were subjected to far Western blot analysis (Fig. 2, panel  B, lane c) which comigrated with the same mobility as the eIF5-reacting band of eIF2 (Fig. 2, panel B, lane a). In contrast, no eIF5-interacting polypeptide was detected in E. coli cell lysates that did not synthesize the ␤ subunit of eIF2 (Fig. 2,  compare lane b of panel A with lane b of panel B). It should be noted here that bacterially expressed eIF2␤ undergoes proteolytic degradation. The lower molecular weight eIF5-reacting band in panel B (lane c) is very likely an N-terminal fragment of eIF2␤ which is weakly detected in immunoblot analysis (panel A, lane c) as the anti-eIF2 antibodies recognize the N terminus of eIF2␤ very poorly. Additionally, when lysates of E. coli cells overexpressing the ␥ subunit of eIF2 were subjected to far Western blot analysis, 32 P-labeled eIF5 probe did not detect any eIF5-interacting polypeptide band (Fig. 2, compare lane d   FIG. 1. Identification of eIF5-binding proteins in cell-free extracts. Protein fractions were denatured and electrophoresed in a SDS-15% gel, and the resolved proteins were transferred to PVDF membranes as indicated under "Experimental Procedures." The membrane blots were probed with 32 P-labeled eIF5 (ϳ10 6  were boiled in 1 ϫ Laemmli buffer and then separated by SDS-15% polyacrylamide gel electrophoresis followed by transfer to PVDF membranes. In lane a, 4 g of purified rabbit reticulocyte eIF2 was analyzed. In panel A, the membrane blot was subjected to immunoblot analysis using polyclonal anti-eIF2 antibodies as probes. The positions of the ␣, ␤, and ␥ subunits of eIF2 are indicated. A nonspecific immunoreactive band above the position of the ␤ subunit is present in all the lanes. Lower molecular weight immunoreactive polypeptides are likely to be degradation products of bacterially expressed eIF2␤ and eIF2␥. In panel B, the blot was probed with 32 P-labeled eIF5 (10 6 cpm of radioactivity in 20 ml of blotto buffer) in far Western blot analysis as described under "Experimental Procedures." The dried blot was then analyzed by autoradiography. The position of the ␤ subunit, which migrated anomalously with an apparent M r ϭ 54,000 in this gel system, is indicated. Panel C, 32 P-labeled recombinant eIF5 (containing 40,000 cpm of radioactivity) was separately incubated with 4 g of GST (lane b) and with 4 g of GST-eIF2␤ fusion protein (lane c) both immobilized on glutathione-Sepharose beads. Following incubation at 4°C by gentle shaking, reaction mixtures were centrifuged, and the beads were washed as described under "Experimental Procedures," suspended in 1 ϫ Laemmli buffer, and subjected to SDS-15% polyacrylamide gel electrophoresis. In lane a, 32 P-labeled eIF5 alone was electrophoresed. The dried gel was subjected to autoradiography.

(panel A) with lane d (panel B))
. These results demonstrate that eIF5 specifically interacted with the ␤ subunit of eIF2.
The interaction of eIF5 with the non-denatured form of eIF2␤ was explored by cloning the coding region of eIF2␤-cDNA into the pGEX-2T expression vector so as to express eIF2␤ as a GST-eIF2␤ fusion protein in E. coli DH5␣ cells as described under "Experimental Procedures." The GST-eIF2␤ fusion protein was immobilized on glutathione-Sepharose beads and tested for its ability to bind 32 P-labeled eIF5. Fig. 2, panel C, shows that eIF5 was specifically retained on beads coupled to GST-eIF2␤ but was not bound to GSH-Sepharose beads linked to GST alone. Thus, eIF5 forms a specific complex even with the non-denatured form of ␤ subunit of eIF2.
Interaction of Yeast eIF5 with the ␤ Subunit of Yeast eIF2-It is now well established that the basic mechanism of translation initiation is highly conserved from the unicellular yeast to mammals. If the interaction between eIF5 and the ␤ subunit of eIF2 plays an essential role in the function of eIF5 in mammalian translation initiation, such an interaction should also occur in the yeast system. To determine whether yeast eIF5 and yeast eIF2␤ subunit interact with each other, we performed far Western blot analysis of a purified yeast eIF2 preparation using 32 P-labeled yeast eIF5 as probe (see "Experimental Procedures"). Fig. 3, panel A, (lane c), shows that 32 P-labeled eIF5 interacted with a single polypeptide band in the purified yeast eIF2 preparation. This eIF5-interacting polypeptide migrated with the same electrophoretic mobility as the ␤ subunit of yeast eIF2 as determined by Western blot analysis of purified yeast eIF2 using anti-yeast eIF2␤ antibodies as probe (Fig. 3, panel  A, lane b). The position of the ␣ (M r ϭ 34,700), ␤ (M r ϭ 31,600), and ␥ (M r ϭ 57,900) subunits of yeast eIF2 are shown in Fig. 3,  panel A (lane a). Furthermore, in agreement with the functional homology between mammalian and yeast eIF5 in vivo (24) and in vitro (15,24), we observed that 32 P-labeled rat eIF5 also bound yeast eIF2␤ (Fig. 3, panel B) and 32 P-labeled yeast eIF5 bound mammalian eIF2␤ (Fig. 3, panel C). These results suggest that the interaction domains of eIF5 and the ␤ subunit of eIF2 are conserved through evolution.
Interaction of Rat eIF5 with the Deletion Mutants of Human eIF2␤-To map the region of human eIF2␤ involved in the binding of eIF5, appropriate PCR primers were used to amplify shorter fragments of the human eIF2␤ coding sequence. The wild-type eIF2␤ coding region as well as the PCR-amplified eIF2␤ cDNA fragments were cloned into the inducible tac promoter-based pGEX-2T expression vector. Since anti-eIF2␤ antibodies do not recognize the N-terminal region of eIF2␤ (presumably because the polyclonal antibodies were primarily directed against the C-terminal region of eIF2␤), we tagged truncated eIF2␤ fragments to GST to enable easy detection in immunoblots with anti-GST antibodies as probe. The pGEX-2T expression plasmids were introduced into E. coli DH5␣ cells, and the cell extracts of IPTG-induced bacteria were then monitored for expression of GST-fused wild-type and truncated eIF2␤ by Western blot analysis using anti-GST antibodies as probe (Fig. 4, panel B). Initially, the effects of the N-terminal truncations of eIF2␤ on its ability to interact with eIF5 were examined. Fig. 4, panel B, shows the expression pattern of wild-type eIF2␤ as well as the two N-terminal truncated eIF2␤ polypeptides, GST-␤-(27-333) and GST-␤-(154 -333). All three fusion proteins were partially purified through glutathione-Sepharose columns, and during the purification process these proteins underwent extensive degradation. Deletion mutants of eIF2␤ showed anomalous migration similar to the wild-type protein. When these expressed polypeptides were examined for their ability to bind 32 P-labeled eIF5 by far Western blot analysis, the binding ability of the mutant GST-␤-(27-333) was found to be comparable to that of the wild-type GST-eIF2␤ (Fig.  4, panel C). In contrast, when amino acids 1-153 were deleted from eIF2␤, GST-␤-(154 -333) failed to interact with eIF5. These results suggested that the region between amino acids 27 and 154 of eIF2␤ is involved in the binding of eIF5. As expected, GST alone did not bind 32 P-labeled eIF5 (Fig. 4,  panel C).
To delimit further the eIF5-interacting region, we carried out a combination of C-terminal and N-terminal deletions of eIF2␤ as shown in Fig. 4, panel A. When E. coli cell lysates containing equivalent amounts of truncated eIF2␤ polypeptides fused to GST (Fig. 4, panel B) were examined for their ability to bind 32 P-labeled eIF5 (Fig. 4, panel C) the following results were obtained. Deletion of 163 amino acids from the C terminus of eIF2␤ did not affect the ability of the resultant mutant GST-␤-(1-170) fusion protein to bind 32 P-labeled eIF5. Further deletion of amino acids 1-26 from ␤-(1-170) did not affect the binding ability of the fusion protein ␤-  to 32 P-labeled eIF5. However, in the context of ␤-(1-170), when amino acids 1-87 were deleted, the resulting GST-␤-(88 -170) fusion protein interacted very weakly with eIF5 compared with GST-␤- (1-170). The GST-␤-(27-123) fusion protein was also recognized by eIF5 quite efficiently although slightly less than ␤- , whereas the GST-␤-(88 -123) fusion protein did not interact with eIF5. Taken together, these results suggested that the region of eIF2␤ encompassing amino acids 27-87 is sufficient for conferring major eIF5-binding properties to eIF2␤. This was confirmed by showing that GST-␤-(27-89) bound 32 P-labeled eIF5 very efficiently. The eIF5-binding region of eIF2␤ was further delimited by showing that the deletion mutant GST-␤-(68 -95) bound 32 P-labeled eIF5 quite efficiently. Thus, the eIF5-binding region of eIF2␤ is a 22-amino acid stretch that lies between amino acids 68 and 89, whereas the region between amino acids 124 and 170 also had low eIF5 binding activity.
Mutational Analysis of the Conserved Polylysine Stretches in the eIF5-binding Region of eIF2␤-Comparison of the amino acid sequence of human eIF2␤ with those from other species showed the presence of three highly conserved stretches of six to eight lysine residues, designated K1 (amino acids 14 -21), K2 (amino acids 79 -87), and K3 (amino acids 124 -129) in the N-terminal region of eIF2␤ (Fig. 5). The minimal eIF5-binding Unlike mammalian eIF2 where the ␤ subunit migrates slower than the ␥ subunit, the three subunits of yeast eIF2 migrate true to their molecular weights. The polypeptide above the ␥ subunit is a contaminant in the eIF2 preparation. In lanes b and c, the resolved subunits were transferred to a PVDF membrane and then analyzed by immunoblot analysis using monospecific anti-yeast eIF2␤ antibodies as probe (lane b) or by far Western blot analysis using 32 P-labeled purified yeast eIF5 as probe (lane c). Panel B, purified yeast eIF2 (2 g) was subjected to far Western blot analysis using 32 P-labeled recombinant rat eIF5 as a probe. Panel C, purified rabbit reticulocyte eIF2 (4 g) was analyzed by far Western blot using 32 P-labeled purified yeast eIF5 as a probe. region (amino acids 68 -89) of eIF2␤ contains the conserved polylysine stretch K2, whereas the region comprising amino acids 124 -170 which exhibits low eIF5 binding activity contains the conserved polylysine stretch K3 (Fig. 5). To investigate the role of these lysine-rich domains of eIF2␤ in eIF5 binding, we substituted each polylysine stretch with a polyalanine stretch of the same length as described under "Experimental Procedures." Fig. 6, left panel, shows that all the mutant proteins were expressed in comparable amounts. Substitution of the polylysine stretch K2 of eIF2␤ with a polyalanine stretch A2 in the mutant ␤ (K2 3 A2) reduced by about 4-fold (from that for wild-type eIF2␤) the binding affinity of 32 P-labeled eIF5 (Fig. 6, right panel, compare lane a with lane c) as determined by densitometric scanning (not shown). In contrast, the mutants ␤ (K1 3 A1) and ␤ (K3 3 A3) in which the polylysine stretches K1 and K3 have been replaced by polyalanine stretches A1 and A3, respectively, showed no significant change in their ability to bind 32 P-labeled eIF5 (Fig. 6, right  panel, lanes b and d). Furthermore, when both the polylysine stretches K2 and K3 were replaced with the polyalanine stretches A2 and A3, the resulting ␤ mutant ␤ (K2 3 A2, K3 3 A3) had similar reduced binding ability to 32 P-labeled eIF5 as ␤ (K2 3 A2) (Fig. 6, right panel, lane e). These results suggest that the polylysine stretch K2, which lies in the minimal eIF5binding region comprising amino acids 68 -89 of eIF2␤, is required for optimal eIF5 binding. DISCUSSION The GTP/GDP cycle involved in eukaryotic translation initiation resembles many aspects of prokaryotic translation initi- FIG. 4. Interaction between eIF2␤ deletion mutants and eIF5. Panel A, schematic representation of deletion mutants of eIF2␤ expressed as GST fusion proteins in E. coli DH5␣ cells. The three conserved polylysine stretches between amino acids 14 and 21, 79 and 87, and 124 and 129 are designated K1, K2, and K3, respectively, and are indicated by a black bar. The C-terminal region that is conserved between different species of eIF2␤ that spans between amino acids 168 and 333 is shown by a shaded bar. The efficiency with which each mutant protein interacts with 32 P-labeled eIF5 in far Western blot analysis is shown by ϩ and Ϫ. Panel B, the expression of GST fusion proteins of deletion mutants of eIF2␤ (as shown in panel A) was measured by lysing IPTG-induced E. coli DH5␣ cells directly in 1 ϫ Laemmli buffer followed by Western blot analysis of each cell lysate using anti-GST antibodies as probe. Panel C, the same cell-free extracts were subjected to far Western blot analysis using 32 P-labeled recombinant rat eIF5 as probe as explained under "Experimental Procedures." FIG. 5. Conservation of eIF5 binding region in eIF2␤ through evolution. The amino acid sequences of the N-terminal region of human, rabbit, and yeast eIF2␤ that contain the eIF5-binding region are aligned for maximum homology using the program DNASTAR. The sequence of the human eIF2␤ is from Ref. 21 and that of the yeast protein is from Ref. 29. The sequence of the rabbit homologue was obtained from SWISSPROT (accession number P41035). The highly conserved residues between eIF2␤ of all three species are highlighted with dark shading, and the moderately conserved residues are highlighted with light shading. The conserved polylysine stretches, K1, K2, and K3 are indicated. Gaps are represented by broken lines. The minimum region of eIF2␤ required for efficient binding of rat eIF5 is indicated (underlined) with a striped line, and the flanking region required for optimal binding is indicated (underlined) with a bold line.
ation (1,2), prokaryotic and eukaryotic translation elongation (1)(2)(3), as well as protein translocation, G-protein-mediated signal transduction, microtubule assembly, Ras and Rho-mediated neoplastic transformation, Rab group of small GTPasedirected vesicular trafficking, and many other regulated processes (16). An important feature of these GTP-binding proteins is that they all possess intrinsic GTPase activity which is often stimulated by their interaction either with an appro-priate acceptor or a separate protein molecule. For example, in both prokaryotic translation initiation and elongation, the GTP-bound form of IF2 and EF1A mediates the transfer of fMet-tRNA and aminoacyl-tRNAs, respectively, to ribosomes with the concomitant hydrolysis of GTP (1,2). However, it has been observed that both IF2 and EF1A have very low intrinsic GTPase activities that are activated when the GTP-bound form of the proteins interact with 50 S ribosomal subunits that act as effectors of GTPase activity (25,26). Likewise, in eukaryotic translation initiation, eIF2⅐GTP is directly involved in transferring Met-tRNA f to a 40 S ribosomal subunit to form the 40 S initiation complex. However, hydrolysis of GTP is not activated by the interaction of the 60 S ribosomal subunit with the 40 S initiation complex. Rather, prior to the joining of the 60 S ribosomal subunit, eIF5 interacts with the 40 S initiation complex in the absence of 60 S ribosomal subunits to promote GTP hydrolysis (8,9). The key question, therefore, concerning the role of eIF5 in translation initiation is the mechanism by which the interaction of this factor with the 40 S initiation complex causes the hydrolysis of bound GTP.
Previous results on the properties of mammalian eIF5 (8) have shown that eIF5, by itself, does not hydrolyze either free GTP or GTP bound to the Met-tRNA f ⅐eIF2⅐GTP ternary complex in the absence of 40 S ribosomal subunit. These results suggest that eIF5 does not directly interact with GTP to catalyze its hydrolysis. Rather, it is likely that the interaction of eIF5 with the 40 S initiation complex causes a conformational change in the 40 S subunit-bound eIF2 which then acts as a GTPase catalyzing the hydrolysis of bound GTP. In this sense, eIF5 acts as a GTPase-activating protein, and both this initiation factor and possibly 40 S ribosomal subunit act as effectors in GTP hydrolysis catalyzed by eIF2 during translation initia-FIG. 6. Effect of substitution of conserved polylysine stretches with polyalanine stretches in eIF2␤ on its eIF5 binding property. Crude extracts of E. coli BL21 (DE3) cells expressing equivalent levels of wild-type and mutant eIF2␤ proteins were boiled in 1 ϫ Laemmli buffer. Equal amounts of sample were loaded onto two separate gels run simultaneously and then transferred to a PVDF membrane. One blot was analyzed by immunoblotting using polyclonal anti-eIF2 antibodies as probes (left panel), and the other blot was subjected to far Western analysis using 32 P-labeled eIF5 as probe (right panel) as described under "Experimental Procedures." FIG. 7. Sequence homology between the C-terminal region of eIF2␤ and the N-terminal region of eIF5. The amino acid sequences of the C-terminal region of human, rabbit, Drosophila, and yeast eIF2␤ are aligned with the N-terminal region of eIF5 from a wide variety of species using the programs PRODOM and DNASTAR as shown. The sequences of human and yeast eIF2␤ are from Refs. 21 and 29, respectively, and those of rabbit and Drosophila are from SWISSPROT (accession numbers P41035, and P41375, respectively). The sequences of human, rat, and yeast eIF5 are from Refs. 12, 14, and 15, respectively, and those of S. pombe and Phavu were obtained from SWISSPROT (accession numbers Q09689 and P41375, respectively). Highly conserved amino acid residues of eIF2␤ and eIF5 are highlighted with dark shading. The amino acids fully conserved between eIF2␤ and eIF5 are indicated by an asterisk (*), and the conserved cysteine residues of C-4 type zinc finger motif in both eIF2␤ and eIF5 are indicated by a filled circle (q). The consensus sequence between the C-terminal portion of eIF2␤ and the N-terminal portion of eIF5 is also shown. Residues moderately conserved to the consensus are highlighted with light shading. The portion of the G1 domain of eIF5 which falls within this conserved region is underlined.
tion. According to this hypothesis, protein-protein interaction between eIF5 and the 40 S subunit bound-eIF2 may be critical for the hydrolysis of GTP bound to the 40 S initiation complex. Such an interaction between eIF2 and eIF5 has been directly demonstrated (13).
Experiments presented in this paper clearly demonstrate that interaction of eIF5 with eIF2 occurs through the ␤ subunit of eIF2. This observation was somewhat unexpected in view of the recent demonstration that the ␥ subunits of both mammalian and yeast eIF2 contain consensus GTP-binding domains (19,20,27), and this subunit of eIF2 is presumably involved in guanine nucleotide binding although this has not yet been experimentally demonstrated. In proteins of the GTPase superfamily, the subunit that binds GTP also possesses the latent GTPase activity that is activated by its interaction with an effector molecule (16). We, therefore, expected eIF5, the effector protein to interact directly with the ␥ subunit of eIF2. It is likely, however, that interaction of eIF5 with the ␤ subunit of eIF2 induces a conformational change in the eIF2 molecule resulting in the activation of the latent GTPase activity of the ␥ subunit which in conjunction with the 40 S ribosomal subunit can then act as a GTPase hydrolyzing the bound GTP. Further work is clearly necessary to understand the mechanism of eIF5-mediated hydrolysis of GTP bound to the 40 S initiation complex.
We have characterized a 22-amino acid region in the human eIF2␤ (amino acids 68 -89) which appears necessary and sufficient for interaction with eIF5 although the region encompassing amino acids 124 -170 also has some weak eIF5 binding activity and may, in fact, be required for optimal binding of eIF5 to eIF2␤. Experiments presented in this paper also demonstrate that the polylysine stretch K2, 79 KKKKKKTKK 87 , present in the primary eIF5-binding region of eIF2␤ is critical for binding eIF5. The polylysine stretch K3, present in the secondary eIF5-binding region was, however, found to be nonessential for binding. In agreement with these results, the polylysine stretch K2 has been found to be conserved in all species of eIF2␤ so far examined, and the polylysine stretch K3 is absent in wheat eIF2␤ (28). Further mutational analysis in the eIF5-binding region of eIF2␤ is clearly necessary to identify the critical amino acid residues involved in the interaction of eIF5 with eIF2␤. It will also be important to carry out systematic deletion and point mutations in eIF5 and use the mutant proteins both for their ability to bind eIF2␤ as well as in mediating the hydrolysis of GTP bound to the 40 S initiation complex. These studies will then allow us to correlate the binding of eIF5 to eIF2 (through the ␤ subunit) with the function of eIF5 in GTP hydrolysis during translation initiation. These studies are currently underway in our laboratory.
Finally, scanning the available sequence data base for eIF2␤ and eIF5 from a wide variety of eukaryotic species revealed a high degree of sequence conservation between the N-terminal region of eIF5 and the C-terminal region of eIF2␤ (Fig. 7). This sequence conservation spanned over a stretch of about 95 amino acid residues. The significance of this sequence conser-vation between the two proteins which have distinct functions in translation initiation is not completely clear. However, it is tempting to speculate that this conserved region in each protein may be involved in binding to a common component of the translation initiation machinery, perhaps the 40 S ribosomal subunit. Additionally, we also observed that eIF5 from different species has a conserved zinc finger motif of the type Cys-X 2 -Cys-X 19 -Cys-X 2 -Cys (Fig. 7). The presence of a similar zinc finger motif in yeast eIF2␤ has been implicated in ribosomal start site selection during the scanning process in which eIF2 plays an important role (29). The significance of the presence of such a sequence motif in both mammalian and yeast eIF5 is currently unknown.