The eukaryotic translation initiation factor (eIF) eIF2B is a key regulator of mRNA translation, being the guanine nt exchange factor (GEF) responsible for the recycling of the heterotrimeric G-protein, eIF2, which is required to allow translation initiation to occur. Unusually for a GEF, eIF2B is a multi-subunit protein, comprising five different subunits termed α through ε in order of increasing size. eIF2B is subject to tight regulation in the cell and may also serve additional functions. Here we review recent insights into the subunit organization of the mammalian eIF2B complex, gained both from structural studies of the complex and from studies of mutations of eIF2B that result in the neurological disorder leukoencephalopathy with vanishing white matter (VWM). We will also discuss recent data from yeast demonstrating a novel function of the eIF2B complex key for translational regulation.
- eukaryotic initiation factor 2B (eIF2B)
- integrated stress response
- integrated stress response inhibitor (ISRIB)
- protein synthesis
- vanishing white matter
Ternary complex and the initiation of translation
The initiation of eukaryotic protein synthesis is a multi-step process requiring a large number of initiation factors. The function of these factors is to recruit mRNAs to the 40S ribosomal subunit, mediate the scanning of the ribosome along the mRNA to identify the appropriate start codon and to facilitate the recruitment of the 60S ribosomal subunit to allow translation to commence . A key factor in this process is the ternary complex, comprising the heterotrimeric G-protein eukaryotic translation initiation factor 2 (eIF2) bound to GTP and the initiator methionyl-tRNA (tRNAMeti) . Since the ternary complex provides both the means of identifying the start codon and the first amino acid of the translated polypeptide chain, its availability is thought to be rate-determining for translation initiation .
Following identification of the start codon, the GTP bound to eIF2 is hydrolysed to GDP by the GTPase-activator protein (GAP) eIF5. eIF2 is released from the initiation complex as inactive eIF2–GDP which is then recycled to eIF2–GTP by the guanine nt exchange factor (GEF) eIF2B. The activity of eIF2B is crucial for controlling the rate of translation in a cell (Figure 1A) .
eIF2B is a multimeric protein complex
eIF2B comprises five distinct polypeptides, termed α through ε in order of increasing size. The subunits have been grouped by sequence homology and complex formation in the budding yeast Saccharomyces cerevisiae into two groups (Figure 1B) [3–5]. The first group, comprising eIF2Bγ and ε, forms the catalytic sub-complex reflecting the presence of the catalytic domain of the complex at the C-terminal end of eIF2Bε. These two subunits share significant domain and structural homology, containing a nucleotidyl-transferase (NT) homology domain and an isoleucine-rich repeating motif region known as an I-patch, which forms a left-handed β-helical barrel .
eIF2Bα, β and δ form the regulatory sub-complex, so-called because it mediates the regulation of eIF2B by phosphorylation of its substrate eIF2 in response to cellular stress (Figure 1B) . The structure of eIF2Bα has been solved revealing an α-helical bundle towards the N-terminal end and a domain known as a Rossmann fold towards the C-terminal end (Figure 2A) . The significant sequence homology between the three subunits of the regulatory sub-complex suggests that this structure is shared between all three subunits (Figure 1B) [3,4].
Isolation of recombinant complexes shows three stable complexes in mammalian cells: eIF2B(αβγδε), eIF2B(βγδε) and eIF2B(γε) . These complexes differ in activity, with eIF2B(βγδε) and eIF2B(γε) exhibiting approximately 50% and 20% respectively the GEF activity of eIF2B(αβγδε) .
eIF2B and vanishing white matter
Mutations in the genes encoding the eIF2B subunits (EIF2B1-5, encoding eIF2Bα-ε respectively) cause the autosomal recessive inherited neurological disorder leukoencephalopathy with vanishing white matter [VWM, also known as childhood ataxia with central hypomyelination (CACH)] . Patients with this disease exhibit demyelination of the central nervous system white matter, resulting in degradation of the white matter neurons. MRI and post-mortem pathology analyses have shown the formation of large cavities in the brain caused by white matter degradation. Patients exhibit ataxia, spasticity and some mild mental retardation. The disease is distinguished by being both chronic and episodic. Patients will exhibit slow decline punctuated by episodes of very rapid decline caused by neurological stresses including head injury or infection with fever. Often patients will remain undiagnosed until such episodes occur .
‘Classical’ VWM is diagnosed in middle childhood and generally results in patient mortality within a few years of diagnosis. However, a wide variation in the spectrum of VWM severity has been observed, with some patients exhibiting mild, adult-onset disease with very mild symptoms and some patients exhibiting pre-natal VWM resulting in death at only a few months old .
A large number of VWM-associated mutations in EIF2B1–5 have been identified to date, with the majority observed in EIF2B5 . There is a rough association between the genotype of a patient and their disease severity, with some mutations resulting in particularly severe disease. In all cases reported to date, VWM only occurs if a patient is homozygous or compound heterozygous for mutations in the same gene . We and others have analysed the effects of a number of these mutations both on the GEF activity of the complex and its structural formation and substrate interaction [8,9,12–14]. Interestingly, we have found that a number of mutations affect neither activity nor complex formation, a particularly intriguing example being the A391D mutant of eIF2Bδ which results in severe neonatal disease. This suggests some mutations may affect novel functions of eIF2B .
Regulation of eIF2B activity
Direct regulation of the complex
eIF2B is a phosphoprotein and a number of phosphorylation sites have been shown to be important both basal eIF2B activity or for regulation of its function . Although a number of phosphorylation sites in all five subunits have been identified via mass spectrometric screens (see, e.g. www.phosida.org, www.phosphosite.org), sites shown to have an effect on eIF2B activity have only been observed in eIF2Bε. Two sites in the extreme C-terminal of eIF2Bε are required for basal eIF2B GEF activity . Furthermore, two sites on eIF2Bε have been identified that regulate eIF2B activity under different cellular conditions. The first, Ser535 in human eIF2Bε, is phosphorylated by GSK3 (glycogen synthase kinase 3) under conditions of reduced insulin availability and inhibits the activity of the complex . This inhibition requires a priming phosphorylation of Ser539 by the kinase DYRK (dual specificity tyrosine-phosphorylation-regulated kinase) . The second, Ser525, is phosphorylated under conditions of reduced amino acid availability by an unknown kinase and again leads to inhibition of the complex .
Regulation by the integrated stress response
eIF2B activity is inhibited following phosphorylation of the α-subunit of eIF2 at Ser51 in response to a variety of cellular stresses . These pathways, known collectively as the integrated stress response (ISR) involve the four kinases: GCN2 (general control nonderepressible 2), PKR (protein kinase RNA activated), PERK (PKR-like endoplasmic reticulum kinase) and HRI (heme-regulated inhibitor kinase). These kinases are activated in response to different cellular stresses including amino acid deprivation, viral infection and accumulation of unfolded proteins in the endoplasmic reticulum. eIF2 phosphorylation serves to reduce the rate of protein synthesis in the cell through decreasing ternary complex levels. Due to the presence of re-initiating short upstream ORFs (uORFs) in their 5′-UTR, the translation of certain mRNAs, such as that for ATF4 (activating transcription factor 4), is actually increased in response to eIF2α phosphorylation allowing the cell to formulate an appropriate response to the particular stress . Phosphorylation of eIF2 has also been observed in tumours under nutrient of hypoxic stress and has been shown to be cytoprotective and tumour promoting for these cells .
Phosphorylated eIF2 inhibits eIF2B activity through tight binding to the subunits of the regulatory sub-complex, eIF2Bα, β and δ, so that it becomes a competitive inhibitor of eIF2B . Studies in yeast have identified mutations that can disrupt this inhibitory binding and leave the eIF2B complex immune to inhibition [22,23]. Loss of eIF2Bα from the complex prevents inhibition by phosphorylated eIF2 [24,25]. In their paper presenting the structure of eIF2Bα, Hiyama et al.  identified a potential binding pocket for phosphorylated eIF2. We have recently studied mutations of eIF2Bα involved in VWM, including one, N208Y, which disrupts this binding pocket and did not see a difference in inhibition of the complex by phosphorylated eIF2 . This suggests that this pocket may not be crucial for the binding of phosphorylated eIF2.
Recently, a pharmacological compound known as ISRIB (integrated stress response inhibitor) has been identified that is able to inhibit the inhibition of eIF2B by phosphorylated eIF2, without dissociating the subunits from the complex [27,28].
Regulation by subunit composition
As noted above, incomplete eIF2B complexes display reduced GEF activity  and lack of eIF2Bα eliminates sensitivity to eIF2(αP) . Therefore, the level of expression of individual subunits within cells is crucial for determining both the activity of eIF2B and its ability to be regulated by the ISR. For example, cells expressing inadequate levels of eIF2Bα compared with the other subunits would have reduced eIF2B activity, but the complexes would not be susceptible to inhibition by phosphorylated eIF2. We and others have shown that the interaction of eIF2Bα with the rest of the complex is transient and that a heterogeneous mix of eIF2B complexes is present in the cell [28,29]. Therefore, control of eIF2Bα levels could fine-tune a cell's response to ISR stresses.
Evidence exists for a requirement of co-expression of particular eIF2B subunits. We have previously shown, through studies overexpressing the different subunits, that eIF2Bε is less well expressed in the absence of eIF2Bγ. Re-expressing fragments of eIF2Bγ that interact with eIF2Bε rescues eIF2Bε expression . Thus, it appears that levels of the subunits are mutually co-regulated.
eIF2B complex activity can also be affected by different subunit isoforms. Martin et al.  demonstrated that an alternative isoform of eIF2Bδ, with an alternative extended N-terminal region, deemed the complexes insensitive to regulation by phosphorylated eIF2. Other isoforms of eIF2B subunits have been predicted from mRNA libraries, but the occurrence and characteristics of these other isoforms have yet to be studied.
The structural organization of eIF2B
In light of the importance of the eIF2B complex in translational control and also its role in VWM, an understanding of the structural organization of the complex is crucial. A number of recent studies have sought to identify the interactions between the subunits [6,29,31,32].
eIF2B is a heterodecameric protein complex
Traditionally, eIF2B has been described as a heteropentameric protein. The publication of the structure of eIF2Bα suggested potential higher order structure since a homodimer was observed (Figure 2A) . Further work on eIF2Bα by us and Bogorad et al.  demonstrated that the complex does indeed exist as a homodimer in cells . Furthermore, we identified a VWM mutation, V183F, which lies on the predicted interface of the subunits and disrupts their interaction, confirming that the arrangement of the subunits in the crystal structure also occurs in cells (Figure 2A) .
The formation of eIF2Bα dimers suggested that the entire complex may dimerize. Our high MS analyses of purified eIF2B complexes revealed that eIF2B does exist as a heterodecamer, containing two copies of each subunit (Figure 2B). This approach also allowed us to explore the subunit arrangement by applying collision energy to dissociate individual subunits. This demonstrated that eIF2B comprises two eIF2B(βγδε) tetramers linked by an eIF2Bα homodimer to form the decameric complex (Figure 2B). The data also showed that each eIF2Bα monomer contacts both tetramers. Pulldown assays in cells and gel filtration of cell lysates confirmed these data. Investigation of the interaction of different complexes showed that eIF2 binds less well to the tetrameric complexes, which may explain the decreased activity of these complexes .
Further work by Bogorad et al.  examined all the subunits in the regulatory sub-complex. They showed that, unlike eIF2Bα, eIF2Bβ is unable to form homodimers. The homology between the three subunits suggests they have a similar structural core and so they proposed that eIF2Bβ and δ form a heterodimer within the complex . Although this has not been observed in subunits overexpressed in mammalian cells, presumably due to the instability of this complex, we have isolated tagged eIF2Bβδ heterodimers co-expressed in bacterial cells (Noel C. Wortham and Christopher G. Proud, unpublished data), suggesting that such complexes do occur.
Similar work showed that yeast eIF2B is also a decamer, but that loss of eIF2Bα resulted in an octameric complex eIF2B(βγδε). This difference may reflect large sequence insertions present in yeast eIF2Bγ and δ, compared with the mammalian subunits, which may form other links between these subunits in the complex, retaining interactions even in the absence of eIF2Bα .
Recent work on the ISR inhibitor ISRIB has demonstrated that this compound functions by targeting eIF2B and preventing its inhibition by phosphorylated eIF2, while still allowing GEF activity to occur [28,33]. Sekine et al.  identified eIF2Bδ as the subunit responsible for binding ISRIB and demonstrated that mutations of this subunit inhibited the effects of ISRIB. These residues are in the N-terminal region of the subunit, which is not homologous to eIF2Bα or β. Sidrauski et al.  showed that ISRIB increases the proportion of decameric complexes in the cell. Since ISRIB is a symmetrical molecule, it is likely that it binds to two eIF2Bδ subunits in the tetrameric complexes and strengthens the binding of eIF2Bα. It is unknown how this then relates to the change in regulation by phosphorylated eIF2.
Interactions identified by deletion and mutation studies
Further clues to the sites of interaction of the eIF2B subunits have been gleaned from mutation studies, both by using targeted mutations and by examining VWM mutations to determine their effects on eIF2B complex formation [6,8,9,26]. Within the catalytic sub-complex, we demonstrated that the region of eIF2Bε N-terminal to the I-patch domain is required for interaction with eIF2Bγ and that the I-patch domain of this subunit is required for the interaction of the catalytic sub-complex with the regulatory sub-complex, whereas eIF2Bγ required only the region N-terminal to the I-patch for interaction with the rest of the complex (Figure 1B) .
Studies on VWM mutations have also identified sites crucial for subunit interactions within the complex. As discussed above, V183F in eIF2Bα disrupts the formation of homodimers . We have recently shown that other VWM mutations in this subunit disrupt its association with the rest of the complex . We have identified a number of mutations that cause total disruption of the complex or dissociation of single subunits (Table 1). Interestingly, the Ala391 of eIF2Bδ, which is mutated to aspartic acid in one of the most severe forms of VWM, is in the region of eIF2Bα that is homologous to Val183 and so would be predicted to be at the interface between eIF2Bβ and eIF2Bδ [29,31]. In our original study on this mutation, we found no effect on complex formation or activity . However, further investigation demonstrated that this mutation does have an effect on eIF2B complexes in the absence of co-expressed eIF2Bα . Therefore, eIF2Bα homodimers seem likely to strengthen the interaction between eIF2Bβ and eIF2Bδ.
eIF2B as a GDI dissociation factor: a novel function of the complex
The complexity of the eIF2B holoprotein has suggested it may serve additional functions, beyond GEF activity that may be related to translational control. Recent studies by Graham Pavitt's laboratory on yeast eIF2B have identified a novel means of control of guanine nt exchange which prevents uncontrolled GDP to GTP exchange [34,35]. They demonstrated in yeast that eIF5, the GAP for eIF2, dissociates from the initiating ribosome in a complex with eIF2. Within this complex, eIF5 prevents the spontaneous dissociation of GDP, maintaining eIF2 in an inactive state, thus acting as a GDP-dissociation inhibitor (GDI; Figure 1A) . This activity provides an extra layer of translational control, since it ensures that eIF2 can only be activated by eIF2B and not spontaneously. Overexpression of eIF5, which promotes the formation of eIF2–eIF5 complexes, has been shown to antagonize eIF2B GEF activity .
In order to disrupt the eIF5–eIF2 complex to allow GDP–GTP exchange to occur, a separate factor, known as a GDI-dissociation factor (GDF), is required. Jennings et al.  demonstrated that eIF2B possesses this activity through assays showing the dissociation of eIF2 from immobilized eIF2. They examined a variety of subunit combinations to isolate the subunits required for this activity and showed that the catalytic sub-complex, eIF2B(γε) was necessary for GDF activity. Importantly, they showed that phosphorylation of eIF2 does not affect GDF activity. They also identified mutations of eIF2Bγ and ε that can inhibit GDF activity [34,35].
This novel GDF activity has potentially important consequences for VWM pathology. We have identified a number of VWM mutations that do not affect GEF activity or in some cases increase it. One of the mutations they study is the yeast equivalent of the G11V mutation in eIF2Bγ, which causes infantile VWM. Therefore, disruption of GDF activity is likely to be another means by which VWM mutations are able to affect translation.
An understanding of the regulation of eIF2B and how that relates to structure is important for comprehending its role in the regulation of protein synthesis and for identifying the pathological basis of mutations resulting in VWM. The identification of GDF activity and GTP binding by eIF2Bγ provide us with further clues for activities that could be affected by these mutations, whereas the increasing structural information will help us to place the mutations in context and to tease out their effects on the protein.
This work was supported by the UK Biotechnology & Biological Sciences Research Council [grant number BB/J007706/1] to C.G.P.
Translation UK 2015: Held at the University of Aberdeen, U.K., 7–9 July 2015.
Abbreviations: eIF, eukaryotic translation initiation factor; GAP, GTPase-activator protein; GDF, GDI-dissociation factor; GDI, GDP-dissociation inhibitor; GEF, guanine nt exchange factor; ISR, integrated stress response; ISRIB, integrated stress response inhibitor; NT, nucleotidyl-transferase; tRNAMeti, initiator methionyl-tRNA; VWM, vanishing white matter
- © 2015 Authors; published by Portland Press Limited