A Yeast Purification System for Human Translation Initiation Factors eIF2 and eIF2Bε and Their Use in the Diagnosis of CACH/VWM Disease

Recessive inherited mutations in any of five subunits of the general protein synthesis factor eIF2B are responsible for a white mater neurodegenerative disease with a large clinical spectrum. The classical form is called Childhood Ataxia with CNS hypomyelination (CACH) or Vanishing White Matter Leukoencephalopathy (VWM). eIF2B-related disorders affect glial cells, despite the fact that eIF2B is a ubiquitous protein that functions as a guanine-nucleotide exchange factor (GEF) for its partner protein eIF2 in the translation initiation process in all eukaryotic cells. Decreased eIF2B activity measured by a GEF assay in patients’ immortalised lymphocytic cells provides a biochemical diagnostic assay but is limited by the availability of eIF2 protein, which is classically purified from a mammalian cell source by column chromatography. Here we describe the generation of a recombinant expression system to produce purified human eIF2 from yeast cells. We demonstrate that human eIF2 can function in yeast cells in place of the equivalent yeast factor. We purify human eIF2 and the C-terminal domain of human eIF2Bε using affinity chromatography from engineered yeast cells and find that both function in a GEF assay: the first demonstration that this human eIF2Bε domain has GEF function. We show that CACH/VWM mutations within this domain reduce its activity. Finally we demonstrate that the recombinant eIF2 functions similarly to eIF2 purified from rat liver in GEF assays with CACH/VWM eIF2B-mutated patient derived lymphocytic cells.


Introduction
Childhood Ataxia with CNS hypomyelination (CACH) or Vanishing White Matter Leukoencephalopathy (VWM) (OMIM #603896) was described in the 1990s [1,2]. It is a fatal childhood onset white matter disease with a chronic progressive course exacerbated by acute episodes [3,4]. Inherited mutations in any of the five genes encoding subunits of the general protein synthesis initiation factor eIF2B (EIF2B1-5) cause CACH/VWM [3,5]. The subsequent description of a large clinical spectrum of the disease from neonatal to adult onset or even asymptomatic forms led to the concept of eIF2B-related disorders that are recognized by peculiar magnetic resonance imaging (MRI) abnormalities [6]. Well over 100 different, mainly missense, mutations have been presently reported [4]. Their consequences on the eIF2B complex have been demonstrated in yeast and in humans [7,8,9]. Abnormalities in glial cell development have been suggested by studies of patient samples [10,11,12,13,14,15,16] and studies of a mouse model [17]. One recent suggested explanation is that altered expression of splicing regulatory factors in eIF2B mutated glial cells may cause altered splicing regulation of the important myelin proteins PLP and DM20 [18]. However many aspects of the disease are still not understood and no current therapy is available.
eIF2B is well established as a key regulated general translation initiation factor. It functions as a guanine nucleotide exchange factor (GEF) to accelerate the dissociation of GDP from its substrate eIF2NGDP in the first step of translation initiation to form eIF2NGTP [19]. This complex then binds to initiator methionyl tRNA (Met-tRNA i Met ) to deliver it to 40S ribosomes in a reaction that is stimulated by several other translation factors and is required for each translation initiation event on almost all mRNAs [20,21]. During the initiation cycle eIF2-bound GTP is hydrolysed to GDP to inactivate eIF2 and reset the system. eIF2B is regulated by cellular stresses facilitating translational control in a wide variety of settings. Most widely studied is the activation of eIF2a kinases that phosphorylate eIF2 on its alpha subunit at Ser51 [22]. This phosphorylation reaction converts eIF2 from substrate to inhibitor of eIF2B. When phosphorylated at Ser51, eIF2 binds with higher affinity to eIF2B, but without undergoing nucleotide exchange [19]. Genetic and biochemical studies using eIF2B from yeast identified mutations in three subunits of eIF2B (eIF2Babd) that disrupt this regulation and define a potential regulatory interface between eIF2 and eIF2B that is critical for regulation [23]. Subsequent experiments verified that equivalent mutations in mammalian eIF2B also disrupt eIF2B regulation confirming conservation of this regulatory mechanism [24,25]. eIF2B activity is measured by a GEF assay that was first established in the 1980s [26,27]. eIF2 forms a stable complex with GDP in the presence of physiological concentrations of Mg 2+ ions. It requires significant amounts of eIF2 protein purified from rat liver, rabbit reticulocytes or mammalian cell lysates by rounds of conventional column chromatography [27,28]. Pure eIF2 mixed with radiolabelled GDP in the presence of Mg 2+ acts as a substrate and eIF2B is added along with excess unlabelled nucleotide. eIF2B promotes release of labelled GDP and this is assayed by monitoring the progressive decline in labelled eIF2 captured on protein-binding filters. Two versions of the assay have been used; employing either purified eIF2B proteins or extracts from cultured cells. Both formats of the assay have been useful in defining that the largest subunit of eIF2B (eIF2Be) is the only one that possesses GEF activity, and that this activity is enhanced by complex formation with the other four subunits. Catalytic activity has been shown for yeast, drosophila and mammalian proteins [29,30,31]. In addition experiments using purified yeast proteins showed that the C-terminal ,200 amino acids contained all the elements necessary for minimal exchange function. This region was termed the catalytic domain ecat [32,33].
The cell extract format of the assay has been used to assess eIF2B activity in immortalised lymphocytic cells isolated from blood samples of CACH/VWM patients [9,10,34]. This provides a biochemical diagnostic assay to complement MRI and genetic analyses. The adoption of this GEF assay for diagnostic purposes is limited by the availability of purified eIF2 protein. We, and others, have previously developed a yeast cell expression system to overexpress and purify epitope tagged yeast eIF2, which was subsequently used for in vitro studies [29,35]. Generating recombinant systems for mammalian eIF2 has proved challenging because eIF2 possesses three different subunits in a 1:1:1 complex, and because popular expression host cells, including Escherichia coli, appear refractory to expressing significant amounts of the gamma subunit. We therefore decided to develop a recombinant yeast cell system to purify active human eIF2 protein.

Human eIF2 Subunits can Complement the Function of the Equivalent Yeast Gene
The translation initiation factor eIF2 performs critical roles in the initiation and control of protein synthesis in eukaryotic cells. eIF2 is composed of three non-identical subunits and must interact with GDP, GTP, Met-tRNA i Met , eIF3, eIF5, eIF2B and 40S ribosomes to perform its functions [35,36,37] as well several eIF2a protein kinases for regulation. eIF2 has an archaeal homologue, but is not found in eubacteria [38]. eIF2 subunits are highly conserved between yeast and mammals, including humans ( Figure 1A). We decided to develop a yeast system as an expression vehicle for human eIF2 (heIF2). As a first step we obtained cDNA clones and subcloned them into yeast expression vectors, under the control of conditional (galactose carbon source inducible) yeast promoters and bearing short terminal epitope tags ( Figure 1B and Materials and Methods). Each yeast eIF2 gene is essential (SUI2, SUI3, and GCD11 encoding eIF2a-c respectively). Strains bearing individual gene deletions covered by a plasmid copy of the yeast gene were used to assess the function of the human expression clones. Each human vector was introduced into its corresponding yeast deletion strain by transformation and plasmid shuffling was used to evict the covering plasmid. We found that plasmids bearing EIF2S1 encoding heIF2a complemented a sui2D ( Figure 1C, compare lanes 2 and 3 with lane 1) and grew as well as wild type yeast. Similarly plasmids bearing EIF2S2 encoding heIF2b complemented sui3D ( Figure 1C, compare lanes 5 and 6 with lane 4). However our initial EIF2S3 plasmids (encoding heIF2c) could not complement gcd11D (data not shown). As EIF2S3 contains a significant number of codons rarely used in yeast we obtained a commercially synthesised yeast-codonoptimized clone and subcloned this into a similar compatible yeast expression plasmid. This could complement the gcd11D strain, but the resulting strain grew poorly ( Figure 1C, lane 8). Western blotting confirmed both deletion of endogenous yeast genes and expression of the corresponding heIF2 subunits ( Figure 1D). Because eIF2 functions as a heterotrimer, in these complemented cells heterologous eIF2 complexes should form, each with one human and two yeast subunits.
We were concerned by the slow growth of the EIF2S3 complemented strain ( Figure 1C, lane 8), as this may indicate that the codon-optimized cDNA is not fully functional. Several explanations are possible. Firstly, heIF2c may not be expressed at a high enough level to form sufficient eIF2 complexes for rapid growth. Expression of heIF2c did not alter the expression levels of the yeast a and b subunits ( Figure 1D). We observed that different EIF2S3 transformants grew at different rates. When the expression level of heIF2c was examined in a selection of these cells, we consistently found that transformants expressing the highest heIF2c levels grew more slowly than those with lower expression (Figure 2A and data not shown). These results are therefore more consistent with the idea that the slow-growth phenotype is related to excess levels of heIF2c. Excess free heIF2c may bind to and sequester one or more interacting factors into non-or partially functional complexes. This second idea however is unlikely as the slow-growth phenotype is recessive. heIF2c is only slow growing in the absence of yeast GCD11 ( Figure 2B and data not shown).
Because eIF2c is the 'core' subunit that binds to both eIF2a and b [38], excess heIF2c may form some ac or bc complexes and thereby reduce the level of full abc complexes. If so, then reducing excess heIF2c may ameliorate complex disruption and improve growth as observed. Alternatively, because eIF2 must interact with many yeast factors including translation factors (eIF2B, eIF5, eIF3) Met-tRNA i Met and 40S ribosomes, heIF2c may not interact effectively with one or more of these. We assessed whether transforming the heIF2c strain with additional copies of yeast eIF2-interacting factors could complement the growth phenotype by mass action. A panel of high copy plasmids expressing tRNA i Met or combinations of subunits of eIF2B, eIF3 or eIF5 was transformed into the gcd11D strain expressing heIF2c (see Methods for plasmids assessed). Only excess eIF5 reproducibly suppressed the slow-growth phenotype ( Figure 2B and data not shown). Western blotting confirmed that slow-growth suppression in these cells was not caused by reduced heIF2c expression ( Figure 2C). eIF5 interacts with both eIF2b and c [36] as part of the multifactor complex [39], 43S pre-initiation complex and free eIF2 [40,41]. It has GTPase acceleration and GDP-dissociation inhibitor functions [42]. While not conclusive, perhaps impaired contact between the hybrid eIF2 and yeast eIF5 affects eIF2 or eIF5 function and that this could be suppressed by mass action. For example excess eIF5 may prevent its premature loss from initiating ribosomes. Alternatively it may stabilise eIF2bc interactions within the eIF2 complex. Because eIF5 interacts with many initiation factors, several other explanations where the suppressing effect is less direct are also possible to envisage. As the main focus of our study was to generate an expression system to purify active human eIF2, we decided not to investigate these heteromeric yeast-human hybrid complexes further and instead focussed on coexpression of all three human subunits.

Co-expression of All Three heIF2 Subunits Complements a Triple Yeast eIF2 Gene Deletion Strain
Dever and colleagues recently reported construction of a triple eIF2 gene deletion yeast strain complemented by a single plasmid expressing the yeast genes [43]. To assess if co-expression of all three heIF2 subunits could completely replace the yeast factor, we modified this strain to create a trp1 selectable marker (see Materials and Methods) and transformed in two heIF2 plasmids. One plasmid co-expressed both heIF2ab and a second plasmid expressed heIF2c. Plasmid shuffling with FOA generated yeast strains entirely supported by heIF2 ( Figure 3A, lower panel). As expected, growth was carbon source dependent. When expression of heIF2 genes from P GAL promoters was repressed by glucose, heIF2-dependent strains failed to grow ( Figure 3A, upper panel). Western blotting confirmed that the yeast genes had been deleted and that higher expression of heIF2 conferred a slower rate of growth ( Figure 3B). In conclusion, our heIF2 is functional and can replace the yeast protein in vivo.

Purification of Human eIF2 and Catalytic Domains
Because we appended small epitope tags to each heIF2 subunit, we used affinity chromatography to purify the protein complex. We employed a yeast strain deleted for the only yeast eIF2 kinase GCN2. This means that unlike the proteins purified from mammalian sources, our recombinant eIF2 is homogeneously unphosphorylated at the key regulatory site, Ser 51 of the alpha subunit. Purification by a single step using Flag M2 affinity resin to bind heIF2a, or a single nickel agarose step to bind heIF2b was not sufficient to purify heIF2 ( Figure 4A). The single step Flag affinity purification recovered a mixture of of abc trimers, excess eIF2a and residual contaminating proteins ( Figure 4A lanes 2-7). Similarly a single-step his 6 purification captured a mixture of abc trimers and excess eIF2b ( Figure 4A, lanes 8-12 and 4B, lanes 2-6). These results are consistent with the known structure of the archaeal homologue of eIF2, aIF2abc where both aIF2a and b each separately bind to aIF2c [38] and with the idea that expression of heIF2c in our cells is limiting, so that excess free heIF2a and b subunits are formed. We therefore adopted a twostep purification strategy employing nickel agarose, followed by Flag resin ( Figure 4B) to purify heterotrimeric eIF2 free from excess a and b subunits ( Figure 4B, lane 11). Because eIF2c expression levels were limiting in our system we transformed in a second plasmid to boost the amounts of the full eIF2 complex. With this strategy we obtained eIF2 that was approximately 90% pure ( Figure 4C).
We wished to assess nucleotide binding and exchange with our heIF2 protein. The nucleotide exchange factor eIF2B is specific for eIF2. eIF2B is composed of five distinct subunits. However previous work has shown that the largest subunit alone retains catalytic function and studies in yeast showed that the carboxy terminal domain alone (yeast residues 518-712) is the catalytic domain, termed ecat [32]. Subsequently, deletion of residues 549-596 within human eIF2Be resulted in a protein that could form eIF2B complexes, but had no GEF activity in vitro showing this region is critical for human eIF2B GEF function [8]. In addition X-ray crystallographic structure determination has shown that the human equivalent domain adopts the same stacked, paired ahelical structure as the yeast ecat [33,44]. We therefore predicted that a construct bearing the equivalent residues of human eIF2Be, residues 533-721 would comprise the human eIF2B catalytic domain (h2Becat). A yeast codon-optimized h2Becat cDNA was synthesized and expressed with tandem Flag and polyhistidine tags from a galactose inducible promoter vector in a suitable yeast strain host and the same purification scheme devised for heIF2 was used ( Figure 4C). To assess the functions of purified heIF2 and h2Becat, we set up a standard nucleotide exchange assay with radiolabelled GDP (see Materials and Methods). In the absence of eIF2B, but in the presence of excess unlabelled GDP, heIF2 bound [ 3 H]GDP at 10-30uC with MgCl 2 from 0.2-2 mM. This indicates that the nucleotide binding function of heIF2 is intact. When h2Becat was added, this stimulated release of [ 3 H]GDP with first order kinetics ( Figure 4C). This confirms that nucleotide binding to eIF2 is reversible and that residues 533-721 of human eIF2Be contains the catalytic domain.

Catalytic Domain CACH/VWM Mutants have Reduced GEF Activity
As stated in the introduction, mutations in eIF2B cause a genetic disease. Mutations have been found in all five subunits, including several within the catalytic domain. Catalytic domain mutations include missense alleles: P604S [45], I649T [46], E650K [47], M608I [48] and W684S [49] and small deletions: D610-613 [9] D666-672 [47]. In all patients reported thus far, these alleles occur as compound heterozygotes with other mutations contributing to the disease pathology and the measured eIF2B activity from patient cells. Using a mammalian cell expression system, the Proud laboratory has pioneered in vitro biochemical analyses of eIF2B CACH/VWM mutants. Typically HEK293 cells are transformed with vectors overexpressing all five subunits of eIF2B, which are affinity purified and the resulting complexes analysed for complex formation and GEF activity. For example when analysed as part of 5-subunit complex W628R reduced activity to ,20% [8], I649T to ,40% and E650K to ,30% of wild type [49]. The latter two also apparently reduced association with the alpha (and possibly delta) subunits of eIF2B [49] indicating possible reduced complex integrity that could contribute to reduced activity. The D610-613 mutant dramatically reduced expression or subunit stability and activity to 25% [9].
To further test our recombinant heIF2 and h2Becat, we introduced several mutations into our h2Becat expression vector and purified these to homogeneity ( Figure 5). We included mutations that had been analysed as part of 5-subunit eIF2B complexes (W628R, I649T and E650K) as well as mutations not analysed previously (P604S and D666-672). All mutant forms significantly reduced GEF activity ( Figure 5), validating that our heIF2 and h2Becat behave as expected. It was perhaps surprising that the D666-672 mutant had only a modest reduction in activity, when compared with missense mutations that might be predicted to have a smaller impact on the overall structure. Examining the locations of the mutated residues on the human catalytic domain structure [33,44] suggests that most of the missense mutations affect residues predominantly buried internally within the structure. Macromolecular modelling of the impact of the D666-672 mutant, suggests that the normal domain fold can be adopted, but that there is local disruption of part of one a-helix only (data not shown). The affected helix does not contribute to the surface regions of ecat identified previously as critical for direct eIF2 binding and activity [50]. This observation appears to fit with the modest reduction in GEF activity observed for h2Becat, but does not rule out that it may have a more significant defect in vivo in the context of the full five subunit protein complex.

heIF2 Functions in GEF Assays with Extracts from Patientderived Cells
Extracts from patient-derived cells have been used as a source of eIF2B to provide biochemical diagnosis for eIF2B related disorders [10,34]. To assess whether our heIF2 would perform as a substrate in GEF assays with patient cells, we performed assays with lymphoblastoid cell line extracts from 18 CACH/VWM patients and compared the results to those obtained previously with the rat eIF2 substrate classically used for this diagnostic assay (Table 1). Our panel of cell lines included mutations in four of the five eIF2B subunits and those previously shown to have a range of disease severity from severe, early onset to classical/mild forms of disease [34].
Firstly, we observed a decreased GEF activity (#87.2%) using the heIF2 substrate for the patient cell lines which exhibited a decreased GEF activity under the previously described diagnostic threshold (77.5%) using rat eIF2 as substrate [34]. The GEF activity values using the heIF2 substrate are marginally higher and show more variability (with a mean SD = 9.1% using heIF2 in comparison to 4.8% using rat eIF2 with a correlation factor between the two assays of r = 0.78. The disease-diagnostic threshold used with this substrate should therefore be revised from 77.5% to 89.9%. Using this new value, the diagnostic impact of the GEF assay is identical to the assay using rat eIF2: the patients with GEF activity ,77.5% with rat eIF2 exacerbated also a GEF activity ,89.9% with yeast heIF2 (Table 1). There are several possible reasons that may explain the observed minor difference between the two sources of eIF2. Firstly, there are minor sequence variations between the two species. Secondly there may be differences in post-translational modifications between the two preparations. For example, we assume that eIF2 from rat liver will be partially phosphorylated at ser51 by endogenous eIF2 alpha kinases, while heIF2 from yeast is unphosphorylated. Phosphorylated eIF2 inhibits eIF2B reducing the observed activity. Specific mutations in the a, b and d subunits alter the sensitivity of eIF2B to inhibition by phosphorylated eIF2 [23,24,25]. However it is not yet known whether disease-causing mutations alter eIF2B sensitivity to eIF2 phosphorylation.
Cells from two patients tested here (432-1 and 1074-1) exhibited normal GEF activity using the rat eIF2 substrate. Here lower GEF activity was found with the heIF2 substrate, taking patient 432-1 just below the proposed diagnostic threshold (Table 1). A recent report also identified two severe CACH/VWM patients with eIF2B mutations, but no apparent defect in eIF2B activity [49]. However the eIF2B GEF activity for those patients was measured only in primary fibroblasts. Further studies are needed to assess the diagnostic value of measuring eIF2B GEF activity in fibroblasts. Discrepancies have been previously reported between measuring eIF2B defects in immortalised lymphocytic cells compared to primary fibroblasts [51]. In summary, the recombinant heIF2 performs as well as the previously used substrate in this assay and measurement of reduced GEF activity with heIF2 and extracts from immortalised lymphocytic cells is diagnostic for eIF2Bmediated disease.
In conclusion, we describe a strategy for production of heterotrimeric recombinant human eIF2. We show that it can functionally replace yeast eIF2 in vivo and demonstrate its utility as a diagnostic tool for measuring the impact of eIF2B mutations that cause human disease. Employing recombinant heIF2 from yeast will remove the need to use animal tissue sources to obtain purified eIF2 for these purposes.

Plasmid Constructions
Human eIF2 cDNA expression. Human cDNAs encoding EIF2S1 (eIF2a) and EIF2S2 (eIF2b) were cloned with N-terminal Flag or his6 epitope tags respectively into the pBEVY series of plasmids downstream of strong yeast promoters. These constructs failed to express any protein and on resequencing had acquired mutations within the promoter sequences. As the ORFs were intact, they were subcloned into pBEVY-GT, a multicopy yeast expression plasmid bearing divergent promoters from the GAL1 and GAL10 genes (P GAL1 and P GAL10 ) that we have recently used with success elsewhere [52,53]. The cDNAs were cloned singly as Kpn1-EcoRI fragments downstream of P GAL1 creating pAV1901 (EIF2S2-his 6 ) and pAV1907 (EIF2S1-Flag). In addition EIF2S1-Flag was cloned downstream of P GAL10 as a BamHI-Pst1 fragment in pAV1901 to create pAV1905 (EIF2S2-his 6 and EIF2S1-Flag). EIF2S3 (eIF2c) cDNA was codon-optimized and synthesized (GeneScript USA Inc.) and subcloned as a Xba1-Sal1 fragment into pBEVY-GL creating pAV1970 and pBEVY-GU creating pAV1974. All constructs were confirmed by DNA sequence analysis. Further details are shown in Table 2.  Human eIF2Be catalytic domain. The EIF2B5 (eIF2Be) catalytic domain (amino acids 533-721, termed here h2Becat) was identified by sequence alignment of translated ORF from Genbank (NG_015826) with the yeast Saccharomyces cerevisie Gcd6p catalytic domain [32,33]. The cDNA was codon optimised and custom synthesized (GeneScript USA Inc) with the 39UTR of GCD6. This was subcloned using MluI and BamHI into plasmid pAV1427 in frame with 5' tandem Flag and his 6 tags downstream of the P GAL1 promoter creating plasmid pAV2075. Quikchange (Agilent) site-directed mutagenesis kit was used to introduce mutations into the GeneScript clone prior to subcloning into pAV1427 to create the plasmids pAV2095-99, each identical to pAV1075 except for the site-directed changes shown in Table 2. All constructs were confirmed by DNA sequence analysis.
Activity measured with extracts from patient lymphoblastoid cell lines (lymphoblasts) as a source of eIF2B was performed with heIF2 as described previously for eIF2 purified from rat liver [7,25] with the following modifications: the use of 1 mCi [ 3 H]GDP (4.5 Ci mmol 21 ) for eIF2N[ 3 H]GDP binary complex formation, incubation of this mixture at 30uC for 30 min (instead of 10 min), and the [ 3 H]GDP dissociation kinetics was monitored every 5 min (instead of every 2 min: from 0 to 15 min). Such comparative analyses were performed at least in triplicate for cells from 18 patients and matched controls.