Critical Contacts between the Eukaryotic Initiation Factor 2B (eIF2B) Catalytic Domain and both eIF2β and -2γ Mediate Guanine Nucleotide Exchange^▿,‡

S. cerevisiae strains and genetic methods.

Previously described yeast strains GP3750 (MATa ura3-52::HIS4-lacZ leu2-3-112 ino1 gcn2Δ::hisG gcd6Δ [GCD6 URA3]), GP3889 (MATa ura3-52 leu2-3-112 trp1-Δ63 gcn2Δ pep4::LEU2), GP4115 (MATa ura3-52::HIS4-lacZ leu2-3-112 ino1 gcn2Δ::hisG gcd6Δ [high-copy-number {hc} eIF2 LEU2]) (21), and their derivatives were transformed with plasmids by the lithium acetate method (1). Plasmid shuffling employed 5-fluoroorotic acid (5-FOA) (1). 3-Aminotriazole (3-AT) was added where indicated. An analysis of growth on selective media was achieved by growing liquid cell cultures to an A₆₀₀ of 0.3, from which fivefold dilutions were made. Three microliters of each dilution was then spotted onto the appropriate medium and incubated at 30°C unless indicated.

Plasmids.

Saccharomyces cerevisiae eIF2Bɛ is encoded by the GCD6 gene. eIF2α, -β, and -γ are encoded by the SUI2, SUI3, and GCD11 genes, respectively. For consistency in the Results and Discussion sections, all proteins/mutants are referred to as eIF2 or eIF2Bɛ derivatives, but here yeast genetic nomenclature is used to specify DNA constructs. pAV1767 [GCD6 CEN LEU2] and several site-directed mutant derivatives (pAV1762 [gcd6-E548A CEN LEU2], pAV1763 [gcd6-R555A], pAV1764 [gcd6-D564A], pAV1768 [gcd6-L570A], pAV1782 [gcd6-E569A], and pAV1783 [gcd6-L568A]) were described previously (9). Other gcd6 plasmids used here include pAV1427 [p_GALFH₆-GCD6 URA3 2μm] (where FH₆ represents tandem FLAG (DYKDDDDK) and polyhistidine (HHHHHH) epitope tags and p_GAL represents the galactose-inducible GAL1 promoter), pAV1458 [p_GALFH₆-gcd6-S576N URA3; 2μm], pAV1513 [GCD6 CEN URA3], pAV1524 [gcd6-F250L CEN URA3], pAV1582 [gcd6-T552I CEN URA3], and pAV1586 [gcd6-S576N CEN URA3] (22). Plasmids overexpressing different combinations of yeast eIF2 subunits from YEp24-derived plasmids (eIF2β, p927 [SUI3 URA3; 2μm]; eIF2γ, p1781 [GCD11 URA3; 2μm]; and eIF2αβγ, p1780 [SUI2 SUI3 GCD11 URA3; 2μm]) have been described previously (13), as were plasmids expressing GCN2, (p722 [GCN2 CEN URA3]) (42) and gcn2-507 (p561 [gcn2-507 CEN URA3]). gcn2-507 is a weakened mutant allele caused by a two-codon (E-L) insertion at residue 1246 within the HisRS domain of GCN2 (41).

Mutant plasmids made for this study by the QuikChange method (Stratagene) were pAV1800 [gcd6-E569D CEN LEU2], pAV1813 [gcd6-E569R], pAV1814 [gcd6-W699A], and pAV1815 [gcd6-E569K], using mutagenesis oligonucleotides (data not shown). The DNA sequence of GCD6 was confirmed after each reaction. Mutants with phenotypes were subcloned into pRS316 (URA3 CEN) to create pAV1781 [gcd6-E569A], pAV1816 [gcd6-E569D], pAV1817 [gcd6-E569R], pAV1818 [gcd6-W699A], and pAV1819 [gcd6-E569K]. To overexpress the minimal “catalytic” domain (ɛ^cat), residues 518 to 712 were subcloned as PCR fragments using MluI and SpeI cut fragments ligated into a derivative of the p_GAL1 expression plasmid pEMBLyex4 containing the FH₆ epitope tags, pAV1411 [p_GAL1GCN2 2μm URA3 leu2-d], replacing the GCN2 DNA with the indicated ɛ^cat allele (21). The plasmids generated were pAV1689 [p_GAL1FH₆-ɛ^cat 2μm URA3 leu2-d], pAV1833 [p_GAL1FH₆-ɛ^cat-L568A 2μm URA3 leu2-d], pAV1834 [p_GAL1FH₆-ɛ^cat-E569A 2μm URA3 leu2-d], pAV1835 [p_GAL1FH₆- ɛ^cat-E569D 2μm URA3 leu2-d], pAV1836 [p_GAL1FH₆-ɛ^cat-E569K 2μm URA3 leu2-d], pAV1837 [p_GAL1FH₆-ɛ^cat-E569R 2μm URA3 leu2-d], and pAV1838 [p_GAL1FH₆-ɛ^cat-W699A 2μm URA3 leu2-d].

β-Galactosidase assay.

Activity from a HIS4-lacZ fusion integrated at ura3-52 was assessed as described previously (12).

Purification of proteins.

Yeast eIF2 was purified as described previously (7). FLAG affinity purification of eIF2Bɛ proteins and minimal fragments from yeast cells was described in detail recently (27).

Guanine nucleotide exchange assay.

This assay was performed as described previously (7). Reactions were carried out at 0°C.

FH₆-ɛ^cat and eIF2 interaction assays. (i) interaction between prepurified proteins.

Two hundred nanomolar of FH₆-ɛ^cat, FH₆-ɛ^cat-E569A, or FLAG peptide was immobilized on preequilibrated anti-Flag M2 resin (20 μl; Sigma) in binding buffer (100 mM KCl, 20 mM Tris-HCl, pH 7.5, 2 mM MgCl₂, 1 μg bovine serum albumin [BSA], 0.1% Triton X-100, 5 mM β-mercaptoethanol, and 1× complete EDTA-free protease inhibitors [Roche]) to a final volume of 100 μl at 4°C for 2 h. After being washed with binding buffer (3× 100 μl), purified eIF2 was added to each reaction mixture at a concentration of 40, 20, 10, or 5 nM and incubated for 2 h at 4°C and washed as before. Finally, the resin was resuspended in Laemmli sample loading buffer and denatured for 5 min at 100°C before 3 μl was loaded onto a 10% polyacrylamide-sodium dodecyl sulfate (SDS) gel. Bound proteins were detected by immunoblotting (using anti-eIF2α rabbit polyclonal or anti-M2-FLAG mouse monoclonal primary antisera [Sigma], followed by horseradish peroxidase-linked secondary antibody and enhanced chemiluminescence [GE Healthcare]).

(ii) Salt dependence of eIF2Bɛ-eIF2 interaction.

Reactions were performed as described above except that the binding and wash buffers contained 10 μg BSA and the indicated concentrations of NaCl in place of KCl. Ten microliters of EZview anti-FLAG resin (Sigma) and 20 nM of purified eIF2 were used.

(iii) ɛ^cat-eIF2 subunit interaction assays.

A total of 500 μg extract from GP3889 yeast cells overexpressing wild-type FH₆-ɛ^cat or the indicated missense mutant was immobilized on 10 μl anti-M2-FLAG EZview resin (Sigma) in binding buffer BB-0.5 [20 mM HEPES, pH 7.4, 0.5 M CH₃COOK, 3 mM (CH₃COO)₂Mg, 5 mM β-mercaptoethanol, 0.1% Triton X-100, 1× complete EDTA-free protease inhibitor cocktail (Roche), 1 mM phenylmethylsulfonyl fluoride, 0.7 μg/ml pepstatin, 1 μg/ml leupeptin, 5 μg/ml aprotinin] for 2 h at 4°C. Control reaction mixtures included 500 μg BSA and 6 μg of 3× FLAG peptide in place of FH₆-ɛ^cat extracts. Collected resin was washed once in BB-0.5 and twice in BB-0.1 (as in BB-0.5 but with 0.1 M CH₃COOK). Each resin pellet containing immobilized FH₆-ɛ^cat was incubated with 250 μg extract from strain GP3889 overexpressing either eIF2β or eIF2γ alone, eIF2αβγ, or a vector control in 100 μl BB-0.1 at 4°C for 2 h. After three washes in BB-0.1 resin, bound complexes were eluted into Laemmli sample loading buffer at 95°C and resolved by electrophoresis on 10% polyacrylamide-SDS gels.

eIF2Bɛ E569A mutant is catalytically inactive in vitro.

Previously we presented genetic evidence that within the eIF2Bɛ coding region, a single amino acid substitution at glutamic acid 569 to alanine (E569A) was lethal in an otherwise wild-type cell when present as the only source of eIF2Bɛ. In contrast, similar alanine substitutions at other highly conserved residues within this region had no obvious deleterious effects (highlighted in the sequence alignment in Fig. 1A). As GCD6, the gene encoding eIF2Bɛ in yeast, is essential, we used a genetic trick to overcome this lethality by cooverexpressing all three subunits of eIF2 and a gene (termed hcTC). This strain (GP4115) bypasses the otherwise essential function of eIF2B but remains extremely slow growing. When mutated eIF2Bɛ^E569A was expressed in this strain, we were able to determine that the mutant protein was expressed normally, but as the growth rate of the resulting cells remained unchanged, we concluded that the mutant retained no function in vivo (9).

We hypothesized that the E569A mutant was not functional in vivo, either because it could not interact with eIF2 or because the protein had lost a charged surface residue (Fig. 1B) that was critically important for directly mediating the guanine-nucleotide exchange reaction. As eIF2B is a large, multisubunit, multidomain protein with many eIF2 interaction sites (28), we assumed that any defect might be more readily apparent in a minimal system. We therefore decided to focus our attention on the previously defined minimal eIF2Bɛ domain responsible for stimulating nucleotide exchange in vitro, which extends from residues 518 to 712, here called ɛ^cat (21). An in vitro guanine nucleotide exchange assay was carried out to assess the rate of release of ³H-GDP from an eIF2-³H-GDP binary complex. As previously reported, purified wild-type, FLAG-tagged ɛ^cat promoted the release of ³H-GDP, while the mutant protein ɛ^cat-E569A could not enhance nucleotide release over the intrinsic dissociation rate (Fig. 2A). The ɛ^cat structure is composed of four pairs of α-helices stacked in HEAT repeats (Fig. 1B) (9). We used circular dichroism (CD) spectroscopy to assess whether the introduced mutation altered the overall secondary structure of the purified protein. The results (see Fig. S1 in the supplemental material) show that the wild-type and mutated CD spectra are indistinguishable and confirm that the α-helical nature has not been altered by the mutation.

To assess whether catalytic inactivity was caused by reduced substrate (eIF2) binding activity, we used protein affinity chromatography to assess the binding of various eIF2 concentrations to excess FLAG-immobilized ɛ^cat and ɛ^cat-E569A. After extensive washing, the extent of immobilized ɛ^cat/eIF2 complexes was assessed by immunoblotting. We found that ɛ^cat and ɛ^cat-E569A bound similar substrate amounts over the range of eIF2 concentrations tested (Fig. 2B). Together with previously published experiments, these results are consistent with the idea that the E569A mutation reduces guanine-nucleotide exchange activity to undetectable levels without disrupting the normal secondary structure or the ability to interact with eIF2. We conclude that the mutation has altered a residue that is critical for directly stimulating the guanine nucleotide exchange.

Alanine mutations to residues surrounding E569 do not impair GCN4 activation.

There was a dramatic contrast between the effects of the E569A mutation, which was lethal in vivo and inactive in vitro, and those of the other alanine mutations we introduced into residues surrounding E569 that exhibited no growth phenotype at 30°C in our initial studies (9). This was especially unanticipated because we included the adjacent residues L568 and L570 in this analysis, and the mutation S576N impairs eIF2B function dramatically (22). We decided to examine the nonlethal mutants in more detail in an attempt to reveal any minor phenotypes. As in previous studies, we exploited the fact that down-regulation of eIF2B activity is necessary for the activation of GCN4 translation to assess any minor defect in eIF2B function. As described in the introduction, GCN4 is a sensitive indicator of eIF2B function. Under conditions of histidine starvation induced by 3-AT, gcn2Δ cells expressing wild-type eIF2B are unable to induce GCN4 expression via the activation of Gcn2p and fail to grow (see Fig. S2A in the supplemental material). Mutations in eIF2B that reduce eIF2B activity enhance growth of cells on 3-AT medium as the eIF2B defect lowers TC (for an example, see reference 31). In this assay, none of the mutants tested conferred enhanced growth on 3-AT. Transforming these cells with a plasmid encoding wild-type GCN2 or a mutant kinase with reduced activity (called gcn2-507) produced the expected growth on 3-AT medium for wild-type eIF2B cells and an identical growth pattern for each mutant tested. As the cells contain a β-galactosidase fusion to the Gcn4p-target gene HIS4, the expression levels of HIS4-lacZ were assessed in each strain as a quantitative measure of GCN4 activity (see Fig. S2B in the supplemental material). Consistent with the plating results, no significant differences were observed between strains in these assays; all strains show a GCN2-dependent and 3-AT-dependent increase in HIS4 activity. These results show that none of the mutant strains tested exhibit a phenotype different from that of the wild-type strain at the optimal growth temperature for yeast (30°C). These observations make the results obtained with the E569A mutant even more striking.

Genetic analysis reveals that eIF2Bɛ-L568A has a cold-sensitive phenotype.

In the course of our studies, we transformed yeast strains bearing either the wild type or the eIF2Bɛ-S576N slow-growing mutant with compatible plasmids, each expressing an alanine-substituted eIF2Bɛ mutant. Thus, in these strains there were two sources of eIF2Bɛ, each encoded on a centromeric plasmid. In the presence of wild-type eIF2B, all eIF2Bɛ-alanine substitutions were recessive, but when combined with eIF2Bɛ-S576N, both E569A and L568A conferred a dominant reduction in growth rate (Fig. 3A). We reasoned that in cells expressing both the S576N and the E569A alleles of eIF2Bɛ, each was competing for inclusion within the eIF2B complex and that the inclusion of the nonfunctional ɛ^E569A reduced the concentration of active eIF2B complexes containing ɛ^S576N. Because in this strain, eIF2B activity is limiting for growth, any further reduction in eIF2B activity resulted in an enhancement of the growth defect. Surprisingly, we found that ɛ^L568A also caused a dominant growth phenotype similar to that of ɛ^E569A (Fig. 3A). This was the first genetic evidence that ɛ^L568A is a reduced-function allele.

In our hunt for further phenotypes, we screened strains with single eIF2B mutations at low (15°C) and high (36°C) growth temperatures and found that ɛ^L568A conferred a reduced growth rate at 15°C, while all other alleles tested behaved like the wild type (data not shown). β-Galactosidase assays of the HIS4-lacZ reporter using extracts from cells grown at 15°C revealed a twofold increase in lacZ activity in L568A mutant cells (Fig. 3B), consistent with the growth defect. These data show that L568 is an important residue involved in the nucleotide exchange reaction, especially during low-temperature growth.

Cold sensitivity is specific to mutations within the catalytic domain.

In light of the cold sensitivity of the L568A allele, the growth at low temperature of other previously described, reduced-function eIF2Bɛ missense mutants was assessed. These were F250L, T552I, and S576N, of which F250L is outside the catalytic domain. Serial dilutions taken from liquid cell cultures growing in exponential phase were spotted onto yeast extract-peptone-dextrose (YPD) medium and grown at either 30°C or 15°C (Fig. 3C). At 30°C, T552I has a moderate slow-growth phenotype, while F250L and S576N are more severe. All strains grew slower at 15°C, but we noted that relative to the wild-type strain, those bearing mutations within the catalytic domain grew slower at 15°C, with the strain carrying S576N having the most extreme phenotype observed. However, the F250L mutant was relatively unaffected by low temperature. These similar phenotypes suggested that these mutations in ɛ^cat helices I and II all affected eIF2B activity via a similar mechanism. Immunoblotting extracts from cells grown at 15°C confirmed that the phenotypes were not caused by a reduction in protein expression or stability of the cold-sensitive alleles (Fig. 3D).

The interaction between ɛ^S576N and eIF2 is salt sensitive.

It has been suggested by several authors that cold-sensitive phenotypes are caused by protein-protein interaction defects (29). This would indicate that either an eIF2B subunit interaction defect or an eIF2B/eIF2 interaction defect is likely. In a previous work, neither was found (22). To assess the eIF2/eIF2B interaction more rigorously, we examined the salt sensitivity of binding. eIF2 and eIF2B copurify from cells in 100 mM salt-containing buffers but are dissociated in buffers containing 500 mM salt (11). To assess whether the mutants may have a defect in eIF2/eIF2Bɛ complex formation, we examined the interaction of ɛ^S576N and eIF2 and compared it with that of wild-type ɛ and eIF2 in buffers containing increasing NaCl concentrations. We chose this mutant as it appeared to be the most cold sensitive in vivo. Differences in the extent of eIF2/eIF2Bɛ complex formation were detected (Fig. 4). A sharp reduction of eIF2 binding to ɛ^S576N was observed at 150 mM NaCl (a loss of ∼66% compared with the ∼25% loss from wild-type ɛ at this concentration). This result provided evidence that the cold-sensitive eIF2Bɛ alleles may be partially defective in eIF2 binding.

Substitution of other amino acids at position E569.

As described above, E569 was the only residue that when mutated caused the most severe phenotype in that E569A eliminated eIF2B nucleotide exchange function, a result observed previously with only alleles lacking the catalytic domain entirely (21). To examine the role of the charged side chain of E569 more closely, we used SDM to introduce either a conservative Asp or a charge reversal Arg or Lys residue in place of the Glu residue within the full eIF2Bɛ coding region. When introduced into eIF2BɛΔ yeast (strain GP3750) on a centromeric plasmid and subjected to plasmid shuffling on medium containing 5-fluoroorotic acid (5-FOA), all mutations, including the conservative E569D, were found to be lethal (Fig. 5A). When introduced into the slow-growing eIF2BɛΔ hcTC strain (GP4115) as the sole source of eIF2Bɛ, only the conservative E569D allele indicated any residual function, as it partially rescued the doubling time of the strain from 6 to 5 h (Fig. 5B). This made E569D identical in growth behavior to the previously described lethal allele N249K (Fig. 5B). Although lethal, purified five-subunit eIF2B containing N249K did retain some nucleotide exchange activity in vitro that was consistent with these results (22). Immunoblotting confirmed that none of the mutations introduced affected the expression of the resulting protein (Fig. 5C). As E569D is a very modest change to the protein, shortening the side chain of this residue by a single -CH₂- group, we were surprised by this dramatic reduction in function.

W699 is a third critical residue in vivo.

When examining interactions between eIF2Bɛ and its substrate, we have used intact eIF2 heterotrimer as a source of eIF2. Hinnebusch and colleagues found that multiple alanine substitutions of conserved residues within the AA boxes at the C terminus of eIF2Bɛ disrupted its interaction with repeated lysine clusters within the eIF2β subunit (6). The subsequent ɛ^cat crystal structure indicated that many of the AA box residues mutated in that study are important for the structural integrity of the domain but that the conserved tryptophan residue at 699 is exposed on the surface (Fig. 1B) and may therefore be important in mediating protein-protein interactions (9). We therefore decided to investigate the role of W699 and used SDM to change this to alanine. In vivo W699A behaved identically to E569A in that it was unable to support the growth of the eIF2BɛΔ strain (Fig. 5A), could not rescue growth of the eIF2BɛΔ/hcTC strain (Fig. 5B), and did not destabilize the protein (Fig. 5C). This analysis therefore identified a second residue essential for eIF2B function in vivo, situated on an opposite surface of the ɛ^cat structure ∼40 Å from E569 (Fig. 1B).

Effects of L568, E569, and W699 mutations on interactions between ɛ^cat and eIF2 subunits.

As stated above, previous work has indicated a critical interaction between eIF2β and eIF2Bɛ. More recently, Alone and Dever showed that both eIF5 and eIF2Bɛ could interact directly with the nucleotide binding G-domain of eIF2γ (2). They further showed that the ɛ^cat domain was sufficient for the interaction. This provided the first evidence that eIF2B might function by direct interaction with eIF2γ rather than via a more indirect mechanism of action.

To examine the interaction of ɛ^cat and mutated versions with eIF2 subunits, we developed an assay to measure the interaction between wild-type and mutant forms of ɛ^cat and eIF2β and eIF2γ. Because eIF2α interacts with the eIF2Bαβδ subcomplex (as opposed to ɛ^cat), direct binding between ɛ^cat and eIF2α was not assessed. FLAG-tagged versions of ɛ^cat were purified to homogeneity. As these mutants lack catalytic function, we assessed the CD spectrum of each and found that all mutants retained an α-helical structure identical to that of the wild-type protein (see Fig. S1 in the supplemental material), indicating that each mutation does not adversely affect the protein's overall helical structure. We developed an interaction assay whereby ɛ^cat was immobilized on anti-FLAG affinity resin and then incubated with a yeast extract (GP3889) overexpressing different eIF2 subunits (eIF2αβγ together, eIF2β alone, or eIF2γ alone) or bearing an empty plasmid vector control (Fig. 6A). After extensive washing, eIF2 subunits remaining bound to the resin were assessed by SDS-polyacrylamide gel electrophoresis and immunoblotting. Control reactions where FLAG peptide was immobilized showed that the eIF2 subunit interactions were specific (Fig. 6C to F, lanes 1). In the absence of artificial overexpression, both wild-type ɛ^cat and E569A alleles bound significant endogenous eIF2 (Fig. 6C, lanes 2 and 3), while the other mutations introduced at E569, L568A, and W699A all significantly reduced eIF2 binding. The overexpression of all eIF2 subunits (∼10-fold) (Fig. 6B and D) enhanced binding between the eIF2 subunits and wild-type ɛ^cat or all ɛ mutants except W699A. Similarly, excess eIF2β alone enhanced binding between eIF2β and all forms of ɛ^cat assessed. But in this experiment, W699A was the only mutant tested with significantly reduced eIF2β binding, suggesting that W699 is important for the eIF2β/ɛ^cat interaction (Fig. 6B and E). In contrast, excess eIF2γ enhanced binding to the wild type and E569A but had only a minor influence on interactions between eIF2γ and the other E569 alleles, W699A and L568A (Fig. 6B and F). Taken together, these data show that the impaired binding of the L568A and E569 alleles to eIF2 (Fig. 6C, lanes 5 to 8) is primarily due to the disruption of the eIF2γ interface, while the W699A allele disrupts both eIF2β and eIF2γ interactions (Fig. 6E and F, lanes 4). These results reveal that both eIF2β and eIF2γ interfaces with eIF2Bɛ^cat are critical for nucleotide exchange.

Evidence that E569 is the principal catalytic residue.

Our initial genetic analysis suggested that E569 was critical for nucleotide exchange in vivo. An alanine substitution at this residue within the full-length protein failed to support the growth of yeast (9). Here we extended our studies and found that the purified ɛ^cat domain (residues 518 to 712) containing the E569A substitution was inactive in our exchange assay but did not significantly alter the interaction with the eIF2 heterotrimer (Fig. 2 and 6). Acidic residues have been found important in other GEFs, including EF-Ts (asp80) (35), Sec7 (glu156) (30), and RopGEF8 (glu338) (36). Interestingly, a charge reversal mutation in ARNO (E156K) resulted in a stabilized inactive Arf-GDP/Sec7 complex (30). Although all GEFs are structurally diverse and ɛ^cat shares no homology with ARNO, we hypothesized that an equivalent charge reversal in eIF2B might also stabilize an eIF2/eIF2B complex. So we introduced changes at E569, including the conservative Asp and the charge reversal Lys or Arg residues. All three additional substitutions at E569 were also lethal. In our most sensitive in vivo assay for eIF2B function, cells lacking eIF2B activity can be kept alive by cooverexpressing eIF2 and TC (Fig. 5B) (17, 21). These cells are extremely slow growing, but reintroducing functional eIF2B alleles on plasmids rescues growth. In this assay, only the conservative E569D allele modestly improved growth, consistent with the notion that it retained some function. We were surprised by the severity of this phenotype, which provides additional strong evidence for the importance of E569.

To examine any defects in eIF2 binding, we developed an assay that allowed us to monitor interactions with both eIF2β and eIF2γ. Binding between eIF2β and eIF2Bɛ is mediated by three runs of lysine residues (K boxes) in the eIF2β N terminus and the AA box regions of eIF2Bɛ (6). This interface is also critical for interaction between eIF2 and eIF5. In addition, it was recently shown that the ɛ^cat domain interacted directly with a glutathione S-transferase-eIF2γ G domain fusion in vitro, while glutathione S-transferase alone did not (2). In contrast to the effects of charge reversals on the Arf-GDP/Sec7 complex (30), we found that E569D/K/R substitutions all destabilized the interaction with eIF2γ (Fig. 6F) but retained the ability to interact with eIF2β (Fig. 6E). It is likely that the eIF2γ-interaction defect was important for the overall reduced affinity for the eIF2 heterotrimer observed with these mutants (Fig. 6C and D). In contrast, the neutral E569A mutation retained the ability to interact with eIF2 (Fig. 2 and 6). Thus, the E569A mutant retains interaction but lacks any activity, while the E569D mutant interaction with eIF2 is weakened but retains some minimal GEF activity in vivo. In silico modeling of the effects of the E569D mutation on the ɛ^cat structure suggests that the center of the negative charge at residue 569 moves so that it is ∼3 to 4 Å further from the adjacent charged surface residues E548 and R555 and instead ∼2 to 4 Å closer to H561, T565, L568, and L573 side chains. This may therefore form altered interactions that destabilize any contributions to overall eIF2γ binding made by these surface residues. Overall, these data are consistent with the idea that the E569/eIF2γ interaction is essential for the nucleotide exchange reaction.

Mutations altering residues surrounding E569 are cold sensitive.

We also found that the eIF2Bɛ mutation L568A, T552I, or S576N conferred cold sensitive growth at 15°C (Fig. 3C). The L568A allele was defective at only a low temperature and increased GCN4-dependent HIS3 expression under non-amino acid starvation conditions at a low temperature (Fig. 3B) but not at the normal growth temperature (data not shown). This result suggested either that yeast cells require higher eIF2B activity at this low growth temperature or that the low temperature exacerbates the mutant protein activity in some manner. As growth of the F250L allele was not exacerbated relative to that of the wild type, the latter idea is more likely. It has been noted that cold-sensitive mutations are indicative of protein-protein interaction defects (29). We examined this possibility for both S576N and L568A alleles. The interaction of S576N with eIF2 was more sensitive to elevated salt conditions (Fig. 4), suggesting that ionic interactions are disrupted more easily in this mutant. The eIF2-ɛ^cat-L568A interaction was also weakened. The interaction defects observed with this mutant were similar to that with eIF2-ɛ^cat-E569D described above, indicating that the eIF2γ/ɛ^cat interaction was most disrupted. We speculate that because our protein-protein interaction assays were performed at 4°C and the mutations are cold sensitive, the interaction defect observed may have been enhanced by the assay conditions. These observations further support the idea that ɛ^cat helices I and II make contacts with eIF2γ that are critical for nucleotide exchange.

W699A disrupts multiple interactions between eIF2 and ɛ^cat.

Our data show that W699 is critical for interaction with both eIF2β and eIF2γ. W699 is highly conserved among both eIF2Bɛ and eIF5 sequences and part of the AA box 2 motif. The mutation of seven conserved residues within this AA box (termed the gcd6-7A mutation, which included W699A) was shown to weaken the interaction between eIF2β and eIF2B. Surprisingly perhaps, the multiple changes in the 7A allele did not result in a lethal phenotype but did lower eIF2B activity in vivo as indicated by the Gcn2p-independent activation of GCN4 (6). The eIF2Bɛ/eIF2β interaction was shown to be dependent upon conserved repeated lysine residues within the eIF2β N terminus (termed K boxes 1 to 3). The equivalent eIF5-7A mutation had similar effects on function and eIF5/eIF2β interaction, leading to the idea that GEF and GTPase-activating protein share a common interaction site on eIF2β (6). The crystal structures of both ɛ^cat and eIF5-CTD confirm that they share a highly similar helical HEAT repeat structure and that the ɛW699 (W391 in yeast eIF5) side chain is exposed on the surface at one end of the protein, predicting that it would be important for eIF2 interaction (8, 9, 40) (Fig. 1B). Our finding that a single W699A mutation completely inactivated eIF2B in vivo (Fig. 5) confirmed its importance but made extensive genetic analysis challenging. As W699A failed to enhance growth in the strain kept alive by overexpression of TC (GP4115) (Fig. 5B), this result provides strong genetic evidence that eIF2B function is eliminated by this mutation. Our binding studies show that the mutation within the context of the ɛ^cat domain dramatically impairs binding to eIF2αβγ complexes as well as to either eIF2β or eIF2γ alone (Fig. 6). W699A was the only mutation tested that reduced eIF2β binding (Fig. 6E), demonstrating that this surface region is critical for eIF2β interaction, while the helix I/II surface is not.

W699A also dramatically impaired interaction with eIF2γ (Fig. 6F). This unexpected result indicates that this mutation disrupts ɛ^cat interaction with both eIF2β and eIF2γ interfaces. As W699 is located approximately 40 Å from E569 at the opposite end of the structure (Fig. 7), this result suggests that ɛ^cat makes extensive interactions with eIF2γ. One explanation for W699 being critical for both eIF2β and eIF2γ binding is an analogy with eIF5-CTD, where adjacent surface regions interact with eIF2β and eIF3 (43). As W699 is located at one end of the domain, perhaps this residue is important for the initial contact with both subunits.

Implications for exchange mechanism.

Together, our data show the importance of two regions of ɛ^cat for interaction with eIF2β and γ and catalysis of nucleotide exchange. Mutations within ɛ^cat helix II disrupt interaction with eIF2γ, thus clearly establishing, for the first time, a direct interaction between the GDP binding subunit of eIF2 and residues of eIF2B that are critical for nucleotide exchange. The E569A mutant is effectively catalytically inactive in vitro and in vivo but did not disrupt eIF2β or -γ or -αβγ binding (Fig. 2 and 6), providing strong evidence that E569 is critically important for nucleotide exchange.

Costructures between G proteins and their GEFs exist for small G proteins and translation elongation factors, and the structural rearrangements in the G protein have been reviewed previously by others (10, 35). For these GEFs, binding promotes rearrangement of the phosphate binding P-loop (GXXXXGKS/T) and the switch regions (Sw1 and Sw2) that are important for magnesium ion and phosphate binding. The rearrangements in the P-loop region particularly reduce the Mg²⁺-GDP affinity that facilitates nucleotide release. A glutamate residue has been identified as critical for nucleotide exchange in several GEFs, including the Sec7 domain of ARNO. ARNO Glu156 interacts with and distorts the P-loop of the G protein ARF by binding Lys30 within the P-loop. ARNO also makes contacts with and stabilizes the conformation of Sw2 (30).

In contrast to the small G proteins, the translation elongation factor G proteins EF-Tu and eEF1A possess three domains, a G domain and two additional domains (domains 2 and 3) that play roles in GEF binding. EF-Ts makes extensive contacts with both the G domain and domain 3 of EF-Tu (26), while eEF1Bα interacts with both the G domain and domain 2 of eEF1A (4). eIF2γ is highly homologous to both elongation factors, similarly possessing domains 2 and 3, but is further complicated by additional eIF2 subunits (eIF2α and eIF2β). A recently determined structure for an archaeal translation initiation factor 2βγ (aIF2βγ) complex reveals that aIF2β sits astride the aIF2γ G domain over the zinc binding region and extends to the Sw1 region (34). Similarly X-ray structure and mutagenesis studies show that eIF2α interacts with eIF2γ domain 2 (32) (44). Using these data and that for other GEF/G protein interactions, we propose a model to explain our observations on the importance of E569 and W699 in ɛ^cat (Fig. 7). We suggest that ɛ^cat binds with W699 across an eIF2β/γ interface and that E569, in conjunction with other residues, remodels the P-loop and the Sw1/Sw2 regions to facilitate nucleotide exchange. However, it is possible that eIF2B instead reduces the affinity of eIF2γ for the GDP guanine base. Prior genetic and biochemical analyses of eIF2γ alleles suggest this because mutations altering single Sw1 or P-loop residues were found to impair interactions rather than nucleotide binding. Instead, alteration of the guanine ring binding residue K250 within the conserved NKXD loop enhanced GDP/GTP release and mimicked eIF2B function (15, 16, 24). The genetic tools generated here should facilitate further molecular studies to address these alternative ideas for the mechanism of eIF2B-mediated nucleotide exchange.