eIF2β is critical for eIF5-mediated GDP-dissociation inhibitor activity and translational control

Yeast genetics and cell culture

Strains used and generated in this work are derivatives of S288c and are described in Supplementary Table S1. Cell growth was performed at 30°C in standard yeast rich (YPD), synthetic defined minimal (SD) or synthetic complete (SC) media lacking nutrients for plasmid selection. All media have 2% glucose as a carbon source (25). 3-Amino-1,2,4-triazole (3AT) (Fluka) was added where indicated. For serial dilution growth assays, cells were grown to A₆₀₀ = 0.6, diluted to 0.1 and 2–3 μl of 10-fold dilutions were spotted on the agar surface. Plasmid shuffling to select for loss of URA3 plasmids employed 5-fluoro-orotic acid (5-FOA; Sigma-Aldrich), while loss of LEU2 plasmids relied on passive loss during unselected growth and subsequent screening. Mating type switching employed HO plasmid pAV1621 (pHO1.5) followed by 5-FOA, sporulation, tetrad dissection and mating type testing (25). gcd6Δ sui3Δ double deletion strains (GP7216-19) were made by polymerase chain reaction (PCR) amplifying gcd6Δ::KanMX cassette from a BY4743 gcd6Δ::KanMX/GCD6 diploid strain (GP4146) and transforming the DNA fragment into sui3Δ gcn2Δ [SUI3-Flag TRP1] [GCD6 URA3] strains to generate G418 resistant cells unable to lose the covering GCD6 plasmid upon 5-FOA selection.

Plasmid constructions

Plasmids used are described in Supplementary Table S2. The V5 epitope was added to the N-terminus of eIF2 (SUI3) to create pAV2322 in two steps. First tagged SUI3 was made by custom synthesis (Epoch life Sciences) of a SUI3 gene 5′ fragment inserting the V5 epitope (GKPIPNPLLGLDST) between the second and third codon. This also introduced a SpeI restriction site. Next XmaI and MluI were used to cut and insert this fragment into the same sites of the high copy plasmid pAV1726 (16) to generate a LEU2 plasmid with all three eIF2 genes V5-SUI3, SUI2 and his₆-GCD11 each expressed under their native promoters.

The E189K mutation was introduced into pAV2322 and also pAV2443 (KAB258) (26) to create pAV2451 and pAV2452 by Quikchange site-directed mutagenesis (Agilent Technologies) using the oligonucleotides SUI3E189KFOR (GCATAGATCTCCGAAACATTTGATTCAATATCTCTTCGC) and SUI3E189KREV (GAATCAAATGTTTCGGAGATCTATGCAATTTTTCGGCG). Colonies were screened with BspE1 digestion and the SUI3 ORF in the constructs was verified by Sanger DNA sequence analysis.

Genomic DNA sequencing

GP3755 and GP3763 genomic DNA was sequenced using a SOLiD4 sequencer (Life Technologies) at the University of Manchester genomics technologies facility using standard kits and protocols. To generate paired-end reads (F3 [start of DNA fragment] = 50 bp and F5 [end of DNA fragment] = 35 bp) the CSFASTA and QUAL files for each sample were converted to FASTQ format using the solid2fastq script (BFAST software) (27,28). The reads were quality filtered using SOLiD_preprocess_filter_v2.pl (29) (settings used -x y -p 3 -q 22 -y y -e 10 -d 9). The reads were mapped to the Saccharomyces cerevisiae reference genome ‘sacCer3’ obtained from UCSC (30) (http://genome.ucsc.edu/), which includes the assembled chromosomes, the mitochondrial genome and the two micron plasmid. The F3 reads were mapped using BFAST ‘match’ and the F5 reads were mapped using BFAST + BWA 0.7.0.a ‘bwaaln’. Mapping was completed as single-ended using the F3 reads only, or the individually mapped F3 and F5 reads were both used for paired-end mapping. In both cases the ‘-a 2’ flag of the post-process step was used to obtain uniquely mapping alignments. The F3 reads of properly paired reads were obtained using ‘samtools view -f 66’ (31). SRMA v0.1.15 was used to realign reads to correct for erroneous mapping caused by the presence of INDELs at the end of reads. Variant detection was performed using samtools v0.1.18, using mpileup and bcftools without BAQ computation (32,33). Candidate variants where filtered using vcfutils.pl requiring at least 20 high-quality read matches and a maximum coverage calculated as three or five times the median coverage (including 0 counts). Variants were also filtered to remove any with QUAL (variant quality) <40.

Verification sif285-1 is sui3-E189K

Genomic DNA was isolated from yeast strains GP3771-4 using standard procedures (25) and the SUI3 ORF amplified by PCR using primers eIF2betaF1 (ATGTCCTCCGATTTAGCTGC) and eIF2betaR286 (TCACATTCTCCTTCTCTTACC). PCR products were screened by BspE1 digestion and verified by Sanger DNA sequence analysis.

Sequence alignment

Alignment generated using Clustal X (34) and the following eIF2β sequences (termed EIF2S2, SUI3 or TIF212) from Genbank: S. cerevisiae (CAA97959.1), Candida albicans (XP_715210.1), Candida glabrata (XP_446841.1), Cryptococcus neoformans (AAW43820.1), Schizosaccharomyces pombe (CAB11076.2), Caenorhabditis elegans (CAB00049.1), Homo sapiens (CAG33015.1), Mus musculus (AAI13767.1), Arabidopsis thaliana (AAK29672.1), Oryza sativa (NP_001060649.1), Glycine max (XP_006575455.1).

Protein synthesis

The rate of protein synthesis was measured in exponential phase cells in SC-uracil-methionine at 30°C in a water bath. Once cells (10 ml cultures in 50 ml flasks) reached A₆₀₀ = 0.1, 160 μl methionine mix was added (1 μl [³⁵S] methionine (1175 Ci/mmol Tran35S-label, MP Biomedicals) in 50 μM methionine). 100 μl samples were removed immediately and every 20 min over a 2 h period and were immediately precipitated with ice-cold 20% trichloroacetic acid (TCA). Total protein incorporation was measured by scintillation counting of TCA-precipitates absorbed onto GF/C filters (Whatman) washed with ice-cold acetone.

For Polysome profile analysis yeast strains were grown to an A₆₀₀ = 0.7 in 100 ml YPD medium. For glucose starvation, cells were collected by rapid centrifugation at room temperature and resuspended in pre-warmed YP for 10 min prior to final harvest. Extracts were prepared in 50 μg/ml cycloheximide and layered onto 15–50% sucrose gradients as described (35). The gradients were sedimented via centrifugation at 40 000 rpm for 2.5 h using a SW41 rotor (Beckman). Sucrose density gradient fractionation was performed on an ISCO gradient fractionator and the A₂₅₄ was measured continuously to give the traces shown.

β-galactosidase assay

Measurement of β-galactosidase levels from a HIS4-LacZ fusion genomically-integrated at ura3-52 was performed exactly as described (36) using whole protein extracts from cells grown in SC medium lacking histidine and uracil or following addition of 10 mM 3AT for 6 h. Units are nmole o-nitrophenyl-β-D-galacto-pyranoside cleaved/min/μg total protein ± SE, n = 3.

FLAG immune precipitation and western blotting

FLAG (Sigma-Aldrich) immune precipitation and western blotting were performed as described recently (23,37,38) using whole cell extracts to capture FLAG-eIF2β and associated proteins. Bound proteins were probed using specific antibodies and quantitative IR western blot detection was performed using IRDye^® 800CW goat anti-rabbit IgG or IRDye^® 680RD goat anti-mouse IgG with an Odyssey Fc imager (Li-Cor).

GST pull-down assay

Purified GST-eIF5 or GST alone (200 pmol) and eIF2 (200 pmol) were incubated in 500 μl of interaction buffer (30 mM HEPES pH 7.5, 100 mM KCl, 0.1 mM ethylenediaminetetraacetic acid (EDTA), 1 mM 1,4-dithiothreitol (DTT), 2.5 mM MgCl₂) with 25 μl of glutathione-Sepharose beads for 1 h at 4°C. Beads were washed three times with 500 μl of interaction buffer and bound proteins were eluted by boiling in protein-sample buffer. Western blotting was performed as above.

Protein purification

eIF2 was purified from yeast strains GP7145 and GP7146 expressing V5-sui3-E189K or the V5-SUI3 derivatives of eIF2β by sequential nickel affinity, HiTrap Heparin and HiTrap Q sepharose chromatography as described (14). To obtain apo–eIF2 free from nucleotide, eIF2 was dialysed with EDTA (30 mM HEPES, 100 mM KCl, 1 mM DTT, 1 mM EDTA, 10% glycerol, pH 7.4) then with magnesium (30 mM HEPES, 100 mM KCl, 1 mM DTT, 0.1 mM MgCl₂, 10% glycerol, pH 7.4). GST-eIF5 was purified from Escherichia coli as described (23). eIF2B was purified from yeast strain GP5949, as described (39).

Steady state fluorescence

To assay nucleotide affinity, fluorescence intensity of 100 nM of BODIPY-FL-GDP or BODIPY-FL-GTP (Thermo Fisher Scientific) in 180 μl of assay buffer (30 mM HEPES, 100 mM KCl, 10 mM MgCl₂, pH 7.4) was measured using a Fluoromax-4 spectrophotometer (Horiba) (490 nm excitation, 509 nm emission). Yeast Met–tRNA_i was assayed in a similar manner using 20 nM BOP-N-Met–tRNA_i (tRNA probes) but with the addition of 1 mM GTP. Change in fluorescence intensity was measured upon addition of increasing amounts of apo–eIF2, incubating for 5 min at room temperature each time. Each measurement was blanked against a control without nucleotide to account for any affect of eIF2 and data were corrected for dilution effects caused by volume addition and normalized to starting values before being fitted to a single site binding model: y = 1 + [(ΔF_max - 1)*(x/(x + K_d))].

GDP dissociation assay

Fluorescent eIF2•BODIPY-GDP binary complex was formed by incubating apo–eIF2 with a 2× excess of BODIPY-FL-GDP (Thermo Fisher Scientific) for 20 min at room temperature. Excess nucleotide was removed by passing through a G-50 sephadex column (GE Healthcare). Labelling efficiency was calculated to exceed 90%. To measure GDP release, 20 nM eIF2•BODIPY-GDP was quickly mixed with 1 mM of unlabelled GDP (±eIF2B and ±GST-eIF5) in 180 μl of assay buffer (30 mM HEPES, 100 mM KCl, 10 mM MgCl₂, pH 7.4) and fluorescence intensity was continuously measured using a Fluoromax-4 spectrophotometer (Horiba) (490 nm excitation, 509 nm emission, 0.1 s integration time). Experimental data were fitted to exponential dissociation curves to determine the rate constants (K_off) at each eIF2B concentration. K_1/2 and K_max values were determined from curve fitting y = [(K_max × x)/(K_1/2 + x)] + c. Yeast eIF2B was recently shown to be a functional dimer (40), so likely interacts with eIF2 as a 2:2 complex. For consistency with prior studies, eIF2B and eIF2 concentrations stated and equations used refer to 5-subunit and 3-subunit monomeric complexes respectively.

sif285-1 is a spontaneous growth suppressor of eIF2B mutations

A spontaneous suppressor of the slow-growth of one yeast eIF2Bϵ mutated strain (gcd6-F250L) was noted following routine sub-culture on solid medium (Supplementary Figure S1A). The suppressor-bearing strain carries a deletion of the essential GCD6 gene and is complemented with a plasmid borne gcd6-F250L allele. Following transformation and plasmid shuffling the growth-suppressor strain showed a clear difference in growth rate compared to wild-type (WT) cells when both were combined with the gcd6-F250L mutation which grow slowly in the WT background and grow well in the growth-suppressor strain. The unknown mutation was hence initially named sif285-1 for suppressor of eIF2B5 and the WT version termed SIF285 (Figure 1A and Supplementary Figure S1B). This name is used in the description of our analyses here prior to the identification the molecular defect causing the mutation.

Measurements of growth-rates in liquid cultures, total protein synthesis by ³⁵S-methionine incorporation and polysome profile analyses showed the defect in eIF2B activity conferred by the F250L mutation affected general protein synthesis in an otherwise WT strain, as expected from previous studies (41). In contrast, these defects were largely suppressed by the presence of the sif285-1 mutation (Figure 1A–C). sif285-1 was also able to restore faster growth to all slow-growing gcd6 mutations tested (Supplementary Figure S1C), including suppressing the lethality of gcd6-N249K, a Gcd6 mutation which retains residual GEF activity that is insufficient to permit growth of otherwise WT cells (41). However, sif285-1 could not rescue a complete deletion of GCD6. Mating to create a/α diploid cells showed that the sif285-1 phenotype was semi-dominant (Supplementary Figure S1D). Sporulation and tetrad dissection showed the fast-growing phenotype segregated 2:2 (Supplementary Figure S1E), indicating that sif285-1 was likely a single gene alteration.

Mutations in multiple factors can cause a loss of translational control, including mRNA-decay pathway mutants (42). To assess if sif285-1 is generally resistant to translational control responses, we monitored the initial cell responses to glucose starvation by polysome profile analysis following 10 min glucose withdrawal, as this response is independent of eIF2B inhibition (43). sif285-1 mutant cell translation remains glucose-withdrawal sensitive, suggesting sif285-1 suppression has at least some eIF2B specificity (Figure 1D). Because yeast eIF2B mutations constitutively derepress GCN4 translation and the Gcn4-mediated general amino acid control response (GAAC) independent of Gcn2 (termed a Gcd^– phenotype), we assessed resistance of sif285-1 cells to 3AT, a competitive inhibitor of histidine biosynthesis that activates GAAC via the signalling pathway (box in Figure 1F) (17). sif285-1 could not prevent the GCN2 independent 3AT resistance of gcn2Δ gcd6-F250L cells in a replica plating assay (Figure 1E). To examine if sif285-1 cells alter the Gcn2-dependent GAAC response, we transformed these cells with a WT GCN2 plasmid and repeated our growth analyses. sif285-1 GCN2 cell growth was impaired by 3AT, indicating that sif285-1 was a GAAC non-derepressible mutant (a Gcn^– phenotype, Figure 1F). This observation was confirmed by analysing the levels of β-galactosidase expressed from a genomically integrated HIS4-LacZ fusion present in these strains. HIS4 transcription is derepressed ∼2-fold by GAAC via increased GCN4 translation in the presence of 3AT (17). As expected, gcd6-F250L cells exhibit high unregulated HIS4 expression. In contrast, HIS4-LacZ expression is not derepressed by 3AT treatment in sif285-1 cells bearing GCD6, while regulated HIS4 expression is restored in gcd6-F250L sif285-1 double mutant cells (Figure 1F). In summary our data suggest that the sif285-1 mutation permits cells to grow with reduced requirement for eIF2B that alters the threshold for induction of the GAAC response, but does not confer complete insensitivity to eIF2B. Because mutations in human eIF2B that lead to VWM usually impair eIF2B GEF activity and can cause heightened ATF4 expression (44,45), we decided it would be of interest to characterize the sif285-1 mutation further. Standard complementation experiments showed the sif285-1 mutation was not in any of the five eIF2B subunits.

sif285-1 is E189K in eIF2β (SUI3)

To identify the sif285-1 defect we isolated genomic DNA from the original sif285-1 strain (GP3563) and its parent (strain GP3755) and sequenced it by SOLiD4 next generation sequencing (ABI). Following mapping to the reference S. cerevisiae full genome (UCSC version sacCer3), the reads were screened for single nucleotide variants (SNV) and deletions using standard informatics approaches. SNV analysis revealed only seven clear differences between the WT and sif285-1 genomes that had at least 20 high-quality mapped reads (Supplementary Table S3). Only two SNVs corresponded to missense mutations; the first change gcd6-F250L was known and expected, while the second caused an E189K change in SUI3. SUI3 encodes for the eIF2β subunit: a known interacting partner of Gcd6/eIF2Bϵ (26), and was therefore the likely sif285-1 candidate. The missense mutation was predicted to eliminate a BspE1 restriction site in SUI3. The SUI3 ORF was amplified by PCR from the strains and both BspE1 digestion (Supplementary Figure S2) and Sanger DNA sequencing (Figure 2A) verified that only strains bearing sif285-1 carry the sui3-E189K mutation.

We performed a series of experiments to confirm that the slow-growth-suppression and Gcn^– phenotypes of sif285-1 were allelic with an E189K mutation in SUI3. Firstly, both CEN and two micron plasmids bearing WT SUI3 were transformed into WT and sif285-1 strains. The fast growth rate of sif285-1 gcd6-F250L double mutant suppressor strains is complemented by additional copies of SUI3. Thus slower growth is restored to sif285-1 cells bearing gcd6-F250L and SUI3 plasmids (Figure 2B; compare rows 8 and 12 with row 4), as expected if sif285-1 is a mutation in SUI3. Secondly, site-directed mutagenesis introduced sui3-E189K onto a plasmid that was shuffled into a ‘clean’ sui3Δ strain as the sole source of eIF2β. The resulting strain grew as WT, indicating that like sif285-1, sui3-E189K does not confer a significant growth defect (Figure 2C). Thirdly, examining GAAC in these strains revealed that sui3-E189K confers a Gcn^– phenotype and fails to derepress HIS4-lacZ expression in response to 3AT treatment (Figure 2C) analogous to the sif285-1 results shown in Figure 1F. Finally we created a gcd6Δ sui3Δ double knock out strain and complemented each deletion with either WT or mutant versions of each gene. These strains re-created the original slow-growth phenotype of gcd6-F250L and its suppression by sui3-E189K (compare Figure 2D with Figure 1A). Taken together, these data are fully consistent with sui3-E189K being the cause of the sif285-1 suppressor phenotype.

sui3-E189 conservation

The region of eIF2β around E189 is highly conserved across diverse eukaryotes (Figure 3A). E189 itself is acidic (E or D) in plants and lower eukaryotes, while the equivalent residue is already a lysine in animal sequences including humans. The significance of this evolutionary difference is not clear. E189 is located away from mutations studied previously that exhibit different phenotypes and interaction defects (Figure 3B). The eIF2β amino-terminal poly-lysine blocks (termed ‘K-boxes’) are important for stable interactions with both eIF5 and eIF2Bϵ (26). Nuclear magnetic resonance studies showed K-box 2 binds directly to the eIF5 carboxy-terminal domain (CTD) (46). The eIF2β-CTD residue S264 has been implicated in stabilizing GTP binding and regulating GTP hydrolysis as S264Y (SUI3-2) permits enhanced initiation at a non-AUG start codons (47).

Recent structural analysis of the yeast PIC included structural determination for eIF2β 127-270 as part of eIF2•GTP•Met–tRNA_i, complex bound with other initiation factors and a short AUG containing mRNA to the 40S ribosome (48). In this model the eIF2β E189 side chain projects from the surface of eIF2β, permitting potential ligand modulation. In the structural model, conserved residues close to eIF2β E189 appear to stabilize Met–tRNA_i binding to eIF2•GTP, while residues 218-220 contact eIF1A and ribosomal protein uS19 interacts with residues 166-169 at a different eIF2β surface. These data indicate that eIF2β makes multiple important contacts within the PIC (48) and the relative positioning of E189 suggests that E189K could alter ligand binding to eIF2 (Figure 3C). We therefore decided to perform a series of experiments to determine the biochemical impact of E189K on eIF2 ligand interactions.

E189K does not affect GDP, GTP or tRNA affinity for eIF2

WT and mutant eIF2 (eIF2^βE189K) were purified from an overexpression yeast strain engineered so that the introduced plasmid was the sole source of eIF2β (Supplementary Table S1 and Figure S3A). During purification incubation with EDTA was used to remove residual bound magnesium and nucleotide, as described previously (3). To assess nucleotide and Met–tRNA_i binding affinities we monitored the change in fluorescence upon binding of eIF2 to ligands labelled with BODIPY derivatives at room temperature (Figure 4A). These experiments show that BODIPY-GDP and BOP-N-Met–tRNA_i affinities for WT eIF2 are within the ranges measured previously by other techniques (K_{d GDP} = 19.1 nM and K_{d Met–tRNAi} = 2.8 nM) (3,49). Measured BODIPY-GTP affinity is tighter than when GTP-binding was measured by filter-binding assays (3,49), so that the difference between GDP and GTP affinities is significantly less than noted previously. Perhaps surprisingly, the affinities for each ligand are unchanged by the βE189K mutation (Figure 4A). Next, we pre-bound BODIPY-GDP to eIF2 or eIF2^βE189K, and purified these complexes away from unbound BODIPY-GDP. By incubating each complex with excess unlabelled GDP, we monitored the release of BODIPY-GDP and determined its off-rate. The K_off BODIPY-GDP was identical for both WT and eIF2^βE189K (WT K_off = 0.292 ± 0.025 min⁻¹ and eIF2^βE189K K_off = 0.293 ± 0.015 min⁻¹; Figure 4B, 0 nM eIF2B). These results demonstrate that eIF2β-E189K mutation does not simply alter the binding of either nucleotide or Met–tRNA_i. Hence eIF2^βE189K does not share defects in ligand interactions described previously for other eIF2 mutations (47,49).

E189K modestly reduces eIF2B GEF activity, but not PIC interactions

Because the E189K mutation suppresses the slow-growth of eIF2B mutants with reduced GEF activity, one possible mechanism of mutant action was the enhancement of eIF2B activity. We tested this idea by adding increasing concentrations of purified eIF2B (39) to purified eIF2•BODIPY-GDP to determine K and K_max values for eIF2B GEF. The measured K_max for WT eIF2 was ∼14-fold faster than the rate determined previously with a filter-binding assay (14). This likely reflects that these fluorescent assays were performed at room temperature rather than 0°C used previously. Surprisingly, the rate of nucleotide exchange from eIF2^βE189K complexes was modestly impaired relative to that with WT eIF2, increasing K from 51.7 to 92.6 nM eIF2B (Figure 4B). This experiment identified the first clear difference between WT and eIF2^βE189K complexes, but because impairment of eIF2B function by E189K should exacerbate rather than suppress the growth defect of eIF2B mutants, this observation cannot readily account for the growth suppression phenotype uncovered.

As outlined above, eIF2 contacts other ligands within the PIC including eIF5, 40S ribosomes, eIF3 and eIF1. To examine eIF2–protein interactions more widely, we performed anti-FLAG immunoprecipitation of FLAG-eIF2β from WT and E189K cells and assessed the ability of known eIF2-interacting partners to co-associate by quantitative western blotting. The only significant defect observed was a small, but statistically significant, 20% reduction in association of eIF2B subunits with eIF2β-E189K observed upon quantification of the signals (Figure 4C). Together with the modest reduction in activity seen in the GEF assays, these results suggest that one consequence of the eIF2β mutation is impaired eIF2B binding and activity. In contrast, binding of eIF2 to PIC factors eIF5, eIF3, eIF1 and 40S ribosomes appears unchanged (Figure 4C), consistent with the growth characteristics of eIF2β-E189K cells. These data support the idea that the E189K mutation does not significantly, if at all, interfere with eIF2 functions in PIC formation, and scanning/AUG recognition, but does modestly reduce eIF2B interaction and activity.

eIF2β-E189K eliminates eIF5 GDI activity

We found previously that eIF5 mutants that impair its GDI activity can partially suppress slow-growth phenotypes of eIF2Bϵ mutations and confer a Gcn^– phenotype (14,23) in a manner similar to that seen with eIF2^βE189K and eIF2B alleles, including rescue of the lethality of the gcd6-N249K mutation (Supplementary Figure S1C). This suggests a link between eIF2^βE189K and eIF5 GDI mutants. However, the eIF5 GDI mutants characterized act, at least in part, by reducing the affinity of eIF5 for eIF2 (23), whereas the eIF2^βE189K mutation does not reduce eIF5 interaction in vivo (Figure 4C) or in vitro (Supplementary Figure S3B). These observations imply that reduced eIF5 binding to eIF2 does not explain the suppressor phenotype. We therefore decided to assess eIF5 GDI activity directly. We monitored eIF5 GDI activity by incubating eIF2•BODIPY-GDP (20 nM) with purified eIF5 (0–80 nM) and recorded the fluorescence change upon dissociation of BODIPY-GDP from eIF2. GDP-binding to WT eIF2 was stabilized by eIF5 in line with previous observations implying that a 1:1 eIF2•GDP/eIF5 complex forms that stabilizes GDP binding to eIF2 (23). Importantly GDP binding to eIF2^βE189K was not stabilized by eIF5 even at excess eIF5 levels (Figure 5A). Hence the experiment shows that eIF2^βE189K eliminates eIF5 GDI activity and this finding represents the likely cause contributing to the original gcd6-F250L slow-growth suppression phenotype observed in sif285-1/sui3-E189K mutant cells (Figure 1).

To examine the impact of eIF5 on eIF2B GDF and GEF activities, we monitored BODIPY-GDP release from eIF2 pre-bound to eIF5 over a wide range of eIF2B concentrations. In line with previous observations (14) GDP release from WT eIF2 was inhibited at low eIF2B concentrations by eIF5 GDI. At higher eIF2B levels GDP release is unaffected by eIF5 because excess eIF2B can fully displace eIF5, nullifying GDI (compare blue [+eIF5] with black [–eIF5] in Figure 5B, left panel). In contrast, even at low eIF2B concentrations the kinetics of GDP-release from eIF2^βE189K are completely unaffected by the presence of eIF5 (compare green with red symbols in Figure 5B, zoomed right panel).

eIF2 is between 6-12–fold in excess over eIF2B in vivo (22,50). When eIF2B is mutated or eIF2α phosphorylated at serine 51, eIF2B activity is reduced such that it is limiting for translation initiation (14,41). In our assays at low levels of active eIF2B (below 8 nM eIF2B in Figure 5B, zoomed right panel) the rate of GDP release from eIF5/eIF2^βE189K is accelerated relative to eIF5/eIF2 WT (Figure 5B, right panel, compare green and blue symbols). Our assays contained 20 nM each of eIF2•GDP and eIF5, so eIF2B concentrations below 8 nM likely reflect factor ratios in vivo. The increased rate of GDP release from eIF2^βE189K observed with 5 nM or less eIF2B is statistically significant (points marked * or † in Figure 5B, right panel) and is fully consistent with the slow-growth suppression phenotype observed in cells where low eIF2B GEF activity normally limits growth-rate (Figures 1 and 2). To conclude, our data indicate that eIF2β-E189K eliminates eIF5 GDI activity and thereby facilitates faster rates of nucleotide exchange when eIF2B activity is limiting. This is consistent with our observations of enhanced protein synthesis rates and enhanced rates of cell growth in eIF2β-E189K/sif285-1 cells bearing defective eIF2B mutants with lower eIF2B GEF activity.

Physiologically eIF2B activity is reduced in vivo by phosphorylation of eIF2α at serine 51. A prediction of our genetic (Gcn^– phenotype) and biochemical results (loss of GDI activity) is that eIF2β-E189K should reduce the sensitivity of cells to very high levels of phosphorylated eIF2. We tested this hypothesis by transforming cells with plasmids expressing one of two constitutively-active alleles of the Gcn2 kinase (GCN2^C) that each cause high levels of eIF2(αP) and result in severe slow-growth (51). In accord with our predictions, eIF2β-E189K reverted the slow-growth of both GCN2^C mutants tested and lowered the aberrantly high expression of the HIS4-LacZ reporter seen in SUI3 GCN2^Ccells (Figure 5C). Our data demonstrate that eIF2β–E189 and GDI are critical for the normal tight control of translation and GAAC by eIF2(αP).