Protein synthesis (also termed mRNA translation) is a key step in the expression of a cell's genetic information, in which the information contained within the coding region of the mRNA is used to direct the synthesis of the new protein, a process that is catalysed by the ribosome. Protein synthesis must be tightly controlled, to ensure the right proteins are made in the right amounts at the right time, and must be accurate, to avoid errors that could lead to the production of defective and potentially damaging proteins. In addition to the ribosome, protein synthesis also requires proteins termed translation factors, which mediate specific steps of the process. The first major stage of mRNA translation is termed ‘initiation’ and involves the recruitment of the ribosome to the mRNA and the identification of the correct start codon to commence translation. In eukaryotic cells, this process requires a set of eIFs (eukaryotic initiation factors). During the second main stage of translation, ‘elongation’, the ribosome traverses the coding region of the mRNA, assembling the new polypeptide: this process requires eEFs (eukaryotic elongation factors). Control of eEF2 is important in certain neurological processes. It is now clear that defects in eIFs or in their control can give rise to a number of diseases. This paper provides an overview of translation initiation and its control mechanisms, particularly those examined in neuronal cells. A major focus concerns an inherited neurological condition termed VHM (vanishing white matter) or CACH (childhood ataxia with central nervous system hypomyelination). VWM/CACH is caused by mutations in the translation initiation factor, eIF2B, a component of the basal translational machinery in all cells.
- autism spectrum disorder (ASD)
- childhood ataxia with central nervous system hypomyelination (CACH)
- eukaryotic initiation factor 2B (eIF2B)
- protein synthesis
- vanishing white matter (VWM)
Translation initiation and its control initiation mechanism
Eukaryotic translation initiation typically relies on a defined set of protein–protein and protein–RNA interactions that select and prime an mRNA for protein synthesis. The goal is to correctly position an 80S ribosome complex at the mRNA AUG initiator codon with tRNAiMet (initiator methionyl-tRNA) bound so that the ribosome, aided by elongation factors, can then begin to decode the ORF (open reading frame) and synthesize the protein. The current model for this has been derived from mechanistic studies of the pathway using model mRNAs, e.g. globin mRNA, combined with genetic and biochemical experiments, especially those from model eukaryotes such as Saccharomyces cerevisiae . These studies have determined many details of the initiation pathway and highlighted that it is tightly controlled. tRNAiMet is delivered to the small (40S) subunit as part of a ternary complex with eIF2 (eukaryotic initiation factor 2) and GTP. Other factors (eIFs 1, 1A, 3 and 5) join to form a 43S pre-initiation complex. mRNA recruitment to the 43S complex requires the mRNA 5′-cap-binding complex, eIF4F (comprising eIFs 4E, 4G and 4A). This interacts with the cap of the chosen mRNA to direct binding of the 43S complex near the mRNA 5′-end . Locating the AUG codon normally requires scanning, i.e. movement of the ribosomal complex along the mRNA in search of the AUG codon. On AUG codon recognition by tRNAiMet, the eIF2-bound GTP is hydrolysed to GDP in a reaction stimulated by eIF5. This releases inactive eIF2•GDP from the 40S complex, probably still in a complex with eIF5 . Probably most initiation factors are released at this time. eIF5B, a second GTP-binding factor, stimulates 60S joining, in conjunction with eIF1A. This leaves the 80S ribosome with AUG-bound tRNAiMet at the P site ready to accept the first elongation factor–tRNA complex. An outline of this process is shown in Figure 1.
The role of eIF2B in initiation
eIF2 released from the ribosome is bound to GDP. GDP must be replaced by GTP to enable ternary complex formation. As eIF2 has a higher affinity for GDP , eIF2B functions to promote guanine nucleotide exchange (Figure 1). The eIF2•GTP product is not stable unless tRNAiMet joins to form the ternary complex . This is one of the rate-limiting steps of translation initiation. tRNAiMet is in excess, making the eIF2B-catalysed step rate-limiting for translation initiation. Compared with other GEFs (guanine-nucleotide-exchange factors) in other systems, eIF2B is remarkably complex containing five subunits. One reason for this complexity is probably the fact that eIF2B is regulated by diverse signals. As eIF2B mutations cause VWM/CACH (childhood ataxia with central nervous system hypomyelination) disease, this is described in further detail below.
Translational control: role of eIF2B inhibition
eIF2B is controlled by several mechanisms, particularly in response to cellular stress in diverse organisms from yeast to humans. One well-studied and evolutionarily conserved mechanism is via phosphorylation of the α subunit of eIF2. Several different protein kinases are activated in response to diverse stresses (for a review, see ). Once active, all phosphorylate an evolutionarily conserved serine residue (Ser51) in the N-terminal domain of the α subunit. Phosphorylated eIF2 has a higher affinity for eIF2B [7,8], but instead of promoting nucleotide exchange, the increased affinity inhibits the nucleotide-exchange reaction and therefore impairs the formation of the eIF2 ternary complex required for all translation initiation events (Figure 1).
The kinases that regulate eIF2/2B in this way include the HRI (haem-regulated inhibitor of translation) that primarily functions in reticulocytes to balance globin synthesis with haem availability and PKR (protein kinase R) whose transcription is IFN (interferon) induced and the protein is activated by binding dsRNAs (double-stranded RNAs), as part of the cellular innate antiviral immunity response. Many viruses have evolved diverse mechanisms to counteract PKR activation and enable productive infection of cells (see  for a review). High-level activation of HCR (haem-controlled repressor) and PKR in cells can dramatically inhibit almost all protein synthesis in the affected cells. In contrast, two other eIF2Ks (eIF2 kinases) appear to moderate protein synthesis more subtly, to enable activation of specific stress-responsive genes that are controlled at the level of translation. The PERK [PKR-like ER (endoplasmic reticulum) kinase]  is an ER resident protein that responds to the stress of accumulated unfolded proteins in the ER, PERK activation limits translation of ER destined proteins, while signals sent to the nucleus enhance the synthesis of ER chaperones, e.g. BiP (immunoglobulin heavy-chain-binding protein). Thus control of eIF2B activity is a key component of the cellular UPR (unfolded protein response). Finally, yeast S. cerevisiae Gcn2p (general control non-derepressible 2) and its mammalian counterpart mGCN2 (mammalian GCN2)  respond to limitation of one or more amino acids. Thus many stress signals can regulate eIF2B activity by signalling via eIF2Ks and phosphorylating eIF2α. mGCN2 has also been shown to regulate aspects of long-term memory (see below).
Down-regulation of eIF2B activity by these eIF2Ks enables two responses: first, overall protein synthesis is inhibited. This can be either global (affecting most proteins synthesized in cells) or (perhaps in the case of PERK) restricted to ER-localized synthesis of client proteins. Secondly, in an apparent paradox, phosphorylation of eIF2, and inhibition of eIF2B actually enhances the translation of specific mRNAs. This was first described for the yeast mRNA GCN4 (general control non-derepressible 4), where four short uORFs (upstream ORFs) in the 5′ leader region restrict the flow of translating ribosomes to the GCN4 ORF under nutrient complete conditions. However, on amino acid starvation, Gcn2p activation limits ternary complex levels, enabling ribosomes that have translated uORF1 to bypass the inhibitory uORFs (uORF2–uORF4) and to translate GCN4 instead. This elevates Gcn4p protein levels ~10-fold. Gcn4p activates transcription of many genes including 73-amino-acid biosynthetic enzymes that enable yeast cells to adapt to the amino acid starvation conditions imposed . The human mRNA for ATF4 (activating transcription factor 4, a transcription regulator) possesses two uORFs and its translation is regulated by a similar mechanism following PERK or mGCN2 activation [6,13,14]. Control of ATF4 translation appears critical in neuronal and glial cells as well as other tissues and this is explored further below.
Translational control: 4E-BPs (eIF4E-binding proteins)
A second major control point in translation initiation is through regulation of factors binding to the mRNA 5′-cap. eIF4E, the cap-binding factor, interacts with the scaffolding protein eIF4G to bridge interactions between mRNAs and the 40S ribosome. A family of regulatory factors termed 4E-BPs compete with eIF4G for binding to a shared surface on eIF4E and thereby prevent recruitment of ribosomes (Figure 1) . In many cases, phosphorylation of specific 4E-BPs regulates their recruitment to eIF4E. There are also now several examples where specific mRNAs are targeted for control by binding to additional mRNA-binding proteins, often recruited to the mRNA's 3′-UTR (3′-untranslated region), which form closed-loop complexes with specific 4E-BPs. These regulatory events have been studied extensively during early embryo development and more recently in neuronal cells, and can impart localized control of specific transcripts. This is important for establishing cell polarity.
The importance of translational control for neuronal function in health and disease
Many lines of evidence show that controlling mRNA translation, especially the translation of specific mRNAs, plays an important role in the function and adaptation of the nervous system. Controlling gene expression at the level of translation rather than transcription has important advantages, two of which are especially important for neural cells. First, switching on or off translation of specific mRNAs allows much more rapid response than would be possible if one modulated the transcription of the corresponding genes: the mRNA is already available, and can be turned on or off almost instantaneously. Secondly, it is especially relevant for neuronal cells, given the remoteness of synapses/dendrites/axons from the cell body (where the nucleus is), as regulating translation allows localized control over the production of specific proteins, e.g. at a particular synapse in response to incoming signals from neurotransmitters.
Translational controls in long-term memory
Kang and Schuman  showed that local protein synthesis is required for long-term memory. Several recent studies have found that factors known to regulate translation in many cell types have important roles in regulating learning and memory. These processes have been studied in a number of model organisms including Drosophila, Aplysia and in the mouse. Here, we will illustrate a few examples from mutant rodent studies and refer readers to other excellent recent reviews for more in-depth information [16,17]. Memories are typically divided into short-term, lasting a few hours, and long-term, which can last for an entire lifetime. Long-term memory formation requires new protein synthesis, and is now believed to be dependent on the translational regulation of specific mRNAs. Memory has been studied at the cellular level, by studying LTP (long-term potentiation), whereby synaptic strength changes are measured following electrical stimulation of excised brain slices. LTP appears to be an appropriate model for memory as there is good agreement between results from LTP experiments and other memory tests . LTP is divided into early and late phases: E-LTP (early LTP) requires pre-existing proteins, whereas L-LTP (late LTP) needs synthesis of new proteins. L-LTP is induced by several repetitions of electrical stimulations, each separated by a few minutes, and may persist for many hours. Whole animal studies monitor responses in many tests including fear conditioning tests and the Morris water maze. Studies have shown that factors regulating both the most commonly regulated steps of translation initiation are important for memory formation: the eIF2K mGCN2 and 4E-BP2 provide two clear examples. Both proteins are enriched in mammalian brains [11,19].
mGCN2 in dietary sensing and long-term memory
mGCN2 is the only eIF2K that is evolutionarily conserved from yeast to mammals. mGCN2 primarily responds to amino acid starvation in multiple organisms and tissues and in rodents was first shown to have a neuron-specific role in dietary sensing, monitoring the dietary content of essential amino acids. Half of the amino acids required for protein synthesis cannot be synthesized by mammals and amino acid levels are maintained by regular ingestion of balanced proteins in the diet. This leads mammals to reject diets deficient in essential amino acids within a 20 min time frame. The amino acid sensing occurs intracellularly within neurons of the anterior piriform cortex and requires mGCN2-mediated detection of de-acylated tRNA and phosphorylation of eIF2α to activate translation of the transcription factor c-Jun in rats and mice [20,21].
mGCN2 was subsequently shown to have an additional role in LTP and memory formation in the hippocampus. Thus, in LTP experiments, mGCN2−/− mice had aberrant responses to neural stimulation. A single stimulation induced a strong and sustained (L-LTP) in mGCN2−/− hippocampal slices, unlike wild-type mice. In contrast, a stimulation that elicited L-LTP in wild-type slices failed to stimulate L-LTP in mGCN2−/− slices. mGCN2−/− mice behaved similarly in a Morris water maze: after low-level training, their spatial memory was enhanced, but was then impaired after intense training; again the reverse of the response seen with wild-type mice. This has been linked to translational control of a second target mRNA: ATF4, as activated mGCN2 stimulates mRNA translation of ATF4, an antagonist of CREB (cAMP-response-element-binding protein). In the hippocampus of mGCN2−/− mice, the expression of ATF4 was reduced, while CREB activity was elevated . Similar results were found for heterozygous knock-in mice where the phosphorylation site on eIF2α (Ser51) was changed to alanine .
Regulation of eIF4E: memory and disease
The second commonly regulated step in protein synthesis initiation is the mRNA selection step. This is controlled by the mRNA 5′-cap-binding factor eIF4E, which interacts with eIF4G to enable ribosome recruitment. Several studies show that this step is also critical for memory and disease. Mice with a homozygous knockout mutation in the 4E-BP, 4E-BP2, show defects in learning and memory that are similar to mGCN2−/− mice described above. For example, 4E-BP2-knockout mice (4E-BP2−/−) displayed impaired L-LTP and spatial learning in both Morris water maze and contextual fear conditioning. The results are consistent with loss of an inhibitor, resulting in activation of protein synthesis . Further experiments have extended the range of memory types affected by the 4E-BP2−/− mutation to include motor memory, working memory and associative memory for aversive taste . Thus mutations affecting regulation of both major control points for translation initiation (either ternary complex formation or mRNA selection) impact on the stimulus required to elicit long-term memories.
Further evidence for the importance of appropriate regulation of eIF4E for normal brain development comes from a number of lines of evidence linking the dysregulation of eIF4E to autism [ASD (autism spectrum disorder); OMIM no. 209850]. The fragile X mental syndrome (OMIM no. 300624) arises due to expansion of CGG repeats within the FMR1 gene, which encodes the FMRP (fragile-X mental retardation protein). This can result in the loss of expression of FMRP, which is normally found in all tissues but at high levels in the brain (and testes). Loss of FMRP leads to the fragile X syndrome, which is characterized by defects in the maturation of synapses and is associated with features of autism/ASD. FMRP is an RNA-binding protein that is thought to repress the translation of specific mRNAs, although the mechanism by which it exerts this effect has been elusive . Recent studies have shown that FMRP interacts with a protein termed CYFIP1 (cytoplasmic FMRP interacting-protein 1), which also binds eIF4E . It binds to eIF4E in such a way that it blocks eIF4E's ability to interact with the scaffold factor, eIF4G, thereby interfering with the recruitment of ribosomes to the mRNA. This both provides a mechanism by which FMRP can repress the translation of specific mRNAs and indicates that correct control of eIF4E is important for normal brain development.
As discussed above, the activity of eIF4E is also inhibited by 4E-BPs: these proteins are phosphorylated by mTORC1 [mTOR (mammalian target of rapamycin) complex 1], leading to their release from eIF4E, thereby activating eIF4E (Figure 1). The condition TSC (tuberous sclerosis complex; OMIM no. 191100) arises due to loss or loss of function of the TSC proteins that inhibit mTORC1 signalling and thus normally repress eIF4E function. TSC is associated with a number of neurological features often including autism .
A recent report describes a chromosome translocation which occurs in children with autistic symptoms, from two unrelated families, and which results in a nucleotide insertion in the promoter of the EIF4E gene . This enhances the activity of the EIF4E promoter and binding of a transcription factor, raising the so far unsubstantiated possibility that these children express elevated levels of eIF4E. Taken together, these findings suggest that the dysregulation of eIF4E (increased activity) contributes to the development of autistic features in TSC patients.
Translation elongation and LTD (long-term depression)
Like LTP, LTD (another type of synaptic plasticity) also requires ongoing protein synthesis. But which protein products are important in these processes? A good deal of attention has focused on Arc/Arc3.1, whose synthesis is induced by activation of NMDA (N-methyl-D-aspartate) receptors. Arc/Arc3.1 localizes at active synapses and is involved in regulating the expression of AMPA (α-amino-3-hydroxy-5-methylisoxazole-4-propionic acid) receptors and in LTD . FMRP (which affects translation initiation) plays a role in regulating the translation of the Arc/Arc3.1 mRNA . Recent results demonstrate that the regulation of translation elongation is also important for proper control of Arc/Arc3.1 expression.
eEF2 (eukaryotic elongation factor 2) is inhibited by phosphorylation catalysed by a specific CaMK (Ca2+/calmodulin-dependent protein kinase), eEF2K (eEF2 kinase). Phosphorylation of eEF2 will decrease the rate of elongation  and, since eEF2 is needed for all elongation effects, this should decrease the translation of all mRNAs. In mice lacking eEF2K, both the regulation of Arc/Arc3.1 translation and the development of LTD are disrupted . How can this be explained? It is suggested that decreasing the rate of elongation allows inefficiently translated mRNAs to acquire ribosomes and undergo translation, because elongation rather than initiation becomes the limiting process . In support of this idea, treating cells with low concentrations of cycloheximide, which also decreases the rate of elongation, leads to increased synthesis of certain proteins , including that of Arc/Arc3.1 . Neurotransmitters may activate eEF2K by triggering rises in Ca2+-ion levels, thus inducing the phosphorylation of eEF2 and allowing the up-regulation of the translation of mRNAs such as that encoding Arc/Arc3.1. It will be important to identify other mRNAs that are regulated by similar mechanisms.
eIF2B mutations cause VWM/CACH disease
VWM/CACH (OMIM no. 603896) is caused by mutations in the genes for eIF2B. As described above, eIF2B is generally termed a GEF. It mediates the conversion of the inactive form of its substrate, eIF2•GDP, into the active form eIF2•GTP by stimulating the release of GDP, thus allowing GTP to replace it. This step of translation initiation is subjected to control (Figure 1 and text above) and this may be critical to understanding why eIF2B mutations cause this disease. eIF2B consists of five non-identical subunits α–ε of which the largest, eIF2Bε, contains the catalytic domain [34,35]. eIF2B's five subunits form two distinct types of the subcomplex: eIF2Bγ and eIF2Bε, which show extensive mutual sequence similarity, form a catalytic subcomplex that displays high GEF activity, but is not regulated by phosphorylated eIF2 (eIF2[αP]) . The other three subunits, eIF2α, eIF2Bβ and eIF2Bδ (which also display sequence similarities), confer control by eIF2[αP] and are thus termed the regulatory subcomplex .
The activity of mammalian eIF2B can also be regulated in other ways, including through changes in its own phosphorylation, e.g., in response to hormones such as insulin [through the GSK3 (glycogen synthase kinase 3) , which phosphorylates eIF2Bε] or amino acids, which regulate the phosphorylation of a different site in eIF2Bε . In yeast, metabolites termed fusel alcohols also control eIF2B activity , although the mechanism of regulation is not yet clear. Finally, the response to volatile anaesthetics has implicated involvement of eIF2B, both in yeast and in the mammalian systems studied. However, again, the mechanisms are not yet fully understood [39,40].
Linkage analysis of patients with VWM/CACH and their families in the laboratory of Professor Marjo van der Knaap (Free University Amsterdam, Amsterdam, The Netherlands) led to the initial discovery that the disease was caused by mutations in EIF2B5, the gene encoding eIF2Bε . The same group then demonstrated that, in fact, mutations in the genes for any eIF2B subunit could lead to this disease . Subsequent studies in laboratories across the world have uncovered many more mutations that cause VWM/CACH in diverse human populations. Although many mutations in EIF2B genes other than EIF2B5 are now known to cause this condition, such mutations arise most frequently in EIF2B5. Tables 1–3 summarize the extent of published information available about eIF2B mutations that cause VWM/CACH. To date 120 different mutations have been described; 83% (99 different mutations) are missense mutations that cause a single amino acid to be altered to another amino acid (Tables 1 and 2) and the remainder are a mixture of premature nonsense mutations, ones that cause a frame shift and others that alter splicing (Tables 1 and 3). All affected individuals have two altered copies of a single EIF2B gene (autosomal recessive inheritance). No one individual has yet been found with changes in two different EIF2B genes. At least one mutated allele must be a missense change. This is not surprising since eIF2B performs an essential function and all subunits are likely to be essential in humans. In yeast all genes except GCN3, encoding eIF2Bα, are essential.
One feature of the disease is that there is a wide variation in clinical presentation  from severe congenital forms  and those that manifest in early infancy  to classical childhood onset and milder forms where the disease is not recognized until adulthood [46,47] (Table 4). In general, the earlier the disease manifests, the more rapid it progresses. The disease is always fatal. Diagnosis is now by MRI (magnetic resonance imaging), which reveals a diffuse cerebral hemispheric leucoencephalopathy in which increasing areas of the abnormal white matter have a signal intensity close to that of CSF (cerebral spinal fluid) . Genetic diagnosis is also possible, but due to the wide range of potential mutations described (Tables 2 and 3) this is not necessarily a trivial task. Many of the case reports describing the identification of mutations also report clinical progress enabling a correlation between individual mutations and disease phenotype (Tables 2 and 4) for many described mutations. However, in most of the cases, the mutations are not homozygous, making it difficult to assign a clinical score unambiguously.
eIF2B subunits are homologous at the primary sequence level to other proteins whose three-dimensional structures have been determined. As these proteins are mainly enzymes involved in other pathways, the significance of this similarity for eIF2B function is not clear. Nevertheless, the identified similarity has enabled the generation of structural models for each individual eIF2B subunit. We used a subset of models available at MODBASE  to generate models of each eIF2B subunit and then indicated on the secondary structure cartoons the location of each missense mutation (Figure 2). This clearly demonstrates that mutations can fall anywhere within the largest subunit eIF2Bε, except the primary catalytic domain, which remains largely spared. Although fewer in number, the mutations in the homologous eIF2Bγ subunit are found in both structural domains. In contrast, the mutations found in the other subunits do show some clustering. All three subunits share extensive homology with each other and all α and β, and most δ, mutations cluster within the C-terminal globular domains. In addition, the eIF2Bδ mutations are predicted to cluster along one side of the protein only (Figure 2). The significance of the clustering is not yet clear; however, one idea based on the available biochemical evidence (see below) is that some of these mutations alter residues critical for subunit interactions within eIF2B. Further work is required to address whether the idea is valid.
Effects of VWM mutations on eIF2B structure and function
Richardson et al.  studied the effects of a number of mutations in EIF2B2 and EIF2B5 on the activity and assembly of the eIF2B complex. By exploiting the similarity between eIF2B subunits from diverse eukaryotic species (see, for example, Table 2), the Pavitt group was able to bring their knowledge and tools developed to analyse yeast eIF2B to study the effects of selected mutations on eIF2B in a yeast cellular model system. The genetic system available enabled the removal of the endogenous wild-type protein and its replacement with a single mutant. In general, the EIF2B5 mutations studied had modest effects on global translation and cell growth. Biochemical studies showed that several alleles reduced the steady-state expression of eIF2Bε protein. This was assumed to reduce the number of functional complexes available in cells as this defect was sufficient to promote an altered stress response, by activating translation of GCN4 and its transcriptional target HIS3. Thus, in these cells expressing only VWM/CACH mutant, eIF2B gene expression was reprogrammed as if the cells were responding to the stress of amino acid starvation, even when amino acids were provided . Two of the EIF2B2 mutations studied (equivalent to V316D and G329V) did not alter the steady-state expression of protein, but did cause reduced eIF2B complex formation or stability. In the V316D mutant (yeast eIF2Bβ[V341D]), this complex instability was extreme and resulted in reduced measured nucleotide exchange activity that was assumed to be responsible for the observed severe global translation and growth inhibition as well as constitutively elevated expression of GCN4 . This study concluded that VWM/CACH mutations reduced eIF2B activity and that reduced numbers of functional eIF2B complexes was a likely cause for the disease.
Li et al.  studied the effects of a number of mutations in EIF2B2 and EIF2B5 on the activity and assembly of the eIF2B complex, creating the mutations in the human cDNAs and studying the effects in human cells. In agreement with Richardson et al. , they found that the V316D mutation in eIF2Bβ affected the assembly of the eIF2B holocomplex. Only the T91A mutation in eIF2Bε actually impaired formation of eIF2B complexes, but all the EIF2B5 mutations tested decreased GEF activity, the deficit varying between approx. 30% and >80%. Despite residing in a subunit of the regulatory subcomplex, not the catalytic one, the eIF2Bβ mutation E213G also reduced eIF2B activity (by approx. 50%) . For V316D, its defect in complex formation gives rise to an apparent substantial (>95%) decrease in GEF activity when the activity of complexes was assessed.
More recent data from the Proud group have extended this analysis to additional mutations including the ones in other eIF2B subunits, including mutations associated with the most severe (congenital) forms of the disease. These analyses have revealed two important points. First, certain VWM mutations (in EIF2B5) actually result in increased GEF activity, rather than (partial) loss of this function. Secondly, analysis of two mutations in EIF2B4 that both give rise to congenital disease has shown dramatically different effects on eIF2B function. Whereas R483W mutation precludes formation of eIF2B complexes and therefore causes a major reduction in eIF2B activity, the A391D mutation has no effect on either of the parameters (X. Wang, R. Liu and C.G. Proud, unpublished work). Thus there is no correlation between the nature or extent of defects in these aspects of eIF2B function and the severity of the disease they give rise to. One possible explanation for this surprising observation could be that certain mutations affect the behaviour of eIF2B in ways that are not apparent from these biochemical assays. For example, it has been suggested that eIF2B may play additional roles in the translation initiation process, which are not tested in these assays .
One approach to assessing the effects of VWM/CACH mutations in a cellular context is to express mutant eIF2B in a cell type relevant to the pathology of the disease. Since this is a recessive trait, in order to test the effects of this, one first needs to eliminate the cells’ own wild-type eIF2B. Kantor et al.  used short interfering RNAs to knock down expression of endogenous eIF2Bε in a rat oligodendroglial-derived cell line and then expressed the R195H mutant. Cells expressing eIF2Bε[R195H] showed an enhanced ER-stress response even under basal conditions, as judged by the increased expression of ATF4, GADD34 (growth-arrest and DNA-damage-inducible protein 34; which is transcriptionally up-regulated by ATF4) and proteins involved in protein folding within the ER such as BiP and protein disulfide-isomerases. In response to ER stress induced by thapsigargin, there was an exaggerated induction of these proteins as compared with the control cells. This could suggest that this VWM/CACH mutant elicits an ER-stress response when expressed in this cell type: however, two limitations with this study are that the eIF2Bε[R195H] mutant was expressed at much higher levels than the endogenous protein, and that no control was performed in which wild-type human eIF2Bε was expressed at similar levels.
Biochemical studies from patient-derived cells and tissues
An alternative approach adopted by several teams has been to study protein synthesis, stress responses and other parameters in cells derived from VWM/CACH patients. Owing to the difficulty in obtaining brain material from such patients, initial studies utilized more readily available cell types such as fibroblasts and lymphocytes. Two studies showed that eIF2B activity was reduced in lymphocytes from VWM/CACH patients [54,55], without effects on the expression levels of the eIF2B subunits or, in one study, on overall rates of protein synthesis. The results of Fogli et al.  suggested a correlation between the degree of defect in eIF2B activity and the age of onset of the disease. van Kollenburg et al.  observed a blunting of the increase in eIF2α phosphorylation normally observed in response to heat shock, which may be relevant given that deterioration of patients’ symptoms often occurs after an episode of fever. One possible explanation is that the induction of GADD34 (referred to above) is expected to enhance protein phosphatase activity against eIF2[αP], thereby diminishing the increase in phosphorylation. Thus VWM/CACH mutations may both induce an enhanced basal stress response and blunt changes in the phosphorylation of one component involved in stress responses, eIF2α.
Data from fibroblasts obtained from VWM/CACH patients also indicate an enhanced stress response, even though eIF2B activity barely differed between wild-type and VWM/CACH cells . The inhibition of protein synthesis caused by ER stress was also similar for wild-type and mutant cells. Nonetheless, as alluded to already, in VWM/CACH cells, ER stress caused a greater induction of ATF4 and its target CHOP [C/EBP (CCAAT/enhancer-binding protein)-homologous protein] than in wild-type cells. Thus VWM/CACH cells may be predisposed towards a stronger stress response (‘hypersensitized’) as compared with normal cells. The significance of this for the aetiology of VWM/CACH remains to be established, but it could fit well with the observed association between onset or worsening of the disease and stresses such as infections with fever or head trauma. It is thus of particular interest that brain-derived material from VWM/CACH patients also shows increased levels of ATF4 and CHOP, indicators of an activated UPR .
In order to investigate cells more related to the pathology of the disease, Dietrich et al.  developed cultures of cells from the brain of one VWM/CACH patient. These cells contained many oligodendrocytes plus neural precursor cells. Perhaps the most important observations stemming from these studies are (i) that the oligodendrocytes appeared normal, but (ii) that there was an evident defect in the production of astroglial cells from progenitors, as judged from the lack of expression of the astrocytic protein GFAP (glial fibrillary acidic protein), and (iii) that the astrocytes that did form had an aberrant morphology with elongated arms and flatted feet. The results derived from studies of a single patient strongly implicate a deficiency in astroglial cells as a contributing factor to the clinical features of VWM/CACH. There are some similarities between the Dietrich study and the findings of others. A study by Rodriguez et al.  performed before the involvement of eIF2B was determined also observed an imbalance between oligodendrocyte and astrocyte numbers when two patients were analysed post mortem. It has also been found that mice that lack GFAP show features similar to those of VWM/CACH, such as sensitivity to head injury. A separate leucodystrophy with variable age of onset, Alexander disease (OMIM no. 203450), is caused by mutations in the GFAP gene .
However, the importance of oligodendrocyte dysfunction has also been highlighted. Rodriguez et al. found that myelin protein levels and lipid content were reduced , while Wong et al.  described oligodendrocytes with ‘foamy’ cytoplasm. More recently, in a search for a potential diagnostic biomarker for the disease, Vanderver et al. [62,63] analysed CSF samples from patients and controls and identified a reproducible and specific decrease in the ratio of asialotransferrin to transferrin in patient samples. Transferrin has roles in oligodendrocyte maturation and homoeostasis, iron metabolism and oxidative stress. The significance of these last observations for disease pathology remains unclear. However, as the authors observe, these findings raise the possibility of a biochemical diagnostic assay for CACH/VWM disease.
A link between eIF2B and multiple sclerosis?
Multiple sclerosis is a chronic disease of the CNS (central nervous system) with a high level of incidence. The disorder is characterized by inflammatory destruction of oligodendrocytes and myelin sheaths. Some of these features are therefore shared with inherited leucodystrophies including VWM/CACH. IFNγ, a cytokine, is found to be associated with multiple sclerosis inflammation, but its role remains unclear. A recent study using the mouse model for multiple sclerosis, EAE (experimental autoimmune encephalomyelitis), found that adding IFNγ to the CNS activated the eIF2K PERK in oligodendrocytes. This is expected to inhibit eIF2B activity. IFNγ protected oligodendrocyte demyelination and axon damage effects that were diminished in PERK−/− mice . Therefore the integrated stress response prevents demyelination by protecting mature oligodendrocytes against immune-mediated damage. These observations are apparently at odds with those described above that VWM/CACH mutant cells may both induce an enhanced basal stress response and blunt changes in the phosphorylation of eIF2α. In addition, genetic studies examining links between multiple sclerosis and EIF2B5 mutations have not identified EIF2B5 mutations in multiple sclerosis patients [65,66]. Perhaps these differing roles of the UPR can be reconciled if it has differing roles in the developing CNS and in mature oligodendrocytes. Hence, during CNS white matter development, elevated UPR as observed in VWM/CACH studies could be deleterious to glial cell functions. However, once the tissue has formed, elevated UPR in mature oligodendrocytes could protect against multiple sclerosis demyelination. It is clear that further studies will be required to resolve any clinical relevance between eIF2B and multiple sclerosis.
There has been a recent very rapid growth in knowledge about the key role that translational control mechanisms play in neurological processes, such as learning and memory. This suggests that there is much more to learn about the control mechanisms that regulate translation in neuronal cells and about the mRNAs that are regulated in such situations. In the specific case of VWM/CACH, there is a clear need to understand how mutations in eIF2B affect neuronal cell functions, especially within oligodendrocytes. Several research groups are now engaged in developing mutant mouse models expressing distinct VWM/CACH mutant forms of eIF2B, which will hopefully bring new insight into the role of eIF2B in the development of glial cells and the early stages of the disease. Through such knowledge, it may ultimately prove possible to devise ways to manage this serious disease and also to provide insights into common neurological conditions such as multiple sclerosis.
Work in our laboratories is sponsored by grants to G.D.P. from the Biotechnology and Biological Sciences Research Council [grant numbers BB/D000106/1, BB/E002005/1, BB/G012571/1 and BB/F013272/1] and The European Leukodystrophy Association [grant number ELA 2008-03915] and to C.G.P. from the Wellcome Trust [grant number 086688/Z/08/Z], Biotechnology and Biological Sciences Research Council [grant number BB/G008396/1], British Heart Foundation [grant number PG/08/099/26124], European Union and Royal Society [grant number 502011.K501].
Gene Expression in Neuronal Disease: Biochemical Society Focused Meeting held at University of Cardiff, Cardiff, U.K., 16–18 July 2009. Organized and Edited by Nicola Gray (MRC Human Reproductive Sciences Unit, Edinburgh, U.K.), Lesley Jones (Cardiff, U.K.) and Ian Wood (Leeds, U.K.).
Abbreviations: ASD, autism spectrum disorder; ATF4, activating transcription factor 4; BiP, immunoglobulin heavy-chain-binding protein; CACH, childhood ataxia with central nervous system hypomyelination; CHOP, C/EBP (CCAAT/enhancer-binding protein)-homologous protein; CNS, central nervous system; CREB, cAMP-response-element-binding protein; CSF, cerebral spinal fluid; 4E-BP, eIF4E-binding protein; eEF, eukaryotic elongation factor; eEF2K, eEF2 kinase; eIF, eukaryotic initiation factor; eIF2K, eIF2 kinase; eIF2[αP], phosphorylated eIF2; ER, endoplasmic reticulum; FMRP, fragile-X mental retardation protein; GADD34, growth-arrest and DNA-damage-inducible protein 34; Gcn2p, general control non-derepressible 2; GCN4, general control non-derepressible 4; GEF, guanine-nucleotide-exchange factor; GFAP, glial fibrillary acidic protein; IFNγ, interferon γ; LTD, long-term depression; LTP, long-term potentiation; L-LTP, late LTP; mGCN2, mammalian GCN2; mTORC1, mammalian target of rapamycin complex 1; ORF, open reading frame; PKR, protein kinase R; PERK, PKR-like ER kinase; tRNAiMet, initiator methionyl-tRNA; TSC, tuberous sclerosis complex; uORF, upstream ORF; UPR, unfolded protein response; VWM, vanishing white matter
- © The Authors Journal compilation © 2009 Biochemical Society