The eukaryotic translation initiation factor eIF2B is a multi-subunit complex with a crucial role in the regulation of global protein synthesis in the cell. The complex comprises five subunits, termed α through ε in order of increasing size, arranged as a heterodecamer with two copies of each subunit. Regulation of the co-stoichiometric expression of the eIF2B subunits is crucial for the proper function and regulation of the eIF2B complex in cells. We have investigated the control of stoichiometric eIF2B complexes through mutual stabilization of eIF2B subunits.
Our data show that the stable expression of the catalytic eIF2Bε subunit in human cells requires co-expression of eIF2Bγ. Similarly, stable expression of eIF2Bδ requires both eIF2Bβ and eIF2Bγ+ε. The expression of these subunits decreases despite there being no change in either the levels or the translation of their mRNAs. Instead, these subunits are targeted for degradation by the ubiquitin–proteasome system.
The data allow us to propose a model for the formation of stoichiometric eIF2B complexes which can ensure their stoichiometric incorporation into the holocomplex.
Many proteins function as components of multi-polypeptide complexes. Eukaryotic initiation factor 2B (eIF2B) is an excellent example; it comprises five distinct subunits, α–ε, and functions as the guanine nucleotide exchange factor (GEF) for eIF2 . Its ε-subunit (eIF2Bε) contains the GEF catalytic domain , whereas eIF2Bα plays a key role in the regulation of its activity .
The substrate of eIF2B, eIF2 is a heterotrimeric protein which (when bound to GTP) brings the initiator methionyl-tRNA, Met-tRNAiMet, to the 40S ribosomal subunit, to allow it to locate start codons in mRNAs . The subunits of eIF2B are related by sequence homology, dividing them into two groups based on subcomplex formation in yeast: the catalytic subcomplex, comprising eIF2Bγ and the catalytic eIF2Bε subunits; and the regulatory subcomplex, comprising eIF2Bα, β and δ [3–5]. eIF2B is arranged as a heterodecamer, comprising two eIF2B(βγδε) tetramers linked by an eIF2Bα homodimer . Recent modelling work by Bogorad et al.  has predicted that eIF2Bβ and δ form a heterodimer with a similar arrangement to the eIF2Bα homodimer. This was recently confirmed by Kuhle et al. , who solved the crystal structure of the eIF2Bβδ complex from Chaetomium thermophilum.
The physiological importance of eIF2B is underlined by the fact that mutations in the genes for its subunits cause an inherited neurological disorder termed leukoencephalopathy with vanishing white matter (VWM ; also termed childhood ataxia with central hypomyelination, CACH). Mutations in any of the five subunits of eIF2B can cause this disorder, although the highest proportion of mutations arise in EIF2B5, the gene encoding eIF2Bε . So far, it is unclear how mutations in eIF2B subunits lead to VWM as they exert diverse effects on the activity or integrity of human eIF2B [10–12]. For example, some mutations that lead to severe (e.g. pre-natal or immediate post-natal death) have no effect on the in vitro GEF activity of eIF2B .
In mammalian cells, three types of eIF2B complexes can be isolated: eIF2B(αβγδε)2 decamers, eIF2B(βγδε) tetramers and eIF2B(γε) heterodimers . The latter two complexes possess ∼50% and 20%, respectively, of the activity of the decamer . However, since eIF2Bα is required to confer sensitivity of the eIF2B complex to regulation by phosphorylation of eIF2 as a part of the integrated stress response, these less active complexes are immune to regulation by this pathway [13,14]. This leads to lower but relatively uncontrolled eIF2B activity, which is likely to be detrimental to the cell.
It therefore follows that it is crucially important to maintain the correct 1:1 ratio of subunits to achieve appropriate levels of eIF2B activity, as well as for the regulation of its activity via the phosphorylation of its substrate eIF2. We have previously shown that lower levels of eIF2Bγ result in reduced co-expression of eIF2Bε, and that this can be rescued by co-expression of eIF2Bγ fragments that interact with eIF2Bε . Furthermore, decreased expression of eIF2Bδ has been observed in yeast cells lacking eIF2Bβ . In the present study, we further explored these observations in order to understand the regulation of subunit stoichiometry and eIF2B complex formation in mammalian cells. We show that regulation of eIF2Bε and δ is via ubiquitination in the absence of their stabilizing subunits. Using the data from the present study, we propose a model for the sequential formation of the eIF2B subcomplex.
MATERIALS AND METHODS
Chemicals, plasmids, siRNA and antibodies
Chemicals used in the present study were purchased from Sigma, Fisher Scientific or Melford. Plasmids encoding myc- and His6-myc-tagged eIF2B subunits and their mutants have been described previously . siRNAs targeting EIF2B1, EIF2B2 and EIF2B3 as well as a non-silencing control were purchased from Dharmacon (GE Healthcare–catalogue numbers can be found in Supplementary Table S1). Antibodies against eIF2Bα (18010-1-AP), eIF2Bβ (11034-1-AP), eIF2Bδ (1132-1-AP) and myc-tag (60003-2-Ig) were purchased from Proteintech. The antibody against eIF2Bγ (sc-9980) was purchased from Santa Cruz Biotechnology. Anti-eIF2Bε (ARP61329_P050) was purchased from Aviva Systems Biology. The antibody against actin (A1978) was purchased from Sigma–Aldrich.
Cell culture and transfection
HEK293 cells were cultured as previously described . Plasmid transfections were carried out by the calcium phosphate method as previously described using plasmid quantities to provide roughly equal expression of each subunit . siRNA transfections were carried out using Lipofectamine RNAiMax (Thermofisher Scientific) by reverse transfection according to the manufacturer's protocol. Transfection of plasmids for re-expression of subunits following siRNA treatment was carried out using Genejuice (Merck) as outlined in the manufacturer's protocol.
Cell lysis and Western blotting
For standard Western blot analysis of lysates, cells were lysed in 20 mM Hepes–KOH pH 7.6 containing 10% (v/v) glycerol, 50 mM β-glycerophosphate, 0.5% Triton X-100, 50 mM KCl, 0.5 mM sodium orthovanadate, 14.3 mM 2-mercaptoethanol and EDTA-free protease inhibitors (Roche). Lysates were clarified by centrifugation at 4°C and protein concentration was measured by Bradford assay (Bio-Rad Laboratories) using BSA as a concentration standard. Equal amounts of protein were loaded on to 12.5% SDS/PAGE gels and transferred on to nitrocellulose membrane. Following blocking with either 5% (w/v) milk or BSA in TBS containing 0.2% Tween-20, membranes were probed with primary antibody either for 1 h at room temperature or overnight at 4°C. Following washing, membranes were probed with Dylight secondary antibodies (ThermoFisher Scientific) and imaged on a Licor Odyssey® using Imagestudio software (Licor).
HEK293 cells were seeded into 96-well plates and transfected as described above with combinations of subunits and the pGLp vector encoding destabilized firefly luciferase . Luciferase assays were carried out using the luciferase assay system and a Glomax instrument (Promega) following the manufacturer's instructions. The protein content of each well was determined by Bradford assay (Bio-Rad Laboratories) and used to normalize the luminescence of each well.
Polysome analysis of siRNA treated HEK293 cells was carried out as previously described . Each fraction was spiked with 1 ng of kanamycin mRNA from the Improm-II RT kit (Promega) to act as a control for qPCR analysis.
Following transfection, total mRNA was isolated from cells using the RNeasy RNA isolation kit (Qiagen). One microgram of total RNA was used to make cDNA using the Improm-II reverse transcriptase kit (Promega) with oligo dT primers according to the manufacturer's instructions. qPCR was carried out using Precision qPCR master mix (Primerdesign) and a StepOne plus instrument (ThermoFisher scientific) using a two-step procedure. Melt curves of each reaction were also carried out to ensure single products were present. Primer sequences for qPCR can be found in Supplementary Table S2.
Ubiquitination of eIF2B subunits was analysed using a method modified from Tatham et al. . HEK293 cells were transfected with plasmids expressing combinations of myc-tagged eIF2B subunits as described above, with the subunit of interest being His6-tagged, and HA-tagged ubiquitin. Thirty-six hours after transfection, the cells were washed with PBS and the medium was replaced. Cells were treated with either 10 μM MG-132 (Sigma) or the equivalent volume of DMSO and incubated for 16 h. The following day, 0.5 mM iodoacetamide (Sigma) was added to the media for 1 h. Cells were collected by scraping and centrifugation for 5 min at 3000 × g at 4°C, followed by two washes in ice-cold PBS containing 0.5 mM iodoacetamide. During the second wash, 5% of the cell suspension was removed, the cells were pelleted by centrifugation and the resulting pellet was lysed in SDS/PAGE sample buffer [20% (v/v) glycerol, 62.5 mM Tris/HCl pH 6.8, 7% SDS, 0.7 M 2-mercaptoethanol] as an input control. The remaining cell pellet was lysed in lysis buffer containing 6 M guanidine hydrochloride, 100 mM sodium phosphate pH 8.0, and 10 mM Tris/HCl pH 8.0. The cells were incubated at room temperature with rocking for 10 min to ensure complete lysis. Lysates were stored at −20°C for up to 1 month.
To carry out the ubiquitination pull-down assay, 5 mM imidazole and 5 mM 2-mercaptoethanol were added to each lysate. Lysates were sonicated at 20% power for 3×20 s on ice to shear genomic DNA, and centrifuged for 15 min at 3000 g to pellet any large insoluble material. Each lysate was added to 50 μl of Ni2+-NTA agarose (Qiagen), pre-equilibrated in lysis buffer, and incubated with rotation at 4°C overnight. The following day, the resin was pelleted by centrifugation for 2 min at 3000 g at room temperature and washed once with wash buffer 1 (6 M guanidine hydrochloride, 100 mM sodium phosphate pH 8.0, 10 mM Tris/HCl pH 8.0, 1% Triton X-100, 20 mM imidazole, 5 mM 2-mercaptoethanol), once with wash buffer 2 (8 M urea, 100 mM sodium phosphate pH 8.0, 10 mM Tris pH 8.0, 1% Triton X-100, 20 mM imidazole, 5 mM 2-mercaptoethanol) and thrice with wash buffer 3 (8 M urea, 100 mM sodium phosphate pH 6.3, 10 mM Tris pH 6.8, 1% Triton X-100, 20 mM imidazole, 5 mM 2-mercaptoethanol). Following the final wash, the resin was resuspended in SDS/PAGE sample buffer containing 250 mM imidazole, incubated at 70°C for 5 min and analysed by Western blot.
Mass-spectrometric analysis of ubiquitinated eIF2B subunits
Samples from the ubiquitination assay were separated on a 7.5% TGX precast gel (Bio-Rad Laboratories) and stained with InstantBlue stain (Expedeon). Bands corresponding to the ubiquitinated samples were excised and subjected to in situ trypsin digestion using the method of Shevchenko et al. . The resulting peptides were extracted and a ∼1–2 μg sample of each fraction was loaded on to a reverse phase trap column (Symmetry C18, 5 μm, 180 μm × 20 mm, Waters), at a trapping rate of 5 μl/min and washed for 10 min with buffer A prior to analytical nano-LC separation using a C18 Reverse phase column (HSS T3, 1.8 μm, 200 mm × 75 μm, Waters). The eluted peptides were fractionated over a 37 min continuous gradient from 1% acetonitrile+0.1% formic acid up to 80% acetonitrile+0.1% formic acid, at a flow rate of 300 nl/min. Eluted samples were sprayed directly into a Synapt G2-S mass spectrometer (Waters) operating in MSE mode. Data were acquired from 50 to 2000 m/z using alternate low and high collision energy (CE) scans. Low CE was 5 V and elevated CE ramp from 15 to 40 V. Ion mobility was implemented prior to fragmentation using a wave height of 650 m/s and wave velocity of 40 V. The lock mass Glu-fibrinopeptide [(M+2H)+2, m/z=785.8426] was infused at a concentration of 100 fmol/μl with a flow rate of 250 nl/min and acquired every 60 s.
The raw mass spectra were processed using ProteinLynx Global Server Ver 3.0 (Waters), and the data processed to generate reduced charge state and deisotoped precursor and associated product ion peak lists (using thresholds of 125 for low energy and 10 for high energy). These peak lists were searched against the human UniProt sequence database (obtained from UniProt 10/2014). A maximum of one missed cleavage was ‘allowed’ for tryptic digestion with a fixed modification for carboxyamidomethylation of cysteine and the variable modifications of oxidation of methionine and ubiquitination of lysine (detected by diglycine adducts).
Precursor ion and product ion mass tolerances were set at 150 ppm and 0.2 Da, respectively, and the false positive rate was set at 4%.
Statistical analysis was carried out in GraphPad Prism (Graphpad) using the tests described in individual figure legends.
Overexpression of eIF2B subunits
To examine whether omission of one subunit of eIF2B affected the expression levels of any of the others, we expressed a range of combinations of eIF2B subunits in HEK293 cells. Each subunit expressed with a myc-tag, allowing us to easily assess their relative levels of expression by immunoblot analysis of the resulting cell lysates. Omission of eIF2Bα did not reduce the expression of any of the other subunits (Figure 1A; data from multiple experiments are quantified in Figure 1B), consistent with the fact that the other four subunits can form heterotetrameric eIF2B(β–ε) complexes. Omission of any of the subunits from the eIF2B(βγδε) tetramer led to a decrease in the overall expression of the eIF2B subunits by ∼30% (Figure 1B). However, omitting eIF2Bβ led to significantly lower levels of expression of eIF2Bδ, but not eIF2Bα or eIF2Bε (Figure 1B, marked αγδε). Omission of eIF2Bγ caused significantly lower expression of eIF2Bε (Figures 1A and 1B). These changes were not a reflection of deficiencies in the mRNA levels for individual subunits. Indeed, where vectors for single subunits were omitted, we observed actually higher expression of mRNAs for the other subunits, including those most destabilized at the protein level (Figure 1C).
Overexpression of eIF2B subunits in HEK293 cells reveals a requirement of specific subunits for the stable expression of the complex
These data indicate that subunits of eIF2B probably stabilize one another; the fact that eIF2Bγ and eIF2Bε form a binary complex  likely explains why omitting the latter leads to reduced expression of the former. Furthermore, Bogorad et al.  have postulated that eIF2Bβ and δ form a heterodimer, formation of which may stabilize eIF2Bδ. Expression levels of eIF2Bα were not affected by omitting other subunits, probably because it is rather loosely associated with the other subunits (e.g. complexes lacking this subunit can readily be isolated [22–24] and eIF2Bα can be isolated as a separate homodimer ).
Overexpression of subunit combinations lacking eIF2Bβ, γ, δ or ε resulted in the reduced expression of all the subunits compared with eIF2B(αβγδε) or eIF2B(βγδε) (Figure 1B). We hypothesized that overexpression of the most active forms of eIF2B [i.e. either eIF2B(αβγδε) or eIF2B(βγδε)] would increase the rate of their own translation in the cell, whereas incomplete complexes would have a lesser effect. Therefore, we investigated the effect of eIF2B on the overexpression of proteins in cells to see if eIF2B promoted its own expression as well as that of other proteins. To do this, we transfected different combinations of eIF2B subunits used in Figure 1(A) alongside a vector containing firefly luciferase (fLuc). The mRNAs produced from the plasmids overexpressing eIF2B contain short, unstructured 5′-UTRs and a strong Kozak consensus sequence, are efficiently translated. The fLuc plasmid produces an mRNA with both of these features, and so was an ideal reporter. We subsequently lysed the cells and measured the change in luciferase activity in the cells normalized to total protein in the cells to measure the effect of eIF2B overexpression on translation of fLuc. We found that overexpression of eIF2B(αβγδε) or eIF2B(βγδε) led to a large increase in fLuc activity, whereas the other subunit combinations only caused a smaller rise in activity (Figure 1D). We tested levels of fLuc mRNA in these experiments and found no change, indicating that the increase in fLuc levels are due to its increased translation (Supplementary Figure S1). We previously noted that co-expression of translation factors with a reporter gives rise to much larger changes in reporter protein production than in total protein synthesis ; this may reflect differing levels of transient transfection of individual cells, so that cells that receive more reporter plasmid also receive higher levels of the translation factor plasmids under study and which affect reporter protein expression.
Interestingly, overexpression of eIF2B(βγδε) led to a greater increase in luciferase than eIF2B(αβγδε) despite the latter complex being more active than the former . We have previously observed increased levels of phosphorylated eIF2 following overexpression of eIF2B(αβγδε), thought to be caused by the complex protecting eIF2 from dephosphorylation , whereas we observe no such accumulation with eIF2B(βγδε) . The increased levels of phosphorylated eIF2 will inhibit the overexpressed eIF2B(αβγδε), reducing the effect of its overexpression, whereas eIF2B(βγδε) is immune to this inhibition. Therefore, overexpression of the most catalytically active combinations of eIF2B subunits leads to their own increased expression, presumably due to an increased rate of translation of the overexpressed mRNA.
Effects of depletion of endogenous eIF2B subunits
Having shown the destabilization of eIF2Bε and eIF2Bδ in the absence of eIF2Bγ and eIF2Bβ, respectively, for overexpressed subunits, we investigated whether this effect is also true for endogenous subunits. Therefore, we treated HEK293 cells with pooled siRNAs targeting the EIF2B1, EIF2B2 or EIF2B3 mRNAs encoding the eIF2Bα, β and γ subunits, respectively. As expected, siRNA against a given subunit depleted its mRNA levels but also decreased the expression of all mRNAs encoding the eIF2B subunits to ∼60% of the value observed for the non-silencing siRNA (siNS) (Figure 2A). When we examined the protein level of each of the subunits, we found that knockdown of eIF2Bα did not affect protein levels of the other eIF2B subunits (Figures 2B and 2C). Knockdown of eIF2Bβ or γ led to a reduction in expression of all the subunits. However, expression of eIF2Bδ was more reduced than the other subunits when eIF2Bβ was knocked down and expression of eIF2Bε was more reduced than the other subunits when eIF2Bγ was knocked down (Figures 2B and 2C). This is in agreement with the data for the overexpressed subunits (Figures 1A and 1B) and suggests that, despite the reduction in levels of all eIF2B mRNAs, some subunits are selectively destabilized at the protein level. Furthermore, when eIF2Bγ was knocked down, eIF2Bδ levels were again reduced more strongly than eIF2Bα, β and γ, suggesting that loss of eIF2Bγε complexes also destabilize eIF2Bδ levels.
siRNA-mediated knockdown of eIF2Bα, β or γ has destabilizing effects on endogenous eIF2B subunits
To confirm that knockdown of eIF2Bβ and eIF2Bγ is responsible for the changes in subunit expression, and not an off-target effect of the siRNA, we transfected cells treated with individual siRNAs followed by transfection with either an empty plasmid, or a plasmid containing myc-tagged versions of the subunit targeted by the siRNA (Figure 2D). Although the siRNAs used targeted the ORF of each subunit, we found we were able to re-express sufficient amounts of each subunit without needing to make silent mutations to protect them. The approximate levels of re-expression of the subunit compared with the endogenous subunit levels are shown in Supplementary Table S3. In all cases, the targeted subunit was re-expressed to at least endogenous levels or greater (Figure 2D). We found that re-expression of eIF2Bβ or eIF2Bγ led to recovery of levels of eIF2Bδ and eIF2Bε, confirming that the presence of these subunits was sufficient and necessary for the stabilization of eIF2Bδ and ε, respectively (Figure 2D).
Data for polysome profiles and qPCR
Having already shown that the decreases in eIF2Bδ and ε levels are not due to changes in mRNA levels, we investigated whether the rate of translation initiation was affected, for example by the stabilizing subunit interacting with the mRNA of the destabilized subunit.
To assess the effect of knocking down expression of individual subunits on translation initiation, we analysed the proportion of ribosomes in polysomes or non-polysomal material on sucrose density gradients (Figure 3A). siRNA knockdown of eIF2Bβ or eIF2Bγ caused a reduction in polysome levels and an increase in non-polysomal ribosomes (80S subunits) indicating an impairment of translation initiation. In contrast, knocking down eIF2Bα did not affect the proportions of polysomal or non-polysomal particles.
siRNA knockdown of eIF2Bα, β or γ does not affect translation initiation of eIF2Bδ or ε
We extracted RNA from individual fractions from the gradients and analysed the abundance of the EIF2B4 and EIF2B5 mRNAs, and actin as a control, to see if siRNA treatment affected the distribution of these mRNAs on the density gradient (Figure 3B). For example, a shift in the mRNAs to lighter polysomes would indicate impairment of the initiation of their translation. We found that siRNA treatment yielded no change in the distribution of these mRNAs, indicating that reduced expression of their protein products is not due to changes in translation of their mRNAs. Therefore, the mechanism leading to the destabilization of these subunits was likely to be post-translational.
eIF2B subunits are ubiquitinated
A major mechanism responsible for post-translational protein turnover in cells is ubiquitination, which targets proteins for degradation by the proteasome. In order to test this, we transfected HEK293 cells with His6-myc tagged eIF2Bδ or eIF2Bε along with myc-tagged eIF2B subunits in combinations that would either stabilize or destabilize the eIF2Bδ or eIF2Bε polypeptides. We co-transfected a plasmid containing HA-tagged ubiquitin in order to trace changes in ubiquitination. Twenty-four hours after transfection, the cells were treated with the proteasome inhibitor MG-132 overnight in order to prevent degradation of ubiquitinated proteins. The cells were harvested under denaturing conditions. We found that MG-132 treatment of the cells led to significant accumulation of both eIF2Bδ and ε regardless of the co-expressing subunits (Figure 4A, top panel; Figures 4B and 4C) suggesting that proteasome-mediated turnover of these subunits is an important determinant of their levels of expression.
Analysis of ubiquitination of eIF2Bδ and ε
In order to assess the ubiquitination of eIF2B subunits, we isolated the His6-tagged subunits on Ni2+-NTA agarose under denaturing conditions. We observed that in lysates of cells treated with MG-132, but only in the absence of eIF2Bβ or γ, there are additional slower-migrating bands for both eIF2Bδ and ε (Figure 4A, bottom panel). Reprobing the membrane for the HA-tagged ubiquitin confirmed that the higher molecular mass bands for eIF2Bδ and eIF2Bε were due to ubiquitination, which occurred mostly in samples where these subunits would be destabilized (lanes 4 and 8). However, we also observed some ubiquitinated eIF2Bδ in the sample with all five subunits (Figure 4A, lane 2). We suspect that this may be due to excess eIF2Bδ being produced when all five subunits are overexpressed since we observed a significant increase in the levels of eIF2Bδ following MG-132 treatment in both the absence and presence of eIF2Bβ, although the proportional increase in eIF2Bδ levels in the eIF2Bβ was greater (Figure 4B). When all five subunits are expressed, we test their expression levels by Western blotting for their N-terminal myc-tag. However, this only gives us an indication of the steady-state levels of each subunit. The accumulation of ubiquitinated eIF2Bδ even when eIF2Bβ is co-expressed suggests that excess eIF2Bδ is continually degraded as it cannot be incorporated into complexes.
In order to ascertain the sites of ubiquitination, we resolved the pull-downs by SDS/PAGE and cut out the bands corresponding to the ubiquitinated subunits and analysed these by MS, specifically looking for the diglycine signature generated from ubiquitinated lysine residues. We identified a single ubiquitination site at the N-terminus of eIF2Bδ and five sites in eIF2Bε (Figure 4D). Ubiquitination of Lys17 in eIF2Bδ has not been previously reported. Of the sites in eIF2Bε, ubiquitination of Lys103, Lys481 and Lys493 has been previously described [28,29] but Lys108 and Lys420 are novel.
Taken together, these observations indicate (i) that these eIF2B subunits are degraded via the proteasome and (ii) they are likely subject to ubiquitination.
Effect of VWM mutants of eIF2Bβ on the expression of eIF2Bδ
Mutations in the genes encoding the eIF2B subunits cause the autosomal recessive neurological disorder VWM. We have previously shown that certain mutations in eIF2Bβ affect the integrity of eIF2B complexes, in some cases severely . Since decreased association of these variant subunits might affect the stability of other eIF2B polypeptides, we investigated whether expression of these mutant forms of eIF2Bβ affected the co-expression of eIF2Bδ. We chose two mutations, G200V and P291S, which have been shown to affect eIF2B complex formation with different degrees of severity, and one, K273R, which does not affect complex formation . We co-expressed these mutants, along with wild-type eIF2Bβ and a control without eIF2Bβ alongside the other eIF2B subunits. We found no effect on expression of eIF2Bδ (Figure 5A, left-hand panel, quantified in Figure 5B). However, we have previously shown that a mutation in eIF2Bδ (A391D), which was not originally observed to affect eIF2B complexes , did do so in the absence of eIF2Bα . Therefore, we repeated the transfections in the absence of eIF2Bα and found that the G200V mutant, which causes the more severe complex disruption, shows an impaired ability to stabilize eIF2Bδ expression under these conditions (Figure 5A, right-hand panel).
The effect of eIF2Bβ VWM mutants on expression of eIF2Bδ
An explanation for the lack of destabilization of eIF2Bδ when co-expressed with the G200V or P291S mutations in eIF2Bβ could be due to the inter-subunit interactions these mutations disrupt. Neither Gly200 or Pro291 locates to the region thought to form the interface between eIF2Bβ and δ (Supplementary Figure S2, ), suggesting that they may disrupt the interaction of eIF2Bβδ with eIF2Bγε, which would still cause the complex to dissociate, since the eIF2Bβδ heterodimer is unstable in cells. When we overexpressed different combinations of eIF2B subunits, we observed that only the absence of overexpressed eIF2Bβ led to destabilization of overexpressed eIF2Bδ, in a similar way to disruption of the eIF2Bβδ heterodimer (Figures 1A and 1B). This was not the case when we omitted eIF2Bγ or eIF2Bε from the set of overexpressed subunits, although the interaction of the eIF2Bβδ heterodimer with the eIF2Bγε subcomplex would still be impaired under these conditions. Therefore, in these experiments employing overexpressed eIF2B subunits, if the mutation disrupts the interaction between eIF2Bβ and eIF2Bγε, we are less likely to observe a decrease in eIF2Bδ levels. In contrast, if the mutation disrupts the interaction between eIF2Bβ and δ, we would expect to observe a decrease in eIF2Bδ levels, similar to when eIF2Bβ is absent. Therefore, the lack of destabilization of eIF2Bδ co-expressed with these eIF2Bβ mutations is supported by the data in Figure 1.
The difference observed between the complexes in the presence and absence of eIF2Bα suggests that this subunit is responsible for stabilizing the interaction between eIF2Bβ and δ, in agreement with our previous data . In the presence of eIF2Bα, the eIF2Bβδ interaction may be more stable, slowing the rate at which eIF2Bδ is ubiquitinated and degraded. In the absence of eIF2Bα, this stabilization is not available, so eIF2Bδ is more readily degraded when the eIF2Bβδ heterodimer is unable to interact with eIF2Bγε. In both cases, eIF2B complexes could not be isolated in pull-downs, consistent with our previous observations . This may also explain the effect that we observed for the A391D mutation in eIF2Bδ . We show in Figure 2 that knockdown of eIF2Bγ affects the stability of eIF2Bδ, in addition to eIF2Bε, suggesting that under physiological conditions, the levels of eIF2Bγε are important for eIF2Bδ stability, and that disruption of the interaction between eIF2Bβδ and eIF2Bγε is enough to promote the degradation of endogenous eIF2Bδ (Figures 2B and 2C). In cells from VWM patients, the effects of overexpression are obviated. In such cells, the mutations of eIF2Bβ that affect complex formation may destabilize eIF2Bδ expression due to disruption of the eIF2Bβδ and eIF2Bγε interaction. We cannot therefore rule out that these mutations, and others affecting the integrity of eIF2B complexes, may also affect the levels of the individual subunits in cells.
Our data reveal a mechanism to ensure the co-stoichiometric expression of subunits of eIF2B. Given that all of the four larger subunits of this complex are required for the optimal GEF activity of eIF2B, this is likely to be of substantial physiological importance; underexpression of even one of the core non-catalytic subunits would impair overall cellular eIF2B activity and thus general protein synthesis, while leading to inappropriate up-regulation of mRNAs, such as that for ATF4, whose translation is enhanced when eIF2B activity is low .
Our findings lead us to propose the model shown in Figure 6 for the formation of the eIF2B complex in cells. Initially, eIF2Bε binds to eIF2Bγ to form heterodimeric catalytic subcomplexes. Excess eIF2Bε is targeted for degradation by ubiquitination, as supported by the data in Figures 1 and 2. Next, eIF2Bβ and δ interact with the eIF2Bγε heterodimer to form eIF2B(βγδε) tetramers. Excess eIF2Bδ, which has not been incorporated into these complexes, is targeted for degradation, due to reduced levels of either eIF2Bβ or eIF2Bγε. This interpretation is supported by the observation that endogenous eIF2Bδ is destabilized in the absence of both eIF2Bβ and eIF2Bγ. Finally, the tetrameric complexes are joined by an eIF2Bα homodimer to form decamers. Our previous data , and data from the Walter laboratory , which shows increased decamer formation caused by the ISR resistance drug ISRIB, suggest that eIF2B is constantly cycling between decameric and tetrameric forms in the cell.
Model for the stoichiometric assembly of the eIF2B holocomplex
In the present study, we have identified a number of ubiquitination sites in eIF2Bε, and one site in eIF2Bδ. A previous study identified several ubiquitination sites in rat eIF2Bε , a number of which are near a putative PEST motif from residues 452 to 475, predicted by epestfind (emboss.bioinformatics.nl/cgi-bin/emboss/epestfind). Mapping these sites, plus our own identified sites, on to our predicted structure of residues 41–482 of eIF2Bε  (Supplementary Figure S3) shows that the identified sites are preferentially on one face of the protein (red, blue and green highlighted residues). We have previously shown that mutation of Leu149 and Leu150 (highlighted in yellow on the structure) disrupts eIF2Bε binding to eIF2Bγ. These residues, while not on the surface of the protein, do lie towards the face containing the multiple ubiquitination sites we have identified. We therefore hypothesize that insufficient eIF2Bγ leaves this face of the protein exposed and available to be ubiquitinated, thereby targeting the subunit for destruction.
However, a different mechanism is likely to apply to eIF2Bδ. We identified only a single ubiquitination site, at its very N-terminus. The residue is distant from those involved in the predicted interaction with eIF2Bβ . This does not preclude the presence of other ubiquitination sites in the interacting region.
The importance of regulating eIF2B levels in the cell is highlighted by the data in Figure 1D, showing a large increase in translation in cells overexpressing eIF2B. Uncontrolled expression of eIF2B could led to large increases in cellular protein synthesis, which could promote aberrant cell growth leading, for example, to hypertrophy or hyperplasia. Previous studies complement these data by suggesting a role for this mechanism as a means of regulating eIF2B levels and thus mRNA translation in cells. For example, in their study showing decreased protein synthesis in C2C12 cells treated with simvastatin, Tuckow et al.  reported that eIF2Bε protein levels decrease with no fall in the corresponding mRNAs following treatment. Interestingly, they do see a decrease in levels of eIF2Bγ mRNA and protein, which may underlie the change in eIF2Bε protein levels, and thus the drop in protein synthesis. These data suggest that changes in the rates of transcription of the EIF2B2 and EIF2B3 genes (encoding eIF2Bβ and γ), in particular, may play a crucial role in determining the cellular levels of the eIF2B holocomplex.
Our data provide important new insights into the mechanisms by which human cells ensure appropriate levels of the subunits of a heteromultimeric protein, eIF2B, for which incorrect expression levels will affect not only the rate of overall protein synthesis, but also the translation of specific mRNAs (such as that for the transcription factor ATF4). Furthermore, our data also supply further information concerning the still poorly-understood mechanisms by which VWM mutations in EIF2B genes have an impact on the cellular protein synthesis machinery.
Noel Wortham contributed to the conception of the study and experimental design, carried out the majority of the experimental work and co-wrote the manuscript. Joanna Stewart and Sean Harris carried out the gradient qPCR and ubiquitination assays, respectively. Mark Coldwell contributed to the conception of the study and the experimental design. Christopher Proud obtained the funding, contributed to the conception of the study and the experimental design, and co-wrote the manuscript. All authors approved the manuscript.
We thank Drs Paul Skipp and Erika Parkinson (Southampton) for mass spectrometric analyses.
This work was supported by the UK Biotechnology & Biological Sciences Research Council [grant number BB/J007706/1 (to C.G.P.)].
Present address: South Australian Health and Medical Research Institute, North Terrace, Adelaide, SA 5000, Australia.